Snap for 11510257 from b73e8ce9721adac2d07a9e0d844ab767b34eeb0e to simpleperf-release

Change-Id: Ie227a5926aaaffaa1455765c3d62afad69f329b5

198 files changed, 7017 insertions, 2187 deletions

diff --git a/TEST_MAPPING b/TEST_MAPPING
index d7bde8792d..038aaf9ef3 100644
--- a/TEST_MAPPING
+++ b/TEST_MAPPING
@@ -494,6 +494,15 @@
       "name": "art-run-test-2265-checker-select-binary-unary[com.google.android.art.apex]"
     },
     {
+      "name": "art-run-test-2266-checker-remove-empty-ifs[com.google.android.art.apex]"
+    },
+    {
+      "name": "art-run-test-2268-checker-remove-dead-phis[com.google.android.art.apex]"
+    },
+    {
+      "name": "art-run-test-2269-checker-constant-folding-instrinsics[com.google.android.art.apex]"
+    },
+    {
       "name": "art-run-test-300-package-override[com.google.android.art.apex]"
     },
     {
@@ -1328,6 +1337,12 @@
       "name": "art-run-test-843-default-interface[com.google.android.art.apex]"
     },
     {
+      "name": "art-run-test-851-null-instanceof[com.google.android.art.apex]"
+    },
+    {
+      "name": "art-run-test-853-checker-inlining[com.google.android.art.apex]"
+    },
+    {
       "name": "art-run-test-963-default-range-smali[com.google.android.art.apex]"
     },
     {
@@ -1922,6 +1937,15 @@
       "name": "art-run-test-2265-checker-select-binary-unary"
     },
     {
+      "name": "art-run-test-2266-checker-remove-empty-ifs"
+    },
+    {
+      "name": "art-run-test-2268-checker-remove-dead-phis"
+    },
+    {
+      "name": "art-run-test-2269-checker-constant-folding-instrinsics"
+    },
+    {
       "name": "art-run-test-300-package-override"
     },
     {
@@ -2756,6 +2780,12 @@
       "name": "art-run-test-843-default-interface"
     },
     {
+      "name": "art-run-test-851-null-instanceof"
+    },
+    {
+      "name": "art-run-test-853-checker-inlining"
+    },
+    {
       "name": "art-run-test-963-default-range-smali"
     },
     {
@@ -3337,6 +3367,15 @@
       "name": "art-run-test-2265-checker-select-binary-unary"
     },
     {
+      "name": "art-run-test-2266-checker-remove-empty-ifs"
+    },
+    {
+      "name": "art-run-test-2268-checker-remove-dead-phis"
+    },
+    {
+      "name": "art-run-test-2269-checker-constant-folding-instrinsics"
+    },
+    {
       "name": "art-run-test-300-package-override"
     },
     {
@@ -4171,6 +4210,12 @@
       "name": "art-run-test-843-default-interface"
     },
     {
+      "name": "art-run-test-851-null-instanceof"
+    },
+    {
+      "name": "art-run-test-853-checker-inlining"
+    },
+    {
       "name": "art-run-test-963-default-range-smali"
     },
     {
@@ -4253,19 +4298,13 @@
   ],
   "postsubmit": [
     {
-      "name": "art-run-test-2266-checker-remove-empty-ifs"
-    },
-    {
-      "name": "art-run-test-2268-checker-remove-dead-phis"
-    },
-    {
-      "name": "art-run-test-2269-checker-constant-folding-instrinsics"
+      "name": "art-run-test-2247-checker-write-barrier-elimination"
     },
     {
-      "name": "art-run-test-851-null-instanceof"
+      "name": "art-run-test-2273-checker-unreachable-intrinsics"
     },
     {
-      "name": "art-run-test-853-checker-inlining"
+      "name": "art_standalone_dexopt_chroot_setup_tests"
     }
   ]
 }
diff --git a/artd/artd_test.cc b/artd/artd_test.cc
index 79fa7cd4e1..8a704464ea 100644
--- a/artd/artd_test.cc
+++ b/artd/artd_test.cc
@@ -2200,7 +2200,7 @@ TEST_F(ArtdTest, cleanup) {
               },
               {
                   RuntimeArtifactsPath{
-                      .packageName = "com.android.foo", .isa = "arm64", .dexPath = "/a/b/base.apk"},
+                      .packageName = "com.android.foo", .dexPath = "/a/b/base.apk", .isa = "arm64"},
               },
               &aidl_return)
           .isOk());
diff --git a/build/README.md b/build/README.md
index 2c8413075d..7b4e402376 100644
--- a/build/README.md
+++ b/build/README.md
@@ -8,10 +8,6 @@ directory) of platform releases, to ensure it is always available.
 The recommended way to build the ART Module is to use the `master-art` manifest,
 which only has the sources and dependencies required for the module.
 
-Currently it is also possible to build ART directly from sources in a platform
-build, i.e. as has been the traditional way up until Android S. However that
-method is being phased out.
-
 The ART Module is available as a debug variant, `com.android.art.debug.apex`,
 which has extra internal consistency checks enabled, and some debug tools. A
 device cannot have both the non-debug and debug variants installed at once - it
diff --git a/build/boot/Android.bp b/build/boot/Android.bp
index e0bf395872..a2fa55ca42 100644
--- a/build/boot/Android.bp
+++ b/build/boot/Android.bp
@@ -142,6 +142,7 @@ bootclasspath_fragment {
         // result in a build failure due to inconsistent flags.
         package_prefixes: [
             "android.compat",
+            "android.crypto.hpke",
             "com.android.okhttp",
             "com.android.org.bouncycastle",
             "com.android.org.kxml2",
diff --git a/build/makevars.go b/build/makevars.go
index b35ee8561d..00124ec621 100644
--- a/build/makevars.go
+++ b/build/makevars.go
@@ -56,7 +56,7 @@ func makeVarsProvider(ctx android.MakeVarsContext) {
 	sort.Strings(testNames)
 
 	for _, name := range testNames {
-		ctx.Strict("ART_TEST_LIST_"+name, strings.Join(testMap[name], " "))
+		ctx.Strict("ART_TEST_LIST_"+name, strings.Join(android.FirstUniqueStrings(testMap[name]), " "))
 	}
 
 	// Create list of copy commands to install the content of the testcases directory.
diff --git a/compiler/compiler_reflection_test.cc b/compiler/compiler_reflection_test.cc
index d8e2b9e5b9..f3c07db136 100644
--- a/compiler/compiler_reflection_test.cc
+++ b/compiler/compiler_reflection_test.cc
@@ -29,7 +29,6 @@ namespace art HIDDEN {
 class CompilerReflectionTest : public CommonCompilerTest {};
 
 TEST_F(CompilerReflectionTest, StaticMainMethod) {
-  TEST_DISABLED_FOR_RISCV64();
   ScopedObjectAccess soa(Thread::Current());
   jobject jclass_loader = LoadDex("Main");
   StackHandleScope<1> hs(soa.Self());
diff --git a/compiler/driver/compiler_options.h b/compiler/driver/compiler_options.h
index eaf9cc2fdb..6b3e26fc70 100644
--- a/compiler/driver/compiler_options.h
+++ b/compiler/driver/compiler_options.h
@@ -69,7 +69,7 @@ class CompilerOptions final {
   static constexpr size_t kUnsetInlineMaxCodeUnits = -1;
   // We set a lower inlining threshold for baseline to reduce code size and compilation time. This
   // cannot be changed via flags.
-  static constexpr size_t kBaselineInlineMaxCodeUnits = 8;
+  static constexpr size_t kBaselineInlineMaxCodeUnits = 14;
 
   enum class CompilerType : uint8_t {
     kAotCompiler,             // AOT compiler.
diff --git a/compiler/jni/quick/calling_convention.h b/compiler/jni/quick/calling_convention.h
index 2657e943e6..b8b4cc14b1 100644
--- a/compiler/jni/quick/calling_convention.h
+++ b/compiler/jni/quick/calling_convention.h
@@ -324,9 +324,11 @@ class JniCallingConvention : public CallingConvention {
   virtual ArrayRef<const ManagedRegister> CalleeSaveRegisters() const = 0;
 
   // Subset of core callee save registers that can be used for arbitrary purposes after
-  // constructing the JNI transition frame. These should be managed callee-saves as well.
+  // constructing the JNI transition frame. These should be both managed and native callee-saves.
   // These should not include special purpose registers such as thread register.
-  // JNI compiler currently requires at least 3 callee save scratch registers.
+  // JNI compiler currently requires at least 4 callee save scratch registers, except for x86
+  // where we have only 3 such registers but all args are passed on stack, so the method register
+  // is never clobbered by argument moves and does not need to be preserved elsewhere.
   virtual ArrayRef<const ManagedRegister> CalleeSaveScratchRegisters() const = 0;
 
   // Subset of core argument registers that can be used for arbitrary purposes after
@@ -356,10 +358,6 @@ class JniCallingConvention : public CallingConvention {
 
   virtual ~JniCallingConvention() {}
 
-  static constexpr size_t SavedLocalReferenceCookieSize() {
-    return 4u;
-  }
-
   bool IsFastNative() const {
     return is_fast_native_;
   }
diff --git a/compiler/jni/quick/jni_compiler.cc b/compiler/jni/quick/jni_compiler.cc
index 3357a5f8d7..c721825683 100644
--- a/compiler/jni/quick/jni_compiler.cc
+++ b/compiler/jni/quick/jni_compiler.cc
@@ -38,6 +38,7 @@
 #include "entrypoints/quick/quick_entrypoints.h"
 #include "instrumentation.h"
 #include "jni/jni_env_ext.h"
+#include "jni/local_reference_table.h"
 #include "runtime.h"
 #include "thread.h"
 #include "utils/arm/managed_register_arm.h"
@@ -51,19 +52,6 @@
 
 namespace art HIDDEN {
 
-constexpr size_t kIRTCookieSize = JniCallingConvention::SavedLocalReferenceCookieSize();
-
-template <PointerSize kPointerSize>
-static void PushLocalReferenceFrame(JNIMacroAssembler<kPointerSize>* jni_asm,
-                                    ManagedRegister jni_env_reg,
-                                    ManagedRegister saved_cookie_reg,
-                                    ManagedRegister temp_reg);
-template <PointerSize kPointerSize>
-static void PopLocalReferenceFrame(JNIMacroAssembler<kPointerSize>* jni_asm,
-                                   ManagedRegister jni_env_reg,
-                                   ManagedRegister saved_cookie_reg,
-                                   ManagedRegister temp_reg);
-
 template <PointerSize kPointerSize>
 static void SetNativeParameter(JNIMacroAssembler<kPointerSize>* jni_asm,
                                JniCallingConvention* jni_conv,
@@ -306,22 +294,30 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
   // 3. Push local reference frame.
   // Skip this for @CriticalNative methods, they cannot use any references.
   ManagedRegister jni_env_reg = ManagedRegister::NoRegister();
-  ManagedRegister saved_cookie_reg = ManagedRegister::NoRegister();
+  ManagedRegister previous_state_reg = ManagedRegister::NoRegister();
+  ManagedRegister current_state_reg = ManagedRegister::NoRegister();
   ManagedRegister callee_save_temp = ManagedRegister::NoRegister();
   if (LIKELY(!is_critical_native)) {
     // To pop the local reference frame later, we shall need the JNI environment pointer
     // as well as the cookie, so we preserve them across calls in callee-save registers.
     CHECK_GE(callee_save_scratch_regs.size(), 3u);  // At least 3 for each supported architecture.
     jni_env_reg = callee_save_scratch_regs[0];
-    saved_cookie_reg = __ CoreRegisterWithSize(callee_save_scratch_regs[1], kIRTCookieSize);
-    callee_save_temp = __ CoreRegisterWithSize(callee_save_scratch_regs[2], kIRTCookieSize);
+    constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+    previous_state_reg = __ CoreRegisterWithSize(callee_save_scratch_regs[1], kLRTSegmentStateSize);
+    current_state_reg = __ CoreRegisterWithSize(callee_save_scratch_regs[2], kLRTSegmentStateSize);
+    if (callee_save_scratch_regs.size() >= 4) {
+      callee_save_temp = callee_save_scratch_regs[3];
+    }
+    const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kPointerSize);
 
     // Load the JNI environment pointer.
     __ LoadRawPtrFromThread(jni_env_reg, Thread::JniEnvOffset<kPointerSize>());
 
-    // Push the local reference frame.
-    PushLocalReferenceFrame<kPointerSize>(
-        jni_asm.get(), jni_env_reg, saved_cookie_reg, callee_save_temp);
+    // Load the local reference frame states.
+    __ LoadLocalReferenceTableStates(jni_env_reg, previous_state_reg, current_state_reg);
+
+    // Store the current state as the previous state (push the LRT frame).
+    __ Store(jni_env_reg, previous_state_offset, current_state_reg, kLRTSegmentStateSize);
   }
 
   // 4. Make the main native call.
@@ -344,6 +340,7 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
   refs.clear();
   mr_conv->ResetIterator(FrameOffset(current_frame_size));
   main_jni_conv->ResetIterator(FrameOffset(main_out_arg_size));
+  bool check_method_not_clobbered = false;
   if (UNLIKELY(is_critical_native)) {
     // Move the method pointer to the hidden argument register.
     // TODO: Pass this as the last argument, not first. Change ARM assembler
@@ -354,10 +351,16 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
   } else {
     main_jni_conv->Next();    // Skip JNIEnv*.
     FrameOffset method_offset(current_out_arg_size + mr_conv->MethodStackOffset().SizeValue());
-    if (!is_static || main_jni_conv->IsCurrentParamOnStack()) {
+    if (main_jni_conv->IsCurrentParamOnStack()) {
+      // This is for x86 only. The method shall not be clobbered by argument moves
+      // because all arguments are passed on the stack to the native method.
+      check_method_not_clobbered = true;
+      DCHECK(callee_save_temp.IsNoRegister());
+    } else if (!is_static) {
       // The method shall not be available in the `jclass` argument register.
       // Make sure it is available in `callee_save_temp` for the call below.
       // (The old method register can be clobbered by argument moves.)
+      DCHECK(!callee_save_temp.IsNoRegister());
       ManagedRegister new_method_reg = __ CoreRegisterWithSize(callee_save_temp, kRawPointerSize);
       DCHECK(!method_register.IsNoRegister());
       __ Move(new_method_reg, method_register, kRawPointerSize);
@@ -399,6 +402,10 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
     refs.push_back(is_reference ? mr_conv->CurrentParamStackOffset() : kInvalidReferenceOffset);
   }
   DCHECK(!main_jni_conv->HasNext());
+  DCHECK_IMPLIES(check_method_not_clobbered,
+                 std::all_of(dest_args.begin(),
+                             dest_args.end(),
+                             [](const ArgumentLocation& loc) { return !loc.IsRegister(); }));
   __ MoveArguments(ArrayRef<ArgumentLocation>(dest_args),
                    ArrayRef<ArgumentLocation>(src_args),
                    ArrayRef<FrameOffset>(refs));
@@ -525,8 +532,10 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
 
   // 6. Pop local reference frame.
   if (LIKELY(!is_critical_native)) {
-    PopLocalReferenceFrame<kPointerSize>(
-        jni_asm.get(), jni_env_reg, saved_cookie_reg, callee_save_temp);
+    __ StoreLocalReferenceTableStates(jni_env_reg, previous_state_reg, current_state_reg);
+    // For x86, the `callee_save_temp` is not valid, so let's simply change it to one
+    // of the callee save registers that we don't need anymore for all architectures.
+    callee_save_temp = current_state_reg;
   }
 
   // 7. Return from the JNI stub.
@@ -658,8 +667,7 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
         jni_asm->cfi().AdjustCFAOffset(main_out_arg_size);
         __ DecreaseFrameSize(main_out_arg_size);
       }
-      PopLocalReferenceFrame<kPointerSize>(
-          jni_asm.get(), jni_env_reg, saved_cookie_reg, callee_save_temp);
+      __ StoreLocalReferenceTableStates(jni_env_reg, previous_state_reg, current_state_reg);
     }
     DCHECK_EQ(jni_asm->cfi().GetCurrentCFAOffset(), static_cast<int>(current_frame_size));
     __ DeliverPendingException();
@@ -736,40 +744,6 @@ static JniCompiledMethod ArtJniCompileMethodInternal(const CompilerOptions& comp
 }
 
 template <PointerSize kPointerSize>
-static void PushLocalReferenceFrame(JNIMacroAssembler<kPointerSize>* jni_asm,
-                                    ManagedRegister jni_env_reg,
-                                    ManagedRegister saved_cookie_reg,
-                                    ManagedRegister temp_reg) {
-  const size_t kRawPointerSize = static_cast<size_t>(kPointerSize);
-  const MemberOffset jni_env_cookie_offset = JNIEnvExt::LocalRefCookieOffset(kRawPointerSize);
-  const MemberOffset jni_env_segment_state_offset = JNIEnvExt::SegmentStateOffset(kRawPointerSize);
-
-  // Load the old cookie that we shall need to restore.
-  __ Load(saved_cookie_reg, jni_env_reg, jni_env_cookie_offset, kIRTCookieSize);
-
-  // Set the cookie in JNI environment to the current segment state.
-  __ Load(temp_reg, jni_env_reg, jni_env_segment_state_offset, kIRTCookieSize);
-  __ Store(jni_env_reg, jni_env_cookie_offset, temp_reg, kIRTCookieSize);
-}
-
-template <PointerSize kPointerSize>
-static void PopLocalReferenceFrame(JNIMacroAssembler<kPointerSize>* jni_asm,
-                                   ManagedRegister jni_env_reg,
-                                   ManagedRegister saved_cookie_reg,
-                                   ManagedRegister temp_reg) {
-  const size_t kRawPointerSize = static_cast<size_t>(kPointerSize);
-  const MemberOffset jni_env_cookie_offset = JNIEnvExt::LocalRefCookieOffset(kRawPointerSize);
-  const MemberOffset jni_env_segment_state_offset = JNIEnvExt::SegmentStateOffset(kRawPointerSize);
-
-  // Set the current segment state to the current cookie in JNI environment.
-  __ Load(temp_reg, jni_env_reg, jni_env_cookie_offset, kIRTCookieSize);
-  __ Store(jni_env_reg, jni_env_segment_state_offset, temp_reg, kIRTCookieSize);
-
-  // Restore the cookie in JNI environment to the saved value.
-  __ Store(jni_env_reg, jni_env_cookie_offset, saved_cookie_reg, kIRTCookieSize);
-}
-
-template <PointerSize kPointerSize>
 static void SetNativeParameter(JNIMacroAssembler<kPointerSize>* jni_asm,
                                JniCallingConvention* jni_conv,
                                ManagedRegister in_reg) {
diff --git a/compiler/optimizing/code_generator_arm64.cc b/compiler/optimizing/code_generator_arm64.cc
index 5ba26b4754..50b86489ee 100644
--- a/compiler/optimizing/code_generator_arm64.cc
+++ b/compiler/optimizing/code_generator_arm64.cc
@@ -37,6 +37,7 @@
 #include "intrinsics_arm64.h"
 #include "intrinsics_list.h"
 #include "intrinsics_utils.h"
+#include "jit/profiling_info.h"
 #include "linker/linker_patch.h"
 #include "lock_word.h"
 #include "mirror/array-inl.h"
@@ -1306,7 +1307,9 @@ void CodeGeneratorARM64::MaybeIncrementHotness(HSuspendCheck* suspend_check, boo
     __ Bind(&done);
   }
 
-  if (GetGraph()->IsCompilingBaseline() && !Runtime::Current()->IsAotCompiler()) {
+  if (GetGraph()->IsCompilingBaseline() &&
+      GetGraph()->IsUsefulOptimizing() &&
+      !Runtime::Current()->IsAotCompiler()) {
     ProfilingInfo* info = GetGraph()->GetProfilingInfo();
     DCHECK(info != nullptr);
     DCHECK(!HasEmptyFrame());
@@ -2353,14 +2356,12 @@ void InstructionCodeGeneratorARM64::HandleFieldSet(HInstruction* instruction,
       codegen_->StoreNeedsWriteBarrier(field_type, instruction->InputAt(1), write_barrier_kind);
 
   if (needs_write_barrier) {
-    // TODO(solanes): If we do a `HuntForOriginalReference` call to the value in WBE, we will be
-    // able to DCHECK that the write_barrier_kind is kBeingReliedOn when Register(value).IsZero(),
-    // and we could remove the `!Register(value).IsZero()` from below.
-    codegen_->MaybeMarkGCCard(obj,
-                              Register(value),
-                              value_can_be_null &&
-                                  write_barrier_kind == WriteBarrierKind::kEmitNotBeingReliedOn &&
-                                  !Register(value).IsZero());
+    DCHECK_IMPLIES(Register(value).IsZero(),
+                   write_barrier_kind == WriteBarrierKind::kEmitBeingReliedOn);
+    codegen_->MaybeMarkGCCard(
+        obj,
+        Register(value),
+        value_can_be_null && write_barrier_kind == WriteBarrierKind::kEmitNotBeingReliedOn);
   } else if (codegen_->ShouldCheckGCCard(field_type, instruction->InputAt(1), write_barrier_kind)) {
     codegen_->CheckGCCardIsValid(obj);
   }
@@ -3049,9 +3050,8 @@ void InstructionCodeGeneratorARM64::VisitArraySet(HArraySet* instruction) {
     // write barrier when its value is null (without an extra cbz since we already checked if the
     // value is null for the type check). This will be done as a follow-up since it is a runtime
     // optimization that needs extra care.
-    // TODO(solanes): We can also skip it for known zero values which are not relied on i.e. when
-    // we have the Zero register as the value. If we do `HuntForOriginalReference` on the value
-    // we'll resolve this.
+    DCHECK_IMPLIES(Register(value).IsZero(),
+                   write_barrier_kind == WriteBarrierKind::kEmitBeingReliedOn);
     codegen_->MarkGCCard(array);
 
     UseScratchRegisterScope temps(masm);
diff --git a/compiler/optimizing/code_generator_arm_vixl.cc b/compiler/optimizing/code_generator_arm_vixl.cc
index 85f61f5303..0ed3f8b2c9 100644
--- a/compiler/optimizing/code_generator_arm_vixl.cc
+++ b/compiler/optimizing/code_generator_arm_vixl.cc
@@ -36,6 +36,7 @@
 #include "intrinsics_arm_vixl.h"
 #include "intrinsics_list.h"
 #include "intrinsics_utils.h"
+#include "jit/profiling_info.h"
 #include "linker/linker_patch.h"
 #include "mirror/array-inl.h"
 #include "mirror/class-inl.h"
@@ -2302,7 +2303,9 @@ void CodeGeneratorARMVIXL::MaybeIncrementHotness(HSuspendCheck* suspend_check,
     }
   }
 
-  if (GetGraph()->IsCompilingBaseline() && !Runtime::Current()->IsAotCompiler()) {
+  if (GetGraph()->IsCompilingBaseline() &&
+      GetGraph()->IsUsefulOptimizing() &&
+      !Runtime::Current()->IsAotCompiler()) {
     ProfilingInfo* info = GetGraph()->GetProfilingInfo();
     DCHECK(info != nullptr);
     DCHECK(!HasEmptyFrame());
diff --git a/compiler/optimizing/code_generator_riscv64.cc b/compiler/optimizing/code_generator_riscv64.cc
index ed57683e0a..abbd74ac65 100644
--- a/compiler/optimizing/code_generator_riscv64.cc
+++ b/compiler/optimizing/code_generator_riscv64.cc
@@ -2518,8 +2518,7 @@ void InstructionCodeGeneratorRISCV64::HandleFieldSet(HInstruction* instruction,
       codegen_->StoreNeedsWriteBarrier(type, instruction->InputAt(1), write_barrier_kind);
   if (needs_write_barrier) {
     if (value.IsConstant()) {
-      // TODO(solanes): If we do a `HuntForOriginalReference` call to the value in WBE, we will be
-      // able to DCHECK that the write_barrier_kind is kBeingReliedOn.
+      DCHECK_EQ(write_barrier_kind, WriteBarrierKind::kEmitBeingReliedOn);
       codegen_->MarkGCCard(obj);
     } else {
       codegen_->MaybeMarkGCCard(
@@ -2947,9 +2946,8 @@ void InstructionCodeGeneratorRISCV64::VisitArraySet(HArraySet* instruction) {
 
   if (needs_write_barrier) {
     DCHECK_EQ(value_type, DataType::Type::kReference);
+    DCHECK_IMPLIES(value.IsConstant(), value.GetConstant()->IsArithmeticZero());
     const bool storing_constant_zero = value.IsConstant();
-    // TODO(solanes): If we do a `HuntForOriginalReference` call to the value in WBE, we will be
-    // able to DCHECK that the write_barrier_kind is kBeingReliedOn when we have a constant.
     if (!storing_constant_zero) {
       Riscv64Label do_store;
 
@@ -3010,11 +3008,13 @@ void InstructionCodeGeneratorRISCV64::VisitArraySet(HArraySet* instruction) {
       }
     }
 
-    DCHECK_NE(instruction->GetWriteBarrierKind(), WriteBarrierKind::kDontEmit);
+    DCHECK_NE(write_barrier_kind, WriteBarrierKind::kDontEmit);
     // TODO(solanes): The WriteBarrierKind::kEmitNotBeingReliedOn case should be able to skip
     // this write barrier when its value is null (without an extra Beqz since we already checked
     // if the value is null for the type check). This will be done as a follow-up since it is a
     // runtime optimization that needs extra care.
+    DCHECK_IMPLIES(storing_constant_zero,
+                   write_barrier_kind == WriteBarrierKind::kEmitBeingReliedOn);
     codegen_->MarkGCCard(array);
   } else if (codegen_->ShouldCheckGCCard(value_type, instruction->GetValue(), write_barrier_kind)) {
     codegen_->CheckGCCardIsValid(array);
@@ -5763,7 +5763,9 @@ void CodeGeneratorRISCV64::MaybeIncrementHotness(HSuspendCheck* suspend_check,
     __ Bind(&done);
   }
 
-  if (GetGraph()->IsCompilingBaseline() && !Runtime::Current()->IsAotCompiler()) {
+  if (GetGraph()->IsCompilingBaseline() &&
+      GetGraph()->IsUsefulOptimizing() &&
+      !Runtime::Current()->IsAotCompiler()) {
     ProfilingInfo* info = GetGraph()->GetProfilingInfo();
     DCHECK(info != nullptr);
     DCHECK(!HasEmptyFrame());
diff --git a/compiler/optimizing/code_generator_x86.cc b/compiler/optimizing/code_generator_x86.cc
index a61dca3022..21d3492e8a 100644
--- a/compiler/optimizing/code_generator_x86.cc
+++ b/compiler/optimizing/code_generator_x86.cc
@@ -1357,9 +1357,9 @@ void CodeGeneratorX86::MaybeIncrementHotness(HSuspendCheck* suspend_check, bool
     }
   }
 
-  if (GetGraph()->IsCompilingBaseline() && !Runtime::Current()->IsAotCompiler()) {
-    // Note the slow path doesn't save SIMD registers, so if we were to
-    // call it on loop back edge, we would need to fix this.
+  if (GetGraph()->IsCompilingBaseline() &&
+      GetGraph()->IsUsefulOptimizing() &&
+      !Runtime::Current()->IsAotCompiler()) {
     ProfilingInfo* info = GetGraph()->GetProfilingInfo();
     DCHECK(info != nullptr);
     uint32_t address = reinterpret_cast32<uint32_t>(info) +
diff --git a/compiler/optimizing/code_generator_x86_64.cc b/compiler/optimizing/code_generator_x86_64.cc
index db4062b00d..af6c6255e5 100644
--- a/compiler/optimizing/code_generator_x86_64.cc
+++ b/compiler/optimizing/code_generator_x86_64.cc
@@ -1788,7 +1788,9 @@ void CodeGeneratorX86_64::MaybeIncrementHotness(HSuspendCheck* suspend_check, bo
     __ Bind(&overflow);
   }
 
-  if (GetGraph()->IsCompilingBaseline() && !Runtime::Current()->IsAotCompiler()) {
+  if (GetGraph()->IsCompilingBaseline() &&
+      GetGraph()->IsUsefulOptimizing() &&
+      !Runtime::Current()->IsAotCompiler()) {
     ProfilingInfo* info = GetGraph()->GetProfilingInfo();
     DCHECK(info != nullptr);
     CHECK(!HasEmptyFrame());
diff --git a/compiler/optimizing/constant_folding.cc b/compiler/optimizing/constant_folding.cc
index b0a65ca64d..f57d8ade16 100644
--- a/compiler/optimizing/constant_folding.cc
+++ b/compiler/optimizing/constant_folding.cc
@@ -117,12 +117,9 @@ bool HConstantFolding::Run() {
 
 
 void HConstantFoldingVisitor::VisitBasicBlock(HBasicBlock* block) {
-  // Traverse this block's instructions (phis don't need to be
-  // processed) in (forward) order and replace the ones that can be
-  // statically evaluated by a compile-time counterpart.
-  for (HInstructionIterator it(block->GetInstructions()); !it.Done(); it.Advance()) {
-    it.Current()->Accept(this);
-  }
+  // Traverse this block's instructions (phis don't need to be processed) in (forward) order
+  // and replace the ones that can be statically evaluated by a compile-time counterpart.
+  VisitNonPhiInstructions(block);
 }
 
 void HConstantFoldingVisitor::VisitUnaryOperation(HUnaryOperation* inst) {
diff --git a/compiler/optimizing/constructor_fence_redundancy_elimination.cc b/compiler/optimizing/constructor_fence_redundancy_elimination.cc
index 9b3bb91d21..30c33dd5c5 100644
--- a/compiler/optimizing/constructor_fence_redundancy_elimination.cc
+++ b/compiler/optimizing/constructor_fence_redundancy_elimination.cc
@@ -37,8 +37,8 @@ class CFREVisitor final : public HGraphVisitor {
         stats_(stats) {}
 
   void VisitBasicBlock(HBasicBlock* block) override {
-    // Visit all instructions in block.
-    HGraphVisitor::VisitBasicBlock(block);
+    // Visit all non-Phi instructions in the block.
+    VisitNonPhiInstructions(block);
 
     // If there were any unmerged fences left, merge them together,
     // the objects are considered 'published' at the end of the block.
diff --git a/compiler/optimizing/dead_code_elimination.cc b/compiler/optimizing/dead_code_elimination.cc
index 6746771fa4..fe1361c935 100644
--- a/compiler/optimizing/dead_code_elimination.cc
+++ b/compiler/optimizing/dead_code_elimination.cc
@@ -937,7 +937,7 @@ void HDeadCodeElimination::RemoveDeadInstructions() {
     for (HBackwardInstructionIterator phi_it(block->GetPhis()); !phi_it.Done(); phi_it.Advance()) {
       DCHECK(phi_it.Current()->IsPhi());
       HPhi* phi = phi_it.Current()->AsPhi();
-      if (phi->IsDeadAndRemovable()) {
+      if (phi->IsPhiDeadAndRemovable()) {
         block->RemovePhi(phi);
         MaybeRecordStat(stats_, MethodCompilationStat::kRemovedDeadPhi);
       }
diff --git a/compiler/optimizing/graph_checker.cc b/compiler/optimizing/graph_checker.cc
index d60ec06097..b061b401ff 100644
--- a/compiler/optimizing/graph_checker.cc
+++ b/compiler/optimizing/graph_checker.cc
@@ -304,10 +304,11 @@ void GraphChecker::VisitBasicBlock(HBasicBlock* block) {
                             current_block_->GetBlockId()));
     }
     if (current->GetNext() == nullptr && current != block->GetLastInstruction()) {
-      AddError(StringPrintf("The recorded last instruction of block %d does not match "
-                            "the actual last instruction %d.",
-                            current_block_->GetBlockId(),
-                            current->GetId()));
+      AddError(
+          StringPrintf("The recorded last instruction of block %d does not match "
+                       "the actual last instruction %d.",
+                       current_block_->GetBlockId(),
+                       current->GetId()));
     }
     current->Accept(this);
   }
@@ -657,15 +658,18 @@ void GraphChecker::VisitInstruction(HInstruction* instruction) {
   // Ensure an instruction dominates all its uses.
   for (const HUseListNode<HInstruction*>& use : instruction->GetUses()) {
     HInstruction* user = use.GetUser();
-    if (!user->IsPhi() && !instruction->StrictlyDominates(user)) {
-      AddError(StringPrintf("Instruction %s:%d in block %d does not dominate "
-                            "use %s:%d in block %d.",
-                            instruction->DebugName(),
-                            instruction->GetId(),
-                            current_block_->GetBlockId(),
-                            user->DebugName(),
-                            user->GetId(),
-                            user->GetBlock()->GetBlockId()));
+    if (!user->IsPhi() && (instruction->GetBlock() == user->GetBlock()
+                               ? seen_ids_.IsBitSet(user->GetId())
+                               : !instruction->GetBlock()->Dominates(user->GetBlock()))) {
+      AddError(
+          StringPrintf("Instruction %s:%d in block %d does not dominate "
+                       "use %s:%d in block %d.",
+                       instruction->DebugName(),
+                       instruction->GetId(),
+                       current_block_->GetBlockId(),
+                       user->DebugName(),
+                       user->GetId(),
+                       user->GetBlock()->GetBlockId()));
     }
   }
 
@@ -677,22 +681,24 @@ void GraphChecker::VisitInstruction(HInstruction* instruction) {
                           current_block_->GetBlockId()));
   }
 
-  // Ensure an instruction having an environment is dominated by the
-  // instructions contained in the environment.
-  for (HEnvironment* environment = instruction->GetEnvironment();
-       environment != nullptr;
-       environment = environment->GetParent()) {
-    for (size_t i = 0, e = environment->Size(); i < e; ++i) {
-      HInstruction* env_instruction = environment->GetInstructionAt(i);
-      if (env_instruction != nullptr
-          && !env_instruction->StrictlyDominates(instruction)) {
-        AddError(StringPrintf("Instruction %d in environment of instruction %d "
-                              "from block %d does not dominate instruction %d.",
-                              env_instruction->GetId(),
-                              instruction->GetId(),
-                              current_block_->GetBlockId(),
-                              instruction->GetId()));
-      }
+  // Ensure an instruction dominates all its environment uses.
+  for (const HUseListNode<HEnvironment*>& use : instruction->GetEnvUses()) {
+    HInstruction* user = use.GetUser()->GetHolder();
+    if (user->IsPhi()) {
+      AddError(StringPrintf("Phi %d shouldn't have an environment", instruction->GetId()));
+    }
+    if (instruction->GetBlock() == user->GetBlock()
+            ? seen_ids_.IsBitSet(user->GetId())
+            : !instruction->GetBlock()->Dominates(user->GetBlock())) {
+      AddError(
+          StringPrintf("Instruction %s:%d in block %d does not dominate "
+                       "environment use %s:%d in block %d.",
+                       instruction->DebugName(),
+                       instruction->GetId(),
+                       current_block_->GetBlockId(),
+                       user->DebugName(),
+                       user->GetId(),
+                       user->GetBlock()->GetBlockId()));
     }
   }
 
@@ -709,14 +715,15 @@ void GraphChecker::VisitInstruction(HInstruction* instruction) {
       for (HInstructionIterator phi_it(catch_block->GetPhis()); !phi_it.Done(); phi_it.Advance()) {
         HPhi* catch_phi = phi_it.Current()->AsPhi();
         if (environment->GetInstructionAt(catch_phi->GetRegNumber()) == nullptr) {
-          AddError(StringPrintf("Instruction %s:%d throws into catch block %d "
-                                "with catch phi %d for vreg %d but its "
-                                "corresponding environment slot is empty.",
-                                instruction->DebugName(),
-                                instruction->GetId(),
-                                catch_block->GetBlockId(),
-                                catch_phi->GetId(),
-                                catch_phi->GetRegNumber()));
+          AddError(
+              StringPrintf("Instruction %s:%d throws into catch block %d "
+                           "with catch phi %d for vreg %d but its "
+                           "corresponding environment slot is empty.",
+                           instruction->DebugName(),
+                           instruction->GetId(),
+                           catch_block->GetBlockId(),
+                           catch_phi->GetId(),
+                           catch_phi->GetRegNumber()));
         }
       }
     }
@@ -1291,7 +1298,7 @@ bool IsRemovedWriteBarrier(DataType::Type type,
                            HInstruction* value) {
   return write_barrier_kind == WriteBarrierKind::kDontEmit &&
          type == DataType::Type::kReference &&
-         !value->IsNullConstant();
+         !HuntForOriginalReference(value)->IsNullConstant();
 }
 
 void GraphChecker::VisitArraySet(HArraySet* instruction) {
diff --git a/compiler/optimizing/graph_checker.h b/compiler/optimizing/graph_checker.h
index 5704bcec1a..541a9cc3d2 100644
--- a/compiler/optimizing/graph_checker.h
+++ b/compiler/optimizing/graph_checker.h
@@ -30,7 +30,7 @@ namespace art HIDDEN {
 class CodeGenerator;
 
 // A control-flow graph visitor performing various checks.
-class GraphChecker : public HGraphDelegateVisitor {
+class GraphChecker final : public HGraphDelegateVisitor {
  public:
   explicit GraphChecker(HGraph* graph,
                         CodeGenerator* codegen = nullptr,
diff --git a/compiler/optimizing/graph_visualizer.cc b/compiler/optimizing/graph_visualizer.cc
index afbf941355..6696c1ef17 100644
--- a/compiler/optimizing/graph_visualizer.cc
+++ b/compiler/optimizing/graph_visualizer.cc
@@ -195,7 +195,7 @@ class HGraphVisualizerDisassembler {
 /**
  * HGraph visitor to generate a file suitable for the c1visualizer tool and IRHydra.
  */
-class HGraphVisualizerPrinter : public HGraphDelegateVisitor {
+class HGraphVisualizerPrinter final : public HGraphDelegateVisitor {
  public:
   HGraphVisualizerPrinter(HGraph* graph,
                           std::ostream& output,
diff --git a/compiler/optimizing/gvn.cc b/compiler/optimizing/gvn.cc
index a6ca057cfc..9113860387 100644
--- a/compiler/optimizing/gvn.cc
+++ b/compiler/optimizing/gvn.cc
@@ -248,7 +248,7 @@ class ValueSet : public ArenaObject<kArenaAllocGvn> {
   // Iterates over buckets with impure instructions (even indices) and deletes
   // the ones on which 'cond' returns true.
   template<typename Functor>
-  void DeleteAllImpureWhich(Functor cond) {
+  void DeleteAllImpureWhich(Functor&& cond) {
     for (size_t i = 0; i < num_buckets_; i += 2) {
       Node* node = buckets_[i];
       Node* previous = nullptr;
diff --git a/compiler/optimizing/inliner.cc b/compiler/optimizing/inliner.cc
index fd3e787fc8..d7ca17b646 100644
--- a/compiler/optimizing/inliner.cc
+++ b/compiler/optimizing/inliner.cc
@@ -541,6 +541,7 @@ bool HInliner::TryInline(HInvoke* invoke_instruction) {
                        << " statically resolve the target";
     // For baseline compilation, we will collect inline caches, so we should not
     // try to inline using them.
+    outermost_graph_->SetUsefulOptimizing();
     return false;
   }
 
@@ -1552,9 +1553,7 @@ bool HInliner::IsInliningEncouraged(const HInvoke* invoke_instruction,
     return false;
   }
 
-  size_t inline_max_code_units = graph_->IsCompilingBaseline()
-      ? CompilerOptions::kBaselineInlineMaxCodeUnits
-      : codegen_->GetCompilerOptions().GetInlineMaxCodeUnits();
+  size_t inline_max_code_units = codegen_->GetCompilerOptions().GetInlineMaxCodeUnits();
   if (accessor.InsnsSizeInCodeUnits() > inline_max_code_units) {
     LOG_FAIL(stats_, MethodCompilationStat::kNotInlinedCodeItem)
         << "Method " << method->PrettyMethod()
@@ -1565,6 +1564,14 @@ bool HInliner::IsInliningEncouraged(const HInvoke* invoke_instruction,
     return false;
   }
 
+  if (graph_->IsCompilingBaseline() &&
+      accessor.InsnsSizeInCodeUnits() > CompilerOptions::kBaselineInlineMaxCodeUnits) {
+    LOG_FAIL_NO_STAT() << "Reached baseline maximum code unit for inlining  "
+                       << method->PrettyMethod();
+    outermost_graph_->SetUsefulOptimizing();
+    return false;
+  }
+
   if (invoke_instruction->GetBlock()->GetLastInstruction()->IsThrow()) {
     LOG_FAIL(stats_, MethodCompilationStat::kNotInlinedEndsWithThrow)
         << "Method " << method->PrettyMethod()
@@ -2129,6 +2136,7 @@ bool HInliner::CanInlineBody(const HGraph* callee_graph,
         if (depth_ + 1 > maximum_inlining_depth_for_baseline) {
           LOG_FAIL_NO_STAT() << "Reached maximum depth for inlining in baseline compilation: "
                              << depth_ << " for " << callee_graph->GetArtMethod()->PrettyMethod();
+          outermost_graph_->SetUsefulOptimizing();
           return false;
         }
       }
diff --git a/compiler/optimizing/instruction_builder.cc b/compiler/optimizing/instruction_builder.cc
index ca0a8e81af..410d6fd0d0 100644
--- a/compiler/optimizing/instruction_builder.cc
+++ b/compiler/optimizing/instruction_builder.cc
@@ -33,6 +33,7 @@
 #include "intrinsics.h"
 #include "intrinsics_utils.h"
 #include "jit/jit.h"
+#include "jit/profiling_info.h"
 #include "mirror/dex_cache.h"
 #include "oat/oat_file.h"
 #include "optimizing_compiler_stats.h"
diff --git a/compiler/optimizing/instruction_simplifier.cc b/compiler/optimizing/instruction_simplifier.cc
index 5db8235e29..ae778b421a 100644
--- a/compiler/optimizing/instruction_simplifier.cc
+++ b/compiler/optimizing/instruction_simplifier.cc
@@ -170,7 +170,7 @@ bool InstructionSimplifierVisitor::Run() {
     // post order visit, we sometimes need to revisit an instruction index.
     do {
       simplification_occurred_ = false;
-      VisitBasicBlock(block);
+      VisitNonPhiInstructions(block);
       if (simplification_occurred_) {
         didSimplify = true;
       }
@@ -450,15 +450,22 @@ static bool TryReplaceShiftsByConstantWithTypeConversion(HBinaryOperation *instr
   bool is_signed = instruction->IsShr();
   DataType::Type conv_type =
       is_signed ? source_integral_type : DataType::ToUnsigned(source_integral_type);
+
+  DCHECK(DataType::IsTypeConversionImplicit(conv_type, instruction->GetResultType()));
+
   HInstruction* shl_value = shl->GetLeft();
   HBasicBlock *block = instruction->GetBlock();
 
-  HTypeConversion* new_conversion =
-      new (block->GetGraph()->GetAllocator()) HTypeConversion(conv_type, shl_value);
-
-  DCHECK(DataType::IsTypeConversionImplicit(conv_type, instruction->GetResultType()));
+  // We shouldn't introduce new implicit type conversions during simplification.
+  if (DataType::IsTypeConversionImplicit(shl_value->GetType(), conv_type)) {
+    instruction->ReplaceWith(shl_value);
+    instruction->GetBlock()->RemoveInstruction(instruction);
+  } else {
+    HTypeConversion* new_conversion =
+        new (block->GetGraph()->GetAllocator()) HTypeConversion(conv_type, shl_value);
+    block->ReplaceAndRemoveInstructionWith(instruction, new_conversion);
+  }
 
-  block->ReplaceAndRemoveInstructionWith(instruction, new_conversion);
   shl->GetBlock()->RemoveInstruction(shl);
 
   return true;
@@ -614,8 +621,7 @@ bool InstructionSimplifierVisitor::TryReplaceWithRotateConstantPattern(HBinaryOp
   size_t rdist = Int64FromConstant(ushr->GetRight()->AsConstant());
   size_t ldist = Int64FromConstant(shl->GetRight()->AsConstant());
   if (((ldist + rdist) & (reg_bits - 1)) == 0) {
-    ReplaceRotateWithRor(op, ushr, shl);
-    return true;
+    return ReplaceRotateWithRor(op, ushr, shl);
   }
   return false;
 }
@@ -636,6 +642,10 @@ bool InstructionSimplifierVisitor::TryReplaceWithRotateConstantPattern(HBinaryOp
 //    OP   dst, dst, tmp
 // with
 //    Ror  dst, x,   d
+//
+// Requires `d` to be non-zero for the HAdd and HXor case. If `d` is 0 the shifts and rotate are
+// no-ops and the `OP` is never executed. This is fine for HOr since the result is the same, but the
+// result is different for HAdd and HXor.
 bool InstructionSimplifierVisitor::TryReplaceWithRotateRegisterNegPattern(HBinaryOperation* op,
                                                                           HUShr* ushr,
                                                                           HShl* shl) {
@@ -643,11 +653,20 @@ bool InstructionSimplifierVisitor::TryReplaceWithRotateRegisterNegPattern(HBinar
   DCHECK(ushr->GetRight()->IsNeg() || shl->GetRight()->IsNeg());
   bool neg_is_left = shl->GetRight()->IsNeg();
   HNeg* neg = neg_is_left ? shl->GetRight()->AsNeg() : ushr->GetRight()->AsNeg();
-  // And the shift distance being negated is the distance being shifted the other way.
-  if (neg->InputAt(0) == (neg_is_left ? ushr->GetRight() : shl->GetRight())) {
-    ReplaceRotateWithRor(op, ushr, shl);
+  HInstruction* value = neg->InputAt(0);
+
+  // The shift distance being negated is the distance being shifted the other way.
+  if (value != (neg_is_left ? ushr->GetRight() : shl->GetRight())) {
+    return false;
   }
-  return false;
+
+  const bool needs_non_zero_value = !op->IsOr();
+  if (needs_non_zero_value) {
+    if (!value->IsConstant() || value->AsConstant()->IsArithmeticZero()) {
+      return false;
+    }
+  }
+  return ReplaceRotateWithRor(op, ushr, shl);
 }
 
 // Try replacing code looking like (x >>> d OP x << (#bits - d)):
diff --git a/compiler/optimizing/intrinsic_objects.cc b/compiler/optimizing/intrinsic_objects.cc
index 6c799d4132..c625d435ae 100644
--- a/compiler/optimizing/intrinsic_objects.cc
+++ b/compiler/optimizing/intrinsic_objects.cc
@@ -35,7 +35,7 @@ static int32_t FillIntrinsicsObjects(
     ObjPtr<mirror::ObjectArray<mirror::Object>> live_objects,
     int32_t expected_low,
     int32_t expected_high,
-    T type_check,
+    T&& type_check,
     int32_t index)
     REQUIRES_SHARED(Locks::mutator_lock_) {
   ObjPtr<mirror::ObjectArray<mirror::Object>> cache =
diff --git a/compiler/optimizing/intrinsics_arm64.cc b/compiler/optimizing/intrinsics_arm64.cc
index 2ae44cd4b0..56a5186d36 100644
--- a/compiler/optimizing/intrinsics_arm64.cc
+++ b/compiler/optimizing/intrinsics_arm64.cc
@@ -3969,7 +3969,7 @@ template<typename OP>
 void GenerateFP16Round(HInvoke* invoke,
                        CodeGeneratorARM64* const codegen_,
                        MacroAssembler* masm,
-                       const OP roundOp) {
+                       OP&& roundOp) {
   DCHECK(codegen_->GetInstructionSetFeatures().HasFP16());
   LocationSummary* locations = invoke->GetLocations();
   UseScratchRegisterScope scratch_scope(masm);
diff --git a/compiler/optimizing/linearize_test.cc b/compiler/optimizing/linearize_test.cc
index 6f4f2b6cf6..01daa23511 100644
--- a/compiler/optimizing/linearize_test.cc
+++ b/compiler/optimizing/linearize_test.cc
@@ -55,7 +55,6 @@ void LinearizeTest::TestCode(const std::vector<uint16_t>& data,
 }
 
 TEST_F(LinearizeTest, CFG1) {
-  TEST_DISABLED_FOR_RISCV64();
   // Structure of this graph (+ are back edges)
   //            Block0
   //              |
@@ -81,7 +80,6 @@ TEST_F(LinearizeTest, CFG1) {
 }
 
 TEST_F(LinearizeTest, CFG2) {
-  TEST_DISABLED_FOR_RISCV64();
   // Structure of this graph (+ are back edges)
   //            Block0
   //              |
@@ -107,7 +105,6 @@ TEST_F(LinearizeTest, CFG2) {
 }
 
 TEST_F(LinearizeTest, CFG3) {
-  TEST_DISABLED_FOR_RISCV64();
   // Structure of this graph (+ are back edges)
   //            Block0
   //              |
@@ -135,7 +132,6 @@ TEST_F(LinearizeTest, CFG3) {
 }
 
 TEST_F(LinearizeTest, CFG4) {
-  TEST_DISABLED_FOR_RISCV64();
   /* Structure of this graph (+ are back edges)
   //            Block0
   //              |
@@ -166,7 +162,6 @@ TEST_F(LinearizeTest, CFG4) {
 }
 
 TEST_F(LinearizeTest, CFG5) {
-  TEST_DISABLED_FOR_RISCV64();
   /* Structure of this graph (+ are back edges)
   //            Block0
   //              |
@@ -197,7 +192,6 @@ TEST_F(LinearizeTest, CFG5) {
 }
 
 TEST_F(LinearizeTest, CFG6) {
-  TEST_DISABLED_FOR_RISCV64();
   //            Block0
   //              |
   //            Block1
@@ -224,7 +218,6 @@ TEST_F(LinearizeTest, CFG6) {
 }
 
 TEST_F(LinearizeTest, CFG7) {
-  TEST_DISABLED_FOR_RISCV64();
   // Structure of this graph (+ are back edges)
   //            Block0
   //              |
diff --git a/compiler/optimizing/live_ranges_test.cc b/compiler/optimizing/live_ranges_test.cc
index 7e488ba41d..fb1a23eef4 100644
--- a/compiler/optimizing/live_ranges_test.cc
+++ b/compiler/optimizing/live_ranges_test.cc
@@ -47,7 +47,6 @@ HGraph* LiveRangesTest::BuildGraph(const std::vector<uint16_t>& data) {
 }
 
 TEST_F(LiveRangesTest, CFG1) {
-  TEST_DISABLED_FOR_RISCV64();
   /*
    * Test the following snippet:
    *  return 0;
@@ -82,7 +81,6 @@ TEST_F(LiveRangesTest, CFG1) {
 }
 
 TEST_F(LiveRangesTest, CFG2) {
-  TEST_DISABLED_FOR_RISCV64();
   /*
    * Test the following snippet:
    *  var a = 0;
@@ -127,7 +125,6 @@ TEST_F(LiveRangesTest, CFG2) {
 }
 
 TEST_F(LiveRangesTest, CFG3) {
-  TEST_DISABLED_FOR_RISCV64();
   /*
    * Test the following snippet:
    *  var a = 0;
@@ -197,7 +194,6 @@ TEST_F(LiveRangesTest, CFG3) {
 }
 
 TEST_F(LiveRangesTest, Loop1) {
-  TEST_DISABLED_FOR_RISCV64();
   /*
    * Test the following snippet:
    *  var a = 0;
@@ -274,7 +270,6 @@ TEST_F(LiveRangesTest, Loop1) {
 }
 
 TEST_F(LiveRangesTest, Loop2) {
-  TEST_DISABLED_FOR_RISCV64();
   /*
    * Test the following snippet:
    *  var a = 0;
@@ -346,7 +341,6 @@ TEST_F(LiveRangesTest, Loop2) {
 }
 
 TEST_F(LiveRangesTest, CFG4) {
-  TEST_DISABLED_FOR_RISCV64();
   /*
    * Test the following snippet:
    *  var a = 0;
diff --git a/compiler/optimizing/liveness_test.cc b/compiler/optimizing/liveness_test.cc
index 6af07aea4e..0b421cf9e6 100644
--- a/compiler/optimizing/liveness_test.cc
+++ b/compiler/optimizing/liveness_test.cc
@@ -70,7 +70,6 @@ void LivenessTest::TestCode(const std::vector<uint16_t>& data, const char* expec
 }
 
 TEST_F(LivenessTest, CFG1) {
-  TEST_DISABLED_FOR_RISCV64();
   const char* expected =
     "Block 0\n"
     "  live in: (0)\n"
@@ -94,7 +93,6 @@ TEST_F(LivenessTest, CFG1) {
 }
 
 TEST_F(LivenessTest, CFG2) {
-  TEST_DISABLED_FOR_RISCV64();
   const char* expected =
     "Block 0\n"
     "  live in: (0)\n"
@@ -117,7 +115,6 @@ TEST_F(LivenessTest, CFG2) {
 }
 
 TEST_F(LivenessTest, CFG3) {
-  TEST_DISABLED_FOR_RISCV64();
   const char* expected =
     "Block 0\n"  // entry block
     "  live in: (000)\n"
@@ -147,7 +144,6 @@ TEST_F(LivenessTest, CFG3) {
 }
 
 TEST_F(LivenessTest, CFG4) {
-  TEST_DISABLED_FOR_RISCV64();
   // var a;
   // if (0 == 0) {
   //   a = 5;
@@ -196,7 +192,6 @@ TEST_F(LivenessTest, CFG4) {
 }
 
 TEST_F(LivenessTest, CFG5) {
-  TEST_DISABLED_FOR_RISCV64();
   // var a = 0;
   // if (0 == 0) {
   // } else {
@@ -242,7 +237,6 @@ TEST_F(LivenessTest, CFG5) {
 }
 
 TEST_F(LivenessTest, Loop1) {
-  TEST_DISABLED_FOR_RISCV64();
   // Simple loop with one preheader and one back edge.
   // var a = 0;
   // while (a == a) {
@@ -289,7 +283,6 @@ TEST_F(LivenessTest, Loop1) {
 }
 
 TEST_F(LivenessTest, Loop3) {
-  TEST_DISABLED_FOR_RISCV64();
   // Test that the returned value stays live in a preceding loop.
   // var a = 0;
   // while (a == a) {
@@ -337,7 +330,6 @@ TEST_F(LivenessTest, Loop3) {
 
 
 TEST_F(LivenessTest, Loop4) {
-  TEST_DISABLED_FOR_RISCV64();
   // Make sure we support a preheader of a loop not being the first predecessor
   // in the predecessor list of the header.
   // var a = 0;
@@ -390,7 +382,6 @@ TEST_F(LivenessTest, Loop4) {
 }
 
 TEST_F(LivenessTest, Loop5) {
-  TEST_DISABLED_FOR_RISCV64();
   // Make sure we create a preheader of a loop when a header originally has two
   // incoming blocks and one back edge.
   // Bitsets are made of:
@@ -447,7 +438,6 @@ TEST_F(LivenessTest, Loop5) {
 }
 
 TEST_F(LivenessTest, Loop6) {
-  TEST_DISABLED_FOR_RISCV64();
   // Bitsets are made of:
   // (constant0, constant4, constant5, phi in block 2)
   const char* expected =
@@ -499,7 +489,6 @@ TEST_F(LivenessTest, Loop6) {
 
 
 TEST_F(LivenessTest, Loop7) {
-  TEST_DISABLED_FOR_RISCV64();
   // Bitsets are made of:
   // (constant0, constant4, constant5, phi in block 2, phi in block 6)
   const char* expected =
@@ -554,7 +543,6 @@ TEST_F(LivenessTest, Loop7) {
 }
 
 TEST_F(LivenessTest, Loop8) {
-  TEST_DISABLED_FOR_RISCV64();
   // var a = 0;
   // while (a == a) {
   //   a = a + a;
diff --git a/compiler/optimizing/load_store_elimination.cc b/compiler/optimizing/load_store_elimination.cc
index 38c1cfc0d2..80cf9e669b 100644
--- a/compiler/optimizing/load_store_elimination.cc
+++ b/compiler/optimizing/load_store_elimination.cc
@@ -1834,8 +1834,8 @@ void LSEVisitor::VisitBasicBlock(HBasicBlock* block) {
   } else {
     MergePredecessorRecords(block);
   }
-  // Visit instructions.
-  HGraphVisitor::VisitBasicBlock(block);
+  // Visit non-Phi instructions.
+  VisitNonPhiInstructions(block);
 }
 
 bool LSEVisitor::MayAliasOnBackEdge(HBasicBlock* loop_header, size_t idx1, size_t idx2) const {
@@ -2727,7 +2727,7 @@ struct ScopedRestoreHeapValues {
   }
 
   template<typename Func>
-  void ForEachRecord(Func func) {
+  void ForEachRecord(Func&& func) {
     for (size_t blk_id : Range(to_restore_.size())) {
       for (size_t heap_loc : Range(to_restore_[blk_id].size())) {
         LSEVisitor::ValueRecord* vr = &to_restore_[blk_id][heap_loc];
diff --git a/compiler/optimizing/loop_optimization_test.cc b/compiler/optimizing/loop_optimization_test.cc
index 49e3c0418f..7f694fb655 100644
--- a/compiler/optimizing/loop_optimization_test.cc
+++ b/compiler/optimizing/loop_optimization_test.cc
@@ -30,7 +30,6 @@ namespace art HIDDEN {
 class LoopOptimizationTest : public OptimizingUnitTest {
  protected:
   void SetUp() override {
-    TEST_SETUP_DISABLED_FOR_RISCV64();
     OptimizingUnitTest::SetUp();
 
     graph_ = CreateGraph();
@@ -45,7 +44,6 @@ class LoopOptimizationTest : public OptimizingUnitTest {
   }
 
   void TearDown() override {
-    TEST_TEARDOWN_DISABLED_FOR_RISCV64();
     codegen_.reset();
     compiler_options_.reset();
     graph_ = nullptr;
@@ -136,20 +134,17 @@ class LoopOptimizationTest : public OptimizingUnitTest {
 //
 
 TEST_F(LoopOptimizationTest, NoLoops) {
-  TEST_DISABLED_FOR_RISCV64();
   PerformAnalysis();
   EXPECT_EQ("", LoopStructure());
 }
 
 TEST_F(LoopOptimizationTest, SingleLoop) {
-  TEST_DISABLED_FOR_RISCV64();
   AddLoop(entry_block_, return_block_);
   PerformAnalysis();
   EXPECT_EQ("[]", LoopStructure());
 }
 
 TEST_F(LoopOptimizationTest, LoopNest10) {
-  TEST_DISABLED_FOR_RISCV64();
   HBasicBlock* b = entry_block_;
   HBasicBlock* s = return_block_;
   for (int i = 0; i < 10; i++) {
@@ -161,7 +156,6 @@ TEST_F(LoopOptimizationTest, LoopNest10) {
 }
 
 TEST_F(LoopOptimizationTest, LoopSequence10) {
-  TEST_DISABLED_FOR_RISCV64();
   HBasicBlock* b = entry_block_;
   HBasicBlock* s = return_block_;
   for (int i = 0; i < 10; i++) {
@@ -173,7 +167,6 @@ TEST_F(LoopOptimizationTest, LoopSequence10) {
 }
 
 TEST_F(LoopOptimizationTest, LoopSequenceOfNests) {
-  TEST_DISABLED_FOR_RISCV64();
   HBasicBlock* b = entry_block_;
   HBasicBlock* s = return_block_;
   for (int i = 0; i < 10; i++) {
@@ -201,7 +194,6 @@ TEST_F(LoopOptimizationTest, LoopSequenceOfNests) {
 }
 
 TEST_F(LoopOptimizationTest, LoopNestWithSequence) {
-  TEST_DISABLED_FOR_RISCV64();
   HBasicBlock* b = entry_block_;
   HBasicBlock* s = return_block_;
   for (int i = 0; i < 10; i++) {
@@ -223,7 +215,6 @@ TEST_F(LoopOptimizationTest, LoopNestWithSequence) {
 //
 // This is a test for nodes.cc functionality - HGraph::SimplifyLoop.
 TEST_F(LoopOptimizationTest, SimplifyLoopReoderPredecessors) {
-  TEST_DISABLED_FOR_RISCV64();
   // Can't use AddLoop as we want special order for blocks predecessors.
   HBasicBlock* header = new (GetAllocator()) HBasicBlock(graph_);
   HBasicBlock* body = new (GetAllocator()) HBasicBlock(graph_);
@@ -269,7 +260,6 @@ TEST_F(LoopOptimizationTest, SimplifyLoopReoderPredecessors) {
 //
 // This is a test for nodes.cc functionality - HGraph::SimplifyLoop.
 TEST_F(LoopOptimizationTest, SimplifyLoopSinglePreheader) {
-  TEST_DISABLED_FOR_RISCV64();
   HBasicBlock* header = AddLoop(entry_block_, return_block_);
 
   header->InsertInstructionBefore(
diff --git a/compiler/optimizing/nodes.cc b/compiler/optimizing/nodes.cc
index 94f197d8f1..c87c78815b 100644
--- a/compiler/optimizing/nodes.cc
+++ b/compiler/optimizing/nodes.cc
@@ -1693,10 +1693,20 @@ void HGraphVisitor::VisitReversePostOrder() {
 }
 
 void HGraphVisitor::VisitBasicBlock(HBasicBlock* block) {
+  VisitPhis(block);
+  VisitNonPhiInstructions(block);
+}
+
+void HGraphVisitor::VisitPhis(HBasicBlock* block) {
   for (HInstructionIterator it(block->GetPhis()); !it.Done(); it.Advance()) {
-    it.Current()->Accept(this);
+    DCHECK(it.Current()->IsPhi());
+    VisitPhi(it.Current()->AsPhi());
   }
+}
+
+void HGraphVisitor::VisitNonPhiInstructions(HBasicBlock* block) {
   for (HInstructionIterator it(block->GetInstructions()); !it.Done(); it.Advance()) {
+    DCHECK(!it.Current()->IsPhi());
     it.Current()->Accept(this);
   }
 }
diff --git a/compiler/optimizing/nodes.h b/compiler/optimizing/nodes.h
index c862e31de7..4d6b909629 100644
--- a/compiler/optimizing/nodes.h
+++ b/compiler/optimizing/nodes.h
@@ -425,6 +425,7 @@ class HGraph : public ArenaObject<kArenaAllocGraph> {
         cached_current_method_(nullptr),
         art_method_(nullptr),
         compilation_kind_(compilation_kind),
+        useful_optimizing_(false),
         cha_single_implementation_list_(allocator->Adapter(kArenaAllocCHA)) {
     blocks_.reserve(kDefaultNumberOfBlocks);
   }
@@ -742,6 +743,9 @@ class HGraph : public ArenaObject<kArenaAllocGraph> {
   void SetNumberOfCHAGuards(uint32_t num) { number_of_cha_guards_ = num; }
   void IncrementNumberOfCHAGuards() { number_of_cha_guards_++; }
 
+  void SetUsefulOptimizing() { useful_optimizing_ = true; }
+  bool IsUsefulOptimizing() const { return useful_optimizing_; }
+
  private:
   void RemoveDeadBlocksInstructionsAsUsersAndDisconnect(const ArenaBitVector& visited) const;
   void RemoveDeadBlocks(const ArenaBitVector& visited);
@@ -897,6 +901,10 @@ class HGraph : public ArenaObject<kArenaAllocGraph> {
   // directly jump to.
   const CompilationKind compilation_kind_;
 
+  // Whether after compiling baseline it is still useful re-optimizing this
+  // method.
+  bool useful_optimizing_;
+
   // List of methods that are assumed to have single implementation.
   ArenaSet<ArtMethod*> cha_single_implementation_list_;
 
@@ -2440,18 +2448,26 @@ class HInstruction : public ArenaObject<kArenaAllocInstruction> {
   bool IsRemovable() const {
     return
         !DoesAnyWrite() &&
-        !CanThrow() &&
+        // TODO(solanes): Merge calls from IsSuspendCheck to IsControlFlow into one that doesn't
+        // do virtual dispatching.
         !IsSuspendCheck() &&
-        !IsControlFlow() &&
         !IsNop() &&
         !IsParameterValue() &&
         // If we added an explicit barrier then we should keep it.
         !IsMemoryBarrier() &&
-        !IsConstructorFence();
+        !IsConstructorFence() &&
+        !IsControlFlow() &&
+        !CanThrow();
   }
 
   bool IsDeadAndRemovable() const {
-    return IsRemovable() && !HasUses();
+    return !HasUses() && IsRemovable();
+  }
+
+  bool IsPhiDeadAndRemovable() const {
+    DCHECK(IsPhi());
+    DCHECK(IsRemovable()) << " phis are always removable";
+    return !HasUses();
   }
 
   // Does this instruction dominate `other_instruction`?
@@ -8556,6 +8572,9 @@ class HGraphVisitor : public ValueObject {
 #undef DECLARE_VISIT_INSTRUCTION
 
  protected:
+  void VisitPhis(HBasicBlock* block);
+  void VisitNonPhiInstructions(HBasicBlock* block);
+
   OptimizingCompilerStats* stats_;
 
  private:
diff --git a/compiler/optimizing/nodes_shared.h b/compiler/optimizing/nodes_shared.h
index 4b0187d536..d627c6daee 100644
--- a/compiler/optimizing/nodes_shared.h
+++ b/compiler/optimizing/nodes_shared.h
@@ -97,6 +97,10 @@ class HBitwiseNegatedRight final : public HBinaryOperation {
     }
   }
 
+  bool InstructionDataEquals(const HInstruction* other) const override {
+    return op_kind_ == other->AsBitwiseNegatedRight()->op_kind_;
+  }
+
   HConstant* Evaluate(HIntConstant* x, HIntConstant* y) const override {
     return GetBlock()->GetGraph()->GetIntConstant(
         Compute(x->GetValue(), y->GetValue()), GetDexPc());
diff --git a/compiler/optimizing/optimizing_compiler.cc b/compiler/optimizing/optimizing_compiler.cc
index a1c4130bc1..65e8e51712 100644
--- a/compiler/optimizing/optimizing_compiler.cc
+++ b/compiler/optimizing/optimizing_compiler.cc
@@ -905,6 +905,7 @@ CodeGenerator* OptimizingCompiler::TryCompile(ArenaAllocator* allocator,
   }
 
   if (compilation_kind == CompilationKind::kBaseline && compiler_options.ProfileBranches()) {
+    graph->SetUsefulOptimizing();
     // Branch profiling currently doesn't support running optimizations.
     RunRequiredPasses(graph, codegen.get(), dex_compilation_unit, &pass_observer);
   } else {
@@ -917,6 +918,7 @@ CodeGenerator* OptimizingCompiler::TryCompile(ArenaAllocator* allocator,
   // this method already, do it now.
   if (jit != nullptr &&
       compilation_kind == CompilationKind::kBaseline &&
+      graph->IsUsefulOptimizing() &&
       graph->GetProfilingInfo() == nullptr) {
     ProfilingInfoBuilder(
         graph, codegen->GetCompilerOptions(), codegen.get(), compilation_stats_.get()).Run();
@@ -1448,6 +1450,11 @@ bool OptimizingCompiler::JitCompile(Thread* self,
     debug_info = GenerateJitDebugInfo(info);
   }
 
+  if (compilation_kind == CompilationKind::kBaseline &&
+      !codegen->GetGraph()->IsUsefulOptimizing()) {
+    compilation_kind = CompilationKind::kOptimized;
+  }
+
   if (!code_cache->Commit(self,
                           region,
                           method,
diff --git a/compiler/optimizing/prepare_for_register_allocation.h b/compiler/optimizing/prepare_for_register_allocation.h
index 0426f8470b..f53c8a1285 100644
--- a/compiler/optimizing/prepare_for_register_allocation.h
+++ b/compiler/optimizing/prepare_for_register_allocation.h
@@ -30,7 +30,7 @@ class OptimizingCompilerStats;
  * For example it changes uses of null checks and bounds checks to the original
  * objects, to avoid creating a live range for these checks.
  */
-class PrepareForRegisterAllocation : public HGraphDelegateVisitor {
+class PrepareForRegisterAllocation final : public HGraphDelegateVisitor {
  public:
   PrepareForRegisterAllocation(HGraph* graph,
                                const CompilerOptions& compiler_options,
diff --git a/compiler/optimizing/profiling_info_builder.cc b/compiler/optimizing/profiling_info_builder.cc
index 19795f5466..7faf2bf5be 100644
--- a/compiler/optimizing/profiling_info_builder.cc
+++ b/compiler/optimizing/profiling_info_builder.cc
@@ -28,14 +28,12 @@
 namespace art HIDDEN {
 
 void ProfilingInfoBuilder::Run() {
+  DCHECK(GetGraph()->IsUsefulOptimizing());
   DCHECK_EQ(GetGraph()->GetProfilingInfo(), nullptr);
   // Order does not matter.
   for (HBasicBlock* block : GetGraph()->GetReversePostOrder()) {
     // No need to visit the phis.
-    for (HInstructionIteratorHandleChanges inst_it(block->GetInstructions()); !inst_it.Done();
-         inst_it.Advance()) {
-      inst_it.Current()->Accept(this);
-    }
+    VisitNonPhiInstructions(block);
   }
 
   ScopedObjectAccess soa(Thread::Current());
@@ -122,6 +120,12 @@ bool ProfilingInfoBuilder::IsInlineCacheUseful(HInvoke* invoke, CodeGenerator* c
       return false;
     }
   }
+
+  if (!codegen->GetGraph()->IsUsefulOptimizing()) {
+    // Earlier pass knew what the calling target was. No need for an inline
+    // cache.
+    return false;
+  }
   return true;
 }
 
diff --git a/compiler/optimizing/profiling_info_builder.h b/compiler/optimizing/profiling_info_builder.h
index c8dc59a03c..ca1c8dd431 100644
--- a/compiler/optimizing/profiling_info_builder.h
+++ b/compiler/optimizing/profiling_info_builder.h
@@ -28,7 +28,7 @@ class HInliner;
 class InlineCache;
 class ProfilingInfo;
 
-class ProfilingInfoBuilder : public HGraphDelegateVisitor {
+class ProfilingInfoBuilder final : public HGraphDelegateVisitor {
  public:
   ProfilingInfoBuilder(HGraph* graph,
                        const CompilerOptions& compiler_options,
diff --git a/compiler/optimizing/reference_type_propagation.cc b/compiler/optimizing/reference_type_propagation.cc
index 6f44d45ed4..91a6fd5d3a 100644
--- a/compiler/optimizing/reference_type_propagation.cc
+++ b/compiler/optimizing/reference_type_propagation.cc
@@ -316,16 +316,11 @@ bool ReferenceTypePropagation::Run() {
 
 void ReferenceTypePropagation::RTPVisitor::VisitBasicBlock(HBasicBlock* block) {
   // Handle Phis first as there might be instructions in the same block who depend on them.
-  for (HInstructionIterator it(block->GetPhis()); !it.Done(); it.Advance()) {
-    VisitPhi(it.Current()->AsPhi());
-  }
+  VisitPhis(block);
 
   // Handle instructions. Since RTP may add HBoundType instructions just after the
   // last visited instruction, use `HInstructionIteratorHandleChanges` iterator.
-  for (HInstructionIteratorHandleChanges it(block->GetInstructions()); !it.Done(); it.Advance()) {
-    HInstruction* instr = it.Current();
-    instr->Accept(this);
-  }
+  VisitNonPhiInstructions(block);
 
   // Add extra nodes to bound types.
   BoundTypeForIfNotNull(block);
diff --git a/compiler/optimizing/scheduler.cc b/compiler/optimizing/scheduler.cc
index 9b5091a81a..f4cf7b0a49 100644
--- a/compiler/optimizing/scheduler.cc
+++ b/compiler/optimizing/scheduler.cc
@@ -106,35 +106,15 @@ static bool IsArrayAccess(const HInstruction* instruction) {
 }
 
 static bool IsInstanceFieldAccess(const HInstruction* instruction) {
-  return instruction->IsInstanceFieldGet() ||
-         instruction->IsInstanceFieldSet() ||
-         instruction->IsUnresolvedInstanceFieldGet() ||
-         instruction->IsUnresolvedInstanceFieldSet();
+  return instruction->IsInstanceFieldGet() || instruction->IsInstanceFieldSet();
 }
 
 static bool IsStaticFieldAccess(const HInstruction* instruction) {
-  return instruction->IsStaticFieldGet() ||
-         instruction->IsStaticFieldSet() ||
-         instruction->IsUnresolvedStaticFieldGet() ||
-         instruction->IsUnresolvedStaticFieldSet();
-}
-
-static bool IsResolvedFieldAccess(const HInstruction* instruction) {
-  return instruction->IsInstanceFieldGet() ||
-         instruction->IsInstanceFieldSet() ||
-         instruction->IsStaticFieldGet() ||
-         instruction->IsStaticFieldSet();
-}
-
-static bool IsUnresolvedFieldAccess(const HInstruction* instruction) {
-  return instruction->IsUnresolvedInstanceFieldGet() ||
-         instruction->IsUnresolvedInstanceFieldSet() ||
-         instruction->IsUnresolvedStaticFieldGet() ||
-         instruction->IsUnresolvedStaticFieldSet();
+  return instruction->IsStaticFieldGet() || instruction->IsStaticFieldSet();
 }
 
 static bool IsFieldAccess(const HInstruction* instruction) {
-  return IsResolvedFieldAccess(instruction) || IsUnresolvedFieldAccess(instruction);
+  return IsInstanceFieldAccess(instruction) || IsStaticFieldAccess(instruction);
 }
 
 static const FieldInfo* GetFieldInfo(const HInstruction* instruction) {
@@ -165,12 +145,6 @@ bool SideEffectDependencyAnalysis::MemoryDependencyAnalysis::FieldAccessMayAlias
     return false;
   }
 
-  // If either of the field accesses is unresolved.
-  if (IsUnresolvedFieldAccess(instr1) || IsUnresolvedFieldAccess(instr2)) {
-    // Conservatively treat these two accesses may alias.
-    return true;
-  }
-
   // If both fields accesses are resolved.
   size_t instr1_field_access_heap_loc = FieldAccessHeapLocation(instr1);
   size_t instr2_field_access_heap_loc = FieldAccessHeapLocation(instr2);
@@ -201,6 +175,14 @@ bool SideEffectDependencyAnalysis::MemoryDependencyAnalysis::HasMemoryDependency
     return true;
   }
 
+  // Note: Unresolved field access instructions are currently marked as not schedulable.
+  // If we change that, we should still keep in mind that these instructions can throw and
+  // read or write volatile fields and, if static, cause class initialization and write to
+  // arbitrary heap locations, and therefore cannot be reordered with any other field or
+  // array access to preserve the observable behavior. The only exception is access to
+  // singleton members that could actually be reodered across these instructions but we
+  // currently do not analyze singletons here anyway.
+
   if (IsArrayAccess(instr1) && IsArrayAccess(instr2)) {
     return ArrayAccessMayAlias(instr1, instr2);
   }
@@ -566,20 +548,10 @@ void HScheduler::Schedule(HGraph* graph) {
 void HScheduler::Schedule(HBasicBlock* block,
                           const HeapLocationCollector* heap_location_collector) {
   ScopedArenaAllocator allocator(block->GetGraph()->GetArenaStack());
-  ScopedArenaVector<SchedulingNode*> scheduling_nodes(allocator.Adapter(kArenaAllocScheduler));
 
   // Build the scheduling graph.
-  SchedulingGraph scheduling_graph(&allocator, heap_location_collector);
-  for (HBackwardInstructionIterator it(block->GetInstructions()); !it.Done(); it.Advance()) {
-    HInstruction* instruction = it.Current();
-    CHECK_EQ(instruction->GetBlock(), block)
-        << instruction->DebugName()
-        << " is in block " << instruction->GetBlock()->GetBlockId()
-        << ", and expected in block " << block->GetBlockId();
-    SchedulingNode* node = scheduling_graph.AddNode(instruction, IsSchedulingBarrier(instruction));
-    CalculateLatency(node);
-    scheduling_nodes.push_back(node);
-  }
+  auto [scheduling_graph, scheduling_nodes] =
+      BuildSchedulingGraph(block, &allocator, heap_location_collector);
 
   if (scheduling_graph.Size() <= 1) {
     return;
@@ -717,7 +689,8 @@ bool HScheduler::IsSchedulable(const HInstruction* instruction) const {
   //    HNop
   //    HThrow
   //    HTryBoundary
-  //    All volatile field access e.g. HInstanceFieldGet
+  //    All unresolved field access instructions
+  //    All volatile field access instructions, e.g. HInstanceFieldGet
   // TODO: Some of the instructions above may be safe to schedule (maybe as
   // scheduling barriers).
   return instruction->IsArrayGet() ||
@@ -820,8 +793,7 @@ bool HInstructionScheduling::Run(bool only_optimize_loop_blocks,
 #if defined(ART_ENABLE_CODEGEN_arm)
     case InstructionSet::kThumb2:
     case InstructionSet::kArm: {
-      arm::SchedulingLatencyVisitorARM arm_latency_visitor(codegen_);
-      arm::HSchedulerARM scheduler(selector, &arm_latency_visitor);
+      arm::HSchedulerARM scheduler(selector, codegen_);
       scheduler.SetOnlyOptimizeLoopBlocks(only_optimize_loop_blocks);
       scheduler.Schedule(graph_);
       break;
diff --git a/compiler/optimizing/scheduler.h b/compiler/optimizing/scheduler.h
index 299fbc93f3..a9672ea732 100644
--- a/compiler/optimizing/scheduler.h
+++ b/compiler/optimizing/scheduler.h
@@ -497,9 +497,8 @@ class CriticalPathSchedulingNodeSelector : public SchedulingNodeSelector {
 
 class HScheduler {
  public:
-  HScheduler(SchedulingLatencyVisitor* latency_visitor, SchedulingNodeSelector* selector)
-      : latency_visitor_(latency_visitor),
-        selector_(selector),
+  explicit HScheduler(SchedulingNodeSelector* selector)
+      : selector_(selector),
         only_optimize_loop_blocks_(true),
         cursor_(nullptr) {}
   virtual ~HScheduler() {}
@@ -512,6 +511,35 @@ class HScheduler {
   virtual bool IsSchedulingBarrier(const HInstruction* instruction) const;
 
  protected:
+  virtual std::pair<SchedulingGraph, ScopedArenaVector<SchedulingNode*>> BuildSchedulingGraph(
+      HBasicBlock* block,
+      ScopedArenaAllocator* allocator,
+      const HeapLocationCollector* heap_location_collector) = 0;
+
+  template <typename LatencyVisitor>
+  std::pair<SchedulingGraph, ScopedArenaVector<SchedulingNode*>> BuildSchedulingGraph(
+      HBasicBlock* block,
+      ScopedArenaAllocator* allocator,
+      const HeapLocationCollector* heap_location_collector,
+      LatencyVisitor* latency_visitor) ALWAYS_INLINE {
+    SchedulingGraph scheduling_graph(allocator, heap_location_collector);
+    ScopedArenaVector<SchedulingNode*> scheduling_nodes(allocator->Adapter(kArenaAllocScheduler));
+    for (HBackwardInstructionIterator it(block->GetInstructions()); !it.Done(); it.Advance()) {
+      HInstruction* instruction = it.Current();
+      CHECK_EQ(instruction->GetBlock(), block)
+          << instruction->DebugName()
+          << " is in block " << instruction->GetBlock()->GetBlockId()
+          << ", and expected in block " << block->GetBlockId();
+      SchedulingNode* node =
+          scheduling_graph.AddNode(instruction, IsSchedulingBarrier(instruction));
+      latency_visitor->CalculateLatency(node);
+      node->SetLatency(latency_visitor->GetLastVisitedLatency());
+      node->SetInternalLatency(latency_visitor->GetLastVisitedInternalLatency());
+      scheduling_nodes.push_back(node);
+    }
+    return {std::move(scheduling_graph), std::move(scheduling_nodes)};
+  }
+
   void Schedule(HBasicBlock* block, const HeapLocationCollector* heap_location_collector);
   void Schedule(SchedulingNode* scheduling_node,
                 /*inout*/ ScopedArenaVector<SchedulingNode*>* candidates);
@@ -529,13 +557,6 @@ class HScheduler {
   virtual bool IsSchedulable(const HInstruction* instruction) const;
   bool IsSchedulable(const HBasicBlock* block) const;
 
-  void CalculateLatency(SchedulingNode* node) {
-    latency_visitor_->CalculateLatency(node);
-    node->SetLatency(latency_visitor_->GetLastVisitedLatency());
-    node->SetInternalLatency(latency_visitor_->GetLastVisitedInternalLatency());
-  }
-
-  SchedulingLatencyVisitor* const latency_visitor_;
   SchedulingNodeSelector* const selector_;
   bool only_optimize_loop_blocks_;
 
diff --git a/compiler/optimizing/scheduler_arm.cc b/compiler/optimizing/scheduler_arm.cc
index 510a0f5496..3ee6f06b46 100644
--- a/compiler/optimizing/scheduler_arm.cc
+++ b/compiler/optimizing/scheduler_arm.cc
@@ -17,6 +17,7 @@
 #include "scheduler_arm.h"
 
 #include "arch/arm/instruction_set_features_arm.h"
+#include "code_generator_arm_vixl.h"
 #include "code_generator_utils.h"
 #include "common_arm.h"
 #include "heap_poisoning.h"
@@ -29,6 +30,116 @@ namespace arm {
 using helpers::Int32ConstantFrom;
 using helpers::Uint64ConstantFrom;
 
+// AArch32 instruction latencies.
+// We currently assume that all ARM CPUs share the same instruction latency list.
+// The following latencies were tuned based on performance experiments and
+// automatic tuning using differential evolution approach on various benchmarks.
+static constexpr uint32_t kArmIntegerOpLatency = 2;
+static constexpr uint32_t kArmFloatingPointOpLatency = 11;
+static constexpr uint32_t kArmDataProcWithShifterOpLatency = 4;
+static constexpr uint32_t kArmMulIntegerLatency = 6;
+static constexpr uint32_t kArmMulFloatingPointLatency = 11;
+static constexpr uint32_t kArmDivIntegerLatency = 10;
+static constexpr uint32_t kArmDivFloatLatency = 20;
+static constexpr uint32_t kArmDivDoubleLatency = 25;
+static constexpr uint32_t kArmTypeConversionFloatingPointIntegerLatency = 11;
+static constexpr uint32_t kArmMemoryLoadLatency = 9;
+static constexpr uint32_t kArmMemoryStoreLatency = 9;
+static constexpr uint32_t kArmMemoryBarrierLatency = 6;
+static constexpr uint32_t kArmBranchLatency = 4;
+static constexpr uint32_t kArmCallLatency = 5;
+static constexpr uint32_t kArmCallInternalLatency = 29;
+static constexpr uint32_t kArmLoadStringInternalLatency = 10;
+static constexpr uint32_t kArmNopLatency = 2;
+static constexpr uint32_t kArmLoadWithBakerReadBarrierLatency = 18;
+static constexpr uint32_t kArmRuntimeTypeCheckLatency = 46;
+
+class SchedulingLatencyVisitorARM final : public SchedulingLatencyVisitor {
+ public:
+  explicit SchedulingLatencyVisitorARM(CodeGenerator* codegen)
+      : codegen_(down_cast<CodeGeneratorARMVIXL*>(codegen)) {}
+
+  // Default visitor for instructions not handled specifically below.
+  void VisitInstruction([[maybe_unused]] HInstruction*) override {
+    last_visited_latency_ = kArmIntegerOpLatency;
+  }
+
+// We add a second unused parameter to be able to use this macro like the others
+// defined in `nodes.h`.
+#define FOR_EACH_SCHEDULED_ARM_INSTRUCTION(M) \
+  M(ArrayGet, unused)                         \
+  M(ArrayLength, unused)                      \
+  M(ArraySet, unused)                         \
+  M(Add, unused)                              \
+  M(Sub, unused)                              \
+  M(And, unused)                              \
+  M(Or, unused)                               \
+  M(Ror, unused)                              \
+  M(Xor, unused)                              \
+  M(Shl, unused)                              \
+  M(Shr, unused)                              \
+  M(UShr, unused)                             \
+  M(Mul, unused)                              \
+  M(Div, unused)                              \
+  M(Condition, unused)                        \
+  M(Compare, unused)                          \
+  M(BoundsCheck, unused)                      \
+  M(InstanceFieldGet, unused)                 \
+  M(InstanceFieldSet, unused)                 \
+  M(InstanceOf, unused)                       \
+  M(Invoke, unused)                           \
+  M(LoadString, unused)                       \
+  M(NewArray, unused)                         \
+  M(NewInstance, unused)                      \
+  M(Rem, unused)                              \
+  M(StaticFieldGet, unused)                   \
+  M(StaticFieldSet, unused)                   \
+  M(SuspendCheck, unused)                     \
+  M(TypeConversion, unused)
+
+#define FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(M) \
+  M(BitwiseNegatedRight, unused)                 \
+  M(MultiplyAccumulate, unused)                  \
+  M(IntermediateAddress, unused)                 \
+  M(IntermediateAddressIndex, unused)            \
+  M(DataProcWithShifterOp, unused)
+
+#define DECLARE_VISIT_INSTRUCTION(type, unused)  \
+  void Visit##type(H##type* instruction) override;
+
+  FOR_EACH_SCHEDULED_ARM_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
+  FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
+  FOR_EACH_CONCRETE_INSTRUCTION_ARM(DECLARE_VISIT_INSTRUCTION)
+
+#undef DECLARE_VISIT_INSTRUCTION
+
+ private:
+  bool CanGenerateTest(HCondition* cond);
+  void HandleGenerateConditionWithZero(IfCondition cond);
+  void HandleGenerateLongTestConstant(HCondition* cond);
+  void HandleGenerateLongTest(HCondition* cond);
+  void HandleGenerateLongComparesAndJumps();
+  void HandleGenerateTest(HCondition* cond);
+  void HandleGenerateConditionGeneric(HCondition* cond);
+  void HandleGenerateEqualLong(HCondition* cond);
+  void HandleGenerateConditionLong(HCondition* cond);
+  void HandleGenerateConditionIntegralOrNonPrimitive(HCondition* cond);
+  void HandleCondition(HCondition* instr);
+  void HandleBinaryOperationLantencies(HBinaryOperation* instr);
+  void HandleBitwiseOperationLantencies(HBinaryOperation* instr);
+  void HandleShiftLatencies(HBinaryOperation* instr);
+  void HandleDivRemConstantIntegralLatencies(int32_t imm);
+  void HandleFieldSetLatencies(HInstruction* instruction, const FieldInfo& field_info);
+  void HandleFieldGetLatencies(HInstruction* instruction, const FieldInfo& field_info);
+  void HandleGenerateDataProcInstruction(bool internal_latency = false);
+  void HandleGenerateDataProc(HDataProcWithShifterOp* instruction);
+  void HandleGenerateLongDataProc(HDataProcWithShifterOp* instruction);
+
+  // The latency setting for each HInstruction depends on how CodeGenerator may generate code,
+  // latency visitors may query CodeGenerator for such information for accurate latency settings.
+  CodeGeneratorARMVIXL* codegen_;
+};
+
 void SchedulingLatencyVisitorARM::HandleBinaryOperationLantencies(HBinaryOperation* instr) {
   switch (instr->GetResultType()) {
     case DataType::Type::kInt64:
@@ -1153,5 +1264,28 @@ void SchedulingLatencyVisitorARM::VisitTypeConversion(HTypeConversion* instr) {
   }
 }
 
+bool HSchedulerARM::IsSchedulable(const HInstruction* instruction) const {
+  switch (instruction->GetKind()) {
+#define SCHEDULABLE_CASE(type, unused)            \
+    case HInstruction::InstructionKind::k##type:  \
+      return true;
+    FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(SCHEDULABLE_CASE)
+    FOR_EACH_CONCRETE_INSTRUCTION_ARM(SCHEDULABLE_CASE)
+#undef SCHEDULABLE_CASE
+
+    default:
+      return HScheduler::IsSchedulable(instruction);
+  }
+}
+
+std::pair<SchedulingGraph, ScopedArenaVector<SchedulingNode*>> HSchedulerARM::BuildSchedulingGraph(
+    HBasicBlock* block,
+    ScopedArenaAllocator* allocator,
+    const HeapLocationCollector* heap_location_collector) {
+  SchedulingLatencyVisitorARM latency_visitor(codegen_);
+  return HScheduler::BuildSchedulingGraph(
+      block, allocator, heap_location_collector, &latency_visitor);
+}
+
 }  // namespace arm
 }  // namespace art
diff --git a/compiler/optimizing/scheduler_arm.h b/compiler/optimizing/scheduler_arm.h
index cf00fa12a3..25eac1b2c4 100644
--- a/compiler/optimizing/scheduler_arm.h
+++ b/compiler/optimizing/scheduler_arm.h
@@ -18,144 +18,32 @@
 #define ART_COMPILER_OPTIMIZING_SCHEDULER_ARM_H_
 
 #include "base/macros.h"
-#include "code_generator_arm_vixl.h"
 #include "scheduler.h"
 
 namespace art HIDDEN {
-namespace arm {
-// AArch32 instruction latencies.
-// We currently assume that all ARM CPUs share the same instruction latency list.
-// The following latencies were tuned based on performance experiments and
-// automatic tuning using differential evolution approach on various benchmarks.
-static constexpr uint32_t kArmIntegerOpLatency = 2;
-static constexpr uint32_t kArmFloatingPointOpLatency = 11;
-static constexpr uint32_t kArmDataProcWithShifterOpLatency = 4;
-static constexpr uint32_t kArmMulIntegerLatency = 6;
-static constexpr uint32_t kArmMulFloatingPointLatency = 11;
-static constexpr uint32_t kArmDivIntegerLatency = 10;
-static constexpr uint32_t kArmDivFloatLatency = 20;
-static constexpr uint32_t kArmDivDoubleLatency = 25;
-static constexpr uint32_t kArmTypeConversionFloatingPointIntegerLatency = 11;
-static constexpr uint32_t kArmMemoryLoadLatency = 9;
-static constexpr uint32_t kArmMemoryStoreLatency = 9;
-static constexpr uint32_t kArmMemoryBarrierLatency = 6;
-static constexpr uint32_t kArmBranchLatency = 4;
-static constexpr uint32_t kArmCallLatency = 5;
-static constexpr uint32_t kArmCallInternalLatency = 29;
-static constexpr uint32_t kArmLoadStringInternalLatency = 10;
-static constexpr uint32_t kArmNopLatency = 2;
-static constexpr uint32_t kArmLoadWithBakerReadBarrierLatency = 18;
-static constexpr uint32_t kArmRuntimeTypeCheckLatency = 46;
-
-class SchedulingLatencyVisitorARM final : public SchedulingLatencyVisitor {
- public:
-  explicit SchedulingLatencyVisitorARM(CodeGenerator* codegen)
-      : codegen_(down_cast<CodeGeneratorARMVIXL*>(codegen)) {}
-
-  // Default visitor for instructions not handled specifically below.
-  void VisitInstruction([[maybe_unused]] HInstruction*) override {
-    last_visited_latency_ = kArmIntegerOpLatency;
-  }
-
-// We add a second unused parameter to be able to use this macro like the others
-// defined in `nodes.h`.
-#define FOR_EACH_SCHEDULED_ARM_INSTRUCTION(M) \
-  M(ArrayGet, unused)                         \
-  M(ArrayLength, unused)                      \
-  M(ArraySet, unused)                         \
-  M(Add, unused)                              \
-  M(Sub, unused)                              \
-  M(And, unused)                              \
-  M(Or, unused)                               \
-  M(Ror, unused)                              \
-  M(Xor, unused)                              \
-  M(Shl, unused)                              \
-  M(Shr, unused)                              \
-  M(UShr, unused)                             \
-  M(Mul, unused)                              \
-  M(Div, unused)                              \
-  M(Condition, unused)                        \
-  M(Compare, unused)                          \
-  M(BoundsCheck, unused)                      \
-  M(InstanceFieldGet, unused)                 \
-  M(InstanceFieldSet, unused)                 \
-  M(InstanceOf, unused)                       \
-  M(Invoke, unused)                           \
-  M(LoadString, unused)                       \
-  M(NewArray, unused)                         \
-  M(NewInstance, unused)                      \
-  M(Rem, unused)                              \
-  M(StaticFieldGet, unused)                   \
-  M(StaticFieldSet, unused)                   \
-  M(SuspendCheck, unused)                     \
-  M(TypeConversion, unused)
-
-#define FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(M) \
-  M(BitwiseNegatedRight, unused)                 \
-  M(MultiplyAccumulate, unused)                  \
-  M(IntermediateAddress, unused)                 \
-  M(IntermediateAddressIndex, unused)            \
-  M(DataProcWithShifterOp, unused)
-
-#define DECLARE_VISIT_INSTRUCTION(type, unused)  \
-  void Visit##type(H##type* instruction) override;
 
-  FOR_EACH_SCHEDULED_ARM_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
-  FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
-  FOR_EACH_CONCRETE_INSTRUCTION_ARM(DECLARE_VISIT_INSTRUCTION)
+class CodeGenerator;
 
-#undef DECLARE_VISIT_INSTRUCTION
-
- private:
-  bool CanGenerateTest(HCondition* cond);
-  void HandleGenerateConditionWithZero(IfCondition cond);
-  void HandleGenerateLongTestConstant(HCondition* cond);
-  void HandleGenerateLongTest(HCondition* cond);
-  void HandleGenerateLongComparesAndJumps();
-  void HandleGenerateTest(HCondition* cond);
-  void HandleGenerateConditionGeneric(HCondition* cond);
-  void HandleGenerateEqualLong(HCondition* cond);
-  void HandleGenerateConditionLong(HCondition* cond);
-  void HandleGenerateConditionIntegralOrNonPrimitive(HCondition* cond);
-  void HandleCondition(HCondition* instr);
-  void HandleBinaryOperationLantencies(HBinaryOperation* instr);
-  void HandleBitwiseOperationLantencies(HBinaryOperation* instr);
-  void HandleShiftLatencies(HBinaryOperation* instr);
-  void HandleDivRemConstantIntegralLatencies(int32_t imm);
-  void HandleFieldSetLatencies(HInstruction* instruction, const FieldInfo& field_info);
-  void HandleFieldGetLatencies(HInstruction* instruction, const FieldInfo& field_info);
-  void HandleGenerateDataProcInstruction(bool internal_latency = false);
-  void HandleGenerateDataProc(HDataProcWithShifterOp* instruction);
-  void HandleGenerateLongDataProc(HDataProcWithShifterOp* instruction);
-
-  // The latency setting for each HInstruction depends on how CodeGenerator may generate code,
-  // latency visitors may query CodeGenerator for such information for accurate latency settings.
-  CodeGeneratorARMVIXL* codegen_;
-};
+namespace arm {
 
-class HSchedulerARM : public HScheduler {
+class HSchedulerARM final : public HScheduler {
  public:
-  HSchedulerARM(SchedulingNodeSelector* selector,
-                SchedulingLatencyVisitorARM* arm_latency_visitor)
-      : HScheduler(arm_latency_visitor, selector) {}
+  HSchedulerARM(SchedulingNodeSelector* selector, CodeGenerator* codegen)
+      : HScheduler(selector), codegen_(codegen) {}
   ~HSchedulerARM() override {}
 
-  bool IsSchedulable(const HInstruction* instruction) const override {
-#define CASE_INSTRUCTION_KIND(type, unused) case \
-  HInstruction::InstructionKind::k##type:
-    switch (instruction->GetKind()) {
-      FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(CASE_INSTRUCTION_KIND)
-        return true;
-      FOR_EACH_CONCRETE_INSTRUCTION_ARM(CASE_INSTRUCTION_KIND)
-        return true;
-      default:
-        return HScheduler::IsSchedulable(instruction);
-    }
-#undef CASE_INSTRUCTION_KIND
-  }
+  bool IsSchedulable(const HInstruction* instruction) const override;
+
+ protected:
+  std::pair<SchedulingGraph, ScopedArenaVector<SchedulingNode*>> BuildSchedulingGraph(
+      HBasicBlock* block,
+      ScopedArenaAllocator* allocator,
+      const HeapLocationCollector* heap_location_collector) override;
 
  private:
   DISALLOW_COPY_AND_ASSIGN(HSchedulerARM);
+
+  CodeGenerator* const codegen_;
 };
 
 }  // namespace arm
diff --git a/compiler/optimizing/scheduler_arm64.cc b/compiler/optimizing/scheduler_arm64.cc
index 5113cf446d..08b8a3fb78 100644
--- a/compiler/optimizing/scheduler_arm64.cc
+++ b/compiler/optimizing/scheduler_arm64.cc
@@ -23,6 +23,115 @@
 namespace art HIDDEN {
 namespace arm64 {
 
+static constexpr uint32_t kArm64MemoryLoadLatency = 5;
+static constexpr uint32_t kArm64MemoryStoreLatency = 3;
+
+static constexpr uint32_t kArm64CallInternalLatency = 10;
+static constexpr uint32_t kArm64CallLatency = 5;
+
+// AArch64 instruction latency.
+// We currently assume that all arm64 CPUs share the same instruction latency list.
+static constexpr uint32_t kArm64IntegerOpLatency = 2;
+static constexpr uint32_t kArm64FloatingPointOpLatency = 5;
+
+static constexpr uint32_t kArm64DataProcWithShifterOpLatency = 3;
+static constexpr uint32_t kArm64DivDoubleLatency = 30;
+static constexpr uint32_t kArm64DivFloatLatency = 15;
+static constexpr uint32_t kArm64DivIntegerLatency = 5;
+static constexpr uint32_t kArm64LoadStringInternalLatency = 7;
+static constexpr uint32_t kArm64MulFloatingPointLatency = 6;
+static constexpr uint32_t kArm64MulIntegerLatency = 6;
+static constexpr uint32_t kArm64TypeConversionFloatingPointIntegerLatency = 5;
+static constexpr uint32_t kArm64BranchLatency = kArm64IntegerOpLatency;
+
+static constexpr uint32_t kArm64SIMDFloatingPointOpLatency = 10;
+static constexpr uint32_t kArm64SIMDIntegerOpLatency = 6;
+static constexpr uint32_t kArm64SIMDMemoryLoadLatency = 10;
+static constexpr uint32_t kArm64SIMDMemoryStoreLatency = 6;
+static constexpr uint32_t kArm64SIMDMulFloatingPointLatency = 12;
+static constexpr uint32_t kArm64SIMDMulIntegerLatency = 12;
+static constexpr uint32_t kArm64SIMDReplicateOpLatency = 16;
+static constexpr uint32_t kArm64SIMDDivDoubleLatency = 60;
+static constexpr uint32_t kArm64SIMDDivFloatLatency = 30;
+static constexpr uint32_t kArm64SIMDTypeConversionInt2FPLatency = 10;
+
+class SchedulingLatencyVisitorARM64 final : public SchedulingLatencyVisitor {
+ public:
+  // Default visitor for instructions not handled specifically below.
+  void VisitInstruction([[maybe_unused]] HInstruction*) override {
+    last_visited_latency_ = kArm64IntegerOpLatency;
+  }
+
+// We add a second unused parameter to be able to use this macro like the others
+// defined in `nodes.h`.
+#define FOR_EACH_SCHEDULED_COMMON_INSTRUCTION(M)     \
+  M(ArrayGet             , unused)                   \
+  M(ArrayLength          , unused)                   \
+  M(ArraySet             , unused)                   \
+  M(BoundsCheck          , unused)                   \
+  M(Div                  , unused)                   \
+  M(InstanceFieldGet     , unused)                   \
+  M(InstanceOf           , unused)                   \
+  M(LoadString           , unused)                   \
+  M(Mul                  , unused)                   \
+  M(NewArray             , unused)                   \
+  M(NewInstance          , unused)                   \
+  M(Rem                  , unused)                   \
+  M(StaticFieldGet       , unused)                   \
+  M(SuspendCheck         , unused)                   \
+  M(TypeConversion       , unused)                   \
+  M(VecReplicateScalar   , unused)                   \
+  M(VecExtractScalar     , unused)                   \
+  M(VecReduce            , unused)                   \
+  M(VecCnv               , unused)                   \
+  M(VecNeg               , unused)                   \
+  M(VecAbs               , unused)                   \
+  M(VecNot               , unused)                   \
+  M(VecAdd               , unused)                   \
+  M(VecHalvingAdd        , unused)                   \
+  M(VecSub               , unused)                   \
+  M(VecMul               , unused)                   \
+  M(VecDiv               , unused)                   \
+  M(VecMin               , unused)                   \
+  M(VecMax               , unused)                   \
+  M(VecAnd               , unused)                   \
+  M(VecAndNot            , unused)                   \
+  M(VecOr                , unused)                   \
+  M(VecXor               , unused)                   \
+  M(VecShl               , unused)                   \
+  M(VecShr               , unused)                   \
+  M(VecUShr              , unused)                   \
+  M(VecSetScalars        , unused)                   \
+  M(VecMultiplyAccumulate, unused)                   \
+  M(VecLoad              , unused)                   \
+  M(VecStore             , unused)
+
+#define FOR_EACH_SCHEDULED_ABSTRACT_INSTRUCTION(M)   \
+  M(BinaryOperation      , unused)                   \
+  M(Invoke               , unused)
+
+#define FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(M) \
+  M(BitwiseNegatedRight, unused)                 \
+  M(MultiplyAccumulate, unused)                  \
+  M(IntermediateAddress, unused)                 \
+  M(IntermediateAddressIndex, unused)            \
+  M(DataProcWithShifterOp, unused)
+
+#define DECLARE_VISIT_INSTRUCTION(type, unused)  \
+  void Visit##type(H##type* instruction) override;
+
+  FOR_EACH_SCHEDULED_COMMON_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
+  FOR_EACH_SCHEDULED_ABSTRACT_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
+  FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
+  FOR_EACH_CONCRETE_INSTRUCTION_ARM64(DECLARE_VISIT_INSTRUCTION)
+
+#undef DECLARE_VISIT_INSTRUCTION
+
+ private:
+  void HandleSimpleArithmeticSIMD(HVecOperation *instr);
+  void HandleVecAddress(HVecMemoryOperation* instruction, size_t size);
+};
+
 void SchedulingLatencyVisitorARM64::VisitBinaryOperation(HBinaryOperation* instr) {
   last_visited_latency_ = DataType::IsFloatingPointType(instr->GetResultType())
       ? kArm64FloatingPointOpLatency
@@ -348,5 +457,30 @@ void SchedulingLatencyVisitorARM64::VisitVecStore(HVecStore* instr) {
   last_visited_latency_ = kArm64SIMDMemoryStoreLatency;
 }
 
+bool HSchedulerARM64::IsSchedulable(const HInstruction* instruction) const {
+  switch (instruction->GetKind()) {
+#define SCHEDULABLE_CASE(type, unused)       \
+    case HInstruction::InstructionKind::k##type:  \
+      return true;
+    FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(SCHEDULABLE_CASE)
+    FOR_EACH_CONCRETE_INSTRUCTION_ARM64(SCHEDULABLE_CASE)
+    FOR_EACH_SCHEDULED_COMMON_INSTRUCTION(SCHEDULABLE_CASE)
+#undef SCHEDULABLE_CASE
+
+    default:
+      return HScheduler::IsSchedulable(instruction);
+  }
+}
+
+std::pair<SchedulingGraph, ScopedArenaVector<SchedulingNode*>>
+HSchedulerARM64::BuildSchedulingGraph(
+    HBasicBlock* block,
+    ScopedArenaAllocator* allocator,
+    const HeapLocationCollector* heap_location_collector) {
+  SchedulingLatencyVisitorARM64 latency_visitor;
+  return HScheduler::BuildSchedulingGraph(
+      block, allocator, heap_location_collector, &latency_visitor);
+}
+
 }  // namespace arm64
 }  // namespace art
diff --git a/compiler/optimizing/scheduler_arm64.h b/compiler/optimizing/scheduler_arm64.h
index 7ce00e00ab..044aa48a5a 100644
--- a/compiler/optimizing/scheduler_arm64.h
+++ b/compiler/optimizing/scheduler_arm64.h
@@ -23,137 +23,13 @@
 namespace art HIDDEN {
 namespace arm64 {
 
-static constexpr uint32_t kArm64MemoryLoadLatency = 5;
-static constexpr uint32_t kArm64MemoryStoreLatency = 3;
-
-static constexpr uint32_t kArm64CallInternalLatency = 10;
-static constexpr uint32_t kArm64CallLatency = 5;
-
-// AArch64 instruction latency.
-// We currently assume that all arm64 CPUs share the same instruction latency list.
-static constexpr uint32_t kArm64IntegerOpLatency = 2;
-static constexpr uint32_t kArm64FloatingPointOpLatency = 5;
-
-
-static constexpr uint32_t kArm64DataProcWithShifterOpLatency = 3;
-static constexpr uint32_t kArm64DivDoubleLatency = 30;
-static constexpr uint32_t kArm64DivFloatLatency = 15;
-static constexpr uint32_t kArm64DivIntegerLatency = 5;
-static constexpr uint32_t kArm64LoadStringInternalLatency = 7;
-static constexpr uint32_t kArm64MulFloatingPointLatency = 6;
-static constexpr uint32_t kArm64MulIntegerLatency = 6;
-static constexpr uint32_t kArm64TypeConversionFloatingPointIntegerLatency = 5;
-static constexpr uint32_t kArm64BranchLatency = kArm64IntegerOpLatency;
-
-static constexpr uint32_t kArm64SIMDFloatingPointOpLatency = 10;
-static constexpr uint32_t kArm64SIMDIntegerOpLatency = 6;
-static constexpr uint32_t kArm64SIMDMemoryLoadLatency = 10;
-static constexpr uint32_t kArm64SIMDMemoryStoreLatency = 6;
-static constexpr uint32_t kArm64SIMDMulFloatingPointLatency = 12;
-static constexpr uint32_t kArm64SIMDMulIntegerLatency = 12;
-static constexpr uint32_t kArm64SIMDReplicateOpLatency = 16;
-static constexpr uint32_t kArm64SIMDDivDoubleLatency = 60;
-static constexpr uint32_t kArm64SIMDDivFloatLatency = 30;
-static constexpr uint32_t kArm64SIMDTypeConversionInt2FPLatency = 10;
-
-class SchedulingLatencyVisitorARM64 final : public SchedulingLatencyVisitor {
- public:
-  // Default visitor for instructions not handled specifically below.
-  void VisitInstruction([[maybe_unused]] HInstruction*) override {
-    last_visited_latency_ = kArm64IntegerOpLatency;
-  }
-
-// We add a second unused parameter to be able to use this macro like the others
-// defined in `nodes.h`.
-#define FOR_EACH_SCHEDULED_COMMON_INSTRUCTION(M)     \
-  M(ArrayGet             , unused)                   \
-  M(ArrayLength          , unused)                   \
-  M(ArraySet             , unused)                   \
-  M(BoundsCheck          , unused)                   \
-  M(Div                  , unused)                   \
-  M(InstanceFieldGet     , unused)                   \
-  M(InstanceOf           , unused)                   \
-  M(LoadString           , unused)                   \
-  M(Mul                  , unused)                   \
-  M(NewArray             , unused)                   \
-  M(NewInstance          , unused)                   \
-  M(Rem                  , unused)                   \
-  M(StaticFieldGet       , unused)                   \
-  M(SuspendCheck         , unused)                   \
-  M(TypeConversion       , unused)                   \
-  M(VecReplicateScalar   , unused)                   \
-  M(VecExtractScalar     , unused)                   \
-  M(VecReduce            , unused)                   \
-  M(VecCnv               , unused)                   \
-  M(VecNeg               , unused)                   \
-  M(VecAbs               , unused)                   \
-  M(VecNot               , unused)                   \
-  M(VecAdd               , unused)                   \
-  M(VecHalvingAdd        , unused)                   \
-  M(VecSub               , unused)                   \
-  M(VecMul               , unused)                   \
-  M(VecDiv               , unused)                   \
-  M(VecMin               , unused)                   \
-  M(VecMax               , unused)                   \
-  M(VecAnd               , unused)                   \
-  M(VecAndNot            , unused)                   \
-  M(VecOr                , unused)                   \
-  M(VecXor               , unused)                   \
-  M(VecShl               , unused)                   \
-  M(VecShr               , unused)                   \
-  M(VecUShr              , unused)                   \
-  M(VecSetScalars        , unused)                   \
-  M(VecMultiplyAccumulate, unused)                   \
-  M(VecLoad              , unused)                   \
-  M(VecStore             , unused)
-
-#define FOR_EACH_SCHEDULED_ABSTRACT_INSTRUCTION(M)   \
-  M(BinaryOperation      , unused)                   \
-  M(Invoke               , unused)
-
-#define FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(M) \
-  M(BitwiseNegatedRight, unused)                 \
-  M(MultiplyAccumulate, unused)                  \
-  M(IntermediateAddress, unused)                 \
-  M(IntermediateAddressIndex, unused)            \
-  M(DataProcWithShifterOp, unused)
-
-#define DECLARE_VISIT_INSTRUCTION(type, unused)  \
-  void Visit##type(H##type* instruction) override;
-
-  FOR_EACH_SCHEDULED_COMMON_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
-  FOR_EACH_SCHEDULED_ABSTRACT_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
-  FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(DECLARE_VISIT_INSTRUCTION)
-  FOR_EACH_CONCRETE_INSTRUCTION_ARM64(DECLARE_VISIT_INSTRUCTION)
-
-#undef DECLARE_VISIT_INSTRUCTION
-
- private:
-  void HandleSimpleArithmeticSIMD(HVecOperation *instr);
-  void HandleVecAddress(HVecMemoryOperation* instruction, size_t size);
-};
-
 class HSchedulerARM64 : public HScheduler {
  public:
   explicit HSchedulerARM64(SchedulingNodeSelector* selector)
-      : HScheduler(&arm64_latency_visitor_, selector) {}
+      : HScheduler(selector) {}
   ~HSchedulerARM64() override {}
 
-  bool IsSchedulable(const HInstruction* instruction) const override {
-#define CASE_INSTRUCTION_KIND(type, unused) case \
-  HInstruction::InstructionKind::k##type:
-    switch (instruction->GetKind()) {
-      FOR_EACH_SCHEDULED_SHARED_INSTRUCTION(CASE_INSTRUCTION_KIND)
-        return true;
-      FOR_EACH_CONCRETE_INSTRUCTION_ARM64(CASE_INSTRUCTION_KIND)
-        return true;
-      FOR_EACH_SCHEDULED_COMMON_INSTRUCTION(CASE_INSTRUCTION_KIND)
-        return true;
-      default:
-        return HScheduler::IsSchedulable(instruction);
-    }
-#undef CASE_INSTRUCTION_KIND
-  }
+  bool IsSchedulable(const HInstruction* instruction) const override;
 
   // Treat as scheduling barriers those vector instructions whose live ranges exceed the vectorized
   // loop boundaries. This is a workaround for the lack of notion of SIMD register in the compiler;
@@ -169,8 +45,13 @@ class HSchedulerARM64 : public HScheduler {
            instr->IsVecReplicateScalar();
   }
 
+ protected:
+  std::pair<SchedulingGraph, ScopedArenaVector<SchedulingNode*>> BuildSchedulingGraph(
+      HBasicBlock* block,
+      ScopedArenaAllocator* allocator,
+      const HeapLocationCollector* heap_location_collector) override;
+
  private:
-  SchedulingLatencyVisitorARM64 arm64_latency_visitor_;
   DISALLOW_COPY_AND_ASSIGN(HSchedulerARM64);
 };
 
diff --git a/compiler/optimizing/scheduler_test.cc b/compiler/optimizing/scheduler_test.cc
index c2b1fd6f7c..0b020f1460 100644
--- a/compiler/optimizing/scheduler_test.cc
+++ b/compiler/optimizing/scheduler_test.cc
@@ -406,15 +406,13 @@ TEST_F(SchedulerTest, ArrayAccessAliasingARM64) {
 #if defined(ART_ENABLE_CODEGEN_arm)
 TEST_F(SchedulerTest, DependencyGraphAndSchedulerARM) {
   CriticalPathSchedulingNodeSelector critical_path_selector;
-  arm::SchedulingLatencyVisitorARM arm_latency_visitor(/*CodeGenerator*/ nullptr);
-  arm::HSchedulerARM scheduler(&critical_path_selector, &arm_latency_visitor);
+  arm::HSchedulerARM scheduler(&critical_path_selector, /*codegen=*/ nullptr);
   TestBuildDependencyGraphAndSchedule(&scheduler);
 }
 
 TEST_F(SchedulerTest, ArrayAccessAliasingARM) {
   CriticalPathSchedulingNodeSelector critical_path_selector;
-  arm::SchedulingLatencyVisitorARM arm_latency_visitor(/*CodeGenerator*/ nullptr);
-  arm::HSchedulerARM scheduler(&critical_path_selector, &arm_latency_visitor);
+  arm::HSchedulerARM scheduler(&critical_path_selector, /*codegen=*/ nullptr);
   TestDependencyGraphOnAliasingArrayAccesses(&scheduler);
 }
 #endif
diff --git a/compiler/optimizing/ssa_liveness_analysis_test.cc b/compiler/optimizing/ssa_liveness_analysis_test.cc
index 18c945381d..2df0f34c7d 100644
--- a/compiler/optimizing/ssa_liveness_analysis_test.cc
+++ b/compiler/optimizing/ssa_liveness_analysis_test.cc
@@ -31,7 +31,6 @@ namespace art HIDDEN {
 class SsaLivenessAnalysisTest : public OptimizingUnitTest {
  protected:
   void SetUp() override {
-    TEST_SETUP_DISABLED_FOR_RISCV64();
     OptimizingUnitTest::SetUp();
     graph_ = CreateGraph();
     compiler_options_ = CommonCompilerTest::CreateCompilerOptions(kRuntimeISA, "default");
@@ -43,11 +42,6 @@ class SsaLivenessAnalysisTest : public OptimizingUnitTest {
     graph_->SetEntryBlock(entry_);
   }
 
-  void TearDown() override {
-    TEST_TEARDOWN_DISABLED_FOR_RISCV64();
-    OptimizingUnitTest::TearDown();
-  }
-
  protected:
   HBasicBlock* CreateSuccessor(HBasicBlock* block) {
     HGraph* graph = block->GetGraph();
@@ -64,7 +58,6 @@ class SsaLivenessAnalysisTest : public OptimizingUnitTest {
 };
 
 TEST_F(SsaLivenessAnalysisTest, TestReturnArg) {
-  TEST_DISABLED_FOR_RISCV64();
   HInstruction* arg = new (GetAllocator()) HParameterValue(
       graph_->GetDexFile(), dex::TypeIndex(0), 0, DataType::Type::kInt32);
   entry_->AddInstruction(arg);
@@ -85,7 +78,6 @@ TEST_F(SsaLivenessAnalysisTest, TestReturnArg) {
 }
 
 TEST_F(SsaLivenessAnalysisTest, TestAput) {
-  TEST_DISABLED_FOR_RISCV64();
   HInstruction* array = new (GetAllocator()) HParameterValue(
       graph_->GetDexFile(), dex::TypeIndex(0), 0, DataType::Type::kReference);
   HInstruction* index = new (GetAllocator()) HParameterValue(
@@ -155,7 +147,6 @@ TEST_F(SsaLivenessAnalysisTest, TestAput) {
 }
 
 TEST_F(SsaLivenessAnalysisTest, TestDeoptimize) {
-  TEST_DISABLED_FOR_RISCV64();
   HInstruction* array = new (GetAllocator()) HParameterValue(
       graph_->GetDexFile(), dex::TypeIndex(0), 0, DataType::Type::kReference);
   HInstruction* index = new (GetAllocator()) HParameterValue(
diff --git a/compiler/optimizing/stack_map_stream.h b/compiler/optimizing/stack_map_stream.h
index 8c7b1c01a7..a3daa29b4e 100644
--- a/compiler/optimizing/stack_map_stream.h
+++ b/compiler/optimizing/stack_map_stream.h
@@ -110,7 +110,7 @@ class StackMapStream : public DeletableArenaObject<kArenaAllocStackMapStream> {
 
   // Invokes the callback with pointer of each BitTableBuilder field.
   template<typename Callback>
-  void ForEachBitTable(Callback callback) {
+  void ForEachBitTable(Callback&& callback) {
     size_t index = 0;
     callback(index++, &stack_maps_);
     callback(index++, &register_masks_);
diff --git a/compiler/optimizing/write_barrier_elimination.cc b/compiler/optimizing/write_barrier_elimination.cc
index 27348cd87d..537bc09f93 100644
--- a/compiler/optimizing/write_barrier_elimination.cc
+++ b/compiler/optimizing/write_barrier_elimination.cc
@@ -38,14 +38,14 @@ class WBEVisitor final : public HGraphVisitor {
     // We clear the map to perform this optimization only in the same block. Doing it across blocks
     // would entail non-trivial merging of states.
     current_write_barriers_.clear();
-    HGraphVisitor::VisitBasicBlock(block);
+    VisitNonPhiInstructions(block);
   }
 
   void VisitInstanceFieldSet(HInstanceFieldSet* instruction) override {
     DCHECK(!instruction->GetSideEffects().Includes(SideEffects::CanTriggerGC()));
 
     if (instruction->GetFieldType() != DataType::Type::kReference ||
-        instruction->GetValue()->IsNullConstant()) {
+        HuntForOriginalReference(instruction->GetValue())->IsNullConstant()) {
       instruction->SetWriteBarrierKind(WriteBarrierKind::kDontEmit);
       return;
     }
@@ -72,7 +72,7 @@ class WBEVisitor final : public HGraphVisitor {
     DCHECK(!instruction->GetSideEffects().Includes(SideEffects::CanTriggerGC()));
 
     if (instruction->GetFieldType() != DataType::Type::kReference ||
-        instruction->GetValue()->IsNullConstant()) {
+        HuntForOriginalReference(instruction->GetValue())->IsNullConstant()) {
       instruction->SetWriteBarrierKind(WriteBarrierKind::kDontEmit);
       return;
     }
@@ -100,7 +100,7 @@ class WBEVisitor final : public HGraphVisitor {
     }
 
     if (instruction->GetComponentType() != DataType::Type::kReference ||
-        instruction->GetValue()->IsNullConstant()) {
+        HuntForOriginalReference(instruction->GetValue())->IsNullConstant()) {
       instruction->SetWriteBarrierKind(WriteBarrierKind::kDontEmit);
       return;
     }
diff --git a/compiler/trampolines/trampoline_compiler.cc b/compiler/trampolines/trampoline_compiler.cc
index d9f56629ef..b3fc688cc9 100644
--- a/compiler/trampolines/trampoline_compiler.cc
+++ b/compiler/trampolines/trampoline_compiler.cc
@@ -67,7 +67,7 @@ static std::unique_ptr<const std::vector<uint8_t>> CreateTrampoline(
 
       // VIXL will use the destination as a scratch register if
       // the offset is not encodable as an immediate operand.
-      ___ Ldr(temp_reg, MemOperand(r0, JNIEnvExt::SelfOffset(4).Int32Value()));
+      ___ Ldr(temp_reg, MemOperand(r0, JNIEnvExt::SelfOffset(kArmPointerSize).Int32Value()));
       ___ Ldr(pc, MemOperand(temp_reg, offset.Int32Value()));
       break;
     }
@@ -99,7 +99,7 @@ static std::unique_ptr<const std::vector<uint8_t>> CreateTrampoline(
     case kJniAbi:  // Load via Thread* held in JNIEnv* in first argument (X0).
       __ LoadRawPtr(Arm64ManagedRegister::FromXRegister(IP1),
                       Arm64ManagedRegister::FromXRegister(X0),
-                      Offset(JNIEnvExt::SelfOffset(8).Int32Value()));
+                      Offset(JNIEnvExt::SelfOffset(kArm64PointerSize).Int32Value()));
 
       __ JumpTo(Arm64ManagedRegister::FromXRegister(IP1), Offset(offset.Int32Value()),
                 Arm64ManagedRegister::FromXRegister(IP0));
@@ -134,9 +134,7 @@ static std::unique_ptr<const std::vector<uint8_t>> CreateTrampoline(ArenaAllocat
 
   switch (abi) {
     case kJniAbi:  // Load via Thread* held in JNIEnv* in first argument (A0).
-      __ Loadd(tmp,
-               A0,
-               JNIEnvExt::SelfOffset(static_cast<size_t>(kRiscv64PointerSize)).Int32Value());
+      __ Loadd(tmp, A0, JNIEnvExt::SelfOffset(kRiscv64PointerSize).Int32Value());
       __ Loadd(tmp, tmp, offset.Int32Value());
       __ Jr(tmp);
       break;
diff --git a/compiler/utils/arm/jni_macro_assembler_arm_vixl.cc b/compiler/utils/arm/jni_macro_assembler_arm_vixl.cc
index 9c2589138c..6844b1e656 100644
--- a/compiler/utils/arm/jni_macro_assembler_arm_vixl.cc
+++ b/compiler/utils/arm/jni_macro_assembler_arm_vixl.cc
@@ -21,6 +21,8 @@
 
 #include "entrypoints/quick/quick_entrypoints.h"
 #include "indirect_reference_table.h"
+#include "jni/jni_env_ext.h"
+#include "jni/local_reference_table.h"
 #include "lock_word.h"
 #include "thread.h"
 
@@ -1001,7 +1003,7 @@ void ArmVIXLJNIMacroAssembler::DeliverPendingException() {
 }
 
 std::unique_ptr<JNIMacroLabel> ArmVIXLJNIMacroAssembler::CreateLabel() {
-  return std::unique_ptr<JNIMacroLabel>(new ArmVIXLJNIMacroLabel());
+  return std::unique_ptr<JNIMacroLabel>(new (asm_.GetAllocator()) ArmVIXLJNIMacroLabel());
 }
 
 void ArmVIXLJNIMacroAssembler::Jump(JNIMacroLabel* label) {
@@ -1108,5 +1110,35 @@ void ArmVIXLJNIMacroAssembler::Load(ArmManagedRegister dest,
   }
 }
 
+void ArmVIXLJNIMacroAssembler::LoadLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                                             ManagedRegister previous_state_reg,
+                                                             ManagedRegister current_state_reg) {
+  constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+  DCHECK_EQ(kLRTSegmentStateSize, kRegSizeInBytes);
+  const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kArmPointerSize);
+  const MemberOffset current_state_offset = JNIEnvExt::LrtSegmentStateOffset(kArmPointerSize);
+  DCHECK_EQ(previous_state_offset.SizeValue() + kLRTSegmentStateSize,
+            current_state_offset.SizeValue());
+
+  ___ Ldrd(AsVIXLRegister(previous_state_reg.AsArm()),
+           AsVIXLRegister(current_state_reg.AsArm()),
+           MemOperand(AsVIXLRegister(jni_env_reg.AsArm()), previous_state_offset.Int32Value()));
+}
+
+void ArmVIXLJNIMacroAssembler::StoreLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                                              ManagedRegister previous_state_reg,
+                                                              ManagedRegister current_state_reg) {
+  constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+  DCHECK_EQ(kLRTSegmentStateSize, kRegSizeInBytes);
+  const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kArmPointerSize);
+  const MemberOffset current_state_offset = JNIEnvExt::LrtSegmentStateOffset(kArmPointerSize);
+  DCHECK_EQ(previous_state_offset.SizeValue() + kLRTSegmentStateSize,
+            current_state_offset.SizeValue());
+
+  ___ Strd(AsVIXLRegister(previous_state_reg.AsArm()),
+           AsVIXLRegister(current_state_reg.AsArm()),
+           MemOperand(AsVIXLRegister(jni_env_reg.AsArm()), previous_state_offset.Int32Value()));
+}
+
 }  // namespace arm
 }  // namespace art
diff --git a/compiler/utils/arm/jni_macro_assembler_arm_vixl.h b/compiler/utils/arm/jni_macro_assembler_arm_vixl.h
index f6df7f2c53..3f6512c2a8 100644
--- a/compiler/utils/arm/jni_macro_assembler_arm_vixl.h
+++ b/compiler/utils/arm/jni_macro_assembler_arm_vixl.h
@@ -91,6 +91,14 @@ class ArmVIXLJNIMacroAssembler final
   void GetCurrentThread(ManagedRegister dest) override;
   void GetCurrentThread(FrameOffset dest_offset) override;
 
+  // Manipulating local reference table states.
+  void LoadLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                     ManagedRegister previous_state_reg,
+                                     ManagedRegister current_state_reg) override;
+  void StoreLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                      ManagedRegister previous_state_reg,
+                                      ManagedRegister current_state_reg) override;
+
   // Decode JNI transition or local `jobject`. For (weak) global `jobject`, jump to slow path.
   void DecodeJNITransitionOrLocalJObject(ManagedRegister reg,
                                          JNIMacroLabel* slow_path,
diff --git a/compiler/utils/arm64/jni_macro_assembler_arm64.cc b/compiler/utils/arm64/jni_macro_assembler_arm64.cc
index 8ce44b6c63..50f6b4158d 100644
--- a/compiler/utils/arm64/jni_macro_assembler_arm64.cc
+++ b/compiler/utils/arm64/jni_macro_assembler_arm64.cc
@@ -18,6 +18,8 @@
 
 #include "entrypoints/quick/quick_entrypoints.h"
 #include "indirect_reference_table.h"
+#include "jni/jni_env_ext.h"
+#include "jni/local_reference_table.h"
 #include "lock_word.h"
 #include "managed_register_arm64.h"
 #include "offsets.h"
@@ -797,7 +799,7 @@ void Arm64JNIMacroAssembler::DeliverPendingException() {
 }
 
 std::unique_ptr<JNIMacroLabel> Arm64JNIMacroAssembler::CreateLabel() {
-  return std::unique_ptr<JNIMacroLabel>(new Arm64JNIMacroLabel());
+  return std::unique_ptr<JNIMacroLabel>(new (asm_.GetAllocator()) Arm64JNIMacroLabel());
 }
 
 void Arm64JNIMacroAssembler::Jump(JNIMacroLabel* label) {
@@ -979,6 +981,39 @@ void Arm64JNIMacroAssembler::RemoveFrame(size_t frame_size,
   cfi().DefCFAOffset(frame_size);
 }
 
+void Arm64JNIMacroAssembler::LoadLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                                           ManagedRegister previous_state_reg,
+                                                           ManagedRegister current_state_reg) {
+  constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+  DCHECK_EQ(kLRTSegmentStateSize, kWRegSizeInBytes);
+  const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kArm64PointerSize);
+  const MemberOffset current_state_offset = JNIEnvExt::LrtSegmentStateOffset(kArm64PointerSize);
+  DCHECK_EQ(previous_state_offset.SizeValue() + kLRTSegmentStateSize,
+            current_state_offset.SizeValue());
+
+  ___ Ldp(
+      reg_w(previous_state_reg.AsArm64().AsWRegister()),
+      reg_w(current_state_reg.AsArm64().AsWRegister()),
+      MemOperand(reg_x(jni_env_reg.AsArm64().AsXRegister()), previous_state_offset.Int32Value()));
+}
+
+void Arm64JNIMacroAssembler::StoreLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                                            ManagedRegister previous_state_reg,
+                                                            ManagedRegister current_state_reg) {
+  constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+  DCHECK_EQ(kLRTSegmentStateSize, kWRegSizeInBytes);
+  const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kArm64PointerSize);
+  const MemberOffset current_state_offset = JNIEnvExt::LrtSegmentStateOffset(kArm64PointerSize);
+  DCHECK_EQ(previous_state_offset.SizeValue() + kLRTSegmentStateSize,
+            current_state_offset.SizeValue());
+
+  // Set the current segment state together with restoring the cookie.
+  ___ Stp(
+      reg_w(previous_state_reg.AsArm64().AsWRegister()),
+      reg_w(current_state_reg.AsArm64().AsWRegister()),
+      MemOperand(reg_x(jni_env_reg.AsArm64().AsXRegister()), previous_state_offset.Int32Value()));
+}
+
 #undef ___
 
 }  // namespace arm64
diff --git a/compiler/utils/arm64/jni_macro_assembler_arm64.h b/compiler/utils/arm64/jni_macro_assembler_arm64.h
index 2836e0947d..0750c2655f 100644
--- a/compiler/utils/arm64/jni_macro_assembler_arm64.h
+++ b/compiler/utils/arm64/jni_macro_assembler_arm64.h
@@ -93,6 +93,14 @@ class Arm64JNIMacroAssembler final : public JNIMacroAssemblerFwd<Arm64Assembler,
   void GetCurrentThread(ManagedRegister dest) override;
   void GetCurrentThread(FrameOffset dest_offset) override;
 
+  // Manipulating local reference table states.
+  void LoadLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                     ManagedRegister previous_state_reg,
+                                     ManagedRegister current_state_reg) override;
+  void StoreLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                      ManagedRegister previous_state_reg,
+                                      ManagedRegister current_state_reg) override;
+
   // Decode JNI transition or local `jobject`. For (weak) global `jobject`, jump to slow path.
   void DecodeJNITransitionOrLocalJObject(ManagedRegister reg,
                                          JNIMacroLabel* slow_path,
diff --git a/compiler/utils/assembler_test_base.h b/compiler/utils/assembler_test_base.h
index c147217b3d..c5345130b3 100644
--- a/compiler/utils/assembler_test_base.h
+++ b/compiler/utils/assembler_test_base.h
@@ -147,7 +147,7 @@ class AssemblerTestBase : public testing::Test {
                 "--compile",
                 "-target",
                 "riscv64-linux-gnu",
-                "-march=rv64imafdv_zba_zbb",
+                "-march=rv64imafdcv_zba_zbb_zca_zcd_zcb",
                 // Force the assembler to fully emit branch instructions instead of leaving
                 // offsets unresolved with relocation information for the linker.
                 "-mno-relax"};
@@ -175,7 +175,7 @@ class AssemblerTestBase : public testing::Test {
                 "--no-print-imm-hex",
                 "--no-show-raw-insn",
                 // Disassemble Standard Extensions supported by the assembler.
-                "--mattr=+F,+D,+A,+V,+Zba,+Zbb",
+                "--mattr=+F,+D,+A,+C,+V,+Zba,+Zbb,+Zca,+Zcd,+Zcb",
                 "-M",
                 "no-aliases"};
       default:
diff --git a/compiler/utils/jni_macro_assembler.cc b/compiler/utils/jni_macro_assembler.cc
index 7a90a46f51..1806180980 100644
--- a/compiler/utils/jni_macro_assembler.cc
+++ b/compiler/utils/jni_macro_assembler.cc
@@ -38,6 +38,8 @@
 #include "base/globals.h"
 #include "base/memory_region.h"
 #include "gc_root.h"
+#include "jni/jni_env_ext.h"
+#include "jni/local_reference_table.h"
 #include "stack_reference.h"
 
 namespace art HIDDEN {
@@ -129,4 +131,52 @@ template
 void JNIMacroAssembler<PointerSize::k64>::LoadStackReference(ManagedRegister dest,
                                                              FrameOffset offs);
 
+template <PointerSize kPointerSize>
+void JNIMacroAssembler<kPointerSize>::LoadLocalReferenceTableStates(
+    ManagedRegister jni_env_reg,
+    ManagedRegister previous_state_reg,
+    ManagedRegister current_state_reg) {
+  constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+  const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kPointerSize);
+  const MemberOffset current_state_offset = JNIEnvExt::LrtSegmentStateOffset(kPointerSize);
+
+  Load(previous_state_reg, jni_env_reg, previous_state_offset, kLRTSegmentStateSize);
+  Load(current_state_reg, jni_env_reg, current_state_offset, kLRTSegmentStateSize);
+}
+
+template
+void JNIMacroAssembler<PointerSize::k32>::LoadLocalReferenceTableStates(
+    ManagedRegister jni_env_reg,
+    ManagedRegister previous_state_reg,
+    ManagedRegister current_state_reg);
+template
+void JNIMacroAssembler<PointerSize::k64>::LoadLocalReferenceTableStates(
+    ManagedRegister jni_env_reg,
+    ManagedRegister previous_state_reg,
+    ManagedRegister current_state_reg);
+
+template <PointerSize kPointerSize>
+void JNIMacroAssembler<kPointerSize>::StoreLocalReferenceTableStates(
+    ManagedRegister jni_env_reg,
+    ManagedRegister previous_state_reg,
+    ManagedRegister current_state_reg) {
+  constexpr size_t kLRTSegmentStateSize = sizeof(jni::LRTSegmentState);
+  const MemberOffset previous_state_offset = JNIEnvExt::LrtPreviousStateOffset(kPointerSize);
+  const MemberOffset segment_state_offset = JNIEnvExt::LrtSegmentStateOffset(kPointerSize);
+
+  Store(jni_env_reg, previous_state_offset, previous_state_reg, kLRTSegmentStateSize);
+  Store(jni_env_reg, segment_state_offset, current_state_reg, kLRTSegmentStateSize);
+}
+
+template
+void JNIMacroAssembler<PointerSize::k32>::StoreLocalReferenceTableStates(
+    ManagedRegister jni_env_reg,
+    ManagedRegister previous_state_reg,
+    ManagedRegister current_state_reg);
+template
+void JNIMacroAssembler<PointerSize::k64>::StoreLocalReferenceTableStates(
+    ManagedRegister jni_env_reg,
+    ManagedRegister previous_state_reg,
+    ManagedRegister current_state_reg);
+
 }  // namespace art
diff --git a/compiler/utils/jni_macro_assembler.h b/compiler/utils/jni_macro_assembler.h
index 2d52eada08..0ffa50a53b 100644
--- a/compiler/utils/jni_macro_assembler.h
+++ b/compiler/utils/jni_macro_assembler.h
@@ -167,6 +167,17 @@ class JNIMacroAssembler : public DeletableArenaObject<kArenaAllocAssembler> {
   virtual void GetCurrentThread(ManagedRegister dest) = 0;
   virtual void GetCurrentThread(FrameOffset dest_offset) = 0;
 
+  // Manipulating local reference table states.
+  //
+  // These have a default implementation but they can be overridden to use register pair
+  // load/store instructions on architectures that support them (arm, arm64).
+  virtual void LoadLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                             ManagedRegister previous_state_reg,
+                                             ManagedRegister current_state_reg);
+  virtual void StoreLocalReferenceTableStates(ManagedRegister jni_env_reg,
+                                              ManagedRegister previous_state_reg,
+                                              ManagedRegister current_state_reg);
+
   // Decode JNI transition or local `jobject`. For (weak) global `jobject`, jump to slow path.
   virtual void DecodeJNITransitionOrLocalJObject(ManagedRegister reg,
                                                  JNIMacroLabel* slow_path,
@@ -250,7 +261,7 @@ class JNIMacroAssembler : public DeletableArenaObject<kArenaAllocAssembler> {
 //
 // It is only safe to use a label created
 // via JNIMacroAssembler::CreateLabel with that same macro assembler.
-class JNIMacroLabel {
+class JNIMacroLabel : public DeletableArenaObject<kArenaAllocAssembler> {
  public:
   virtual ~JNIMacroLabel() = 0;
 
diff --git a/compiler/utils/riscv64/assembler_riscv64.cc b/compiler/utils/riscv64/assembler_riscv64.cc
index c98b919f41..eeb4537a31 100644
--- a/compiler/utils/riscv64/assembler_riscv64.cc
+++ b/compiler/utils/riscv64/assembler_riscv64.cc
@@ -61,18 +61,6 @@ void Riscv64Assembler::FinalizeCode() {
   finalized_ = true;
 }
 
-void Riscv64Assembler::Emit(uint32_t value) {
-  if (overwriting_) {
-    // Branches to labels are emitted into their placeholders here.
-    buffer_.Store<uint32_t>(overwrite_location_, value);
-    overwrite_location_ += sizeof(uint32_t);
-  } else {
-    // Other instructions are simply appended at the end here.
-    AssemblerBuffer::EnsureCapacity ensured(&buffer_);
-    buffer_.Emit<uint32_t>(value);
-  }
-}
-
 /////////////////////////////// RV64 VARIANTS extension ///////////////////////////////
 
 //////////////////////////////// RV64 "I" Instructions ////////////////////////////////
@@ -792,6 +780,270 @@ void Riscv64Assembler::FClassD(XRegister rd, FRegister rs1) {
 
 /////////////////////////////// RV64 "FD" Instructions  END ///////////////////////////////
 
+/////////////////////////////// RV64 "C" Instructions  START /////////////////////////////
+
+void Riscv64Assembler::CLwsp(XRegister rd, int32_t offset) {
+  DCHECK_NE(rd, Zero);
+
+  EmitCI(0b010u, rd, ExtractOffset52_76(offset), 0b10u);
+}
+
+void Riscv64Assembler::CLdsp(XRegister rd, int32_t offset) {
+  DCHECK_NE(rd, Zero);
+
+  EmitCI(0b011u, rd, ExtractOffset53_86(offset), 0b10u);
+}
+
+void Riscv64Assembler::CFLdsp(FRegister rd, int32_t offset) {
+  EmitCI(0b001u, rd, ExtractOffset53_86(offset), 0b10u);
+}
+
+void Riscv64Assembler::CSwsp(XRegister rs2, int32_t offset) {
+  EmitCSS(0b110u, ExtractOffset52_76(offset), rs2, 0b10u);
+}
+
+void Riscv64Assembler::CSdsp(XRegister rs2, int32_t offset) {
+  EmitCSS(0b111u, ExtractOffset53_86(offset), rs2, 0b10u);
+}
+
+void Riscv64Assembler::CFSdsp(FRegister rs2, int32_t offset) {
+  EmitCSS(0b101u, ExtractOffset53_86(offset), rs2, 0b10u);
+}
+
+void Riscv64Assembler::CLw(XRegister rd_s, XRegister rs1_s, int32_t offset) {
+  EmitCM(0b010u, ExtractOffset52_6(offset), rs1_s, rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CLd(XRegister rd_s, XRegister rs1_s, int32_t offset) {
+  EmitCM(0b011u, ExtractOffset53_76(offset), rs1_s, rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CFLd(FRegister rd_s, XRegister rs1_s, int32_t offset) {
+  EmitCM(0b001u, ExtractOffset53_76(offset), rs1_s, rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CSw(XRegister rs2_s, XRegister rs1_s, int32_t offset) {
+  EmitCM(0b110u, ExtractOffset52_6(offset), rs1_s, rs2_s, 0b00u);
+}
+
+void Riscv64Assembler::CSd(XRegister rs2_s, XRegister rs1_s, int32_t offset) {
+  EmitCM(0b111u, ExtractOffset53_76(offset), rs1_s, rs2_s, 0b00u);
+}
+
+void Riscv64Assembler::CFSd(FRegister rs2_s, XRegister rs1_s, int32_t offset) {
+  EmitCM(0b101u, ExtractOffset53_76(offset), rs1_s, rs2_s, 0b00u);
+}
+
+void Riscv64Assembler::CLi(XRegister rd, int32_t imm) {
+  DCHECK_NE(rd, Zero);
+  DCHECK(IsInt<6>(imm));
+
+  EmitCI(0b010u, rd, EncodeInt6(imm), 0b01u);
+}
+
+void Riscv64Assembler::CLui(XRegister rd, uint32_t nzimm6) {
+  DCHECK_NE(rd, Zero);
+  DCHECK_NE(rd, SP);
+  DCHECK(IsImmCLuiEncodable(nzimm6));
+
+  EmitCI(0b011u, rd, nzimm6 & MaskLeastSignificant<uint32_t>(6), 0b01u);
+}
+
+void Riscv64Assembler::CAddi(XRegister rd, int32_t nzimm) {
+  DCHECK_NE(rd, Zero);
+  DCHECK_NE(nzimm, 0);
+
+  EmitCI(0b000u, rd, EncodeInt6(nzimm), 0b01u);
+}
+
+void Riscv64Assembler::CAddiw(XRegister rd, int32_t imm) {
+  DCHECK_NE(rd, Zero);
+
+  EmitCI(0b001u, rd, EncodeInt6(imm), 0b01u);
+}
+
+void Riscv64Assembler::CAddi16Sp(int32_t nzimm) {
+  DCHECK_NE(nzimm, 0);
+  DCHECK(IsAligned<16>(nzimm));
+
+  uint32_t unzimm = static_cast<uint32_t>(nzimm);
+
+  // nzimm[9]
+  uint32_t imms1 =  BitFieldExtract(unzimm, 9, 1);
+  // nzimm[4|6|8:7|5]
+  uint32_t imms0 = (BitFieldExtract(unzimm, 4, 1) << 4) |
+                   (BitFieldExtract(unzimm, 6, 1) << 3) |
+                   (BitFieldExtract(unzimm, 7, 2) << 1) |
+                    BitFieldExtract(unzimm, 5, 1);
+
+  EmitCI(0b011u, SP, BitFieldInsert(imms0, imms1, 5, 1), 0b01u);
+}
+
+void Riscv64Assembler::CAddi4Spn(XRegister rd_s, uint32_t nzuimm) {
+  DCHECK_NE(nzuimm, 0u);
+  DCHECK(IsAligned<4>(nzuimm));
+  DCHECK(IsUint<10>(nzuimm));
+
+  // nzuimm[5:4|9:6|2|3]
+  uint32_t uimm = (BitFieldExtract(nzuimm, 4, 2) << 6) |
+                  (BitFieldExtract(nzuimm, 6, 4) << 2) |
+                  (BitFieldExtract(nzuimm, 2, 1) << 1) |
+                   BitFieldExtract(nzuimm, 3, 1);
+
+  EmitCIW(0b000u, uimm, rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CSlli(XRegister rd, int32_t shamt) {
+  DCHECK_NE(shamt, 0);
+  DCHECK_NE(rd, Zero);
+
+  EmitCI(0b000u, rd, shamt, 0b10u);
+}
+
+void Riscv64Assembler::CSrli(XRegister rd_s, int32_t shamt) {
+  DCHECK_NE(shamt, 0);
+  DCHECK(IsUint<6>(shamt));
+
+  EmitCBArithmetic(0b100u, 0b00u, shamt, rd_s, 0b01u);
+}
+
+void Riscv64Assembler::CSrai(XRegister rd_s, int32_t shamt) {
+  DCHECK_NE(shamt, 0);
+  DCHECK(IsUint<6>(shamt));
+
+  EmitCBArithmetic(0b100u, 0b01u, shamt, rd_s, 0b01u);
+}
+
+void Riscv64Assembler::CAndi(XRegister rd_s, int32_t imm) {
+  DCHECK(IsInt<6>(imm));
+
+  EmitCBArithmetic(0b100u, 0b10u, imm, rd_s, 0b01u);
+}
+
+void Riscv64Assembler::CMv(XRegister rd, XRegister rs2) {
+  DCHECK_NE(rd, Zero);
+  DCHECK_NE(rs2, Zero);
+
+  EmitCR(0b1000u, rd, rs2, 0b10u);
+}
+
+void Riscv64Assembler::CAdd(XRegister rd, XRegister rs2) {
+  DCHECK_NE(rd, Zero);
+  DCHECK_NE(rs2, Zero);
+
+  EmitCR(0b1001u, rd, rs2, 0b10u);
+}
+
+void Riscv64Assembler::CAnd(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100011u, rd_s, 0b11u, rs2_s, 0b01u);
+}
+
+void Riscv64Assembler::COr(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100011u, rd_s, 0b10u, rs2_s, 0b01u);
+}
+
+void Riscv64Assembler::CXor(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100011u, rd_s, 0b01u, rs2_s, 0b01u);
+}
+
+void Riscv64Assembler::CSub(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100011u, rd_s, 0b00u, rs2_s, 0b01u);
+}
+
+void Riscv64Assembler::CAddw(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100111u, rd_s, 0b01u, rs2_s, 0b01u);
+}
+
+void Riscv64Assembler::CSubw(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100111u, rd_s, 0b00u, rs2_s, 0b01u);
+}
+
+// "Zcb" Standard Extension, part of "C", opcode = 0b00, 0b01, funct3 = 0b100.
+
+void Riscv64Assembler::CLbu(XRegister rd_s, XRegister rs1_s, int32_t offset) {
+  EmitCAReg(0b100000u, rs1_s, EncodeOffset0_1(offset), rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CLhu(XRegister rd_s, XRegister rs1_s, int32_t offset) {
+  DCHECK(IsUint<2>(offset));
+  DCHECK_ALIGNED(offset, 2);
+  EmitCAReg(0b100001u, rs1_s, BitFieldExtract<uint32_t>(offset, 1, 1), rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CLh(XRegister rd_s, XRegister rs1_s, int32_t offset) {
+  DCHECK(IsUint<2>(offset));
+  DCHECK_ALIGNED(offset, 2);
+  EmitCAReg(0b100001u, rs1_s, 0b10 | BitFieldExtract<uint32_t>(offset, 1, 1), rd_s, 0b00u);
+}
+
+void Riscv64Assembler::CSb(XRegister rs2_s, XRegister rs1_s, int32_t offset) {
+  EmitCAReg(0b100010u, rs1_s, EncodeOffset0_1(offset), rs2_s, 0b00u);
+}
+
+void Riscv64Assembler::CSh(XRegister rs2_s, XRegister rs1_s, int32_t offset) {
+  DCHECK(IsUint<2>(offset));
+  DCHECK_ALIGNED(offset, 2);
+  EmitCAReg(0b100011u, rs1_s, BitFieldExtract<uint32_t>(offset, 1, 1), rs2_s, 0b00u);
+}
+
+void Riscv64Assembler::CZext_b(XRegister rd_rs1_s) {
+  EmitCAImm(0b100111u, rd_rs1_s, 0b11u, 0b000u, 0b01u);
+}
+
+void Riscv64Assembler::CSext_b(XRegister rd_rs1_s) {
+  EmitCAImm(0b100111u, rd_rs1_s, 0b11u, 0b001u, 0b01u);
+}
+
+void Riscv64Assembler::CZext_h(XRegister rd_rs1_s) {
+  EmitCAImm(0b100111u, rd_rs1_s, 0b11u, 0b010u, 0b01u);
+}
+
+void Riscv64Assembler::CSext_h(XRegister rd_rs1_s) {
+  EmitCAImm(0b100111u, rd_rs1_s, 0b11u, 0b011u, 0b01u);
+}
+
+void Riscv64Assembler::CZext_w(XRegister rd_rs1_s) {
+  EmitCAImm(0b100111u, rd_rs1_s, 0b11u, 0b100u, 0b01u);
+}
+
+void Riscv64Assembler::CNot(XRegister rd_rs1_s) {
+  EmitCAImm(0b100111u, rd_rs1_s, 0b11u, 0b101u, 0b01u);
+}
+
+void Riscv64Assembler::CMul(XRegister rd_s, XRegister rs2_s) {
+  EmitCAReg(0b100111u, rd_s, 0b10u, rs2_s, 0b01u);
+}
+
+void Riscv64Assembler::CJ(int32_t offset) { EmitCJ(0b101u, offset, 0b01u); }
+
+void Riscv64Assembler::CJr(XRegister rs1) {
+  DCHECK_NE(rs1, Zero);
+
+  EmitCR(0b1000u, rs1, Zero, 0b10u);
+}
+
+void Riscv64Assembler::CJalr(XRegister rs1) {
+  DCHECK_NE(rs1, Zero);
+
+  EmitCR(0b1001u, rs1, Zero, 0b10u);
+}
+
+void Riscv64Assembler::CBeqz(XRegister rs1_s, int32_t offset) {
+  EmitCBBranch(0b110u, offset, rs1_s, 0b01u);
+}
+
+void Riscv64Assembler::CBnez(XRegister rs1_s, int32_t offset) {
+  EmitCBBranch(0b111u, offset, rs1_s, 0b01u);
+}
+
+void Riscv64Assembler::CEbreak() { EmitCR(0b1001u, Zero, Zero, 0b10u); }
+
+void Riscv64Assembler::CNop() { EmitCI(0b000u, Zero, 0u, 0b01u); }
+
+void Riscv64Assembler::CUnimp() { Emit16(0x0u); }
+
+/////////////////////////////// RV64 "C" Instructions  END ///////////////////////////////
+
 ////////////////////////////// RV64 "Zba" Instructions  START /////////////////////////////
 
 void Riscv64Assembler::AddUw(XRegister rd, XRegister rs1, XRegister rs2) {
@@ -916,6 +1168,18 @@ void Riscv64Assembler::Rev8(XRegister rd, XRegister rs1) {
   EmitR(0x35, 0x18, rs1, 0x5, rd, 0x13);
 }
 
+void Riscv64Assembler::ZbbSextB(XRegister rd, XRegister rs1) {
+  EmitR(0x30, 0x4, rs1, 0x1, rd, 0x13);
+}
+
+void Riscv64Assembler::ZbbSextH(XRegister rd, XRegister rs1) {
+  EmitR(0x30, 0x5, rs1, 0x1, rd, 0x13);
+}
+
+void Riscv64Assembler::ZbbZextH(XRegister rd, XRegister rs1) {
+  EmitR(0x4, 0x0, rs1, 0x4, rd, 0x3b);
+}
+
 /////////////////////////////// RV64 "Zbb" Instructions  END //////////////////////////////
 
 /////////////////////////////// RVV "VSet" Instructions  START ////////////////////////////
@@ -5060,7 +5324,7 @@ void Riscv64Assembler::FLoadd(FRegister rd, Literal* literal) {
 
 void Riscv64Assembler::Unimp() {
   // TODO(riscv64): use 16-bit zero C.UNIMP once we support compression
-  Emit(0xC0001073);
+  Emit32(0xC0001073);
 }
 
 /////////////////////////////// RV64 MACRO Instructions END ///////////////////////////////
@@ -5542,7 +5806,7 @@ void Riscv64Assembler::FinalizeLabeledBranch(Riscv64Label* label) {
     // Branch forward (to a following label), distance is unknown.
     // The first branch forward will contain 0, serving as the terminator of
     // the list of forward-reaching branches.
-    Emit(label->position_);
+    Emit32(label->position_);
     length--;
     // Now make the label object point to this branch
     // (this forms a linked list of branches preceding this label).
@@ -5775,7 +6039,7 @@ void Riscv64Assembler::PromoteBranches() {
       DCHECK(!overwriting_);
       overwriting_ = true;
       overwrite_location_ = first_literal_location;
-      Emit(0);  // Illegal instruction.
+      Emit32(0);  // Illegal instruction.
       overwriting_ = false;
       // Increase target addresses in literal and address loads by 4 bytes in order for correct
       // offsets from PC to be generated.
@@ -5838,7 +6102,7 @@ void Riscv64Assembler::EmitJumpTables() {
         CHECK_EQ(buffer_.Load<uint32_t>(overwrite_location_), 0x1abe1234u);
         // The table will contain target addresses relative to the table start.
         uint32_t offset = GetLabelLocation(target) - start;
-        Emit(offset);
+        Emit32(offset);
       }
     }
 
diff --git a/compiler/utils/riscv64/assembler_riscv64.h b/compiler/utils/riscv64/assembler_riscv64.h
index b67a6508d9..1696251bf6 100644
--- a/compiler/utils/riscv64/assembler_riscv64.h
+++ b/compiler/utils/riscv64/assembler_riscv64.h
@@ -490,6 +490,64 @@ class Riscv64Assembler final : public Assembler {
   void FClassS(XRegister rd, FRegister rs1);
   void FClassD(XRegister rd, FRegister rs1);
 
+  // "C" Standard Extension, Compresseed Instructions
+  void CLwsp(XRegister rd, int32_t offset);
+  void CLdsp(XRegister rd, int32_t offset);
+  void CFLdsp(FRegister rd, int32_t offset);
+  void CSwsp(XRegister rs2, int32_t offset);
+  void CSdsp(XRegister rs2, int32_t offset);
+  void CFSdsp(FRegister rs2, int32_t offset);
+
+  void CLw(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CLd(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CFLd(FRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CSw(XRegister rs2_s, XRegister rs1_s, int32_t offset);
+  void CSd(XRegister rs2_s, XRegister rs1_s, int32_t offset);
+  void CFSd(FRegister rs2_s, XRegister rs1_s, int32_t offset);
+
+  void CLi(XRegister rd, int32_t imm);
+  void CLui(XRegister rd, uint32_t nzimm6);
+  void CAddi(XRegister rd, int32_t nzimm);
+  void CAddiw(XRegister rd, int32_t imm);
+  void CAddi16Sp(int32_t nzimm);
+  void CAddi4Spn(XRegister rd_s, uint32_t nzuimm);
+  void CSlli(XRegister rd, int32_t shamt);
+  void CSrli(XRegister rd_s, int32_t shamt);
+  void CSrai(XRegister rd_s, int32_t shamt);
+  void CAndi(XRegister rd_s, int32_t imm);
+  void CMv(XRegister rd, XRegister rs2);
+  void CAdd(XRegister rd, XRegister rs2);
+  void CAnd(XRegister rd_s, XRegister rs2_s);
+  void COr(XRegister rd_s, XRegister rs2_s);
+  void CXor(XRegister rd_s, XRegister rs2_s);
+  void CSub(XRegister rd_s, XRegister rs2_s);
+  void CAddw(XRegister rd_s, XRegister rs2_s);
+  void CSubw(XRegister rd_s, XRegister rs2_s);
+
+  // "Zcb" Standard Extension, part of "C", opcode = 0b00, 0b01, funct3 = 0b100.
+  void CLbu(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CLhu(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CLh(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CSb(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CSh(XRegister rd_s, XRegister rs1_s, int32_t offset);
+  void CZext_b(XRegister rd_rs1_s);
+  void CSext_b(XRegister rd_rs1_s);
+  void CZext_h(XRegister rd_rs1_s);
+  void CSext_h(XRegister rd_rs1_s);
+  void CZext_w(XRegister rd_rs1_s);
+  void CNot(XRegister rd_rs1_s);
+  void CMul(XRegister rd_s, XRegister rs2_s);
+
+  void CJ(int32_t offset);
+  void CJr(XRegister rs1);
+  void CJalr(XRegister rs1);
+  void CBeqz(XRegister rs1_s, int32_t offset);
+  void CBnez(XRegister rs1_s, int32_t offset);
+
+  void CEbreak();
+  void CNop();
+  void CUnimp();
+
   // "Zba" Standard Extension, opcode = 0x1b, 0x33 or 0x3b, funct3 and funct7 varies.
   void AddUw(XRegister rd, XRegister rs1, XRegister rs2);
   void Sh1Add(XRegister rd, XRegister rs1, XRegister rs2);
@@ -500,9 +558,9 @@ class Riscv64Assembler final : public Assembler {
   void Sh3AddUw(XRegister rd, XRegister rs1, XRegister rs2);
   void SlliUw(XRegister rd, XRegister rs1, int32_t shamt);
 
-  // "Zbb" Standard Extension, opcode = 0x13, 0x1b or 0x33, funct3 and funct7 varies.
-  // Note: We do not support 32-bit sext.b, sext.h and zext.h from the Zbb extension.
-  // (Neither does the clang-r498229's assembler which we currently test against.)
+  // "Zbb" Standard Extension, opcode = 0x13, 0x1b, 0x33 or 0x3b, funct3 and funct7 varies.
+  // Note: 32-bit sext.b, sext.h and zext.h from the Zbb extension are explicitly
+  // prefixed with "Zbb" to differentiate them from the utility macros.
   void Andn(XRegister rd, XRegister rs1, XRegister rs2);
   void Orn(XRegister rd, XRegister rs1, XRegister rs2);
   void Xnor(XRegister rd, XRegister rs1, XRegister rs2);
@@ -524,6 +582,9 @@ class Riscv64Assembler final : public Assembler {
   void Roriw(XRegister rd, XRegister rs1, int32_t shamt);
   void OrcB(XRegister rd, XRegister rs1);
   void Rev8(XRegister rd, XRegister rs1);
+  void ZbbSextB(XRegister rd, XRegister rs1);
+  void ZbbSextH(XRegister rd, XRegister rs1);
+  void ZbbZextH(XRegister rd, XRegister rs1);
 
   ////////////////////////////// RISC-V Vector Instructions  START ///////////////////////////////
   enum class LengthMultiplier : uint32_t {
@@ -1758,6 +1819,13 @@ class Riscv64Assembler final : public Assembler {
   // and emit branches.
   void FinalizeCode() override;
 
+  template <typename Reg>
+  static inline bool IsShortReg(Reg reg) {
+    static_assert(std::is_same_v<Reg, XRegister> || std::is_same_v<Reg, FRegister>);
+    uint32_t uv = enum_cast<uint32_t>(reg) - 8u;
+    return IsUint<3>(uv);
+  }
+
   // Returns the current location of a label.
   //
   // This function must be used instead of `Riscv64Label::GetPosition()`
@@ -1952,7 +2020,23 @@ class Riscv64Assembler final : public Assembler {
   void PatchCFI();
 
   // Emit data (e.g. encoded instruction or immediate) to the instruction stream.
-  void Emit(uint32_t value);
+  template <typename T>
+  void Emit(T value) {
+    static_assert(std::is_same_v<T, uint32_t> || std::is_same_v<T, uint16_t>,
+                  "Only Integer types are allowed");
+    if (overwriting_) {
+      // Branches to labels are emitted into their placeholders here.
+      buffer_.Store<T>(overwrite_location_, value);
+      overwrite_location_ += sizeof(T);
+    } else {
+      // Other instructions are simply appended at the end here.
+      AssemblerBuffer::EnsureCapacity ensured(&buffer_);
+      buffer_.Emit<T>(value);
+    }
+  }
+
+  void Emit16(uint32_t value) { Emit(dchecked_integral_cast<uint16_t>(value)); }
+  void Emit32(uint32_t value) { Emit(value); }
 
   // Adjust base register and offset if needed for load/store with a large offset.
   void AdjustBaseAndOffset(XRegister& base, int32_t& offset, ScratchRegisterScope& srs);
@@ -2030,9 +2114,92 @@ class Riscv64Assembler final : public Assembler {
     return funct6 << 1 | enum_cast<uint32_t>(vm);
   }
 
-  static constexpr uint32_t EncodeInt5(const int32_t imm) {
-    DCHECK(IsInt<5>(imm));
-    return static_cast<uint32_t>(imm) & 0b11111u;
+  template <unsigned kWidth>
+  static constexpr uint32_t EncodeIntWidth(const int32_t imm) {
+    DCHECK(IsInt<kWidth>(imm));
+    return static_cast<uint32_t>(imm) & MaskLeastSignificant<uint32_t>(kWidth);
+  }
+
+  static constexpr uint32_t EncodeInt5(const int32_t imm) { return EncodeIntWidth<5>(imm); }
+  static constexpr uint32_t EncodeInt6(const int32_t imm) { return EncodeIntWidth<6>(imm); }
+
+  template <typename Reg>
+  static constexpr uint32_t EncodeShortReg(const Reg reg) {
+    DCHECK(IsShortReg(reg));
+    return enum_cast<uint32_t>(reg) - 8u;
+  }
+
+  // Rearrange given offset in the way {offset[0] | offset[1]}
+  static constexpr uint32_t EncodeOffset0_1(int32_t offset) {
+    uint32_t u_offset = static_cast<uint32_t>(offset);
+    DCHECK(IsUint<2>(u_offset));
+
+    return u_offset >> 1 | (u_offset & 1u) << 1;
+  }
+
+  // Rearrange given offset, scaled by 4, in the way {offset[5:2] | offset[7:6]}
+  static constexpr uint32_t ExtractOffset52_76(int32_t offset) {
+    DCHECK(IsAligned<4>(offset)) << "Offset should be scalable by 4";
+
+    uint32_t u_offset = static_cast<uint32_t>(offset);
+    DCHECK(IsUint<6 + 2>(u_offset));
+
+    uint32_t imm_52 = BitFieldExtract(u_offset, 2, 4);
+    uint32_t imm_76 = BitFieldExtract(u_offset, 6, 2);
+
+    return BitFieldInsert(imm_76, imm_52, 2, 4);
+  }
+
+  // Rearrange given offset, scaled by 8, in the way {offset[5:3] | offset[8:6]}
+  static constexpr uint32_t ExtractOffset53_86(int32_t offset) {
+    DCHECK(IsAligned<8>(offset)) << "Offset should be scalable by 8";
+
+    uint32_t u_offset = static_cast<uint32_t>(offset);
+    DCHECK(IsUint<6 + 3>(u_offset));
+
+    uint32_t imm_53 = BitFieldExtract(u_offset, 3, 3);
+    uint32_t imm_86 = BitFieldExtract(u_offset, 6, 3);
+
+    return BitFieldInsert(imm_86, imm_53, 3, 3);
+  }
+
+  // Rearrange given offset, scaled by 4, in the way {offset[5:2] | offset[6]}
+  static constexpr uint32_t ExtractOffset52_6(int32_t offset) {
+    DCHECK(IsAligned<4>(offset)) << "Offset should be scalable by 4";
+
+    uint32_t u_offset = static_cast<uint32_t>(offset);
+    DCHECK(IsUint<5 + 2>(u_offset));
+
+    uint32_t imm_52 = BitFieldExtract(u_offset, 2, 4);
+    uint32_t imm_6  = BitFieldExtract(u_offset, 6, 1);
+
+    return BitFieldInsert(imm_6, imm_52, 1, 4);
+  }
+
+  // Rearrange given offset, scaled by 8, in the way {offset[5:3], offset[7:6]}
+  static constexpr uint32_t ExtractOffset53_76(int32_t offset) {
+    DCHECK(IsAligned<8>(offset)) << "Offset should be scalable by 4";
+
+    uint32_t u_offset = static_cast<uint32_t>(offset);
+    DCHECK(IsUint<5 + 3>(u_offset));
+
+    uint32_t imm_53 = BitFieldExtract(u_offset, 3, 3);
+    uint32_t imm_76 = BitFieldExtract(u_offset, 6, 2);
+
+    return BitFieldInsert(imm_76, imm_53, 2, 3);
+  }
+
+  static constexpr bool IsImmCLuiEncodable(uint32_t uimm) {
+    // Instruction c.lui is odd and its immediate value is a bit tricky
+    // Its value is not a full 32 bits value, but its bits [31:12]
+    // (where the bit 17 marks the sign bit) shifted towards the bottom i.e. bits [19:0]
+    // are the meaningful ones. Since that we want a signed non-zero 6-bit immediate to
+    // keep values in the range [0, 0x1f], and the range [0xfffe0, 0xfffff] for negative values
+    // since the sign bit was bit 17 (which is now bit 5 and replicated in the higher bits too)
+    // Also encoding with immediate = 0 is reserved
+    // For more details please see 16.5 chapter is the specification
+
+    return uimm != 0u && (IsUint<5>(uimm) || IsUint<5>(uimm - 0xfffe0u));
   }
 
   // Emit helpers.
@@ -2053,7 +2220,7 @@ class Riscv64Assembler final : public Assembler {
     DCHECK(IsUint<7>(opcode));
     uint32_t encoding = static_cast<uint32_t>(imm12) << 20 | static_cast<uint32_t>(rs1) << 15 |
                         funct3 << 12 | static_cast<uint32_t>(rd) << 7 | opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // R-type instruction:
@@ -2074,7 +2241,7 @@ class Riscv64Assembler final : public Assembler {
     uint32_t encoding = funct7 << 25 | static_cast<uint32_t>(rs2) << 20 |
                         static_cast<uint32_t>(rs1) << 15 | funct3 << 12 |
                         static_cast<uint32_t>(rd) << 7 | opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // R-type instruction variant for floating-point fused multiply-add/sub (F[N]MADD/ F[N]MSUB):
@@ -2098,7 +2265,7 @@ class Riscv64Assembler final : public Assembler {
                         static_cast<uint32_t>(rs2) << 20 | static_cast<uint32_t>(rs1) << 15 |
                         static_cast<uint32_t>(funct3) << 12 | static_cast<uint32_t>(rd) << 7 |
                         opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // S-type instruction:
@@ -2119,7 +2286,7 @@ class Riscv64Assembler final : public Assembler {
                         static_cast<uint32_t>(rs2) << 20 | static_cast<uint32_t>(rs1) << 15 |
                         static_cast<uint32_t>(funct3) << 12 |
                         (static_cast<uint32_t>(imm12) & 0x1F) << 7 | opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // I-type instruction variant for shifts (SLLI / SRLI / SRAI):
@@ -2144,7 +2311,7 @@ class Riscv64Assembler final : public Assembler {
     uint32_t encoding = funct6 << 26 | static_cast<uint32_t>(imm6) << 20 |
                         static_cast<uint32_t>(rs1) << 15 | funct3 << 12 |
                         static_cast<uint32_t>(rd) << 7 | opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // B-type instruction:
@@ -2166,7 +2333,7 @@ class Riscv64Assembler final : public Assembler {
                         static_cast<uint32_t>(rs2) << 20 | static_cast<uint32_t>(rs1) << 15 |
                         static_cast<uint32_t>(funct3) << 12 |
                         (imm12 & 0xfu) << 8 | (imm12 & 0x400u) >> (10 - 7) | opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // U-type instruction:
@@ -2181,7 +2348,7 @@ class Riscv64Assembler final : public Assembler {
     DCHECK(IsUint<5>(static_cast<uint32_t>(rd)));
     DCHECK(IsUint<7>(opcode));
     uint32_t encoding = imm20 << 12 | static_cast<uint32_t>(rd) << 7 | opcode;
-    Emit(encoding);
+    Emit32(encoding);
   }
 
   // J-type instruction:
@@ -2200,7 +2367,224 @@ class Riscv64Assembler final : public Assembler {
     uint32_t encoding = (imm20 & 0x80000u) << (31 - 19) | (imm20 & 0x03ffu) << 21 |
                         (imm20 & 0x400u) << (20 - 10) | (imm20 & 0x7f800u) << (12 - 11) |
                         static_cast<uint32_t>(rd) << 7 | opcode;
-    Emit(encoding);
+    Emit32(encoding);
+  }
+
+  // Compressed Instruction Encodings
+
+  // CR-type instruction:
+  //
+  //   15    12 11      7 6       2 1 0
+  //   ---------------------------------
+  //   [ . . . | . . . . | . . . . | . ]
+  //   [ func4   rd/rs1      rs2    op ]
+  //   ---------------------------------
+  //
+  void EmitCR(uint32_t funct4, XRegister rd_rs1, XRegister rs2, uint32_t opcode) {
+    DCHECK(IsUint<4>(funct4));
+    DCHECK(IsUint<5>(static_cast<uint32_t>(rd_rs1)));
+    DCHECK(IsUint<5>(static_cast<uint32_t>(rs2)));
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t encoding = funct4 << 12 | static_cast<uint32_t>(rd_rs1) << 7 |
+                        static_cast<uint32_t>(rs2) << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  // CI-type instruction:
+  //
+  //   15  13   11      7 6       2 1 0
+  //   ---------------------------------
+  //   [ . . | | . . . . | . . . . | . ]
+  //   [func3 imm rd/rs1     imm    op ]
+  //   ---------------------------------
+  //
+  template <typename Reg>
+  void EmitCI(uint32_t funct3, Reg rd_rs1, uint32_t imm6, uint32_t opcode) {
+    DCHECK(IsUint<3>(funct3));
+    DCHECK(IsUint<5>(static_cast<uint32_t>(rd_rs1)));
+    DCHECK(IsUint<6>(imm6));
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t immH1 = BitFieldExtract(imm6, 5, 1);
+    uint32_t immL5 = BitFieldExtract(imm6, 0, 5);
+
+    uint32_t encoding =
+        funct3 << 13 | immH1 << 12 | static_cast<uint32_t>(rd_rs1) << 7 | immL5 << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  // CSS-type instruction:
+  //
+  //   15  13 12        7 6       2 1 0
+  //   ---------------------------------
+  //   [ . . | . . . . . | . . . . | . ]
+  //   [func3     imm6      rs2     op ]
+  //   ---------------------------------
+  //
+  template <typename Reg>
+  void EmitCSS(uint32_t funct3, uint32_t offset6, Reg rs2, uint32_t opcode) {
+    DCHECK(IsUint<3>(funct3));
+    DCHECK(IsUint<6>(offset6));
+    DCHECK(IsUint<5>(static_cast<uint32_t>(rs2)));
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t encoding = funct3 << 13 | offset6 << 7 | static_cast<uint32_t>(rs2) << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  // CIW-type instruction:
+  //
+  //   15  13 12            5 4   2 1 0
+  //   ---------------------------------
+  //   [ . . | . . . . . . . | . . | . ]
+  //   [func3     imm8         rd'  op ]
+  //   ---------------------------------
+  //
+  void EmitCIW(uint32_t funct3, uint32_t imm8, XRegister rd_s, uint32_t opcode) {
+    DCHECK(IsUint<3>(funct3));
+    DCHECK(IsUint<8>(imm8));
+    DCHECK(IsShortReg(rd_s)) << rd_s;
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t encoding = funct3 << 13 | imm8 << 5 | EncodeShortReg(rd_s) << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  // CL/S-type instruction:
+  //
+  //   15  13 12  10 9  7 6 5 4   2 1 0
+  //   ---------------------------------
+  //   [ . . | . . | . . | . | . . | . ]
+  //   [func3  imm   rs1' imm rds2' op ]
+  //   ---------------------------------
+  //
+  template <typename Reg>
+  void EmitCM(uint32_t funct3, uint32_t imm5, XRegister rs1_s, Reg rd_rs2_s, uint32_t opcode) {
+    DCHECK(IsUint<3>(funct3));
+    DCHECK(IsUint<5>(imm5));
+    DCHECK(IsShortReg(rs1_s)) << rs1_s;
+    DCHECK(IsShortReg(rd_rs2_s)) << rd_rs2_s;
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t immH3 = BitFieldExtract(imm5, 2, 3);
+    uint32_t immL2 = BitFieldExtract(imm5, 0, 2);
+
+    uint32_t encoding = funct3 << 13 | immH3 << 10 | EncodeShortReg(rs1_s) << 7 | immL2 << 5 |
+                        EncodeShortReg(rd_rs2_s) << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  // CA-type instruction:
+  //
+  //   15         10 9  7 6 5 4   2 1 0
+  //   ---------------------------------
+  //   [ . . . . . | . . | . | . . | . ]
+  //   [    funct6 rds1' funct2 rs2' op]
+  //   ---------------------------------
+  //
+  void EmitCA(
+      uint32_t funct6, XRegister rd_rs1_s, uint32_t funct2, uint32_t rs2_v, uint32_t opcode) {
+    DCHECK(IsUint<6>(funct6));
+    DCHECK(IsShortReg(rd_rs1_s)) << rd_rs1_s;
+    DCHECK(IsUint<2>(funct2));
+    DCHECK(IsUint<3>(rs2_v));
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t encoding =
+        funct6 << 10 | EncodeShortReg(rd_rs1_s) << 7 | funct2 << 5  | rs2_v << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  void EmitCAReg(
+      uint32_t funct6, XRegister rd_rs1_s, uint32_t funct2, XRegister rs2_s, uint32_t opcode) {
+    DCHECK(IsShortReg(rs2_s)) << rs2_s;
+    EmitCA(funct6, rd_rs1_s, funct2, EncodeShortReg(rs2_s), opcode);
+  }
+
+  void EmitCAImm(
+      uint32_t funct6, XRegister rd_rs1_s, uint32_t funct2, uint32_t funct3, uint32_t opcode) {
+    EmitCA(funct6, rd_rs1_s, funct2, funct3, opcode);
+  }
+
+  // CB-type instruction:
+  //
+  //   15  13 12  10 9  7 6       2 1 0
+  //   ---------------------------------
+  //   [ . . | . . | . . | . . . . | . ]
+  //   [func3 offset rs1'   offset  op ]
+  //   ---------------------------------
+  //
+  void EmitCB(uint32_t funct3, int32_t offset8, XRegister rd_rs1_s, uint32_t opcode) {
+    DCHECK(IsUint<3>(funct3));
+    DCHECK(IsUint<8>(offset8));
+    DCHECK(IsShortReg(rd_rs1_s)) << rd_rs1_s;
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t offsetH3 = BitFieldExtract<uint32_t>(offset8, 5, 3);
+    uint32_t offsetL5 = BitFieldExtract<uint32_t>(offset8, 0, 5);
+
+    uint32_t encoding =
+        funct3 << 13 | offsetH3 << 10 | EncodeShortReg(rd_rs1_s) << 7 | offsetL5 << 2 | opcode;
+    Emit16(encoding);
+  }
+
+  // Wrappers for EmitCB with different imm bit permutation
+  void EmitCBBranch(uint32_t funct3, int32_t offset, XRegister rs1_s, uint32_t opcode) {
+    DCHECK(IsInt<9>(offset));
+    DCHECK_ALIGNED(offset, 2);
+
+    uint32_t u_offset = static_cast<uint32_t>(offset);
+
+    // offset[8|4:3]
+    uint32_t offsetH3 = (BitFieldExtract(u_offset, 8, 1) << 2) |
+                         BitFieldExtract(u_offset, 3, 2);
+    // offset[7:6|2:1|5]
+    uint32_t offsetL5 = (BitFieldExtract(u_offset, 6, 2) << 3) |
+                        (BitFieldExtract(u_offset, 1, 2) << 1) |
+                         BitFieldExtract(u_offset, 5, 1);
+
+    EmitCB(funct3, BitFieldInsert(offsetL5, offsetH3, 5, 3), rs1_s, opcode);
+  }
+
+  void EmitCBArithmetic(
+      uint32_t funct3, uint32_t funct2, uint32_t imm, XRegister rd_s, uint32_t opcode) {
+    uint32_t imm_5 = BitFieldExtract(imm, 5, 1);
+    uint32_t immH3 = BitFieldInsert(funct2, imm_5, 2, 1);
+    uint32_t immL5 = BitFieldExtract(imm, 0, 5);
+
+    EmitCB(funct3, BitFieldInsert(immL5, immH3, 5, 3), rd_s, opcode);
+  }
+
+  // CJ-type instruction:
+  //
+  //   15  13 12                  2 1 0
+  //   ---------------------------------
+  //   [ . . | . . . . . . . . . . | . ]
+  //   [func3    jump target 11     op ]
+  //   ---------------------------------
+  //
+  void EmitCJ(uint32_t funct3, int32_t offset, uint32_t opcode) {
+    DCHECK_ALIGNED(offset, 2);
+    DCHECK(IsInt<12>(offset)) << offset;
+    DCHECK(IsUint<3>(funct3));
+    DCHECK(IsUint<2>(opcode));
+
+    uint32_t uoffset = static_cast<uint32_t>(offset);
+    // offset[11|4|9:8|10|6|7|3:1|5]
+    uint32_t jumpt = (BitFieldExtract(uoffset, 11, 1) << 10) |
+                     (BitFieldExtract(uoffset, 4, 1) << 9)   |
+                     (BitFieldExtract(uoffset, 8, 2) << 7)   |
+                     (BitFieldExtract(uoffset, 10, 1) << 6)  |
+                     (BitFieldExtract(uoffset, 6, 1) << 5)   |
+                     (BitFieldExtract(uoffset, 7, 1) << 4)   |
+                     (BitFieldExtract(uoffset, 1, 3) << 1)   |
+                      BitFieldExtract(uoffset, 5, 1);
+
+    DCHECK(IsUint<11>(jumpt));
+
+    uint32_t encoding = funct3 << 13 | jumpt << 2 | opcode;
+    Emit16(encoding);
   }
 
   ArenaVector<Branch> branches_;
diff --git a/compiler/utils/riscv64/assembler_riscv64_test.cc b/compiler/utils/riscv64/assembler_riscv64_test.cc
index a8c1fbd9be..87c7641576 100644
--- a/compiler/utils/riscv64/assembler_riscv64_test.cc
+++ b/compiler/utils/riscv64/assembler_riscv64_test.cc
@@ -52,33 +52,39 @@ class AssemblerRISCV64Test : public AssemblerTest<Riscv64Assembler,
 
   InstructionSet GetIsa() override { return InstructionSet::kRiscv64; }
 
-  // Clang's assembler takes advantage of certain extensions for emitting constants with `li`
-  // but our assembler does not. For now, we use a simple `-march` to avoid the divergence.
-  // TODO(riscv64): Implement these more efficient patterns in assembler.
-  void SetUseSimpleMarch(bool value) {
-    use_simple_march_ = value;
-  }
+  class ScopedMarchOverride {
+   public:
+    ScopedMarchOverride(AssemblerRISCV64Test* test, const std::string& march)
+        : test_(test), old_override_(test->march_override_) {
+      test->march_override_ = march;
+    }
+
+    ~ScopedMarchOverride() {
+      test_->march_override_ = old_override_;
+    }
+
+   private:
+    AssemblerRISCV64Test* const test_;
+    std::optional<std::string> const old_override_;
+  };
+
+  class ScopedCSuppression {
+   public:
+    explicit ScopedCSuppression(AssemblerRISCV64Test* test)
+        : smo_(test, "-march=rv64imafdv_zba_zbb") {}
+
+   private:
+    ScopedMarchOverride smo_;
+  };
 
   std::vector<std::string> GetAssemblerCommand() override {
     std::vector<std::string> result = Base::GetAssemblerCommand();
-    if (use_simple_march_) {
+    if (march_override_.has_value()) {
       auto it = std::find_if(result.begin(),
                              result.end(),
                              [](const std::string& s) { return StartsWith(s, "-march="); });
       CHECK(it != result.end());
-      *it = "-march=rv64imafdv";
-    }
-    return result;
-  }
-
-  std::vector<std::string> GetDisassemblerCommand() override {
-    std::vector<std::string> result = Base::GetDisassemblerCommand();
-    if (use_simple_march_) {
-      auto it = std::find_if(result.begin(),
-                             result.end(),
-                             [](const std::string& s) { return StartsWith(s, "--mattr="); });
-      CHECK(it != result.end());
-      *it = "--mattr=+F,+D,+A,+V";
+      *it = march_override_.value();
     }
     return result;
   }
@@ -167,6 +173,20 @@ class AssemblerRISCV64Test : public AssemblerTest<Riscv64Assembler,
     return ArrayRef<const XRegister>(kXRegisters);
   }
 
+  ArrayRef<const XRegister> GetRegistersShort() {
+    static constexpr XRegister kXRegistersShort[] = {
+        S0,
+        S1,
+        A0,
+        A1,
+        A2,
+        A3,
+        A4,
+        A5,
+    };
+    return ArrayRef<const XRegister>(kXRegistersShort);
+  }
+
   ArrayRef<const FRegister> GetFPRegisters() override {
     static constexpr FRegister kFRegisters[] = {
         FT0,
@@ -213,6 +233,20 @@ class AssemblerRISCV64Test : public AssemblerTest<Riscv64Assembler,
     return ArrayRef<const VRegister>(kVRegisters);
   }
 
+  ArrayRef<const FRegister> GetFPRegistersShort() {
+    static constexpr FRegister kFRegistersShort[] = {
+        FS0,
+        FS1,
+        FA0,
+        FA1,
+        FA2,
+        FA3,
+        FA4,
+        FA5,
+    };
+    return ArrayRef<const FRegister>(kFRegistersShort);
+  }
+
   std::string GetSecondaryRegisterName(const XRegister& reg) override {
     CHECK(secondary_register_names_.find(reg) != secondary_register_names_.end());
     return secondary_register_names_[reg];
@@ -954,6 +988,186 @@ class AssemblerRISCV64Test : public AssemblerTest<Riscv64Assembler,
     return str;
   }
 
+  template <typename Reg, typename Imm>
+  std::string RepeatCTemplateRegImm(void (Riscv64Assembler::*f)(Reg, Imm),
+                                    ArrayRef<const Reg> registers,
+                                    std::string (Base::*GetName)(const Reg&),
+                                    int imm_bits,
+                                    int shift,
+                                    bool no_zero_imm,
+                                    const std::string& fmt) {
+    auto imms = CreateImmediateValuesBits(abs(imm_bits), /*as_uint=*/imm_bits > 0, shift);
+
+    CHECK(f != nullptr);
+    std::string str;
+    for (Reg reg : registers) {
+      for (int64_t imm_raw : imms) {
+        if (no_zero_imm && imm_raw == 0) {
+          continue;
+        }
+
+        Imm imm = CreateImmediate(imm_raw);
+        (GetAssembler()->*f)(reg, imm);
+
+        std::string base = fmt;
+        ReplaceReg(REG_TOKEN, (this->*GetName)(reg), &base);
+        ReplaceImm(imm, /*bias=*/0, /*multiplier=*/1, &base);
+        str += base;
+        str += "\n";
+      }
+    }
+    return str;
+  }
+
+  template <typename Imm>
+  std::string RepeatCRImm(void (Riscv64Assembler::*f)(XRegister, Imm),
+                          bool is_short,
+                          bool no_zero_reg,
+                          bool no_zero_imm,
+                          int imm_bits,
+                          int shift,
+                          const std::string& fmt) {
+    auto regs = is_short ? GetRegistersShort() : GetRegisters();
+    if (no_zero_reg) {
+      CHECK(!is_short);
+      CHECK_EQ(regs[0], Zero);
+      regs = regs.SubArray(1);
+    }
+    return RepeatCTemplateRegImm(
+        f, regs, &AssemblerRISCV64Test::GetRegisterName, imm_bits, shift, no_zero_imm, fmt);
+  }
+
+  template <typename Imm>
+  std::string RepeatCFImm(void (Riscv64Assembler::*f)(FRegister, Imm),
+                          int imm_bits,
+                          int shift,
+                          const std::string& fmt) {
+    auto regs = GetFPRegisters();
+    return RepeatCTemplateRegImm(
+        f, regs, &AssemblerRISCV64Test::GetFPRegName, imm_bits, shift, /*no_zero_imm=*/false, fmt);
+  }
+
+  template <typename Reg1>
+  std::string RepeatTemplatedShortRegistersImm(void (Riscv64Assembler::*f)(Reg1,
+                                                                           XRegister,
+                                                                           int32_t),
+                                               ArrayRef<const Reg1> reg1_registers,
+                                               std::string (Base::*GetName1)(const Reg1&),
+                                               int imm_bits,
+                                               int shift,
+                                               bool no_zero_imm,
+                                               const std::string& fmt) {
+    CHECK(f != nullptr);
+    auto imms = CreateImmediateValuesBits(abs(imm_bits), imm_bits > 0, shift);
+    std::string str;
+    for (Reg1 reg1 : reg1_registers) {
+      for (XRegister reg2 : GetRegistersShort()) {
+        for (int64_t imm_raw : imms) {
+          if (no_zero_imm && imm_raw == 0) {
+            continue;
+          }
+
+          int32_t imm = CreateImmediate(imm_raw);
+          (GetAssembler()->*f)(reg1, reg2, imm);
+
+          std::string base = fmt;
+          ReplaceReg(REG1_TOKEN, (this->*GetName1)(reg1), &base);
+          ReplaceReg(REG2_TOKEN, GetRegisterName(reg2), &base);
+          ReplaceImm(imm, /*bias=*/0, /*multiplier=*/1, &base);
+          str += base;
+          str += "\n";
+        }
+      }
+    }
+    return str;
+  }
+
+  std::string RepeatCRRImm(void (Riscv64Assembler::*f)(XRegister, XRegister, int32_t),
+                           int imm_bits,
+                           int shift,
+                           const std::string& fmt) {
+    return RepeatTemplatedShortRegistersImm(f,
+                                            GetRegistersShort(),
+                                            &AssemblerRISCV64Test::GetRegisterName,
+                                            imm_bits,
+                                            shift,
+                                            /*no_zero_imm=*/false,
+                                            fmt);
+  }
+
+  std::string RepeatCFRImm(void (Riscv64Assembler::*f)(FRegister, XRegister, int32_t),
+                           int imm_bits,
+                           int shift,
+                           const std::string& fmt) {
+    return RepeatTemplatedShortRegistersImm(f,
+                                            GetFPRegistersShort(),
+                                            &AssemblerRISCV64Test::GetFPRegName,
+                                            imm_bits,
+                                            shift,
+                                            /*no_zero_imm=*/false,
+                                            fmt);
+  }
+
+  std::string RepeatCRRShort(void (Riscv64Assembler::*f)(XRegister, XRegister),
+                             const std::string& fmt) {
+    return RepeatTemplatedRegisters(f,
+                                    GetRegistersShort(),
+                                    GetRegistersShort(),
+                                    &AssemblerRISCV64Test::GetRegisterName,
+                                    &AssemblerRISCV64Test::GetRegisterName,
+                                    fmt);
+  }
+
+  std::string RepeatCRRNonZero(void (Riscv64Assembler::*f)(XRegister, XRegister),
+                               const std::string& fmt) {
+    auto regs = GetRegisters();
+    CHECK_EQ(regs[0], Zero);
+    auto regs_no_zero = regs.SubArray(1);
+    return RepeatTemplatedRegisters(f,
+                                    regs_no_zero,
+                                    regs_no_zero,
+                                    &AssemblerRISCV64Test::GetRegisterName,
+                                    &AssemblerRISCV64Test::GetRegisterName,
+                                    fmt);
+  }
+
+  std::string RepeatCRShort(void (Riscv64Assembler::*f)(XRegister), const std::string& fmt) {
+    return RepeatTemplatedRegister(
+        f, GetRegistersShort(), &AssemblerRISCV64Test::GetRegisterName, fmt);
+  }
+
+  template <typename Imm>
+  std::string RepeatImm(void (Riscv64Assembler::*f)(Imm),
+                        bool no_zero_imm,
+                        int imm_bits,
+                        int shift,
+                        const std::string& fmt) {
+    auto imms = CreateImmediateValuesBits(abs(imm_bits), imm_bits > 0, shift);
+    std::string str;
+    for (int64_t imm_raw : imms) {
+      if (no_zero_imm && imm_raw == 0) {
+        continue;
+      }
+
+      Imm imm = CreateImmediate(imm_raw);
+      (GetAssembler()->*f)(imm);
+
+      std::string base = fmt;
+      ReplaceImm(imm, /*bias=*/0, /*multiplier=*/1, &base);
+      str += base;
+      str += "\n";
+    }
+
+    return str;
+  }
+
+  std::string RepeatRNoZero(void (Riscv64Assembler::*f)(XRegister), const std::string& fmt) {
+    auto regs = GetRegisters();
+    CHECK_EQ(regs[0], Zero);
+    return RepeatTemplatedRegister(
+        f, regs.SubArray(1), &AssemblerRISCV64Test::GetRegisterName, fmt);
+  }
+
   template <typename Reg1, typename Reg2>
   std::string RepeatTemplatedRegistersRoundingMode(
       void (Riscv64Assembler::*f)(Reg1, Reg2, FPRoundingMode),
@@ -1973,12 +2187,13 @@ class AssemblerRISCV64Test : public AssemblerTest<Riscv64Assembler,
   std::map<XRegister, std::string, RISCV64CpuRegisterCompare> secondary_register_names_;
 
   std::unique_ptr<const Riscv64InstructionSetFeatures> instruction_set_features_;
-  bool use_simple_march_ = false;
+  std::optional<std::string> march_override_;
 };
 
 TEST_F(AssemblerRISCV64Test, Toolchain) { EXPECT_TRUE(CheckTools()); }
 
 TEST_F(AssemblerRISCV64Test, Lui) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRIb(&Riscv64Assembler::Lui, 20, "lui {reg}, {imm}"), "Lui");
 }
 
@@ -1987,21 +2202,25 @@ TEST_F(AssemblerRISCV64Test, Auipc) {
 }
 
 TEST_F(AssemblerRISCV64Test, Jal) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-19, 2" to "-20, 1" for "C" Standard Extension.
   DriverStr(RepeatRIbS(&Riscv64Assembler::Jal, -19, 2, "jal {reg}, {imm}\n"), "Jal");
 }
 
 TEST_F(AssemblerRISCV64Test, Jalr) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-11, 2" to "-12, 1" for "C" Standard Extension.
   DriverStr(RepeatRRIb(&Riscv64Assembler::Jalr, -12, "jalr {reg1}, {reg2}, {imm}\n"), "Jalr");
 }
 
 TEST_F(AssemblerRISCV64Test, Beq) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-11, 2" to "-12, 1" for "C" Standard Extension.
   DriverStr(RepeatRRIbS(&Riscv64Assembler::Beq, -11, 2, "beq {reg1}, {reg2}, {imm}\n"), "Beq");
 }
 
 TEST_F(AssemblerRISCV64Test, Bne) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-11, 2" to "-12, 1" for "C" Standard Extension.
   DriverStr(RepeatRRIbS(&Riscv64Assembler::Bne, -11, 2, "bne {reg1}, {reg2}, {imm}\n"), "Bne");
 }
@@ -2027,50 +2246,62 @@ TEST_F(AssemblerRISCV64Test, Bgeu) {
 }
 
 TEST_F(AssemblerRISCV64Test, Lb) {
+  // Note: There is no 16-bit instruction for `Lb()`.
   DriverStr(RepeatRRIb(&Riscv64Assembler::Lb, -12, "lb {reg1}, {imm}({reg2})"), "Lb");
 }
 
 TEST_F(AssemblerRISCV64Test, Lh) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Lh, -12, "lh {reg1}, {imm}({reg2})"), "Lh");
 }
 
 TEST_F(AssemblerRISCV64Test, Lw) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Lw, -12, "lw {reg1}, {imm}({reg2})"), "Lw");
 }
 
 TEST_F(AssemblerRISCV64Test, Ld) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Ld, -12, "ld {reg1}, {imm}({reg2})"), "Ld");
 }
 
 TEST_F(AssemblerRISCV64Test, Lbu) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Lbu, -12, "lbu {reg1}, {imm}({reg2})"), "Lbu");
 }
 
 TEST_F(AssemblerRISCV64Test, Lhu) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Lhu, -12, "lhu {reg1}, {imm}({reg2})"), "Lhu");
 }
 
 TEST_F(AssemblerRISCV64Test, Lwu) {
+  // Note: There is no 16-bit instruction for `Lwu()`.
   DriverStr(RepeatRRIb(&Riscv64Assembler::Lwu, -12, "lwu {reg1}, {imm}({reg2})"), "Lwu");
 }
 
 TEST_F(AssemblerRISCV64Test, Sb) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Sb, -12, "sb {reg1}, {imm}({reg2})"), "Sb");
 }
 
 TEST_F(AssemblerRISCV64Test, Sh) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Sh, -12, "sh {reg1}, {imm}({reg2})"), "Sh");
 }
 
 TEST_F(AssemblerRISCV64Test, Sw) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Sw, -12, "sw {reg1}, {imm}({reg2})"), "Sw");
 }
 
 TEST_F(AssemblerRISCV64Test, Sd) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Sd, -12, "sd {reg1}, {imm}({reg2})"), "Sd");
 }
 
 TEST_F(AssemblerRISCV64Test, Addi) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Addi, -12, "addi {reg1}, {reg2}, {imm}"), "Addi");
 }
 
@@ -2091,26 +2322,32 @@ TEST_F(AssemblerRISCV64Test, Ori) {
 }
 
 TEST_F(AssemblerRISCV64Test, Andi) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Andi, -12, "andi {reg1}, {reg2}, {imm}"), "Andi");
 }
 
 TEST_F(AssemblerRISCV64Test, Slli) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Slli, 6, "slli {reg1}, {reg2}, {imm}"), "Slli");
 }
 
 TEST_F(AssemblerRISCV64Test, Srli) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Srli, 6, "srli {reg1}, {reg2}, {imm}"), "Srli");
 }
 
 TEST_F(AssemblerRISCV64Test, Srai) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Srai, 6, "srai {reg1}, {reg2}, {imm}"), "Srai");
 }
 
 TEST_F(AssemblerRISCV64Test, Add) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Add, "add {reg1}, {reg2}, {reg3}"), "Add");
 }
 
 TEST_F(AssemblerRISCV64Test, Sub) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Sub, "sub {reg1}, {reg2}, {reg3}"), "Sub");
 }
 
@@ -2123,14 +2360,17 @@ TEST_F(AssemblerRISCV64Test, Sltu) {
 }
 
 TEST_F(AssemblerRISCV64Test, Xor) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Xor, "xor {reg1}, {reg2}, {reg3}"), "Xor");
 }
 
 TEST_F(AssemblerRISCV64Test, Or) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Or, "or {reg1}, {reg2}, {reg3}"), "Or");
 }
 
 TEST_F(AssemblerRISCV64Test, And) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::And, "and {reg1}, {reg2}, {reg3}"), "And");
 }
 
@@ -2147,6 +2387,7 @@ TEST_F(AssemblerRISCV64Test, Sra) {
 }
 
 TEST_F(AssemblerRISCV64Test, Addiw) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRIb(&Riscv64Assembler::Addiw, -12, "addiw {reg1}, {reg2}, {imm}"), "Addiw");
 }
 
@@ -2166,10 +2407,12 @@ TEST_F(AssemblerRISCV64Test, Sraiw) {
 }
 
 TEST_F(AssemblerRISCV64Test, Addw) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Addw, "addw {reg1}, {reg2}, {reg3}"), "Addw");
 }
 
 TEST_F(AssemblerRISCV64Test, Subw) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Subw, "subw {reg1}, {reg2}, {reg3}"), "Subw");
 }
 
@@ -2191,6 +2434,7 @@ TEST_F(AssemblerRISCV64Test, Ecall) {
 }
 
 TEST_F(AssemblerRISCV64Test, Ebreak) {
+  ScopedCSuppression scs(this);
   __ Ebreak();
   DriverStr("ebreak\n", "Ebreak");
 }
@@ -2241,6 +2485,7 @@ TEST_F(AssemblerRISCV64Test, FenceI) {
 }
 
 TEST_F(AssemblerRISCV64Test, Mul) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRRR(&Riscv64Assembler::Mul, "mul {reg1}, {reg2}, {reg3}"), "Mul");
 }
 
@@ -2433,18 +2678,22 @@ TEST_F(AssemblerRISCV64Test, Csrrci) {
 }
 
 TEST_F(AssemblerRISCV64Test, FLw) {
+  // Note: 16-bit variants of `flw` are not available on riscv64.
   DriverStr(RepeatFRIb(&Riscv64Assembler::FLw, -12, "flw {reg1}, {imm}({reg2})"), "FLw");
 }
 
 TEST_F(AssemblerRISCV64Test, FLd) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatFRIb(&Riscv64Assembler::FLd, -12, "fld {reg1}, {imm}({reg2})"), "FLw");
 }
 
 TEST_F(AssemblerRISCV64Test, FSw) {
+  // Note: 16-bit variants of `fsw` are not available on riscv64.
   DriverStr(RepeatFRIb(&Riscv64Assembler::FSw, 2, "fsw {reg1}, {imm}({reg2})"), "FSw");
 }
 
 TEST_F(AssemblerRISCV64Test, FSd) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatFRIb(&Riscv64Assembler::FSd, 2, "fsd {reg1}, {imm}({reg2})"), "FSd");
 }
 
@@ -2874,7 +3123,384 @@ TEST_F(AssemblerRISCV64Test, FClassD) {
   DriverStr(RepeatrF(&Riscv64Assembler::FClassD, "fclass.d {reg1}, {reg2}"), "FClassD");
 }
 
+TEST_F(AssemblerRISCV64Test, CLwsp) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CLwsp,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/true,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/6,
+                        /*shift=*/2,
+                        "c.lwsp {reg}, {imm}(sp)"),
+            "CLwsp");
+}
+
+TEST_F(AssemblerRISCV64Test, CLdsp) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CLdsp,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/true,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/6,
+                        /*shift=*/3,
+                        "c.ldsp {reg}, {imm}(sp)"),
+            "CLdsp");
+}
+
+TEST_F(AssemblerRISCV64Test, CFLdsp) {
+  DriverStr(RepeatCFImm(
+                &Riscv64Assembler::CFLdsp, /*imm_bits=*/6, /*shift=*/3, "c.fldsp {reg}, {imm}(sp)"),
+            "CFLdsp");
+}
+
+TEST_F(AssemblerRISCV64Test, CSwsp) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CSwsp,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/6,
+                        /*shift=*/2,
+                        "c.swsp {reg}, {imm}(sp)"),
+            "CLwsp");
+}
+
+TEST_F(AssemblerRISCV64Test, CSdsp) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CSdsp,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/6,
+                        /*shift=*/3,
+                        "c.sdsp {reg}, {imm}(sp)"),
+            "CLdsp");
+}
+
+TEST_F(AssemblerRISCV64Test, CFSdsp) {
+  DriverStr(RepeatCFImm(
+                &Riscv64Assembler::CFSdsp, /*imm_bits=*/6, /*shift=*/3, "c.fsdsp {reg}, {imm}(sp)"),
+            "CFLdsp");
+}
+
+TEST_F(AssemblerRISCV64Test, CLw) {
+  DriverStr(RepeatCRRImm(
+                &Riscv64Assembler::CLw, /*imm_bits=*/5, /*shift=*/2, "c.lw {reg1}, {imm}({reg2})"),
+            "CLw");
+}
+
+TEST_F(AssemblerRISCV64Test, CLd) {
+  DriverStr(RepeatCRRImm(
+                &Riscv64Assembler::CLd, /*imm_bits=*/5, /*shift=*/3, "c.ld {reg1}, {imm}({reg2})"),
+            "CLd");
+}
+
+TEST_F(AssemblerRISCV64Test, CFLd) {
+  DriverStr(RepeatCFRImm(&Riscv64Assembler::CFLd,
+                         /*imm_bits=*/5,
+                         /*shift=*/3,
+                         "c.fld {reg1}, {imm}({reg2})"),
+            "CFLd");
+}
+
+TEST_F(AssemblerRISCV64Test, CSw) {
+  DriverStr(RepeatCRRImm(
+                &Riscv64Assembler::CSw, /*imm_bits=*/5, /*shift=*/2, "c.sw {reg1}, {imm}({reg2})"),
+            "CSw");
+}
+
+TEST_F(AssemblerRISCV64Test, CSd) {
+  DriverStr(RepeatCRRImm(
+                &Riscv64Assembler::CSd, /*imm_bits=*/5, /*shift=*/3, "c.sd {reg1}, {imm}({reg2})"),
+            "CSd");
+}
+
+TEST_F(AssemblerRISCV64Test, CFSd) {
+  DriverStr(RepeatCFRImm(&Riscv64Assembler::CFSd,
+                         /*imm_bits=*/5,
+                         /*shift=*/3,
+                         "c.fsd {reg1}, {imm}({reg2})"),
+            "CFSd");
+}
+
+TEST_F(AssemblerRISCV64Test, CLi) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CLi,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/true,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/-6,
+                        /*shift=*/0,
+                        "c.li {reg}, {imm}"),
+            "CLi");
+}
+
+TEST_F(AssemblerRISCV64Test, CLui) {
+  std::string str;
+  auto imms = CreateImmediateValuesBits(/*imm_bits=*/5, /*as_uint=*/true);
+  for (uint32_t v = 0xfffe0; v <= 0xfffff; ++v) {
+    imms.push_back(v);
+  }
+
+  for (XRegister reg : GetRegisters()) {
+    for (int64_t imm_raw : imms) {
+      if (imm_raw == 0) {
+        continue;
+      }
+
+      if (reg == Zero || reg == SP) {
+        continue;
+      }
+
+      uint32_t imm = CreateImmediate(imm_raw);
+      GetAssembler()->CLui(reg, imm);
+
+      std::string base = "c.lui {reg}, {imm}";
+      ReplaceReg(REG_TOKEN, GetRegisterName(reg), &base);
+      ReplaceImm(imm, /*bias=*/0, /*multiplier=*/1, &base);
+      str += base;
+      str += "\n";
+    }
+  }
+
+  DriverStr(str, "CLui");
+}
+
+TEST_F(AssemblerRISCV64Test, CAddi) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CAddi,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/true,
+                        /*no_zero_imm=*/true,
+                        /*imm_bits=*/-6,
+                        /*shift=*/0,
+                        "c.addi {reg}, {imm}"),
+            "CAddi");
+}
+
+TEST_F(AssemblerRISCV64Test, CAddiw) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CAddiw,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/true,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/-6,
+                        /*shift=*/0,
+                        "c.addiw {reg}, {imm}"),
+            "CAddiw");
+}
+
+TEST_F(AssemblerRISCV64Test, CAddi16Sp) {
+  DriverStr(RepeatImm(&Riscv64Assembler::CAddi16Sp,
+                      /*no_zero_imm=*/true,
+                      /*imm_bits=*/-6,
+                      /*shift=*/4,
+                      "c.addi16sp sp, {imm}"),
+            "CAddi16Sp");
+}
+
+TEST_F(AssemblerRISCV64Test, CAddi4Spn) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CAddi4Spn,
+                        /*is_short=*/true,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/true,
+                        /*imm_bits=*/8,
+                        /*shift=*/2,
+                        "c.addi4spn {reg}, sp, {imm}"),
+            "CAddi4Spn");
+}
+
+TEST_F(AssemblerRISCV64Test, CSlli) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CSlli,
+                        /*is_short=*/false,
+                        /*no_zero_reg=*/true,
+                        /*no_zero_imm=*/true,
+                        /*imm_bits=*/6,
+                        /*shift=*/0,
+                        "c.slli {reg}, {imm}"),
+            "CSlli");
+}
+
+TEST_F(AssemblerRISCV64Test, CSRli) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CSrli,
+                        /*is_short=*/true,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/true,
+                        /*imm_bits=*/6,
+                        /*shift=*/0,
+                        "c.srli {reg}, {imm}"),
+            "CSRli");
+}
+
+TEST_F(AssemblerRISCV64Test, CSRai) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CSrai,
+                        /*is_short=*/true,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/true,
+                        /*imm_bits=*/6,
+                        /*shift=*/0,
+                        "c.srai {reg}, {imm}"),
+            "CSRai");
+}
+
+TEST_F(AssemblerRISCV64Test, CAndi) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CAndi,
+                        /*is_short=*/true,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/-6,
+                        /*shift=*/0,
+                        "c.andi {reg}, {imm}"),
+            "CAndi");
+}
+
+TEST_F(AssemblerRISCV64Test, CMv) {
+  DriverStr(RepeatCRRNonZero(&Riscv64Assembler::CMv, "c.mv {reg1}, {reg2}"), "CMv");
+}
+
+TEST_F(AssemblerRISCV64Test, CAdd) {
+  DriverStr(RepeatCRRNonZero(&Riscv64Assembler::CAdd, "c.add {reg1}, {reg2}"), "CAdd");
+}
+
+TEST_F(AssemblerRISCV64Test, CAnd) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::CAnd, "c.and {reg1}, {reg2}"), "CAnd");
+}
+
+TEST_F(AssemblerRISCV64Test, COr) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::COr, "c.or {reg1}, {reg2}"), "COr");
+}
+
+TEST_F(AssemblerRISCV64Test, CXor) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::CXor, "c.xor {reg1}, {reg2}"), "CXor");
+}
+
+TEST_F(AssemblerRISCV64Test, CSub) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::CSub, "c.sub {reg1}, {reg2}"), "CSub");
+}
+
+TEST_F(AssemblerRISCV64Test, CAddw) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::CAddw, "c.addw {reg1}, {reg2}"), "CAddw");
+}
+
+TEST_F(AssemblerRISCV64Test, CSubw) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::CSubw, "c.subw {reg1}, {reg2}"), "CSubw");
+}
+
+TEST_F(AssemblerRISCV64Test, CLbu) {
+  DriverStr(RepeatCRRImm(&Riscv64Assembler::CLbu,
+                         /*imm_bits=*/2,
+                         /*shift=*/0,
+                         "c.lbu {reg1}, {imm}({reg2})"),
+            "CLbu");
+}
+
+TEST_F(AssemblerRISCV64Test, CLhu) {
+  DriverStr(RepeatCRRImm(&Riscv64Assembler::CLhu,
+                         /*imm_bits=*/1,
+                         /*shift=*/1,
+                         "c.lhu {reg1}, {imm}({reg2})"),
+            "CLhu");
+}
+
+TEST_F(AssemblerRISCV64Test, CLh) {
+  DriverStr(RepeatCRRImm(&Riscv64Assembler::CLh,
+                         /*imm_bits=*/1,
+                         /*shift=*/1,
+                         "c.lh {reg1}, {imm}({reg2})"),
+            "CLh");
+}
+
+TEST_F(AssemblerRISCV64Test, CSb) {
+  DriverStr(RepeatCRRImm(&Riscv64Assembler::CSb,
+                         /*imm_bits=*/2,
+                         /*shift=*/0,
+                         "c.sb {reg1}, {imm}({reg2})"),
+            "CSb");
+}
+
+TEST_F(AssemblerRISCV64Test, CSh) {
+  DriverStr(RepeatCRRImm(&Riscv64Assembler::CSh,
+                         /*imm_bits=*/1,
+                         /*shift=*/1,
+                         "c.sh {reg1}, {imm}({reg2})"),
+            "CSh");
+}
+
+TEST_F(AssemblerRISCV64Test, CZext_b) {
+  DriverStr(RepeatCRShort(&Riscv64Assembler::CZext_b, "c.zext.b {reg}"), "CZext_b");
+}
+
+TEST_F(AssemblerRISCV64Test, CSext_b) {
+  DriverStr(RepeatCRShort(&Riscv64Assembler::CSext_b, "c.sext.b {reg}"), "CSext_b");
+}
+
+TEST_F(AssemblerRISCV64Test, CZext_h) {
+  DriverStr(RepeatCRShort(&Riscv64Assembler::CZext_h, "c.zext.h {reg}"), "CZext_h");
+}
+
+TEST_F(AssemblerRISCV64Test, CSext_h) {
+  DriverStr(RepeatCRShort(&Riscv64Assembler::CSext_h, "c.sext.h {reg}"), "CSext_h");
+}
+
+TEST_F(AssemblerRISCV64Test, CZext_w) {
+  DriverStr(RepeatCRShort(&Riscv64Assembler::CZext_w, "c.zext.w {reg}"), "CZext_w");
+}
+
+TEST_F(AssemblerRISCV64Test, CNot) {
+  DriverStr(RepeatCRShort(&Riscv64Assembler::CNot, "c.not {reg}"), "CNot");
+}
+
+TEST_F(AssemblerRISCV64Test, CMul) {
+  DriverStr(RepeatCRRShort(&Riscv64Assembler::CMul, "c.mul {reg1}, {reg2}"), "CMul");
+}
+
+TEST_F(AssemblerRISCV64Test, CJ) {
+  DriverStr(
+      RepeatImm(
+          &Riscv64Assembler::CJ, /*no_zero_imm=*/false, /*imm_bits=*/-11, /*shift=*/1, "c.j {imm}"),
+      "CJ");
+}
+
+TEST_F(AssemblerRISCV64Test, CJr) {
+  DriverStr(RepeatRNoZero(&Riscv64Assembler::CJr, "c.jr {reg}"), "CJr");
+}
+
+TEST_F(AssemblerRISCV64Test, CJalr) {
+  DriverStr(RepeatRNoZero(&Riscv64Assembler::CJalr, "c.jalr {reg}"), "CJalr");
+}
+
+TEST_F(AssemblerRISCV64Test, CBeqz) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CBeqz,
+                        /*is_short=*/true,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/-8,
+                        /*shift=*/1,
+                        "c.beqz {reg}, {imm}"),
+            "CBeqz");
+}
+
+TEST_F(AssemblerRISCV64Test, CBnez) {
+  DriverStr(RepeatCRImm(&Riscv64Assembler::CBnez,
+                        /*is_short=*/true,
+                        /*no_zero_reg=*/false,
+                        /*no_zero_imm=*/false,
+                        /*imm_bits=*/-8,
+                        /*shift=*/1,
+                        "c.bnez {reg}, {imm}"),
+            "CBnez");
+}
+
+TEST_F(AssemblerRISCV64Test, CEbreak) {
+  __ CEbreak();
+  DriverStr("c.ebreak", "CEbreak");
+}
+
+TEST_F(AssemblerRISCV64Test, CNop) {
+  __ CNop();
+  DriverStr("c.nop", "CNop");
+}
+
+TEST_F(AssemblerRISCV64Test, CUnimp) {
+  __ CUnimp();
+  DriverStr("c.unimp", "CUnimp");
+}
+
 TEST_F(AssemblerRISCV64Test, AddUw) {
+  ScopedCSuppression scs(this);  // Avoid `c.zext.w`.
   DriverStr(RepeatRRR(&Riscv64Assembler::AddUw, "add.uw {reg1}, {reg2}, {reg3}"), "AddUw");
 }
 
@@ -2991,6 +3617,21 @@ TEST_F(AssemblerRISCV64Test, Rev8) {
   DriverStr(RepeatRR(&Riscv64Assembler::Rev8, "rev8 {reg1}, {reg2}"), "Rev8");
 }
 
+TEST_F(AssemblerRISCV64Test, ZbbSextB) {
+  ScopedCSuppression scs(this);
+  DriverStr(RepeatRR(&Riscv64Assembler::ZbbSextB, "sext.b {reg1}, {reg2}"), "ZbbSextB");
+}
+
+TEST_F(AssemblerRISCV64Test, ZbbSextH) {
+  ScopedCSuppression scs(this);
+  DriverStr(RepeatRR(&Riscv64Assembler::ZbbSextH, "sext.h {reg1}, {reg2}"), "ZbbSextH");
+}
+
+TEST_F(AssemblerRISCV64Test, ZbbZextH) {
+  ScopedCSuppression scs(this);
+  DriverStr(RepeatRR(&Riscv64Assembler::ZbbZextH, "zext.h {reg1}, {reg2}"), "ZbbZextH");
+}
+
 // Vector Instructions
 
 TEST_F(AssemblerRISCV64Test, VSetvl) {
@@ -7203,22 +7844,25 @@ TEST_F(AssemblerRISCV64Test, VId_v) {
 
 // Pseudo instructions.
 TEST_F(AssemblerRISCV64Test, Nop) {
+  ScopedCSuppression scs(this);
   __ Nop();
   DriverStr("addi zero,zero,0", "Nop");
 }
 
 TEST_F(AssemblerRISCV64Test, Li) {
-  SetUseSimpleMarch(true);
+  ScopedMarchOverride smo(this, "-march=rv64imafd");
   TestLoadConst64("Li",
                   /*can_use_tmp=*/ false,
                   [&](XRegister rd, int64_t value) { __ Li(rd, value); });
 }
 
 TEST_F(AssemblerRISCV64Test, Mv) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRR(&Riscv64Assembler::Mv, "addi {reg1}, {reg2}, 0"), "Mv");
 }
 
 TEST_F(AssemblerRISCV64Test, Not) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRR(&Riscv64Assembler::Not, "xori {reg1}, {reg2}, -1"), "Not");
 }
 
@@ -7231,6 +7875,7 @@ TEST_F(AssemblerRISCV64Test, NegW) {
 }
 
 TEST_F(AssemblerRISCV64Test, SextB) {
+  ScopedCSuppression scs(this);
   // Note: SEXT.B from the Zbb extension is not supported.
   DriverStr(RepeatRR(&Riscv64Assembler::SextB,
                      "slli {reg1}, {reg2}, 56\n"
@@ -7239,6 +7884,7 @@ TEST_F(AssemblerRISCV64Test, SextB) {
 }
 
 TEST_F(AssemblerRISCV64Test, SextH) {
+  ScopedCSuppression scs(this);
   // Note: SEXT.H from the Zbb extension is not supported.
   DriverStr(RepeatRR(&Riscv64Assembler::SextH,
                      "slli {reg1}, {reg2}, 48\n"
@@ -7247,14 +7893,17 @@ TEST_F(AssemblerRISCV64Test, SextH) {
 }
 
 TEST_F(AssemblerRISCV64Test, SextW) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRR(&Riscv64Assembler::SextW, "addiw {reg1}, {reg2}, 0\n"), "SextW");
 }
 
 TEST_F(AssemblerRISCV64Test, ZextB) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRR(&Riscv64Assembler::ZextB, "andi {reg1}, {reg2}, 255"), "ZextB");
 }
 
 TEST_F(AssemblerRISCV64Test, ZextH) {
+  ScopedCSuppression scs(this);
   // Note: ZEXT.H from the Zbb extension is not supported.
   DriverStr(RepeatRR(&Riscv64Assembler::ZextH,
                      "slli {reg1}, {reg2}, 48\n"
@@ -7263,6 +7912,7 @@ TEST_F(AssemblerRISCV64Test, ZextH) {
 }
 
 TEST_F(AssemblerRISCV64Test, ZextW) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRR(&Riscv64Assembler::ZextW,
                      "slli {reg1}, {reg2}, 32\n"
                      "srli {reg1}, {reg1}, 32"),
@@ -7310,11 +7960,13 @@ TEST_F(AssemblerRISCV64Test, FNegD) {
 }
 
 TEST_F(AssemblerRISCV64Test, Beqz) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-11, 2" to "-12, 1" for "C" Standard Extension.
   DriverStr(RepeatRIbS(&Riscv64Assembler::Beqz, -11, 2, "beq {reg}, zero, {imm}\n"), "Beqz");
 }
 
 TEST_F(AssemblerRISCV64Test, Bnez) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-11, 2" to "-12, 1" for "C" Standard Extension.
   DriverStr(RepeatRIbS(&Riscv64Assembler::Bnez, -11, 2, "bne {reg}, zero, {imm}\n"), "Bnez");
 }
@@ -7360,6 +8012,7 @@ TEST_F(AssemblerRISCV64Test, Bleu) {
 }
 
 TEST_F(AssemblerRISCV64Test, J) {
+  ScopedCSuppression scs(this);
   // TODO(riscv64): Change "-19, 2" to "-20, 1" for "C" Standard Extension.
   DriverStr(RepeatIbS<int32_t>(&Riscv64Assembler::J, -19, 2, "j {imm}\n"), "J");
 }
@@ -7370,18 +8023,22 @@ TEST_F(AssemblerRISCV64Test, JalRA) {
 }
 
 TEST_F(AssemblerRISCV64Test, Jr) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatR(&Riscv64Assembler::Jr, "jr {reg}\n"), "Jr");
 }
 
 TEST_F(AssemblerRISCV64Test, JalrRA) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatR(&Riscv64Assembler::Jalr, "jalr {reg}\n"), "JalrRA");
 }
 
 TEST_F(AssemblerRISCV64Test, Jalr0) {
+  ScopedCSuppression scs(this);
   DriverStr(RepeatRR(&Riscv64Assembler::Jalr, "jalr {reg1}, {reg2}\n"), "Jalr0");
 }
 
 TEST_F(AssemblerRISCV64Test, Ret) {
+  ScopedCSuppression scs(this);
   __ Ret();
   DriverStr("ret\n", "Ret");
 }
@@ -7434,6 +8091,7 @@ TEST_F(AssemblerRISCV64Test, Csrci) {
 }
 
 TEST_F(AssemblerRISCV64Test, LoadConst32) {
+  ScopedCSuppression scs(this);
   // `LoadConst32()` emits the same code sequences as `Li()` for 32-bit values.
   ScratchRegisterScope srs(GetAssembler());
   srs.ExcludeXRegister(TMP);
@@ -7442,13 +8100,14 @@ TEST_F(AssemblerRISCV64Test, LoadConst32) {
 }
 
 TEST_F(AssemblerRISCV64Test, LoadConst64) {
-  SetUseSimpleMarch(true);
+  ScopedMarchOverride smo(this, "-march=rv64imafd");
   TestLoadConst64("LoadConst64",
                   /*can_use_tmp=*/ true,
                   [&](XRegister rd, int64_t value) { __ LoadConst64(rd, value); });
 }
 
 TEST_F(AssemblerRISCV64Test, AddConst32) {
+  ScopedCSuppression scs(this);
   auto emit_op = [&](XRegister rd, XRegister rs1, int64_t value) {
     __ AddConst32(rd, rs1, dchecked_integral_cast<int32_t>(value));
   };
@@ -7456,7 +8115,7 @@ TEST_F(AssemblerRISCV64Test, AddConst32) {
 }
 
 TEST_F(AssemblerRISCV64Test, AddConst64) {
-  SetUseSimpleMarch(true);
+  ScopedMarchOverride smo(this, "-march=rv64imafd");
   auto emit_op = [&](XRegister rd, XRegister rs1, int64_t value) {
     __ AddConst64(rd, rs1, value);
   };
@@ -7464,38 +8123,47 @@ TEST_F(AssemblerRISCV64Test, AddConst64) {
 }
 
 TEST_F(AssemblerRISCV64Test, BcondForward3KiB) {
+  ScopedCSuppression scs(this);
   TestBcondForward("BcondForward3KiB", 3 * KB, "1", GetPrintBcond());
 }
 
 TEST_F(AssemblerRISCV64Test, BcondForward3KiBBare) {
+  ScopedCSuppression scs(this);
   TestBcondForward("BcondForward3KiB", 3 * KB, "1", GetPrintBcond(), /*is_bare=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, BcondBackward3KiB) {
+  ScopedCSuppression scs(this);
   TestBcondBackward("BcondBackward3KiB", 3 * KB, "1", GetPrintBcond());
 }
 
 TEST_F(AssemblerRISCV64Test, BcondBackward3KiBBare) {
+  ScopedCSuppression scs(this);
   TestBcondBackward("BcondBackward3KiB", 3 * KB, "1", GetPrintBcond(), /*is_bare=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, BcondForward5KiB) {
+  ScopedCSuppression scs(this);
   TestBcondForward("BcondForward5KiB", 5 * KB, "1", GetPrintBcondOppositeAndJ("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, BcondBackward5KiB) {
+  ScopedCSuppression scs(this);
   TestBcondBackward("BcondBackward5KiB", 5 * KB, "1", GetPrintBcondOppositeAndJ("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, BcondForward2MiB) {
+  ScopedCSuppression scs(this);
   TestBcondForward("BcondForward2MiB", 2 * MB, "1", GetPrintBcondOppositeAndTail("2", "3"));
 }
 
 TEST_F(AssemblerRISCV64Test, BcondBackward2MiB) {
+  ScopedCSuppression scs(this);
   TestBcondBackward("BcondBackward2MiB", 2 * MB, "1", GetPrintBcondOppositeAndTail("2", "3"));
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13Forward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Forward("BeqA0A1MaxOffset13Forward",
                      MaxOffset13ForwardDistance() - /*BEQ*/ 4u,
                      "1",
@@ -7503,6 +8171,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13ForwardBare) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Forward("BeqA0A1MaxOffset13ForwardBare",
                      MaxOffset13ForwardDistance() - /*BEQ*/ 4u,
                      "1",
@@ -7511,6 +8180,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13ForwardBare) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13Backward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Backward("BeqA0A1MaxOffset13Forward",
                       MaxOffset13BackwardDistance(),
                       "1",
@@ -7518,6 +8188,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13BackwardBare) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Backward("BeqA0A1MaxOffset13ForwardBare",
                       MaxOffset13BackwardDistance(),
                       "1",
@@ -7526,6 +8197,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset13BackwardBare) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset13Forward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Forward("BeqA0A1OverMaxOffset13Forward",
                      MaxOffset13ForwardDistance() - /*BEQ*/ 4u + /*Exceed max*/ 4u,
                      "1",
@@ -7533,6 +8205,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset13Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset13Backward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Backward("BeqA0A1OverMaxOffset13Forward",
                       MaxOffset13BackwardDistance() + /*Exceed max*/ 4u,
                       "1",
@@ -7540,6 +8213,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset13Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset21Forward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Forward("BeqA0A1MaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u,
                      "1",
@@ -7547,6 +8221,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset21Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset21Backward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Backward("BeqA0A1MaxOffset21Backward",
                       MaxOffset21BackwardDistance() - /*BNE*/ 4u,
                       "1",
@@ -7554,6 +8229,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1MaxOffset21Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset21Forward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Forward("BeqA0A1OverMaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u + /*Exceed max*/ 4u,
                      "1",
@@ -7561,6 +8237,7 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset21Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset21Backward) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1Backward("BeqA0A1OverMaxOffset21Backward",
                       MaxOffset21BackwardDistance() - /*BNE*/ 4u + /*Exceed max*/ 4u,
                       "1",
@@ -7568,15 +8245,18 @@ TEST_F(AssemblerRISCV64Test, BeqA0A1OverMaxOffset21Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1AlmostCascade) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1MaybeCascade("BeqA0A1AlmostCascade", /*cascade=*/ false, GetPrintBcond());
 }
 
 TEST_F(AssemblerRISCV64Test, BeqA0A1Cascade) {
+  ScopedCSuppression scs(this);
   TestBeqA0A1MaybeCascade(
       "BeqA0A1AlmostCascade", /*cascade=*/ true, GetPrintBcondOppositeAndJ("1"));
 }
 
 TEST_F(AssemblerRISCV64Test, BcondElimination) {
+  ScopedCSuppression scs(this);
   Riscv64Label label;
   __ Bind(&label);
   __ Nop();
@@ -7591,6 +8271,7 @@ TEST_F(AssemblerRISCV64Test, BcondElimination) {
 }
 
 TEST_F(AssemblerRISCV64Test, BcondUnconditional) {
+  ScopedCSuppression scs(this);
   Riscv64Label label;
   __ Bind(&label);
   __ Nop();
@@ -7609,54 +8290,67 @@ TEST_F(AssemblerRISCV64Test, BcondUnconditional) {
 }
 
 TEST_F(AssemblerRISCV64Test, JalRdForward3KiB) {
+  ScopedCSuppression scs(this);
   TestJalRdForward("JalRdForward3KiB", 3 * KB, "1", GetPrintJalRd());
 }
 
 TEST_F(AssemblerRISCV64Test, JalRdForward3KiBBare) {
+  ScopedCSuppression scs(this);
   TestJalRdForward("JalRdForward3KiB", 3 * KB, "1", GetPrintJalRd(), /*is_bare=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, JalRdBackward3KiB) {
+  ScopedCSuppression scs(this);
   TestJalRdBackward("JalRdBackward3KiB", 3 * KB, "1", GetPrintJalRd());
 }
 
 TEST_F(AssemblerRISCV64Test, JalRdBackward3KiBBare) {
+  ScopedCSuppression scs(this);
   TestJalRdBackward("JalRdBackward3KiB", 3 * KB, "1", GetPrintJalRd(), /*is_bare=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, JalRdForward2MiB) {
+  ScopedCSuppression scs(this);
   TestJalRdForward("JalRdForward2MiB", 2 * MB, "1", GetPrintCallRd("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, JalRdBackward2MiB) {
+  ScopedCSuppression scs(this);
   TestJalRdBackward("JalRdBackward2MiB", 2 * MB, "1", GetPrintCallRd("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, JForward3KiB) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("JForward3KiB", 3 * KB, "1", GetEmitJ(), GetPrintJ());
 }
 
 TEST_F(AssemblerRISCV64Test, JForward3KiBBare) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("JForward3KiB", 3 * KB, "1", GetEmitJ(/*is_bare=*/ true), GetPrintJ());
 }
 
 TEST_F(AssemblerRISCV64Test, JBackward3KiB) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("JBackward3KiB", 3 * KB, "1", GetEmitJ(), GetPrintJ());
 }
 
 TEST_F(AssemblerRISCV64Test, JBackward3KiBBare) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("JBackward3KiB", 3 * KB, "1", GetEmitJ(/*is_bare=*/ true), GetPrintJ());
 }
 
 TEST_F(AssemblerRISCV64Test, JForward2MiB) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("JForward2MiB", 2 * MB, "1", GetEmitJ(), GetPrintTail("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, JBackward2MiB) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("JBackward2MiB", 2 * MB, "1", GetEmitJ(), GetPrintTail("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, JMaxOffset21Forward) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("JMaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u,
                      "1",
@@ -7665,6 +8359,7 @@ TEST_F(AssemblerRISCV64Test, JMaxOffset21Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, JMaxOffset21ForwardBare) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("JMaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u,
                      "1",
@@ -7673,6 +8368,7 @@ TEST_F(AssemblerRISCV64Test, JMaxOffset21ForwardBare) {
 }
 
 TEST_F(AssemblerRISCV64Test, JMaxOffset21Backward) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("JMaxOffset21Backward",
                       MaxOffset21BackwardDistance(),
                       "1",
@@ -7681,6 +8377,7 @@ TEST_F(AssemblerRISCV64Test, JMaxOffset21Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, JMaxOffset21BackwardBare) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("JMaxOffset21Backward",
                       MaxOffset21BackwardDistance(),
                       "1",
@@ -7689,6 +8386,7 @@ TEST_F(AssemblerRISCV64Test, JMaxOffset21BackwardBare) {
 }
 
 TEST_F(AssemblerRISCV64Test, JOverMaxOffset21Forward) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("JOverMaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u + /*Exceed max*/ 4u,
                      "1",
@@ -7697,6 +8395,7 @@ TEST_F(AssemblerRISCV64Test, JOverMaxOffset21Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, JOverMaxOffset21Backward) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("JMaxOffset21Backward",
                       MaxOffset21BackwardDistance() + /*Exceed max*/ 4u,
                       "1",
@@ -7705,22 +8404,27 @@ TEST_F(AssemblerRISCV64Test, JOverMaxOffset21Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, CallForward3KiB) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("CallForward3KiB", 3 * KB, "1", GetEmitJal(), GetPrintJal());
 }
 
 TEST_F(AssemblerRISCV64Test, CallBackward3KiB) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("CallBackward3KiB", 3 * KB, "1", GetEmitJal(), GetPrintJal());
 }
 
 TEST_F(AssemblerRISCV64Test, CallForward2MiB) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("CallForward2MiB", 2 * MB, "1", GetEmitJal(), GetPrintCall("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, CallBackward2MiB) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("CallBackward2MiB", 2 * MB, "1", GetEmitJal(), GetPrintCall("2"));
 }
 
 TEST_F(AssemblerRISCV64Test, CallMaxOffset21Forward) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("CallMaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u,
                      "1",
@@ -7729,6 +8433,7 @@ TEST_F(AssemblerRISCV64Test, CallMaxOffset21Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, CallMaxOffset21Backward) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("CallMaxOffset21Backward",
                       MaxOffset21BackwardDistance(),
                       "1",
@@ -7737,6 +8442,7 @@ TEST_F(AssemblerRISCV64Test, CallMaxOffset21Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, CallOverMaxOffset21Forward) {
+  ScopedCSuppression scs(this);
   TestBuncondForward("CallOverMaxOffset21Forward",
                      MaxOffset21ForwardDistance() - /*J*/ 4u + /*Exceed max*/ 4u,
                      "1",
@@ -7745,6 +8451,7 @@ TEST_F(AssemblerRISCV64Test, CallOverMaxOffset21Forward) {
 }
 
 TEST_F(AssemblerRISCV64Test, CallOverMaxOffset21Backward) {
+  ScopedCSuppression scs(this);
   TestBuncondBackward("CallMaxOffset21Backward",
                       MaxOffset21BackwardDistance() + /*Exceed max*/ 4u,
                       "1",
@@ -7753,66 +8460,82 @@ TEST_F(AssemblerRISCV64Test, CallOverMaxOffset21Backward) {
 }
 
 TEST_F(AssemblerRISCV64Test, Loadb) {
+  ScopedCSuppression scs(this);  // Suppress 16-bit instructions for address formation.
   TestLoadStoreArbitraryOffset("Loadb", "lb", &Riscv64Assembler::Loadb, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Loadh) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Loadh", "lh", &Riscv64Assembler::Loadh, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Loadw) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Loadw", "lw", &Riscv64Assembler::Loadw, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Loadd) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Loadd", "ld", &Riscv64Assembler::Loadd, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Loadbu) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Loadbu", "lbu", &Riscv64Assembler::Loadbu, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Loadhu) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Loadhu", "lhu", &Riscv64Assembler::Loadhu, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Loadwu) {
+  ScopedCSuppression scs(this);  // Suppress 16-bit instructions for address formation.
   TestLoadStoreArbitraryOffset("Loadwu", "lwu", &Riscv64Assembler::Loadwu, /*is_store=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, Storeb) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Storeb", "sb", &Riscv64Assembler::Storeb, /*is_store=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, Storeh) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Storeh", "sh", &Riscv64Assembler::Storeh, /*is_store=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, Storew) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Storew", "sw", &Riscv64Assembler::Storew, /*is_store=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, Stored) {
+  ScopedCSuppression scs(this);
   TestLoadStoreArbitraryOffset("Stored", "sd", &Riscv64Assembler::Stored, /*is_store=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, FLoadw) {
+  ScopedCSuppression scs(this);  // Suppress 16-bit instructions for address formation.
   TestFPLoadStoreArbitraryOffset("FLoadw", "flw", &Riscv64Assembler::FLoadw);
 }
 
 TEST_F(AssemblerRISCV64Test, FLoadd) {
+  ScopedCSuppression scs(this);
   TestFPLoadStoreArbitraryOffset("FLoadd", "fld", &Riscv64Assembler::FLoadd);
 }
 
 TEST_F(AssemblerRISCV64Test, FStorew) {
+  ScopedCSuppression scs(this);  // Suppress 16-bit instructions for address formation.
   TestFPLoadStoreArbitraryOffset("FStorew", "fsw", &Riscv64Assembler::FStorew);
 }
 
 TEST_F(AssemblerRISCV64Test, FStored) {
+  ScopedCSuppression scs(this);
   TestFPLoadStoreArbitraryOffset("FStored", "fsd", &Riscv64Assembler::FStored);
 }
 
 TEST_F(AssemblerRISCV64Test, Unimp) {
+  ScopedCSuppression scs(this);
   __ Unimp();
   DriverStr("unimp\n", "Unimp");
 }
@@ -7842,14 +8565,17 @@ TEST_F(AssemblerRISCV64Test, LoadLabelAddress) {
 }
 
 TEST_F(AssemblerRISCV64Test, LoadLiteralWithPaddingForLong) {
+  ScopedCSuppression scs(this);
   TestLoadLiteral("LoadLiteralWithPaddingForLong", /*with_padding_for_long=*/ true);
 }
 
 TEST_F(AssemblerRISCV64Test, LoadLiteralWithoutPaddingForLong) {
+  ScopedCSuppression scs(this);
   TestLoadLiteral("LoadLiteralWithoutPaddingForLong", /*with_padding_for_long=*/ false);
 }
 
 TEST_F(AssemblerRISCV64Test, JumpTable) {
+  ScopedCSuppression scs(this);
   std::string expected;
   expected += EmitNops(sizeof(uint32_t));
   Riscv64Label targets[4];
diff --git a/compiler/utils/riscv64/jni_macro_assembler_riscv64.cc b/compiler/utils/riscv64/jni_macro_assembler_riscv64.cc
index 9d3a29d252..00e1f54d03 100644
--- a/compiler/utils/riscv64/jni_macro_assembler_riscv64.cc
+++ b/compiler/utils/riscv64/jni_macro_assembler_riscv64.cc
@@ -548,7 +548,7 @@ void Riscv64JNIMacroAssembler::DeliverPendingException() {
 }
 
 std::unique_ptr<JNIMacroLabel> Riscv64JNIMacroAssembler::CreateLabel() {
-  return std::unique_ptr<JNIMacroLabel>(new Riscv64JNIMacroLabel());
+  return std::unique_ptr<JNIMacroLabel>(new (asm_.GetAllocator()) Riscv64JNIMacroLabel());
 }
 
 void Riscv64JNIMacroAssembler::Jump(JNIMacroLabel* label) {
diff --git a/compiler/utils/riscv64/jni_macro_assembler_riscv64_test.cc b/compiler/utils/riscv64/jni_macro_assembler_riscv64_test.cc
index be6feeb9de..9717930d58 100644
--- a/compiler/utils/riscv64/jni_macro_assembler_riscv64_test.cc
+++ b/compiler/utils/riscv64/jni_macro_assembler_riscv64_test.cc
@@ -46,6 +46,18 @@ class JniMacroAssemblerRiscv64Test : public AssemblerTestBase {
  protected:
   InstructionSet GetIsa() override { return InstructionSet::kRiscv64; }
 
+  std::vector<std::string> GetAssemblerCommand() override {
+    std::vector<std::string> result = AssemblerTestBase::GetAssemblerCommand();
+    if (march_override_.has_value()) {
+      auto it = std::find_if(result.begin(),
+                             result.end(),
+                             [](const std::string& s) { return StartsWith(s, "-march="); });
+      CHECK(it != result.end());
+      *it = march_override_.value();
+    }
+    return result;
+  }
+
   void DriverStr(const std::string& assembly_text, const std::string& test_name) {
     assembler_.FinalizeCode();
     size_t cs = assembler_.CodeSize();
@@ -76,6 +88,9 @@ class JniMacroAssemblerRiscv64Test : public AssemblerTestBase {
   MallocArenaPool pool_;
   ArenaAllocator allocator_;
   Riscv64JNIMacroAssembler assembler_;
+
+  // TODO: Implement auto-compression and remove this override.
+  std::optional<std::string> march_override_ = "-march=rv64imafdv_zba_zbb";
 };
 
 TEST_F(JniMacroAssemblerRiscv64Test, StackFrame) {
diff --git a/compiler/utils/x86/jni_macro_assembler_x86.cc b/compiler/utils/x86/jni_macro_assembler_x86.cc
index dfdbc183f1..3ee0530dc2 100644
--- a/compiler/utils/x86/jni_macro_assembler_x86.cc
+++ b/compiler/utils/x86/jni_macro_assembler_x86.cc
@@ -542,7 +542,7 @@ void X86JNIMacroAssembler::DeliverPendingException() {
 }
 
 std::unique_ptr<JNIMacroLabel> X86JNIMacroAssembler::CreateLabel() {
-  return std::unique_ptr<JNIMacroLabel>(new X86JNIMacroLabel());
+  return std::unique_ptr<JNIMacroLabel>(new (asm_.GetAllocator()) X86JNIMacroLabel());
 }
 
 void X86JNIMacroAssembler::Jump(JNIMacroLabel* label) {
diff --git a/compiler/utils/x86_64/jni_macro_assembler_x86_64.cc b/compiler/utils/x86_64/jni_macro_assembler_x86_64.cc
index e9e6dbdae7..1b9cfa640e 100644
--- a/compiler/utils/x86_64/jni_macro_assembler_x86_64.cc
+++ b/compiler/utils/x86_64/jni_macro_assembler_x86_64.cc
@@ -610,7 +610,7 @@ void X86_64JNIMacroAssembler::DeliverPendingException() {
 }
 
 std::unique_ptr<JNIMacroLabel> X86_64JNIMacroAssembler::CreateLabel() {
-  return std::unique_ptr<JNIMacroLabel>(new X86_64JNIMacroLabel());
+  return std::unique_ptr<JNIMacroLabel>(new (asm_.GetAllocator()) X86_64JNIMacroLabel());
 }
 
 void X86_64JNIMacroAssembler::Jump(JNIMacroLabel* label) {
diff --git a/dex2oat/dex2oat_image_test.cc b/dex2oat/dex2oat_image_test.cc
index 9ba540b973..e276638372 100644
--- a/dex2oat/dex2oat_image_test.cc
+++ b/dex2oat/dex2oat_image_test.cc
@@ -250,7 +250,6 @@ TEST_F(Dex2oatImageTest, TestModesAndFilters) {
 }
 
 TEST_F(Dex2oatImageTest, TestExtension) {
-  TEST_DISABLED_FOR_RISCV64();
   std::string error_msg;
   MemMap reservation = ReserveCoreImageAddressSpace(&error_msg);
   ASSERT_TRUE(reservation.IsValid()) << error_msg;
diff --git a/dex2oat/dex2oat_test.cc b/dex2oat/dex2oat_test.cc
index 9c6a381ac3..4a8a4f3c65 100644
--- a/dex2oat/dex2oat_test.cc
+++ b/dex2oat/dex2oat_test.cc
@@ -1498,7 +1498,6 @@ TEST_F(Dex2oatVerifierAbort, HardFail) {
 class Dex2oatDedupeCode : public Dex2oatTest {};
 
 TEST_F(Dex2oatDedupeCode, DedupeTest) {
-  TEST_DISABLED_FOR_RISCV64();
   // Use MyClassNatives. It has lots of native methods that will produce deduplicate-able code.
   std::unique_ptr<const DexFile> dex(OpenTestDexFile("MyClassNatives"));
   std::string out_dir = GetScratchDir();
diff --git a/dex2oat/driver/compiler_driver_test.cc b/dex2oat/driver/compiler_driver_test.cc
index a82de055a9..759426a1d3 100644
--- a/dex2oat/driver/compiler_driver_test.cc
+++ b/dex2oat/driver/compiler_driver_test.cc
@@ -261,7 +261,6 @@ class CompilerDriverProfileTest : public CompilerDriverTest {
 };
 
 TEST_F(CompilerDriverProfileTest, ProfileGuidedCompilation) {
-  TEST_DISABLED_FOR_RISCV64();
   Thread* self = Thread::Current();
   jobject class_loader;
   {
diff --git a/dex2oat/verifier_deps_test.cc b/dex2oat/verifier_deps_test.cc
index 0b72382b21..a1e93b750e 100644
--- a/dex2oat/verifier_deps_test.cc
+++ b/dex2oat/verifier_deps_test.cc
@@ -584,7 +584,6 @@ TEST_F(VerifierDepsTest, VerifyDeps) {
 }
 
 TEST_F(VerifierDepsTest, CompilerDriver) {
-  TEST_DISABLED_FOR_RISCV64();
   SetupCompilerDriver();
 
   // Test both multi-dex and single-dex configuration.
diff --git a/disassembler/disassembler_riscv64.cc b/disassembler/disassembler_riscv64.cc
index b92d1287b2..09e9faf7d7 100644
--- a/disassembler/disassembler_riscv64.cc
+++ b/disassembler/disassembler_riscv64.cc
@@ -76,11 +76,21 @@ class DisassemblerRiscv64::Printer {
     kOpCFG = 0b111,
   };
 
+  class ScopedNewLinePrinter {
+    std::ostream& os_;
+
+   public:
+    explicit ScopedNewLinePrinter(std::ostream& os) : os_(os) {}
+    ~ScopedNewLinePrinter() { os_ << '\n'; }
+  };
+
   static const char* XRegName(uint32_t regno);
   static const char* FRegName(uint32_t regno);
   static const char* VRegName(uint32_t regno);
   static const char* RoundingModeName(uint32_t rm);
 
+  // Regular instruction immediate utils
+
   static int32_t Decode32Imm12(uint32_t insn32) {
     uint32_t sign = (insn32 >> 31);
     uint32_t imm12 = (insn32 >> 20);
@@ -99,12 +109,89 @@ class DisassemblerRiscv64::Printer {
     return static_cast<int32_t>(imm) - static_cast<int32_t>(bit11 << 12);  // Sign-extend.
   }
 
+  // Compressed instruction immediate utils
+
+  // Extracts the offset from a compressed instruction
+  // where `offset[5:3]` is in bits `[12:10]` and `offset[2|6]` is in bits `[6:5]`
+  static uint32_t Decode16CMOffsetW(uint32_t insn16) {
+    DCHECK(IsUint<16>(insn16));
+    return BitFieldExtract(insn16, 5, 1) << 6 | BitFieldExtract(insn16, 10, 3) << 3 |
+           BitFieldExtract(insn16, 6, 1) << 2;
+  }
+
+  // Extracts the offset from a compressed instruction
+  // where `offset[5:3]` is in bits `[12:10]` and `offset[7:6]` is in bits `[6:5]`
+  static uint32_t Decode16CMOffsetD(uint32_t insn16) {
+    DCHECK(IsUint<16>(insn16));
+    return BitFieldExtract(insn16, 5, 2) << 6 | BitFieldExtract(insn16, 10, 3) << 3;
+  }
+
+  // Re-orders raw immediatate into real value
+  // where `imm[5:3]` is in bits `[5:3]` and `imm[8:6]` is in bits `[2:0]`
+  static uint32_t Uimm6ToOffsetD16(uint32_t uimm6) {
+    DCHECK(IsUint<6>(uimm6));
+    return (BitFieldExtract(uimm6, 3, 3) << 3) | (BitFieldExtract(uimm6, 0, 3) << 6);
+  }
+
+  // Re-orders raw immediatate to form real value
+  // where `imm[5:2]` is in bits `[5:2]` and `imm[7:6]` is in bits `[1:0]`
+  static uint32_t Uimm6ToOffsetW16(uint32_t uimm6) {
+    DCHECK(IsUint<6>(uimm6));
+    return (BitFieldExtract(uimm6, 2, 4) << 2) | (BitFieldExtract(uimm6, 0, 2) << 6);
+  }
+
+  // Re-orders raw immediatate to form real value
+  // where `imm[1]` is in bit `[0]` and `imm[0]` is in bit `[1]`
+  static uint32_t Uimm2ToOffset10(uint32_t uimm2) {
+    DCHECK(IsUint<2>(uimm2));
+    return (uimm2 >> 1) | (uimm2 & 0x1u) << 1;
+  }
+
+  // Re-orders raw immediatate to form real value
+  // where `imm[1]` is in bit `[0]` and `imm[0]` is `0`
+  static uint32_t Uimm2ToOffset1(uint32_t uimm2) {
+    DCHECK(IsUint<2>(uimm2));
+    return (uimm2 & 0x1u) << 1;
+  }
+
+  template <size_t kWidth>
+  static constexpr int32_t SignExtendBits(uint32_t bits) {
+    static_assert(kWidth < BitSizeOf<uint32_t>());
+    const uint32_t sign_bit = (bits >> kWidth) & 1u;
+    return static_cast<int32_t>(bits) - static_cast<int32_t>(sign_bit << kWidth);
+  }
+
+  // Extracts the immediate from a compressed instruction
+  // where `imm[5]` is in bit `[12]` and `imm[4:0]` is in bits `[6:2]`
+  // and performs sign-extension if required
+  template <typename T>
+  static T Decode16Imm6(uint32_t insn16) {
+    DCHECK(IsUint<16>(insn16));
+    static_assert(std::is_integral_v<T>, "T must be integral");
+    const T bits =
+        BitFieldInsert(BitFieldExtract(insn16, 2, 5), BitFieldExtract(insn16, 12, 1), 5, 1);
+    const T checked_bits = dchecked_integral_cast<T>(bits);
+    if (std::is_unsigned_v<T>) {
+      return checked_bits;
+    }
+    return SignExtendBits<6>(checked_bits);
+  }
+
+  // Regular instruction register utils
+
   static uint32_t GetRd(uint32_t insn32) { return (insn32 >> 7) & 0x1fu; }
   static uint32_t GetRs1(uint32_t insn32) { return (insn32 >> 15) & 0x1fu; }
   static uint32_t GetRs2(uint32_t insn32) { return (insn32 >> 20) & 0x1fu; }
   static uint32_t GetRs3(uint32_t insn32) { return insn32 >> 27; }
   static uint32_t GetRoundingMode(uint32_t insn32) { return (insn32 >> 12) & 7u; }
 
+  // Compressed instruction register utils
+
+  static uint32_t GetRs1Short16(uint32_t insn16) { return BitFieldExtract(insn16, 7, 3) + 8u; }
+  static uint32_t GetRs2Short16(uint32_t insn16) { return BitFieldExtract(insn16, 2, 3) + 8u; }
+  static uint32_t GetRs1_16(uint32_t insn16) { return BitFieldExtract(insn16, 7, 5); }
+  static uint32_t GetRs2_16(uint32_t insn16) { return BitFieldExtract(insn16, 2, 5); }
+
   void PrintBranchOffset(int32_t offset);
   void PrintLoadStoreAddress(uint32_t rs1, int32_t offset);
 
@@ -1548,10 +1635,319 @@ void DisassemblerRiscv64::Printer::Dump32(const uint8_t* insn) {
 
 void DisassemblerRiscv64::Printer::Dump16(const uint8_t* insn) {
   uint32_t insn16 = static_cast<uint32_t>(insn[0]) + (static_cast<uint32_t>(insn[1]) << 8);
+  ScopedNewLinePrinter nl(os_);
   CHECK_NE(insn16 & 3u, 3u);
-  // TODO(riscv64): Disassemble instructions from the "C" extension.
-  os_ << disassembler_->FormatInstructionPointer(insn)
-      << StringPrintf(": %04x    \t<unknown16>\n", insn16);
+  os_ << disassembler_->FormatInstructionPointer(insn) << StringPrintf(": %04x    \t", insn16);
+
+  uint32_t funct3 = BitFieldExtract(insn16, 13, 3);
+  int32_t offset = -1;
+
+  switch (insn16 & 3u) {
+    case 0b00u:  // Quadrant 0
+      switch (funct3) {
+        case 0b000u:
+          if (insn16 == 0u) {
+            os_ << "c.unimp";
+          } else {
+            uint32_t nzuimm = BitFieldExtract(insn16, 5, 8);
+            if (nzuimm != 0u) {
+              uint32_t decoded =
+                  BitFieldExtract(nzuimm, 0, 1) << 3 | BitFieldExtract(nzuimm, 1, 1) << 2 |
+                  BitFieldExtract(nzuimm, 2, 4) << 6 | BitFieldExtract(nzuimm, 6, 2) << 4;
+              os_ << "c.addi4spn " << XRegName(GetRs2Short16(insn16)) << ", sp, " << decoded;
+            } else {
+              os_ << "<unknown16>";
+            }
+          }
+          return;
+        case 0b001u:
+          offset = Decode16CMOffsetD(insn16);
+          os_ << "c.fld " << FRegName(GetRs2Short16(insn16));
+          break;
+        case 0b010u:
+          offset = Decode16CMOffsetW(insn16);
+          os_ << "c.lw " << XRegName(GetRs2Short16(insn16));
+          break;
+        case 0b011u:
+          offset = Decode16CMOffsetD(insn16);
+          os_ << "c.ld " << XRegName(GetRs2Short16(insn16));
+          break;
+        case 0b100u: {
+          uint32_t opcode2 = BitFieldExtract(insn16, 10, 3);
+          uint32_t imm = BitFieldExtract(insn16, 5, 2);
+          switch (opcode2) {
+            case 0b000:
+              offset = Uimm2ToOffset10(imm);
+              os_ << "c.lbu " << XRegName(GetRs2Short16(insn16));
+              break;
+            case 0b001:
+              offset = Uimm2ToOffset1(imm);
+              os_ << (BitFieldExtract(imm, 1, 1) == 0u ? "c.lhu " : "c.lh ");
+              os_ << XRegName(GetRs2Short16(insn16));
+              break;
+            case 0b010:
+              offset = Uimm2ToOffset10(imm);
+              os_ << "c.sb " << XRegName(GetRs2Short16(insn16));
+              break;
+            case 0b011:
+              if (BitFieldExtract(imm, 1, 1) == 0u) {
+                offset = Uimm2ToOffset1(imm);
+                os_ << "c.sh " << XRegName(GetRs2Short16(insn16));
+                break;
+              }
+              FALLTHROUGH_INTENDED;
+            default:
+              os_ << "<unknown16>";
+              return;
+          }
+          break;
+        }
+        case 0b101u:
+          offset = Decode16CMOffsetD(insn16);
+          os_ << "c.fsd " << FRegName(GetRs2Short16(insn16));
+          break;
+        case 0b110u:
+          offset = Decode16CMOffsetW(insn16);
+          os_ << "c.sw " << XRegName(GetRs2Short16(insn16));
+          break;
+        case 0b111u:
+          offset = Decode16CMOffsetD(insn16);
+          os_ << "c.sd " << XRegName(GetRs2Short16(insn16));
+          break;
+        default:
+          UNREACHABLE();
+      }
+      os_ << ", ";
+      PrintLoadStoreAddress(GetRs1Short16(insn16), offset);
+      return;
+    case 0b01u:  // Quadrant 1
+      switch (funct3) {
+        case 0b000u: {
+          uint32_t rd = GetRs1_16(insn16);
+          if (rd == 0) {
+            if (Decode16Imm6<uint32_t>(insn16) != 0u) {
+              os_ << "<hint16>";
+            } else {
+              os_ << "c.nop";
+            }
+          } else {
+            int32_t imm = Decode16Imm6<int32_t>(insn16);
+            if (imm != 0) {
+              os_ << "c.addi " << XRegName(rd) << ", " << imm;
+            } else {
+              os_ << "<hint16>";
+            }
+          }
+          break;
+        }
+        case 0b001u: {
+          uint32_t rd = GetRs1_16(insn16);
+          if (rd != 0) {
+            os_ << "c.addiw " << XRegName(rd) << ", " << Decode16Imm6<int32_t>(insn16);
+          } else {
+            os_ << "<unknown16>";
+          }
+          break;
+        }
+        case 0b010u: {
+          uint32_t rd = GetRs1_16(insn16);
+          if (rd != 0) {
+            os_ << "c.li " << XRegName(rd) << ", " << Decode16Imm6<int32_t>(insn16);
+          } else {
+            os_ << "<hint16>";
+          }
+          break;
+        }
+        case 0b011u: {
+          uint32_t rd = GetRs1_16(insn16);
+          uint32_t imm6_bits = Decode16Imm6<uint32_t>(insn16);
+          if (imm6_bits != 0u) {
+            if (rd == 2) {
+              int32_t nzimm =
+                  BitFieldExtract(insn16, 6, 1) << 4 | BitFieldExtract(insn16, 2, 1) << 5 |
+                  BitFieldExtract(insn16, 5, 1) << 6 | BitFieldExtract(insn16, 3, 2) << 7 |
+                  BitFieldExtract(insn16, 12, 1) << 9;
+              os_ << "c.addi16sp sp, " << SignExtendBits<10>(nzimm);
+            } else if (rd != 0) {
+              // sign-extend bits and mask with 0xfffff as llvm-objdump does
+              uint32_t mask = MaskLeastSignificant<uint32_t>(20);
+              os_ << "c.lui " << XRegName(rd) << ", " << (SignExtendBits<6>(imm6_bits) & mask);
+            } else {
+              os_ << "<hint16>";
+            }
+          } else {
+            os_ << "<unknown16>";
+          }
+          break;
+        }
+        case 0b100u: {
+          uint32_t funct2 = BitFieldExtract(insn16, 10, 2);
+          switch (funct2) {
+            case 0b00: {
+              int32_t nzuimm = Decode16Imm6<uint32_t>(insn16);
+              if (nzuimm != 0) {
+                os_ << "c.srli " << XRegName(GetRs1Short16(insn16)) << ", " << nzuimm;
+              } else {
+                os_ << "<hint16>";
+              }
+              break;
+            }
+            case 0b01: {
+              int32_t nzuimm = Decode16Imm6<uint32_t>(insn16);
+              if (nzuimm != 0) {
+                os_ << "c.srai " << XRegName(GetRs1Short16(insn16)) << ", " << nzuimm;
+              } else {
+                os_ << "<hint16>";
+              }
+              break;
+            }
+            case 0b10:
+              os_ << "c.andi " << XRegName(GetRs1Short16(insn16)) << ", "
+                  << Decode16Imm6<int32_t>(insn16);
+              break;
+            case 0b11: {
+              constexpr static const char* mnemonics[] = {
+                  "c.sub", "c.xor", "c.or", "c.and", "c.subw", "c.addw", "c.mul", nullptr
+              };
+              uint32_t opc = BitFieldInsert(
+                  BitFieldExtract(insn16, 5, 2), BitFieldExtract(insn16, 12, 1), 2, 1);
+              DCHECK(IsUint<3>(opc));
+              const char* mnem = mnemonics[opc];
+              if (mnem != nullptr) {
+                os_ << mnem << " " << XRegName(GetRs1Short16(insn16)) << ", "
+                    << XRegName(GetRs2Short16(insn16));
+              } else {
+                constexpr static const char* zbc_mnemonics[] = {
+                    "c.zext.b", "c.sext.b", "c.zext.h", "c.sext.h",
+                    "c.zext.w", "c.not",    nullptr,    nullptr,
+                };
+                mnem = zbc_mnemonics[BitFieldExtract(insn16, 2, 3)];
+                if (mnem != nullptr) {
+                  os_ << mnem << " " << XRegName(GetRs1Short16(insn16));
+                } else {
+                  os_ << "<unknown16>";
+                }
+              }
+              break;
+            }
+            default:
+              UNREACHABLE();
+          }
+          break;
+        }
+        case 0b101u: {
+          int32_t disp = BitFieldExtract(insn16, 3, 3) << 1  | BitFieldExtract(insn16, 11, 1) << 4 |
+                         BitFieldExtract(insn16, 2, 1) << 5  | BitFieldExtract(insn16, 7,  1) << 6 |
+                         BitFieldExtract(insn16, 6, 1) << 7  | BitFieldExtract(insn16, 9,  2) << 8 |
+                         BitFieldExtract(insn16, 8, 1) << 10 | BitFieldExtract(insn16, 12, 1) << 11;
+          os_ << "c.j ";
+          PrintBranchOffset(SignExtendBits<12>(disp));
+          break;
+        }
+        case 0b110u:
+        case 0b111u: {
+          int32_t disp = BitFieldExtract(insn16, 3, 2) << 1 | BitFieldExtract(insn16, 10, 2) << 3 |
+                         BitFieldExtract(insn16, 2, 1) << 5 | BitFieldExtract(insn16, 5,  2) << 6 |
+                         BitFieldExtract(insn16, 12, 1) << 8;
+
+          os_ << (funct3 == 0b110u ? "c.beqz " : "c.bnez ");
+          os_ << XRegName(GetRs1Short16(insn16)) << ", ";
+          PrintBranchOffset(SignExtendBits<9>(disp));
+          break;
+        }
+        default:
+          UNREACHABLE();
+      }
+      break;
+    case 0b10u:  // Quadrant 2
+      switch (funct3) {
+        case 0b000u: {
+          uint32_t nzuimm = Decode16Imm6<uint32_t>(insn16);
+          uint32_t rd = GetRs1_16(insn16);
+          if (rd == 0 || nzuimm == 0) {
+            os_ << "<hint16>";
+          } else {
+            os_ << "c.slli " << XRegName(rd) << ", " << nzuimm;
+          }
+          return;
+        }
+        case 0b001u: {
+          offset = Uimm6ToOffsetD16(Decode16Imm6<uint32_t>(insn16));
+          os_ << "c.fldsp " << FRegName(GetRs1_16(insn16));
+          break;
+        }
+        case 0b010u: {
+          uint32_t rd = GetRs1_16(insn16);
+          if (rd != 0) {
+            offset = Uimm6ToOffsetW16(Decode16Imm6<uint32_t>(insn16));
+            os_ << "c.lwsp " << XRegName(GetRs1_16(insn16));
+          } else {
+            os_ << "<unknown16>";
+            return;
+          }
+          break;
+        }
+        case 0b011u: {
+          uint32_t rd = GetRs1_16(insn16);
+          if (rd != 0) {
+            offset = Uimm6ToOffsetD16(Decode16Imm6<uint32_t>(insn16));
+            os_ << "c.ldsp " << XRegName(GetRs1_16(insn16));
+          } else {
+            os_ << "<unknown16>";
+            return;
+          }
+          break;
+        }
+        case 0b100u: {
+          uint32_t rd_rs1 = GetRs1_16(insn16);
+          uint32_t rs2 = GetRs2_16(insn16);
+          uint32_t b12 = BitFieldExtract(insn16, 12, 1);
+          if (b12 == 0) {
+            if (rd_rs1 != 0 && rs2 != 0) {
+              os_ << "c.mv " << XRegName(rd_rs1) << ", " << XRegName(rs2);
+            } else if (rd_rs1 != 0) {
+              os_ << "c.jr " << XRegName(rd_rs1);
+            } else if (rs2 != 0) {
+              os_ << "<hint16>";
+            } else {
+              os_ << "<unknown16>";
+            }
+          } else {
+            if (rd_rs1 != 0 && rs2 != 0) {
+              os_ << "c.add " << XRegName(rd_rs1) << ", " << XRegName(rs2);
+            } else if (rd_rs1 != 0) {
+              os_ << "c.jalr " << XRegName(rd_rs1);
+            } else if (rs2 != 0) {
+              os_ << "<hint16>";
+            } else {
+              os_ << "c.ebreak";
+            }
+          }
+          return;
+        }
+        case 0b101u:
+          offset = BitFieldExtract(insn16, 7, 3) << 6 | BitFieldExtract(insn16, 10, 3) << 3;
+          os_ << "c.fsdsp " << FRegName(GetRs2_16(insn16));
+          break;
+        case 0b110u:
+          offset = BitFieldExtract(insn16, 7, 2) << 6 | BitFieldExtract(insn16, 9, 4) << 2;
+          os_ << "c.swsp " << XRegName(GetRs2_16(insn16));
+          break;
+        case 0b111u:
+          offset = BitFieldExtract(insn16, 7, 3) << 6 | BitFieldExtract(insn16, 10, 3) << 3;
+          os_ << "c.sdsp " << XRegName(GetRs2_16(insn16));
+          break;
+        default:
+          UNREACHABLE();
+      }
+
+      os_ << ", ";
+      PrintLoadStoreAddress(/* sp */ 2, offset);
+
+      break;
+    default:
+      UNREACHABLE();
+  }
 }
 
 void DisassemblerRiscv64::Printer::Dump2Byte(const uint8_t* data) {
diff --git a/imgdiag/imgdiag_test.cc b/imgdiag/imgdiag_test.cc
index 3ea7093278..9dd7953a24 100644
--- a/imgdiag/imgdiag_test.cc
+++ b/imgdiag/imgdiag_test.cc
@@ -115,7 +115,6 @@ TEST_F(ImgDiagTest, ImageDiffPidSelf) {
 // because it's root read-only.
 TEST_F(ImgDiagTest, DISABLED_ImageDiffPidSelf) {
 #endif
-  TEST_DISABLED_FOR_RISCV64();
   // Invoke 'img_diag' against the current process.
   // This should succeed because we have a runtime and so it should
   // be able to map in the boot.art and do a diff for it.
diff --git a/libartbase/base/common_art_test.h b/libartbase/base/common_art_test.h
index 64e251e16c..c490e085cd 100644
--- a/libartbase/base/common_art_test.h
+++ b/libartbase/base/common_art_test.h
@@ -339,24 +339,4 @@ std::vector<pid_t> GetPidByName(const std::string& process_name);
     GTEST_SKIP() << "WARNING: TEST DISABLED FOR MEMORY TOOL WITH HEAP POISONING"; \
   }
 
-#define TEST_DISABLED_FOR_RISCV64()                       \
-  if (kRuntimeISA == InstructionSet::kRiscv64) {          \
-    GTEST_SKIP() << "WARNING: TEST DISABLED FOR RISCV64"; \
-  }
-
-// Don't print messages on setup to avoid getting multiple "test disabled" messages for one test.
-// Setup phase may need to be disabled as some test rely on having boot image / compiler / other
-// things that are not implemented for RISC-V.
-#define TEST_SETUP_DISABLED_FOR_RISCV64()        \
-  if (kRuntimeISA == InstructionSet::kRiscv64) { \
-    GTEST_SKIP();                                \
-  }
-
-// Don't print messages on teardown to avoid getting multiple "test disabled" messages for one test.
-// Teardown phase may need to be disabled to match the disabled setup phase for some tests.
-#define TEST_TEARDOWN_DISABLED_FOR_RISCV64()     \
-  if (kRuntimeISA == InstructionSet::kRiscv64) { \
-    GTEST_SKIP();                                \
-  }
-
 #endif  // ART_LIBARTBASE_BASE_COMMON_ART_TEST_H_
diff --git a/libartbase/base/safe_map.h b/libartbase/base/safe_map.h
index fa13fe0f68..c489bfd866 100644
--- a/libartbase/base/safe_map.h
+++ b/libartbase/base/safe_map.h
@@ -149,7 +149,7 @@ class SafeMap {
   }
 
   template <typename CreateFn>
-  V& GetOrCreate(const K& k, CreateFn create) {
+  V& GetOrCreate(const K& k, CreateFn&& create) {
     static_assert(std::is_same_v<V, std::invoke_result_t<CreateFn>>,
                   "Argument `create` should return a value of type V.");
     auto lb = lower_bound(k);
diff --git a/libartbase/base/sdk_version.h b/libartbase/base/sdk_version.h
index d39aa95b5d..b955ab0322 100644
--- a/libartbase/base/sdk_version.h
+++ b/libartbase/base/sdk_version.h
@@ -38,6 +38,7 @@ enum class SdkVersion : uint32_t {
   kS     = 31u,
   kS_V2  = 32u,
   kT     = 33u,
+  kU     = 34u,
   kMax   = std::numeric_limits<uint32_t>::max(),
 };
 
diff --git a/libartpalette/apex/palette_test.cc b/libartpalette/apex/palette_test.cc
index c2b7304d8d..63072c491b 100644
--- a/libartpalette/apex/palette_test.cc
+++ b/libartpalette/apex/palette_test.cc
@@ -54,16 +54,10 @@ bool PaletteSetTaskProfilesIsSupported(palette_status_t res) {
 
 }  // namespace
 
-// TODO(riscv64): remove `namespace art` when `TEST_DISABLED_FOR_RISCV64` is no longer needed.
-namespace art {
-
 class PaletteClientTest : public testing::Test {};
 
 TEST_F(PaletteClientTest, SchedPriority) {
-  // On RISC-V tests run in Android-like chroot on a Linux VM => some syscalls work differently.
-  TEST_DISABLED_FOR_RISCV64();
-
-  int32_t tid = ::GetTid();
+  int32_t tid = GetTid();
   int32_t saved_priority;
   EXPECT_EQ(PALETTE_STATUS_OK, PaletteSchedGetPriority(tid, &saved_priority));
 
@@ -87,9 +81,6 @@ TEST_F(PaletteClientTest, Ashmem) {
 #ifndef ART_TARGET_ANDROID
   GTEST_SKIP() << "ashmem is only supported on Android";
 #else
-  // On RISC-V tests run in Android-like chroot on a Linux VM => some syscalls work differently.
-  TEST_DISABLED_FOR_RISCV64();
-
   int fd;
   EXPECT_EQ(PALETTE_STATUS_OK, PaletteAshmemCreateRegion("ashmem-test", 4096, &fd));
   EXPECT_EQ(PALETTE_STATUS_OK, PaletteAshmemSetProtRegion(fd, PROT_READ | PROT_EXEC));
@@ -175,5 +166,3 @@ TEST_F(PaletteClientTest, SetTaskProfilesCpp) {
   }
 #endif
 }
-
-}  // namespace art
diff --git a/libartservice/service/api/system-server-current.txt b/libartservice/service/api/system-server-current.txt
index f316065ed8..05163ebc70 100644
--- a/libartservice/service/api/system-server-current.txt
+++ b/libartservice/service/api/system-server-current.txt
@@ -19,6 +19,7 @@ package com.android.server.art {
     method @NonNull public com.android.server.art.model.DexoptStatus getDexoptStatus(@NonNull com.android.server.pm.PackageManagerLocal.FilteredSnapshot, @NonNull String);
     method @NonNull public com.android.server.art.model.DexoptStatus getDexoptStatus(@NonNull com.android.server.pm.PackageManagerLocal.FilteredSnapshot, @NonNull String, int);
     method public int handleShellCommand(@NonNull android.os.Binder, @NonNull android.os.ParcelFileDescriptor, @NonNull android.os.ParcelFileDescriptor, @NonNull android.os.ParcelFileDescriptor, @NonNull String[]);
+    method public void onApexStaged(@NonNull String[]);
     method public void onBoot(@NonNull String, @Nullable java.util.concurrent.Executor, @Nullable java.util.function.Consumer<com.android.server.art.model.OperationProgress>);
     method public void printShellCommandHelp(@NonNull java.io.PrintWriter);
     method public void removeDexoptDoneCallback(@NonNull com.android.server.art.ArtManagerLocal.DexoptDoneCallback);
diff --git a/libartservice/service/java/com/android/server/art/ArtManagerLocal.java b/libartservice/service/java/com/android/server/art/ArtManagerLocal.java
index 69cba7d860..db82bfd982 100644
--- a/libartservice/service/java/com/android/server/art/ArtManagerLocal.java
+++ b/libartservice/service/java/com/android/server/art/ArtManagerLocal.java
@@ -853,6 +853,25 @@ public final class ArtManagerLocal {
     }
 
     /**
+     * Notifies ART Service that there are apexes staged for installation on next reboot (see
+     * <a href="https://source.android.com/docs/core/ota/apex#apex-manager">the update sequence of
+     * an APEX</a>). ART Service may use this to schedule a pre-reboot dexopt job. This might change
+     * in the future.
+     *
+     * This immediately returns after scheduling the job and doesn't wait for the job to run.
+     *
+     * @param stagedApexModuleNames The <b>module names</b> of the staged apexes, corresponding to
+     *         the directory beneath /apex, e.g., {@code com.android.art} (not the <b>package
+     *         names</b>, e.g., {@code com.google.android.art}).
+     */
+    @SuppressLint("UnflaggedApi") // Flag support for mainline is not available.
+    @RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
+    public void onApexStaged(@NonNull String[] stagedApexModuleNames) {
+        // TODO(b/311377497): Check system requirements.
+        mInjector.getPreRebootDexoptJob().schedule();
+    }
+
+    /**
      * Dumps the dexopt state of all packages in text format for debugging purposes.
      *
      * There are no stability guarantees for the output format.
@@ -1024,6 +1043,17 @@ public final class ArtManagerLocal {
         return mInjector.getBackgroundDexoptJob();
     }
 
+    /**
+     * Should be used by {@link BackgroundDexoptJobService} ONLY.
+     *
+     * @hide
+     */
+    @RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
+    @NonNull
+    PreRebootDexoptJob getPreRebootDexoptJob() {
+        return mInjector.getPreRebootDexoptJob();
+    }
+
     @RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
     @Nullable
     private DexoptResult maybeDowngradePackages(
@@ -1354,6 +1384,7 @@ public final class ArtManagerLocal {
         @Nullable private final PackageManagerLocal mPackageManagerLocal;
         @Nullable private final Config mConfig;
         @Nullable private BackgroundDexoptJob mBgDexoptJob = null;
+        @Nullable private PreRebootDexoptJob mPrDexoptJob = null;
 
         /** For compatibility with S and T. New code should not use this. */
         @Deprecated
@@ -1443,6 +1474,15 @@ public final class ArtManagerLocal {
 
         @RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
         @NonNull
+        public synchronized PreRebootDexoptJob getPreRebootDexoptJob() {
+            if (mPrDexoptJob == null) {
+                mPrDexoptJob = new PreRebootDexoptJob(mContext);
+            }
+            return mPrDexoptJob;
+        }
+
+        @RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
+        @NonNull
         public UserManager getUserManager() {
             return Objects.requireNonNull(mContext.getSystemService(UserManager.class));
         }
diff --git a/libartservice/service/java/com/android/server/art/BackgroundDexoptJob.java b/libartservice/service/java/com/android/server/art/BackgroundDexoptJob.java
index 9aeb2828fb..289c7cd235 100644
--- a/libartservice/service/java/com/android/server/art/BackgroundDexoptJob.java
+++ b/libartservice/service/java/com/android/server/art/BackgroundDexoptJob.java
@@ -41,6 +41,7 @@ import com.android.internal.annotations.GuardedBy;
 import com.android.internal.annotations.VisibleForTesting;
 import com.android.server.LocalManagerRegistry;
 import com.android.server.art.model.ArtFlags;
+import com.android.server.art.model.ArtServiceJobInterface;
 import com.android.server.art.model.Config;
 import com.android.server.art.model.DexoptResult;
 import com.android.server.art.model.OperationProgress;
@@ -59,7 +60,7 @@ import java.util.function.Consumer;
 
 /** @hide */
 @RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
-public class BackgroundDexoptJob {
+public class BackgroundDexoptJob implements ArtServiceJobInterface {
     private static final String TAG = ArtManagerLocal.TAG;
 
     /**
@@ -68,7 +69,7 @@ public class BackgroundDexoptJob {
      */
     private static final String JOB_PKG_NAME = Utils.PLATFORM_PACKAGE_NAME;
     /** An arbitrary number. Must be unique among all jobs owned by the system uid. */
-    private static final int JOB_ID = 27873780;
+    public static final int JOB_ID = 27873780;
 
     @VisibleForTesting public static final long JOB_INTERVAL_MS = TimeUnit.DAYS.toMillis(1);
 
@@ -89,6 +90,7 @@ public class BackgroundDexoptJob {
     }
 
     /** Handles {@link BackgroundDexoptJobService#onStartJob(JobParameters)}. */
+    @Override
     public boolean onStartJob(
             @NonNull BackgroundDexoptJobService jobService, @NonNull JobParameters params) {
         start().thenAcceptAsync(result -> {
@@ -113,6 +115,7 @@ public class BackgroundDexoptJob {
     }
 
     /** Handles {@link BackgroundDexoptJobService#onStopJob(JobParameters)}. */
+    @Override
     public boolean onStopJob(@NonNull JobParameters params) {
         synchronized (this) {
             mLastStopReason = Optional.of(params.getStopReason());
@@ -124,7 +127,7 @@ public class BackgroundDexoptJob {
 
     /** Handles {@link ArtManagerLocal#scheduleBackgroundDexoptJob()}. */
     public @ScheduleStatus int schedule() {
-        if (this != BackgroundDexoptJobService.getJob()) {
+        if (this != BackgroundDexoptJobService.getJob(JOB_ID)) {
             throw new IllegalStateException("This job cannot be scheduled");
         }
 
@@ -164,7 +167,7 @@ public class BackgroundDexoptJob {
 
     /** Handles {@link ArtManagerLocal#unscheduleBackgroundDexoptJob()}. */
     public void unschedule() {
-        if (this != BackgroundDexoptJobService.getJob()) {
+        if (this != BackgroundDexoptJobService.getJob(JOB_ID)) {
             throw new IllegalStateException("This job cannot be unscheduled");
         }
 
diff --git a/libartservice/service/java/com/android/server/art/BackgroundDexoptJobService.java b/libartservice/service/java/com/android/server/art/BackgroundDexoptJobService.java
index 41425ee5cc..8bcc4b98be 100644
--- a/libartservice/service/java/com/android/server/art/BackgroundDexoptJobService.java
+++ b/libartservice/service/java/com/android/server/art/BackgroundDexoptJobService.java
@@ -24,6 +24,7 @@ import android.os.Build;
 import androidx.annotation.RequiresApi;
 
 import com.android.server.LocalManagerRegistry;
+import com.android.server.art.model.ArtServiceJobInterface;
 
 /**
  * Entry point for the callback from the job scheduler. This class is instantiated by the system
@@ -35,16 +36,21 @@ import com.android.server.LocalManagerRegistry;
 public class BackgroundDexoptJobService extends JobService {
     @Override
     public boolean onStartJob(@NonNull JobParameters params) {
-        return getJob().onStartJob(this, params);
+        return getJob(params.getJobId()).onStartJob(this, params);
     }
 
     @Override
     public boolean onStopJob(@NonNull JobParameters params) {
-        return getJob().onStopJob(params);
+        return getJob(params.getJobId()).onStopJob(params);
     }
 
     @NonNull
-    static BackgroundDexoptJob getJob() {
-        return LocalManagerRegistry.getManager(ArtManagerLocal.class).getBackgroundDexoptJob();
+    static ArtServiceJobInterface getJob(int jobId) {
+        if (jobId == BackgroundDexoptJob.JOB_ID) {
+            return LocalManagerRegistry.getManager(ArtManagerLocal.class).getBackgroundDexoptJob();
+        } else if (jobId == PreRebootDexoptJob.JOB_ID) {
+            return LocalManagerRegistry.getManager(ArtManagerLocal.class).getPreRebootDexoptJob();
+        }
+        throw new IllegalArgumentException("Unknown job ID " + jobId);
     }
 }
diff --git a/libartservice/service/java/com/android/server/art/PreRebootDexoptJob.java b/libartservice/service/java/com/android/server/art/PreRebootDexoptJob.java
new file mode 100644
index 0000000000..74bf913e62
--- /dev/null
+++ b/libartservice/service/java/com/android/server/art/PreRebootDexoptJob.java
@@ -0,0 +1,87 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package com.android.server.art;
+
+import static com.android.server.art.model.ArtFlags.ScheduleStatus;
+
+import android.annotation.NonNull;
+import android.app.job.JobParameters;
+import android.content.Context;
+import android.os.Build;
+
+import androidx.annotation.RequiresApi;
+
+import com.android.internal.annotations.VisibleForTesting;
+import com.android.server.art.model.ArtFlags;
+import com.android.server.art.model.ArtServiceJobInterface;
+
+/** @hide */
+@RequiresApi(Build.VERSION_CODES.UPSIDE_DOWN_CAKE)
+public class PreRebootDexoptJob implements ArtServiceJobInterface {
+    private static final String TAG = ArtManagerLocal.TAG;
+
+    /**
+     * "android" is the package name for a <service> declared in
+     * frameworks/base/core/res/AndroidManifest.xml
+     */
+    private static final String JOB_PKG_NAME = Utils.PLATFORM_PACKAGE_NAME;
+    /** An arbitrary number. Must be unique among all jobs owned by the system uid. */
+    public static final int JOB_ID = 27873781;
+
+    @NonNull private final Injector mInjector;
+
+    public PreRebootDexoptJob(@NonNull Context context) {
+        this(new Injector(context));
+    }
+
+    @VisibleForTesting
+    public PreRebootDexoptJob(@NonNull Injector injector) {
+        mInjector = injector;
+    }
+
+    @Override
+    public boolean onStartJob(
+            @NonNull BackgroundDexoptJobService jobService, @NonNull JobParameters params) {
+        // "true" means the job will continue running until `jobFinished` is called.
+        return false;
+    }
+
+    @Override
+    public boolean onStopJob(@NonNull JobParameters params) {
+        // "true" means to execute again in the same interval with the default retry policy.
+        return true;
+    }
+
+    public @ScheduleStatus int schedule() {
+        // TODO(b/311377497): Schedule the job.
+        return ArtFlags.SCHEDULE_SUCCESS;
+    }
+
+    /**
+     * Injector pattern for testing purpose.
+     *
+     * @hide
+     */
+    @VisibleForTesting
+    public static class Injector {
+        @NonNull private final Context mContext;
+
+        Injector(@NonNull Context context) {
+            mContext = context;
+        }
+    }
+}
diff --git a/libartservice/service/java/com/android/server/art/model/ArtServiceJobInterface.java b/libartservice/service/java/com/android/server/art/model/ArtServiceJobInterface.java
new file mode 100644
index 0000000000..569996c566
--- /dev/null
+++ b/libartservice/service/java/com/android/server/art/model/ArtServiceJobInterface.java
@@ -0,0 +1,30 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package com.android.server.art.model;
+
+import android.annotation.NonNull;
+import android.app.job.JobParameters;
+
+import com.android.server.art.BackgroundDexoptJobService;
+
+/** @hide */
+public interface ArtServiceJobInterface {
+    boolean onStartJob(
+            @NonNull BackgroundDexoptJobService jobService, @NonNull JobParameters params);
+
+    boolean onStopJob(@NonNull JobParameters params);
+}
diff --git a/libartservice/service/javatests/com/android/server/art/BackgroundDexoptJobTest.java b/libartservice/service/javatests/com/android/server/art/BackgroundDexoptJobTest.java
index c61461d0c9..85ae079c77 100644
--- a/libartservice/service/javatests/com/android/server/art/BackgroundDexoptJobTest.java
+++ b/libartservice/service/javatests/com/android/server/art/BackgroundDexoptJobTest.java
@@ -16,6 +16,7 @@
 
 package com.android.server.art;
 
+import static com.android.server.art.BackgroundDexoptJob.JOB_ID;
 import static com.android.server.art.model.Config.Callback;
 import static com.android.server.art.model.DexoptResult.DexoptResultStatus;
 import static com.android.server.art.model.DexoptResult.PackageDexoptResult;
@@ -103,7 +104,7 @@ public class BackgroundDexoptJobTest {
         lenient().when(mInjector.getJobScheduler()).thenReturn(mJobScheduler);
 
         mBackgroundDexoptJob = new BackgroundDexoptJob(mInjector);
-        lenient().when(BackgroundDexoptJobService.getJob()).thenReturn(mBackgroundDexoptJob);
+        lenient().when(BackgroundDexoptJobService.getJob(JOB_ID)).thenReturn(mBackgroundDexoptJob);
 
         lenient()
                 .doAnswer(invocation -> {
diff --git a/libdexfile/dex/code_item_accessors-inl.h b/libdexfile/dex/code_item_accessors-inl.h
index b74046071f..4586a0ff7a 100644
--- a/libdexfile/dex/code_item_accessors-inl.h
+++ b/libdexfile/dex/code_item_accessors-inl.h
@@ -202,20 +202,22 @@ template<typename NewLocalVisitor>
 inline bool CodeItemDebugInfoAccessor::DecodeDebugLocalInfo(
     bool is_static,
     uint32_t method_idx,
-    const NewLocalVisitor& new_local) const {
+    NewLocalVisitor&& new_local) const {
   return dex_file_->DecodeDebugLocalInfo(RegistersSize(),
                                          InsSize(),
                                          InsnsSizeInCodeUnits(),
                                          DebugInfoOffset(),
                                          is_static,
                                          method_idx,
-                                         new_local);
+                                         std::forward<NewLocalVisitor>(new_local));
 }
 
 template <typename Visitor>
-inline uint32_t CodeItemDebugInfoAccessor::VisitParameterNames(const Visitor& visitor) const {
+inline uint32_t CodeItemDebugInfoAccessor::VisitParameterNames(Visitor&& visitor) const {
   const uint8_t* stream = dex_file_->GetDebugInfoStream(DebugInfoOffset());
-  return (stream != nullptr) ? DexFile::DecodeDebugInfoParameterNames(&stream, visitor) : 0u;
+  return (stream != nullptr) ?
+             DexFile::DecodeDebugInfoParameterNames(&stream, std::forward<Visitor>(visitor)) :
+             0u;
 }
 
 inline bool CodeItemDebugInfoAccessor::GetLineNumForPc(const uint32_t address,
@@ -233,13 +235,13 @@ inline bool CodeItemDebugInfoAccessor::GetLineNumForPc(const uint32_t address,
 }
 
 template <typename Visitor>
-inline bool CodeItemDebugInfoAccessor::DecodeDebugPositionInfo(const Visitor& visitor) const {
+inline bool CodeItemDebugInfoAccessor::DecodeDebugPositionInfo(Visitor&& visitor) const {
   return dex_file_->DecodeDebugPositionInfo(
       dex_file_->GetDebugInfoStream(DebugInfoOffset()),
       [this](uint32_t idx) {
         return dex_file_->StringDataByIdx(dex::StringIndex(idx));
       },
-      visitor);
+      std::forward<Visitor>(visitor));
 }
 
 }  // namespace art
diff --git a/libdexfile/dex/code_item_accessors.h b/libdexfile/dex/code_item_accessors.h
index 24296c8f1d..5952b2d7ea 100644
--- a/libdexfile/dex/code_item_accessors.h
+++ b/libdexfile/dex/code_item_accessors.h
@@ -164,15 +164,15 @@ class CodeItemDebugInfoAccessor : public CodeItemDataAccessor {
   template<typename NewLocalVisitor>
   bool DecodeDebugLocalInfo(bool is_static,
                             uint32_t method_idx,
-                            const NewLocalVisitor& new_local) const;
+                            NewLocalVisitor&& new_local) const;
 
   // Visit each parameter in the debug information. Returns the line number.
   // The argument of the Visitor is dex::StringIndex.
   template <typename Visitor>
-  uint32_t VisitParameterNames(const Visitor& visitor) const;
+  uint32_t VisitParameterNames(Visitor&& visitor) const;
 
   template <typename Visitor>
-  bool DecodeDebugPositionInfo(const Visitor& visitor) const;
+  bool DecodeDebugPositionInfo(Visitor&& visitor) const;
 
   bool GetLineNumForPc(const uint32_t pc, uint32_t* line_num) const;
 
diff --git a/libdexfile/dex/dex_file-inl.h b/libdexfile/dex/dex_file-inl.h
index 9291f6ee9e..b01b004e5b 100644
--- a/libdexfile/dex/dex_file-inl.h
+++ b/libdexfile/dex/dex_file-inl.h
@@ -427,8 +427,8 @@ bool DexFile::DecodeDebugLocalInfo(uint32_t registers_size,
 
 template<typename DexDebugNewPosition, typename IndexToStringData>
 bool DexFile::DecodeDebugPositionInfo(const uint8_t* stream,
-                                      const IndexToStringData& index_to_string_data,
-                                      const DexDebugNewPosition& position_functor) {
+                                      IndexToStringData&& index_to_string_data,
+                                      DexDebugNewPosition&& position_functor) {
   if (stream == nullptr) {
     return false;
   }
@@ -514,7 +514,7 @@ inline IterationRange<ClassIterator> DexFile::GetClasses() const {
 // Returns the line number
 template <typename Visitor>
 inline uint32_t DexFile::DecodeDebugInfoParameterNames(const uint8_t** debug_info,
-                                                       const Visitor& visitor) {
+                                                       Visitor&& visitor) {
   uint32_t line = DecodeUnsignedLeb128(debug_info);
   const uint32_t parameters_size = DecodeUnsignedLeb128(debug_info);
   for (uint32_t i = 0; i < parameters_size; ++i) {
diff --git a/libdexfile/dex/dex_file.h b/libdexfile/dex/dex_file.h
index a7d288120d..6bc6e4f4ca 100644
--- a/libdexfile/dex/dex_file.h
+++ b/libdexfile/dex/dex_file.h
@@ -787,8 +787,8 @@ class DexFile {
   // Returns false if there is no debugging information or if it cannot be decoded.
   template<typename DexDebugNewPosition, typename IndexToStringData>
   static bool DecodeDebugPositionInfo(const uint8_t* stream,
-                                      const IndexToStringData& index_to_string_data,
-                                      const DexDebugNewPosition& position_functor);
+                                      IndexToStringData&& index_to_string_data,
+                                      DexDebugNewPosition&& position_functor);
 
   const char* GetSourceFile(const dex::ClassDef& class_def) const {
     if (!class_def.source_file_idx_.IsValid()) {
@@ -893,7 +893,7 @@ class DexFile {
 
   template <typename Visitor>
   static uint32_t DecodeDebugInfoParameterNames(const uint8_t** debug_info,
-                                                const Visitor& visitor);
+                                                Visitor&& visitor);
 
   static inline bool StringEquals(const DexFile* df1, dex::StringIndex sidx1,
                                   const DexFile* df2, dex::StringIndex sidx2);
@@ -935,7 +935,17 @@ class DexFile {
   // This is different to the "data section" in the standard dex header.
   ArrayRef<const uint8_t> const data_;
 
-  // Typically the dex file name when available, alternatively some identifying string.
+  // The full absolute path to the dex file, if it was loaded from disk.
+  //
+  // Can also be a path to a multidex container (typically apk), followed by
+  // DexFileLoader.kMultiDexSeparator (i.e. '!') and the file inside the
+  // container.
+  //
+  // On host this may not be an absolute path.
+  //
+  // On device libnativeloader uses this to determine the location of the java
+  // package or shared library, which decides where to load native libraries
+  // from.
   //
   // The ClassLinker will use this to match DexFiles the boot class
   // path to DexCache::GetLocation when loading from an image.
diff --git a/libdexfile/dex/dex_file_loader.h b/libdexfile/dex/dex_file_loader.h
index 6530303542..ec7f8ba977 100644
--- a/libdexfile/dex/dex_file_loader.h
+++ b/libdexfile/dex/dex_file_loader.h
@@ -327,6 +327,17 @@ class DexFileLoader {
   const File* file_ = &kInvalidFile;
   std::optional<File> owned_file_;  // May be used as backing storage for 'file_'.
   std::shared_ptr<DexFileContainer> root_container_;
+
+  // The full absolute path to the dex file, if it was loaded from disk.
+  //
+  // Can also be a path to a multidex container (typically apk), followed by
+  // kMultiDexSeparator and the file inside the container.
+  //
+  // On host this may not be an absolute path.
+  //
+  // On device libnativeloader uses this to determine the location of the java
+  // package or shared library, which decides where to load native libraries
+  // from.
   const std::string location_;
 };
 
diff --git a/libdexfile/dex/dex_file_verifier.cc b/libdexfile/dex/dex_file_verifier.cc
index 8c25b5a4cf..ba3472015d 100644
--- a/libdexfile/dex/dex_file_verifier.cc
+++ b/libdexfile/dex/dex_file_verifier.cc
@@ -20,11 +20,11 @@
 #include <bitset>
 #include <limits>
 #include <memory>
+#include <stack>
 
 #include "android-base/logging.h"
 #include "android-base/macros.h"
 #include "android-base/stringprintf.h"
-
 #include "base/hash_map.h"
 #include "base/leb128.h"
 #include "base/safe_map.h"
@@ -291,9 +291,21 @@ class DexFileVerifier {
   bool CheckStaticFieldTypes(const dex::ClassDef& class_def);
 
   bool CheckPadding(uint32_t aligned_offset, DexFile::MapItemType type);
+
+  // The encoded values, arrays and annotations are allowed to be very deeply nested,
+  // so use heap todo-list instead of stack recursion (the work is done in LIFO order).
+  struct ToDoItem {
+    uint32_t array_size = 0;          // CheckArrayElement.
+    uint32_t annotation_size = 0;     // CheckAnnotationElement.
+    uint32_t last_idx = kDexNoIndex;  // CheckAnnotationElement.
+  };
+  using ToDoList = std::stack<ToDoItem>;
   bool CheckEncodedValue();
   bool CheckEncodedArray();
+  bool CheckArrayElement();
   bool CheckEncodedAnnotation();
+  bool CheckAnnotationElement(/*inout*/ uint32_t* last_idx);
+  bool FlushToDoList();
 
   bool CheckIntraTypeIdItem();
   bool CheckIntraProtoIdItem();
@@ -454,6 +466,9 @@ class DexFileVerifier {
 
   // Class definition indexes, valid only if corresponding `defined_classes_[.]` is true.
   std::vector<uint16_t> defined_class_indexes_;
+
+  // Used by CheckEncodedValue to avoid recursion. Field so we can reuse allocated memory.
+  ToDoList todo_;
 };
 
 template <typename ExtraCheckFn>
@@ -1184,12 +1199,15 @@ bool DexFileVerifier::CheckEncodedValue() {
 
 bool DexFileVerifier::CheckEncodedArray() {
   DECODE_UNSIGNED_CHECKED_FROM(ptr_, size);
+  todo_.emplace(ToDoItem{.array_size = size});
+  return true;
+}
 
-  for (; size != 0u; --size) {
-    if (!CheckEncodedValue()) {
-      failure_reason_ = StringPrintf("Bad encoded_array value: %s", failure_reason_.c_str());
-      return false;
-    }
+// Always called directly from FlushToDoList, which avoids recursion.
+bool DexFileVerifier::CheckArrayElement() {
+  if (!CheckEncodedValue()) {
+    failure_reason_ = StringPrintf("Bad encoded_array value: %s", failure_reason_.c_str());
+    return false;
   }
   return true;
 }
@@ -1201,25 +1219,44 @@ bool DexFileVerifier::CheckEncodedAnnotation() {
   }
 
   DECODE_UNSIGNED_CHECKED_FROM(ptr_, size);
-  uint32_t last_idx = 0;
+  todo_.emplace(ToDoItem{.annotation_size = size, .last_idx = kDexNoIndex});
+  return true;
+}
 
-  for (uint32_t i = 0; i < size; i++) {
-    DECODE_UNSIGNED_CHECKED_FROM(ptr_, idx);
-    if (!CheckIndex(idx, header_->string_ids_size_, "annotation_element name_idx")) {
-      return false;
-    }
+// Always called directly from FlushToDoList, which avoids recursion.
+bool DexFileVerifier::CheckAnnotationElement(/*inout*/ uint32_t* last_idx) {
+  DECODE_UNSIGNED_CHECKED_FROM(ptr_, idx);
+  if (!CheckIndex(idx, header_->string_ids_size_, "annotation_element name_idx")) {
+    return false;
+  }
 
-    if (UNLIKELY(last_idx >= idx && i != 0)) {
-      ErrorStringPrintf("Out-of-order annotation_element name_idx: %x then %x",
-                        last_idx, idx);
-      return false;
-    }
+  if (UNLIKELY(*last_idx >= idx && *last_idx != kDexNoIndex)) {
+    ErrorStringPrintf("Out-of-order annotation_element name_idx: %x then %x", *last_idx, idx);
+    return false;
+  }
+  *last_idx = idx;
 
-    if (!CheckEncodedValue()) {
-      return false;
-    }
+  return CheckEncodedValue();
+}
 
-    last_idx = idx;
+// Keep processing the rest of the to-do list until we are finished or encounter an error.
+bool DexFileVerifier::FlushToDoList() {
+  while (!todo_.empty()) {
+    ToDoItem& item = todo_.top();
+    DCHECK(item.array_size == 0u || item.annotation_size == 0u);
+    if (item.array_size > 0) {
+      item.array_size--;
+      if (!CheckArrayElement()) {
+        return false;
+      }
+    } else if (item.annotation_size > 0) {
+      item.annotation_size--;
+      if (!CheckAnnotationElement(&item.last_idx)) {
+        return false;
+      }
+    } else {
+      todo_.pop();
+    }
   }
   return true;
 }
@@ -1970,7 +2007,8 @@ bool DexFileVerifier::CheckIntraAnnotationItem() {
       return false;
   }
 
-  if (!CheckEncodedAnnotation()) {
+  CHECK(todo_.empty());
+  if (!CheckEncodedAnnotation() || !FlushToDoList()) {
     return false;
   }
 
@@ -2222,7 +2260,8 @@ bool DexFileVerifier::CheckIntraSectionIterate(uint32_t section_count) {
         break;
       }
       case DexFile::kDexTypeEncodedArrayItem: {
-        if (!CheckEncodedArray()) {
+        CHECK(todo_.empty());
+        if (!CheckEncodedArray() || !FlushToDoList()) {
           return false;
         }
         break;
@@ -3465,6 +3504,7 @@ bool DexFileVerifier::Verify() {
     return false;
   }
 
+  CHECK(todo_.empty());  // No unprocessed work left over.
   return true;
 }
 
diff --git a/libelffile/elf/elf_debug_reader.h b/libelffile/elf/elf_debug_reader.h
index fc7ad5654b..05aa339c13 100644
--- a/libelffile/elf/elf_debug_reader.h
+++ b/libelffile/elf/elf_debug_reader.h
@@ -119,7 +119,7 @@ class ElfDebugReader {
   }
 
   template <typename VisitSym>
-  void VisitFunctionSymbols(VisitSym visit_sym) {
+  void VisitFunctionSymbols(VisitSym&& visit_sym) {
     const Elf_Shdr* symtab = GetSection(".symtab");
     const Elf_Shdr* strtab = GetSection(".strtab");
     const Elf_Shdr* text = GetSection(".text");
@@ -135,12 +135,12 @@ class ElfDebugReader {
       }
     }
     if (gnu_debugdata_reader_ != nullptr) {
-      gnu_debugdata_reader_->VisitFunctionSymbols(visit_sym);
+      gnu_debugdata_reader_->VisitFunctionSymbols(std::forward<VisitSym>(visit_sym));
     }
   }
 
   template <typename VisitSym>
-  void VisitDynamicSymbols(VisitSym visit_sym) {
+  void VisitDynamicSymbols(VisitSym&& visit_sym) {
     const Elf_Shdr* dynsym = GetSection(".dynsym");
     const Elf_Shdr* dynstr = GetSection(".dynstr");
     if (dynsym != nullptr && dynstr != nullptr) {
@@ -153,7 +153,7 @@ class ElfDebugReader {
   }
 
   template <typename VisitCIE, typename VisitFDE>
-  void VisitDebugFrame(VisitCIE visit_cie, VisitFDE visit_fde) {
+  void VisitDebugFrame(VisitCIE&& visit_cie, VisitFDE&& visit_fde) {
     const Elf_Shdr* debug_frame = GetSection(".debug_frame");
     if (debug_frame != nullptr) {
       for (size_t offset = 0; offset < debug_frame->sh_size;) {
@@ -169,7 +169,8 @@ class ElfDebugReader {
       }
     }
     if (gnu_debugdata_reader_ != nullptr) {
-      gnu_debugdata_reader_->VisitDebugFrame(visit_cie, visit_fde);
+      gnu_debugdata_reader_->VisitDebugFrame(std::forward<VisitCIE>(visit_cie),
+                                             std::forward<VisitFDE>(visit_fde));
     }
   }
 
diff --git a/libnativeloader/Android.bp b/libnativeloader/Android.bp
index 16449ac745..e9c26c592c 100644
--- a/libnativeloader/Android.bp
+++ b/libnativeloader/Android.bp
@@ -158,6 +158,7 @@ art_cc_test {
         "native_loader_test.cpp",
     ],
     srcs: [
+        "library_namespaces_test.cpp",
         "native_loader_api_test.c",
         "native_loader_test.cpp",
         "open_system_library.cpp",
diff --git a/libnativeloader/library_namespaces.cpp b/libnativeloader/library_namespaces.cpp
index 1e29f4e457..e2b27294f9 100644
--- a/libnativeloader/library_namespaces.cpp
+++ b/libnativeloader/library_namespaces.cpp
@@ -16,23 +16,28 @@
 
 #if defined(ART_TARGET_ANDROID)
 
+#define LOG_TAG "nativeloader"
+
 #include "library_namespaces.h"
 
 #include <dirent.h>
 #include <dlfcn.h>
+#include <stdio.h>
 
+#include <algorithm>
+#include <optional>
 #include <regex>
 #include <string>
+#include <string_view>
 #include <vector>
 
-#include <android-base/file.h>
-#include <android-base/logging.h>
-#include <android-base/macros.h>
-#include <android-base/result.h>
-#include <android-base/strings.h>
-#include <android-base/stringprintf.h>
-#include <nativehelper/scoped_utf_chars.h>
-
+#include "android-base/file.h"
+#include "android-base/logging.h"
+#include "android-base/macros.h"
+#include "android-base/result.h"
+#include "android-base/stringprintf.h"
+#include "android-base/strings.h"
+#include "nativehelper/scoped_utf_chars.h"
 #include "nativeloader/dlext_namespaces.h"
 #include "public_libraries.h"
 #include "utils.h"
@@ -41,20 +46,9 @@ namespace android::nativeloader {
 
 namespace {
 
-constexpr const char* kApexPath = "/apex/";
+using ::android::base::Error;
 
-// The device may be configured to have the vendor libraries loaded to a separate namespace.
-// For historical reasons this namespace was named sphal but effectively it is intended
-// to use to load vendor libraries to separate namespace with controlled interface between
-// vendor and system namespaces.
-constexpr const char* kVendorNamespaceName = "sphal";
-// Similar to sphal namespace, product namespace provides some product libraries.
-constexpr const char* kProductNamespaceName = "product";
-
-// vndk namespace for unbundled vendor apps
-constexpr const char* kVndkNamespaceName = "vndk";
-// vndk_product namespace for unbundled product apps
-constexpr const char* kVndkProductNamespaceName = "vndk_product";
+constexpr const char* kApexPath = "/apex/";
 
 // clns-XX is a linker namespace that is created for normal apps installed in
 // the data partition. To be specific, it is created for the app classloader.
@@ -89,15 +83,8 @@ constexpr const char* kVendorLibPath = "/vendor/" LIB;
 // a symlink to the other.
 constexpr const char* kProductLibPath = "/product/" LIB ":/system/product/" LIB;
 
-const std::regex kVendorDexPathRegex("(^|:)(/system)?/vendor/");
-const std::regex kProductDexPathRegex("(^|:)(/system)?/product/");
-
-// Define origin partition of APK
-using ApkOrigin = enum {
-  APK_ORIGIN_DEFAULT = 0,
-  APK_ORIGIN_VENDOR = 1,   // Includes both /vendor and /system/vendor
-  APK_ORIGIN_PRODUCT = 2,  // Includes both /product and /system/product
-};
+const std::regex kVendorPathRegex("(/system)?/vendor/.*");
+const std::regex kProductPathRegex("(/system)?/product/.*");
 
 jobject GetParentClassLoader(JNIEnv* env, jobject class_loader) {
   jclass class_loader_class = env->FindClass("java/lang/ClassLoader");
@@ -107,22 +94,44 @@ jobject GetParentClassLoader(JNIEnv* env, jobject class_loader) {
   return env->CallObjectMethod(class_loader, get_parent);
 }
 
-ApkOrigin GetApkOriginFromDexPath(const std::string& dex_path) {
-  ApkOrigin apk_origin = APK_ORIGIN_DEFAULT;
-  if (std::regex_search(dex_path, kVendorDexPathRegex)) {
-    apk_origin = APK_ORIGIN_VENDOR;
-  }
-  if (std::regex_search(dex_path, kProductDexPathRegex)) {
-    LOG_ALWAYS_FATAL_IF(apk_origin == APK_ORIGIN_VENDOR,
-                        "Dex path contains both vendor and product partition : %s",
-                        dex_path.c_str());
+}  // namespace
 
-    apk_origin = APK_ORIGIN_PRODUCT;
+ApiDomain GetApiDomainFromPath(const std::string_view path) {
+  if (std::regex_match(path.begin(), path.end(), kVendorPathRegex)) {
+    return API_DOMAIN_VENDOR;
+  }
+  if (is_product_treblelized() && std::regex_match(path.begin(), path.end(), kProductPathRegex)) {
+    return API_DOMAIN_PRODUCT;
   }
-  return apk_origin;
+  return API_DOMAIN_DEFAULT;
 }
 
-}  // namespace
+// Returns the API domain for a ':'-separated list of paths, or an error if they
+// match more than one.
+Result<ApiDomain> GetApiDomainFromPathList(const std::string& path_list) {
+  ApiDomain result = API_DOMAIN_DEFAULT;
+  size_t start_pos = 0;
+  while (true) {
+    size_t end_pos = path_list.find(':', start_pos);
+    ApiDomain api_domain =
+        GetApiDomainFromPath(std::string_view(path_list).substr(start_pos, end_pos));
+    // Allow mixing API_DOMAIN_DEFAULT with any other domain. That's a bit lax,
+    // since the default e.g. includes /data, which strictly speaking is a
+    // separate domain. However, we keep it this way to not risk compat issues
+    // until we actually need all domains.
+    if (api_domain != API_DOMAIN_DEFAULT) {
+      if (result != API_DOMAIN_DEFAULT && result != api_domain) {
+        return Error() << "Path list crosses partition boundaries: " << path_list;
+      }
+      result = api_domain;
+    }
+    if (end_pos == std::string::npos) {
+      break;
+    }
+    start_pos = end_pos + 1;
+  }
+  return result;
+}
 
 void LibraryNamespaces::Initialize() {
   // Once public namespace is initialized there is no
@@ -170,7 +179,7 @@ static const std::string filter_public_libraries(
   }
   std::vector<std::string> filtered;
   std::vector<std::string> orig = android::base::Split(public_libraries, ":");
-  for (const auto& lib : uses_libraries) {
+  for (const std::string& lib : uses_libraries) {
     if (std::find(orig.begin(), orig.end(), lib) != orig.end()) {
       filtered.emplace_back(lib);
     }
@@ -178,32 +187,29 @@ static const std::string filter_public_libraries(
   return android::base::Join(filtered, ":");
 }
 
-Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t target_sdk_version,
-                                                         jobject class_loader, bool is_shared,
-                                                         jstring dex_path_j,
-                                                         jstring java_library_path,
-                                                         jstring java_permitted_path,
-                                                         jstring uses_library_list) {
+Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env,
+                                                         uint32_t target_sdk_version,
+                                                         jobject class_loader,
+                                                         ApiDomain api_domain,
+                                                         bool is_shared,
+                                                         const std::string& dex_path,
+                                                         jstring library_path_j,
+                                                         jstring permitted_path_j,
+                                                         jstring uses_library_list_j) {
   std::string library_path;  // empty string by default.
-  std::string dex_path;
 
-  if (java_library_path != nullptr) {
-    ScopedUtfChars library_path_utf_chars(env, java_library_path);
+  if (library_path_j != nullptr) {
+    ScopedUtfChars library_path_utf_chars(env, library_path_j);
     library_path = library_path_utf_chars.c_str();
   }
 
-  if (dex_path_j != nullptr) {
-    ScopedUtfChars dex_path_chars(env, dex_path_j);
-    dex_path = dex_path_chars.c_str();
-  }
-
   std::vector<std::string> uses_libraries;
-  if (uses_library_list != nullptr) {
-    ScopedUtfChars names(env, uses_library_list);
+  if (uses_library_list_j != nullptr) {
+    ScopedUtfChars names(env, uses_library_list_j);
     uses_libraries = android::base::Split(names.c_str(), ":");
   } else {
-    // uses_library_list could be nullptr when System.loadLibrary is called from a
-    // custom classloader. In that case, we don't know the list of public
+    // uses_library_list_j could be nullptr when System.loadLibrary is called
+    // from a custom classloader. In that case, we don't know the list of public
     // libraries because we don't know which apk the classloader is for. Only
     // choices we can have are 1) allowing all public libs (as before), or 2)
     // not allowing all but NDK libs. Here we take #1 because #2 would surprise
@@ -214,8 +220,6 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
     uses_libraries.emplace_back(LIBRARY_ALL);
   }
 
-  ApkOrigin apk_origin = GetApkOriginFromDexPath(dex_path);
-
   // (http://b/27588281) This is a workaround for apps using custom
   // classloaders and calling System.load() with an absolute path which
   // is outside of the classloader library search path.
@@ -224,8 +228,8 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
   // under /data and /mnt/expand
   std::string permitted_path = kAlwaysPermittedDirectories;
 
-  if (java_permitted_path != nullptr) {
-    ScopedUtfChars path(env, java_permitted_path);
+  if (permitted_path_j != nullptr) {
+    ScopedUtfChars path(env, permitted_path_j);
     if (path.c_str() != nullptr && path.size() > 0) {
       permitted_path = permitted_path + ":" + path.c_str();
     }
@@ -236,13 +240,13 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
 
   std::string system_exposed_libraries = default_public_libraries();
   std::string namespace_name = kClassloaderNamespaceName;
-  ApkOrigin unbundled_app_origin = APK_ORIGIN_DEFAULT;
-  const char* apk_origin_msg = "other apk";  // Only for debug logging.
+  ApiDomain unbundled_app_domain = API_DOMAIN_DEFAULT;
+  const char* api_domain_msg = "other apk";  // Only for debug logging.
 
   if (!is_shared) {
-    if (apk_origin == APK_ORIGIN_VENDOR) {
-      unbundled_app_origin = APK_ORIGIN_VENDOR;
-      apk_origin_msg = "unbundled vendor apk";
+    if (api_domain == API_DOMAIN_VENDOR) {
+      unbundled_app_domain = API_DOMAIN_VENDOR;
+      api_domain_msg = "unbundled vendor apk";
 
       // For vendor apks, give access to the vendor libs even though they are
       // treated as unbundled; the libs and apks are still bundled together in the
@@ -255,9 +259,9 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
 
       // Different name is useful for debugging
       namespace_name = kVendorClassloaderNamespaceName;
-    } else if (apk_origin == APK_ORIGIN_PRODUCT && is_product_treblelized()) {
-      unbundled_app_origin = APK_ORIGIN_PRODUCT;
-      apk_origin_msg = "unbundled product apk";
+    } else if (api_domain == API_DOMAIN_PRODUCT) {
+      unbundled_app_domain = API_DOMAIN_PRODUCT;
+      api_domain_msg = "unbundled product apk";
 
       // Like for vendor apks, give access to the product libs since they are
       // bundled together in the same partition.
@@ -287,20 +291,20 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
       "Configuring %s for %s %s. target_sdk_version=%u, uses_libraries=%s, library_path=%s, "
       "permitted_path=%s",
       namespace_name.c_str(),
-      apk_origin_msg,
+      api_domain_msg,
       dex_path.c_str(),
       static_cast<unsigned>(target_sdk_version),
       android::base::Join(uses_libraries, ':').c_str(),
       library_path.c_str(),
       permitted_path.c_str());
 
-  if (unbundled_app_origin != APK_ORIGIN_VENDOR) {
+  if (unbundled_app_domain != API_DOMAIN_VENDOR) {
     // Extended public libraries are NOT available to unbundled vendor apks, but
     // they are to other apps, including those in system, system_ext, and
     // product partitions. The reason is that when GSI is used, the system
     // partition may get replaced, and then vendor apps may fail. It's fine for
     // product apps, because that partition isn't mounted in GSI tests.
-    auto libs =
+    const std::string libs =
         filter_public_libraries(target_sdk_version, uses_libraries, extended_public_libraries());
     if (!libs.empty()) {
       ALOGD("Extending system_exposed_libraries: %s", libs.c_str());
@@ -320,27 +324,33 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
   bool also_used_as_anonymous = is_main_classloader;
   // Note: this function is executed with g_namespaces_mutex held, thus no
   // racing here.
-  auto app_ns = NativeLoaderNamespace::Create(
-      namespace_name, library_path, permitted_path, parent_ns, is_shared,
-      target_sdk_version < 24 /* is_exempt_list_enabled */, also_used_as_anonymous);
+  Result<NativeLoaderNamespace> app_ns =
+      NativeLoaderNamespace::Create(namespace_name,
+                                    library_path,
+                                    permitted_path,
+                                    parent_ns,
+                                    is_shared,
+                                    target_sdk_version < 24 /* is_exempt_list_enabled */,
+                                    also_used_as_anonymous);
   if (!app_ns.ok()) {
     return app_ns.error();
   }
   // ... and link to other namespaces to allow access to some public libraries
   bool is_bridged = app_ns->IsBridged();
 
-  auto system_ns = NativeLoaderNamespace::GetSystemNamespace(is_bridged);
+  Result<NativeLoaderNamespace> system_ns = NativeLoaderNamespace::GetSystemNamespace(is_bridged);
   if (!system_ns.ok()) {
     return system_ns.error();
   }
 
-  auto linked = app_ns->Link(&system_ns.value(), system_exposed_libraries);
+  Result<void> linked = app_ns->Link(&system_ns.value(), system_exposed_libraries);
   if (!linked.ok()) {
     return linked.error();
   }
 
   for (const auto&[apex_ns_name, public_libs] : apex_public_libraries()) {
-    auto ns = NativeLoaderNamespace::GetExportedNamespace(apex_ns_name, is_bridged);
+    Result<NativeLoaderNamespace> ns =
+        NativeLoaderNamespace::GetExportedNamespace(apex_ns_name, is_bridged);
     // Even if APEX namespace is visible, it may not be available to bridged.
     if (ns.ok()) {
       linked = app_ns->Link(&ns.value(), public_libs);
@@ -351,8 +361,9 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
   }
 
   // Give access to VNDK-SP libraries from the 'vndk' namespace for unbundled vendor apps.
-  if (unbundled_app_origin == APK_ORIGIN_VENDOR && !vndksp_libraries_vendor().empty()) {
-    auto vndk_ns = NativeLoaderNamespace::GetExportedNamespace(kVndkNamespaceName, is_bridged);
+  if (unbundled_app_domain == API_DOMAIN_VENDOR && !vndksp_libraries_vendor().empty()) {
+    Result<NativeLoaderNamespace> vndk_ns =
+        NativeLoaderNamespace::GetExportedNamespace(kVndkNamespaceName, is_bridged);
     if (vndk_ns.ok()) {
       linked = app_ns->Link(&vndk_ns.value(), vndksp_libraries_vendor());
       if (!linked.ok()) {
@@ -362,8 +373,9 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
   }
 
   // Give access to VNDK-SP libraries from the 'vndk_product' namespace for unbundled product apps.
-  if (unbundled_app_origin == APK_ORIGIN_PRODUCT && !vndksp_libraries_product().empty()) {
-    auto vndk_ns = NativeLoaderNamespace::GetExportedNamespace(kVndkProductNamespaceName, is_bridged);
+  if (unbundled_app_domain == API_DOMAIN_PRODUCT && !vndksp_libraries_product().empty()) {
+    Result<NativeLoaderNamespace> vndk_ns =
+        NativeLoaderNamespace::GetExportedNamespace(kVndkProductNamespaceName, is_bridged);
     if (vndk_ns.ok()) {
       linked = app_ns->Link(&vndk_ns.value(), vndksp_libraries_product());
       if (!linked.ok()) {
@@ -373,11 +385,12 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
   }
 
   for (const std::string& each_jar_path : android::base::Split(dex_path, ":")) {
-    auto apex_ns_name = FindApexNamespaceName(each_jar_path);
-    if (apex_ns_name.ok()) {
-      const auto& jni_libs = apex_jni_libraries(*apex_ns_name);
+    std::optional<std::string> apex_ns_name = FindApexNamespaceName(each_jar_path);
+    if (apex_ns_name.has_value()) {
+      const std::string& jni_libs = apex_jni_libraries(apex_ns_name.value());
       if (jni_libs != "") {
-        auto apex_ns = NativeLoaderNamespace::GetExportedNamespace(*apex_ns_name, is_bridged);
+        Result<NativeLoaderNamespace> apex_ns =
+            NativeLoaderNamespace::GetExportedNamespace(apex_ns_name.value(), is_bridged);
         if (apex_ns.ok()) {
           linked = app_ns->Link(&apex_ns.value(), jni_libs);
           if (!linked.ok()) {
@@ -388,12 +401,13 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
     }
   }
 
-  auto vendor_libs = filter_public_libraries(target_sdk_version, uses_libraries,
-                                             vendor_public_libraries());
+  const std::string vendor_libs =
+      filter_public_libraries(target_sdk_version, uses_libraries, vendor_public_libraries());
   if (!vendor_libs.empty()) {
-    auto vendor_ns = NativeLoaderNamespace::GetExportedNamespace(kVendorNamespaceName, is_bridged);
+    Result<NativeLoaderNamespace> vendor_ns =
+        NativeLoaderNamespace::GetExportedNamespace(kVendorNamespaceName, is_bridged);
     // when vendor_ns is not configured, link to the system namespace
-    auto target_ns = vendor_ns.ok() ? vendor_ns : system_ns;
+    Result<NativeLoaderNamespace> target_ns = vendor_ns.ok() ? vendor_ns : system_ns;
     if (target_ns.ok()) {
       linked = app_ns->Link(&target_ns.value(), vendor_libs);
       if (!linked.ok()) {
@@ -402,10 +416,10 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
     }
   }
 
-  auto product_libs = filter_public_libraries(target_sdk_version, uses_libraries,
-                                              product_public_libraries());
+  const std::string product_libs =
+      filter_public_libraries(target_sdk_version, uses_libraries, product_public_libraries());
   if (!product_libs.empty()) {
-    auto target_ns = system_ns;
+    Result<NativeLoaderNamespace> target_ns = system_ns;
     if (is_product_treblelized()) {
       target_ns = NativeLoaderNamespace::GetExportedNamespace(kProductNamespaceName, is_bridged);
     }
@@ -422,8 +436,8 @@ Result<NativeLoaderNamespace*> LibraryNamespaces::Create(JNIEnv* env, uint32_t t
     }
   }
 
-  auto& emplaced = namespaces_.emplace_back(
-      std::make_pair(env->NewWeakGlobalRef(class_loader), *app_ns));
+  std::pair<jweak, NativeLoaderNamespace>& emplaced =
+      namespaces_.emplace_back(std::make_pair(env->NewWeakGlobalRef(class_loader), *app_ns));
   if (is_main_classloader) {
     app_main_namespace_ = &emplaced.second;
   }
@@ -459,7 +473,7 @@ NativeLoaderNamespace* LibraryNamespaces::FindParentNamespaceByClassLoader(JNIEn
   return nullptr;
 }
 
-base::Result<std::string> FindApexNamespaceName(const std::string& location) {
+std::optional<std::string> FindApexNamespaceName(const std::string& location) {
   // Lots of implicit assumptions here: we expect `location` to be of the form:
   // /apex/modulename/...
   //
@@ -473,7 +487,7 @@ base::Result<std::string> FindApexNamespaceName(const std::string& location) {
     std::replace(name.begin(), name.end(), '.', '_');
     return name;
   }
-  return base::Error();
+  return std::nullopt;
 }
 
 }  // namespace android::nativeloader
diff --git a/libnativeloader/library_namespaces.h b/libnativeloader/library_namespaces.h
index 4871528f47..ae1cd88f20 100644
--- a/libnativeloader/library_namespaces.h
+++ b/libnativeloader/library_namespaces.h
@@ -21,20 +21,43 @@
 #error "Not available for host or linux target"
 #endif
 
-#define LOG_TAG "nativeloader"
-
-#include "native_loader_namespace.h"
-
 #include <list>
+#include <optional>
 #include <string>
+#include <string_view>
 
-#include <android-base/result.h>
-#include <jni.h>
+#include "android-base/result.h"
+#include "jni.h"
+#include "native_loader_namespace.h"
 
 namespace android::nativeloader {
 
 using android::base::Result;
 
+// The device may be configured to have the vendor libraries loaded to a separate namespace.
+// For historical reasons this namespace was named sphal but effectively it is intended
+// to use to load vendor libraries to separate namespace with controlled interface between
+// vendor and system namespaces.
+constexpr const char* kVendorNamespaceName = "sphal";
+// Similar to sphal namespace, product namespace provides some product libraries.
+constexpr const char* kProductNamespaceName = "product";
+
+// vndk namespace for unbundled vendor apps
+constexpr const char* kVndkNamespaceName = "vndk";
+// vndk_product namespace for unbundled product apps
+constexpr const char* kVndkProductNamespaceName = "vndk_product";
+
+// API domains, roughly corresponding to partitions. Interdependencies between
+// these must follow API restrictions, while intradependencies do not.
+using ApiDomain = enum {
+  API_DOMAIN_DEFAULT = 0,  // Locations other than those below, in particular for ordinary apps
+  API_DOMAIN_VENDOR = 1,   // Vendor partition
+  API_DOMAIN_PRODUCT = 2,  // Product partition
+};
+
+ApiDomain GetApiDomainFromPath(const std::string_view path);
+Result<ApiDomain> GetApiDomainFromPathList(const std::string& path_list);
+
 // LibraryNamespaces is a singleton object that manages NativeLoaderNamespace
 // objects for an app process. Its main job is to create (and configure) a new
 // NativeLoaderNamespace object for a Java ClassLoader, and to find an existing
@@ -53,10 +76,15 @@ class LibraryNamespaces {
     initialized_ = false;
     app_main_namespace_ = nullptr;
   }
-  Result<NativeLoaderNamespace*> Create(JNIEnv* env, uint32_t target_sdk_version,
-                                        jobject class_loader, bool is_shared, jstring dex_path,
-                                        jstring java_library_path, jstring java_permitted_path,
-                                        jstring uses_library_list);
+  Result<NativeLoaderNamespace*> Create(JNIEnv* env,
+                                        uint32_t target_sdk_version,
+                                        jobject class_loader,
+                                        ApiDomain api_domain,
+                                        bool is_shared,
+                                        const std::string& dex_path,
+                                        jstring library_path_j,
+                                        jstring permitted_path_j,
+                                        jstring uses_library_list_j);
   NativeLoaderNamespace* FindNamespaceByClassLoader(JNIEnv* env, jobject class_loader);
 
  private:
@@ -68,7 +96,7 @@ class LibraryNamespaces {
   std::list<std::pair<jweak, NativeLoaderNamespace>> namespaces_;
 };
 
-Result<std::string> FindApexNamespaceName(const std::string& location);
+std::optional<std::string> FindApexNamespaceName(const std::string& location);
 
 }  // namespace android::nativeloader
 
diff --git a/libnativeloader/library_namespaces_test.cpp b/libnativeloader/library_namespaces_test.cpp
new file mode 100644
index 0000000000..7780418846
--- /dev/null
+++ b/libnativeloader/library_namespaces_test.cpp
@@ -0,0 +1,89 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#if defined(ART_TARGET_ANDROID)
+
+#include "library_namespaces.h"
+
+#include "android-base/result-gmock.h"
+#include "gtest/gtest.h"
+#include "public_libraries.h"
+
+namespace android {
+namespace nativeloader {
+namespace {
+
+using ::android::base::testing::HasError;
+using ::android::base::testing::HasValue;
+using ::android::base::testing::WithMessage;
+using ::testing::StartsWith;
+
+TEST(LibraryNamespacesTest, TestGetApiDomainFromPath) {
+  // GetApiDomainFromPath returns API_DOMAIN_PRODUCT only if the device is
+  // trebleized and has an unbundled product partition.
+  ApiDomain api_domain_product = is_product_treblelized() ? API_DOMAIN_PRODUCT : API_DOMAIN_DEFAULT;
+
+  EXPECT_EQ(GetApiDomainFromPath("/data/somewhere"), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("/system/somewhere"), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("/product/somewhere"), api_domain_product);
+  EXPECT_EQ(GetApiDomainFromPath("/vendor/somewhere"), API_DOMAIN_VENDOR);
+  EXPECT_EQ(GetApiDomainFromPath("/system/product/somewhere"), api_domain_product);
+  EXPECT_EQ(GetApiDomainFromPath("/system/vendor/somewhere"), API_DOMAIN_VENDOR);
+
+  EXPECT_EQ(GetApiDomainFromPath(""), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("/"), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("product/somewhere"), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("/product"), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("/product/"), api_domain_product);
+  EXPECT_EQ(GetApiDomainFromPath(":/product/"), API_DOMAIN_DEFAULT);
+
+  EXPECT_EQ(GetApiDomainFromPath("/data/somewhere:/product/somewhere"), API_DOMAIN_DEFAULT);
+  EXPECT_EQ(GetApiDomainFromPath("/vendor/somewhere:/product/somewhere"), API_DOMAIN_VENDOR);
+  EXPECT_EQ(GetApiDomainFromPath("/product/somewhere:/vendor/somewhere"), api_domain_product);
+}
+
+TEST(LibraryNamespacesTest, TestGetApiDomainFromPathList) {
+  // GetApiDomainFromPath returns API_DOMAIN_PRODUCT only if the device is
+  // trebleized and has an unbundled product partition.
+  ApiDomain api_domain_product = is_product_treblelized() ? API_DOMAIN_PRODUCT : API_DOMAIN_DEFAULT;
+
+  EXPECT_THAT(GetApiDomainFromPathList("/data/somewhere"), HasValue(API_DOMAIN_DEFAULT));
+  EXPECT_THAT(GetApiDomainFromPathList("/system/somewhere"), HasValue(API_DOMAIN_DEFAULT));
+  EXPECT_THAT(GetApiDomainFromPathList("/product/somewhere"), HasValue(api_domain_product));
+  EXPECT_THAT(GetApiDomainFromPathList("/vendor/somewhere"), HasValue(API_DOMAIN_VENDOR));
+  EXPECT_THAT(GetApiDomainFromPathList("/system/product/somewhere"), HasValue(api_domain_product));
+  EXPECT_THAT(GetApiDomainFromPathList("/system/vendor/somewhere"), HasValue(API_DOMAIN_VENDOR));
+
+  EXPECT_THAT(GetApiDomainFromPathList(""), HasValue(API_DOMAIN_DEFAULT));
+  EXPECT_THAT(GetApiDomainFromPathList(":"), HasValue(API_DOMAIN_DEFAULT));
+  EXPECT_THAT(GetApiDomainFromPathList(":/vendor/somewhere"), HasValue(API_DOMAIN_VENDOR));
+  EXPECT_THAT(GetApiDomainFromPathList("/vendor/somewhere:"), HasValue(API_DOMAIN_VENDOR));
+
+  EXPECT_THAT(GetApiDomainFromPathList("/data/somewhere:/product/somewhere"),
+              HasValue(api_domain_product));
+  if (api_domain_product == API_DOMAIN_PRODUCT) {
+    EXPECT_THAT(GetApiDomainFromPathList("/vendor/somewhere:/product/somewhere"),
+                HasError(WithMessage(StartsWith("Path list crosses partition boundaries"))));
+    EXPECT_THAT(GetApiDomainFromPathList("/product/somewhere:/vendor/somewhere"),
+                HasError(WithMessage(StartsWith("Path list crosses partition boundaries"))));
+  }
+}
+
+}  // namespace
+}  // namespace nativeloader
+}  // namespace android
+
+#endif  // ART_TARGET_ANDROID
diff --git a/libnativeloader/native_loader.cpp b/libnativeloader/native_loader.cpp
index 2deb5ef3d0..61925431ef 100644
--- a/libnativeloader/native_loader.cpp
+++ b/libnativeloader/native_loader.cpp
@@ -24,19 +24,21 @@
 #include <algorithm>
 #include <memory>
 #include <mutex>
+#include <optional>
 #include <string>
 #include <vector>
 
-#include <android-base/file.h>
-#include <android-base/macros.h>
-#include <android-base/strings.h>
-#include <android-base/thread_annotations.h>
-#include <nativebridge/native_bridge.h>
-#include <nativehelper/scoped_utf_chars.h>
+#include "android-base/file.h"
+#include "android-base/macros.h"
+#include "android-base/strings.h"
+#include "android-base/thread_annotations.h"
+#include "nativebridge/native_bridge.h"
+#include "nativehelper/scoped_utf_chars.h"
+#include "public_libraries.h"
 
 #ifdef ART_TARGET_ANDROID
-#include <log/log.h>
 #include "library_namespaces.h"
+#include "log/log.h"
 #include "nativeloader/dlext_namespaces.h"
 #endif
 
@@ -46,6 +48,9 @@ namespace {
 
 #if defined(ART_TARGET_ANDROID)
 
+using ::android::base::Result;
+using ::android::nativeloader::LibraryNamespaces;
+
 // NATIVELOADER_DEFAULT_NAMESPACE_LIBS is an environment variable that can be
 // used to list extra libraries (separated by ":") that libnativeloader will
 // load from the default namespace. The libraries must be listed without paths,
@@ -62,21 +67,24 @@ namespace {
 // test libraries that depend on ART internal libraries.
 constexpr const char* kNativeloaderExtraLibs = "nativeloader-extra-libs";
 
-using android::nativeloader::LibraryNamespaces;
-
 std::mutex g_namespaces_mutex;
-LibraryNamespaces* g_namespaces = new LibraryNamespaces;
-NativeLoaderNamespace* g_nativeloader_extra_libs_namespace = nullptr;
-
-android_namespace_t* FindExportedNamespace(const char* caller_location) {
-  auto name = nativeloader::FindApexNamespaceName(caller_location);
-  if (name.ok()) {
-    android_namespace_t* boot_namespace = android_get_exported_namespace(name->c_str());
-    LOG_ALWAYS_FATAL_IF((boot_namespace == nullptr),
-                        "Error finding namespace of apex: no namespace called %s", name->c_str());
-    return boot_namespace;
+LibraryNamespaces* g_namespaces GUARDED_BY(g_namespaces_mutex) = new LibraryNamespaces;
+NativeLoaderNamespace* g_nativeloader_extra_libs_namespace GUARDED_BY(g_namespaces_mutex) = nullptr;
+
+std::optional<NativeLoaderNamespace> FindApexNamespace(const char* caller_location) {
+  std::optional<std::string> name = nativeloader::FindApexNamespaceName(caller_location);
+  if (name.has_value()) {
+    // Native Bridge is never used for APEXes.
+    Result<NativeLoaderNamespace> ns =
+        NativeLoaderNamespace::GetExportedNamespace(name.value(), /*is_bridged=*/false);
+    LOG_ALWAYS_FATAL_IF(!ns.ok(),
+                        "Error finding ns %s for APEX location %s: %s",
+                        name.value().c_str(),
+                        caller_location,
+                        ns.error().message().c_str());
+    return ns.value();
   }
-  return nullptr;
+  return std::nullopt;
 }
 
 Result<void> CreateNativeloaderDefaultNamespaceLibsLink(NativeLoaderNamespace& ns)
@@ -133,26 +141,34 @@ Result<void*> TryLoadNativeloaderExtraLib(const char* path) {
   if (!ns.ok()) {
     return ns.error();
   }
-  return ns.value()->Load(path);
+
+  Result<void*> res = ns.value()->Load(path);
+  ALOGD("Load %s using ns %s from NATIVELOADER_DEFAULT_NAMESPACE_LIBS match: %s",
+        path,
+        ns.value()->name().c_str(),
+        res.ok() ? "ok" : res.error().message().c_str());
+  return res;
 }
 
 Result<NativeLoaderNamespace*> CreateClassLoaderNamespaceLocked(JNIEnv* env,
                                                                 int32_t target_sdk_version,
                                                                 jobject class_loader,
+                                                                nativeloader::ApiDomain api_domain,
                                                                 bool is_shared,
-                                                                jstring dex_path,
-                                                                jstring library_path,
-                                                                jstring permitted_path,
-                                                                jstring uses_library_list)
+                                                                const std::string& dex_path,
+                                                                jstring library_path_j,
+                                                                jstring permitted_path_j,
+                                                                jstring uses_library_list_j)
     REQUIRES(g_namespaces_mutex) {
   Result<NativeLoaderNamespace*> ns = g_namespaces->Create(env,
                                                            target_sdk_version,
                                                            class_loader,
+                                                           api_domain,
                                                            is_shared,
                                                            dex_path,
-                                                           library_path,
-                                                           permitted_path,
-                                                           uses_library_list);
+                                                           library_path_j,
+                                                           permitted_path_j,
+                                                           uses_library_list_j);
   if (!ns.ok()) {
     return ns;
   }
@@ -163,7 +179,7 @@ Result<NativeLoaderNamespace*> CreateClassLoaderNamespaceLocked(JNIEnv* env,
   return ns;
 }
 
-#endif  // #if defined(ART_TARGET_ANDROID)
+#endif  // ART_TARGET_ANDROID
 
 }  // namespace
 
@@ -183,47 +199,86 @@ void ResetNativeLoader() {
 #endif
 }
 
-jstring CreateClassLoaderNamespace(JNIEnv* env, int32_t target_sdk_version, jobject class_loader,
-                                   bool is_shared, jstring dex_path, jstring library_path,
-                                   jstring permitted_path, jstring uses_library_list) {
+// dex_path_j may be a ':'-separated list of paths, e.g. when creating a shared
+// library loader - cf. mCodePaths in android.content.pm.SharedLibraryInfo.
+jstring CreateClassLoaderNamespace(JNIEnv* env,
+                                   int32_t target_sdk_version,
+                                   jobject class_loader,
+                                   bool is_shared,
+                                   jstring dex_path_j,
+                                   jstring library_path_j,
+                                   jstring permitted_path_j,
+                                   jstring uses_library_list_j) {
 #if defined(ART_TARGET_ANDROID)
+  std::string dex_path;
+  if (dex_path_j != nullptr) {
+    ScopedUtfChars dex_path_chars(env, dex_path_j);
+    dex_path = dex_path_chars.c_str();
+  }
+
+  Result<nativeloader::ApiDomain> api_domain = nativeloader::GetApiDomainFromPathList(dex_path);
+  if (!api_domain.ok()) {
+    return env->NewStringUTF(api_domain.error().message().c_str());
+  }
+
   std::lock_guard<std::mutex> guard(g_namespaces_mutex);
   Result<NativeLoaderNamespace*> ns = CreateClassLoaderNamespaceLocked(env,
                                                                        target_sdk_version,
                                                                        class_loader,
+                                                                       api_domain.value(),
                                                                        is_shared,
                                                                        dex_path,
-                                                                       library_path,
-                                                                       permitted_path,
-                                                                       uses_library_list);
+                                                                       library_path_j,
+                                                                       permitted_path_j,
+                                                                       uses_library_list_j);
   if (!ns.ok()) {
     return env->NewStringUTF(ns.error().message().c_str());
   }
+
 #else
-  UNUSED(env, target_sdk_version, class_loader, is_shared, dex_path, library_path, permitted_path,
-         uses_library_list);
+  UNUSED(env,
+         target_sdk_version,
+         class_loader,
+         is_shared,
+         dex_path_j,
+         library_path_j,
+         permitted_path_j,
+         uses_library_list_j);
 #endif
+
   return nullptr;
 }
 
-void* OpenNativeLibrary(JNIEnv* env, int32_t target_sdk_version, const char* path,
-                        jobject class_loader, const char* caller_location, jstring library_path,
-                        bool* needs_native_bridge, char** error_msg) {
+void* OpenNativeLibrary(JNIEnv* env,
+                        int32_t target_sdk_version,
+                        const char* path,
+                        jobject class_loader,
+                        const char* caller_location,
+                        jstring library_path_j,
+                        bool* needs_native_bridge,
+                        char** error_msg) {
 #if defined(ART_TARGET_ANDROID)
-  UNUSED(target_sdk_version);
-
   if (class_loader == nullptr) {
+    // class_loader is null only for the boot class loader (see
+    // IsBootClassLoader call in JavaVMExt::LoadNativeLibrary), i.e. the caller
+    // is in the boot classpath.
     *needs_native_bridge = false;
     if (caller_location != nullptr) {
-      android_namespace_t* boot_namespace = FindExportedNamespace(caller_location);
-      if (boot_namespace != nullptr) {
+      std::optional<NativeLoaderNamespace> ns = FindApexNamespace(caller_location);
+      if (ns.has_value()) {
         const android_dlextinfo dlextinfo = {
             .flags = ANDROID_DLEXT_USE_NAMESPACE,
-            .library_namespace = boot_namespace,
+            .library_namespace = ns.value().ToRawAndroidNamespace(),
         };
         void* handle = android_dlopen_ext(path, RTLD_NOW, &dlextinfo);
-        if (handle == nullptr) {
-          *error_msg = strdup(dlerror());
+        char* dlerror_msg = handle == nullptr ? strdup(dlerror()) : nullptr;
+        ALOGD("Load %s using APEX ns %s for caller %s: %s",
+              path,
+              ns.value().name().c_str(),
+              caller_location,
+              dlerror_msg == nullptr ? "ok" : dlerror_msg);
+        if (dlerror_msg != nullptr) {
+          *error_msg = dlerror_msg;
         }
         return handle;
       }
@@ -244,53 +299,86 @@ void* OpenNativeLibrary(JNIEnv* env, int32_t target_sdk_version, const char* pat
 
     // Fall back to the system namespace. This happens for preloaded JNI
     // libraries in the zygote.
-    // TODO(b/185833744): Investigate if this should fall back to the app main
-    // namespace (aka anonymous namespace) instead.
     void* handle = OpenSystemLibrary(path, RTLD_NOW);
-    if (handle == nullptr) {
-      *error_msg = strdup(dlerror());
+    char* dlerror_msg = handle == nullptr ? strdup(dlerror()) : nullptr;
+    ALOGD("Load %s using system ns (caller=%s): %s",
+          path,
+          caller_location == nullptr ? "<unknown>" : caller_location,
+          dlerror_msg == nullptr ? "ok" : dlerror_msg);
+    if (dlerror_msg != nullptr) {
+      *error_msg = dlerror_msg;
     }
     return handle;
   }
 
   std::lock_guard<std::mutex> guard(g_namespaces_mutex);
-  NativeLoaderNamespace* ns;
-
-  if ((ns = g_namespaces->FindNamespaceByClassLoader(env, class_loader)) == nullptr) {
-    // This is the case where the classloader was not created by ApplicationLoaders
-    // In this case we create an isolated not-shared namespace for it.
-    Result<NativeLoaderNamespace*> isolated_ns =
-        CreateClassLoaderNamespaceLocked(env,
-                                         target_sdk_version,
-                                         class_loader,
-                                         /*is_shared=*/false,
-                                         /*dex_path=*/nullptr,
-                                         library_path,
-                                         /*permitted_path=*/nullptr,
-                                         /*uses_library_list=*/nullptr);
-    if (!isolated_ns.ok()) {
-      *error_msg = strdup(isolated_ns.error().message().c_str());
-      return nullptr;
-    } else {
-      ns = *isolated_ns;
+
+  {
+    NativeLoaderNamespace* ns = g_namespaces->FindNamespaceByClassLoader(env, class_loader);
+    if (ns != nullptr) {
+      *needs_native_bridge = ns->IsBridged();
+      Result<void*> handle = ns->Load(path);
+      ALOGD("Load %s using ns %s from class loader (caller=%s): %s",
+            path,
+            ns->name().c_str(),
+            caller_location == nullptr ? "<unknown>" : caller_location,
+            handle.ok() ? "ok" : handle.error().message().c_str());
+      if (!handle.ok()) {
+        *error_msg = strdup(handle.error().message().c_str());
+        return nullptr;
+      }
+      return handle.value();
     }
   }
 
-  return OpenNativeLibraryInNamespace(ns, path, needs_native_bridge, error_msg);
-#else
+  // This is the case where the classloader was not created by ApplicationLoaders
+  // In this case we create an isolated not-shared namespace for it.
+  const std::string empty_dex_path;
+  Result<NativeLoaderNamespace*> isolated_ns =
+      CreateClassLoaderNamespaceLocked(env,
+                                       target_sdk_version,
+                                       class_loader,
+                                       nativeloader::API_DOMAIN_DEFAULT,
+                                       /*is_shared=*/false,
+                                       empty_dex_path,
+                                       library_path_j,
+                                       /*permitted_path=*/nullptr,
+                                       /*uses_library_list=*/nullptr);
+  if (!isolated_ns.ok()) {
+    ALOGD("Failed to create isolated ns for %s (caller=%s)",
+          path,
+          caller_location == nullptr ? "<unknown>" : caller_location);
+    *error_msg = strdup(isolated_ns.error().message().c_str());
+    return nullptr;
+  }
+
+  *needs_native_bridge = isolated_ns.value()->IsBridged();
+  Result<void*> handle = isolated_ns.value()->Load(path);
+  ALOGD("Load %s using isolated ns %s (caller=%s): %s",
+        path,
+        isolated_ns.value()->name().c_str(),
+        caller_location == nullptr ? "<unknown>" : caller_location,
+        handle.ok() ? "ok" : handle.error().message().c_str());
+  if (!handle.ok()) {
+    *error_msg = strdup(handle.error().message().c_str());
+    return nullptr;
+  }
+  return handle.value();
+
+#else   // !ART_TARGET_ANDROID
   UNUSED(env, target_sdk_version, class_loader, caller_location);
 
   // Do some best effort to emulate library-path support. It will not
   // work for dependencies.
   //
   // Note: null has a special meaning and must be preserved.
-  std::string c_library_path;  // Empty string by default.
-  if (library_path != nullptr && path != nullptr && path[0] != '/') {
-    ScopedUtfChars library_path_utf_chars(env, library_path);
-    c_library_path = library_path_utf_chars.c_str();
+  std::string library_path;  // Empty string by default.
+  if (library_path_j != nullptr && path != nullptr && path[0] != '/') {
+    ScopedUtfChars library_path_utf_chars(env, library_path_j);
+    library_path = library_path_utf_chars.c_str();
   }
 
-  std::vector<std::string> library_paths = base::Split(c_library_path, ":");
+  std::vector<std::string> library_paths = base::Split(library_path, ":");
 
   for (const std::string& lib_path : library_paths) {
     *needs_native_bridge = false;
@@ -323,7 +411,7 @@ void* OpenNativeLibrary(JNIEnv* env, int32_t target_sdk_version, const char* pat
     }
   }
   return nullptr;
-#endif
+#endif  // !ART_TARGET_ANDROID
 }
 
 bool CloseNativeLibrary(void* handle, const bool needs_native_bridge, char** error_msg) {
@@ -351,7 +439,7 @@ void NativeLoaderFreeErrorMessage(char* msg) {
 #if defined(ART_TARGET_ANDROID)
 void* OpenNativeLibraryInNamespace(NativeLoaderNamespace* ns, const char* path,
                                    bool* needs_native_bridge, char** error_msg) {
-  auto handle = ns->Load(path);
+  Result<void*> handle = ns->Load(path);
   if (!handle.ok() && error_msg != nullptr) {
     *error_msg = strdup(handle.error().message().c_str());
   }
@@ -397,4 +485,4 @@ void LinkNativeLoaderNamespaceToExportedNamespaceLibrary(struct NativeLoaderName
 
 #endif  // ART_TARGET_ANDROID
 
-};  // namespace android
+}  // namespace android
diff --git a/libnativeloader/native_loader_lazy.cpp b/libnativeloader/native_loader_lazy.cpp
index 5b82d00560..1c82dc428f 100644
--- a/libnativeloader/native_loader_lazy.cpp
+++ b/libnativeloader/native_loader_lazy.cpp
@@ -35,7 +35,7 @@ void* GetLibHandle() {
 
 template <typename FuncPtr>
 FuncPtr GetFuncPtr(const char* function_name) {
-  auto f = reinterpret_cast<FuncPtr>(dlsym(GetLibHandle(), function_name));
+  FuncPtr f = reinterpret_cast<FuncPtr>(dlsym(GetLibHandle(), function_name));
   LOG_FATAL_IF(f == nullptr, "Failed to get address of %s: %s", function_name, dlerror());
   return f;
 }
diff --git a/libnativeloader/native_loader_namespace.cpp b/libnativeloader/native_loader_namespace.cpp
index 669fa74dc2..cfb84b7d9e 100644
--- a/libnativeloader/native_loader_namespace.cpp
+++ b/libnativeloader/native_loader_namespace.cpp
@@ -52,12 +52,12 @@ std::string GetLinkerError(bool is_bridged) {
 Result<NativeLoaderNamespace> NativeLoaderNamespace::GetExportedNamespace(const std::string& name,
                                                                           bool is_bridged) {
   if (!is_bridged) {
-    auto raw = android_get_exported_namespace(name.c_str());
+    android_namespace_t* raw = android_get_exported_namespace(name.c_str());
     if (raw != nullptr) {
       return NativeLoaderNamespace(name, raw);
     }
   } else {
-    auto raw = NativeBridgeGetExportedNamespace(name.c_str());
+    native_bridge_namespace_t* raw = NativeBridgeGetExportedNamespace(name.c_str());
     if (raw != nullptr) {
       return NativeLoaderNamespace(name, raw);
     }
@@ -69,7 +69,7 @@ Result<NativeLoaderNamespace> NativeLoaderNamespace::GetExportedNamespace(const
 // "system" for those in the Runtime APEX. Try "system" first since
 // "default" always exists.
 Result<NativeLoaderNamespace> NativeLoaderNamespace::GetSystemNamespace(bool is_bridged) {
-  auto ns = GetExportedNamespace(kSystemNamespaceName, is_bridged);
+  Result<NativeLoaderNamespace> ns = GetExportedNamespace(kSystemNamespaceName, is_bridged);
   if (ns.ok()) return ns;
   ns = GetExportedNamespace(kDefaultNamespaceName, is_bridged);
   if (ns.ok()) return ns;
@@ -96,7 +96,7 @@ Result<NativeLoaderNamespace> NativeLoaderNamespace::Create(
   }
 
   // Fall back to the system namespace if no parent is set.
-  auto system_ns = GetSystemNamespace(is_bridged);
+  Result<NativeLoaderNamespace> system_ns = GetSystemNamespace(is_bridged);
   if (!system_ns.ok()) {
     return system_ns.error();
   }
diff --git a/libnativeloader/native_loader_test.cpp b/libnativeloader/native_loader_test.cpp
index 72348ed364..3b05aae06b 100644
--- a/libnativeloader/native_loader_test.cpp
+++ b/libnativeloader/native_loader_test.cpp
@@ -564,7 +564,8 @@ jni com_android_bar libbar.so:libbar2.so
   public com_android_bar libpublic.so
 )";
 
-  auto jni_libs = ParseApexLibrariesConfig(file_content, "jni");
+  Result<std::map<std::string, std::string>> jni_libs =
+      ParseApexLibrariesConfig(file_content, "jni");
   ASSERT_RESULT_OK(jni_libs);
   std::map<std::string, std::string> expected_jni_libs {
     {"com_android_foo", "libfoo.so"},
@@ -572,7 +573,8 @@ jni com_android_bar libbar.so:libbar2.so
   };
   ASSERT_EQ(expected_jni_libs, *jni_libs);
 
-  auto public_libs = ParseApexLibrariesConfig(file_content, "public");
+  Result<std::map<std::string, std::string>> public_libs =
+      ParseApexLibrariesConfig(file_content, "public");
   ASSERT_RESULT_OK(public_libs);
   std::map<std::string, std::string> expected_public_libs {
     {"com_android_bar", "libpublic.so"},
@@ -586,7 +588,7 @@ jni com_android_foo libfoo
 # missing <library list>
 jni com_android_bar
 )";
-  auto result = ParseApexLibrariesConfig(file_content, "jni");
+  Result<std::map<std::string, std::string>> result = ParseApexLibrariesConfig(file_content, "jni");
   ASSERT_FALSE(result.ok());
   ASSERT_EQ("Malformed line \"jni com_android_bar\"", result.error().message());
 }
@@ -598,7 +600,7 @@ public apex2 lib
 # unknown tag
 unknown com_android_foo libfoo
 )";
-  auto result = ParseApexLibrariesConfig(file_content, "jni");
+  Result<std::map<std::string, std::string>> result = ParseApexLibrariesConfig(file_content, "jni");
   ASSERT_FALSE(result.ok());
   ASSERT_EQ("Invalid tag \"unknown com_android_foo libfoo\"", result.error().message());
 }
@@ -608,7 +610,7 @@ TEST(NativeLoaderApexLibrariesConfigParser, RejectInvalidApexNamespace) {
 # apex linker namespace should be mangled ('.' -> '_')
 jni com.android.foo lib
 )";
-  auto result = ParseApexLibrariesConfig(file_content, "jni");
+  Result<std::map<std::string, std::string>> result = ParseApexLibrariesConfig(file_content, "jni");
   ASSERT_FALSE(result.ok());
   ASSERT_EQ("Invalid apex_namespace \"jni com.android.foo lib\"", result.error().message());
 }
@@ -618,7 +620,7 @@ TEST(NativeLoaderApexLibrariesConfigParser, RejectInvalidLibraryList) {
 # library list is ":" separated list of filenames
 jni com_android_foo lib64/libfoo.so
 )";
-  auto result = ParseApexLibrariesConfig(file_content, "jni");
+  Result<std::map<std::string, std::string>> result = ParseApexLibrariesConfig(file_content, "jni");
   ASSERT_FALSE(result.ok());
   ASSERT_EQ("Invalid library_list \"jni com_android_foo lib64/libfoo.so\"", result.error().message());
 }
diff --git a/libnativeloader/public_libraries.cpp b/libnativeloader/public_libraries.cpp
index 87210c8f14..390c2987d6 100644
--- a/libnativeloader/public_libraries.cpp
+++ b/libnativeloader/public_libraries.cpp
@@ -78,7 +78,7 @@ std::string vndk_version_str(bool use_product_vndk) {
 // insert vndk version in every {} placeholder
 void InsertVndkVersionStr(std::string* file_name, bool use_product_vndk) {
   CHECK(file_name != nullptr);
-  auto version = vndk_version_str(use_product_vndk);
+  const std::string version = vndk_version_str(use_product_vndk);
   size_t pos = file_name->find("{}");
   while (pos != std::string::npos) {
     file_name->replace(pos, 2, version);
@@ -94,7 +94,7 @@ Result<std::vector<std::string>> ReadConfig(
     const std::function<Result<bool>(const ConfigEntry& /* entry */)>& filter_fn) {
   std::string file_content;
   if (!base::ReadFileToString(configFile, &file_content)) {
-    return ErrnoError();
+    return ErrnoError() << "Failed to read " << configFile;
   }
   Result<std::vector<std::string>> result = ParseConfig(file_content, filter_fn);
   if (!result.ok()) {
@@ -123,7 +123,7 @@ void ReadExtensionLibraries(const char* dirname, std::vector<std::string>* sonam
             "Error extracting company name from public native library list file path \"%s\"",
             config_file_path.c_str());
 
-        auto ret = ReadConfig(
+        Result<std::vector<std::string>> ret = ReadConfig(
             config_file_path, [&company_name](const struct ConfigEntry& entry) -> Result<bool> {
               if (android::base::StartsWith(entry.soname, "lib") &&
                   android::base::EndsWith(entry.soname, "." + company_name + ".so")) {
@@ -139,8 +139,8 @@ void ReadExtensionLibraries(const char* dirname, std::vector<std::string>* sonam
         if (ret.ok()) {
           sonames->insert(sonames->end(), ret->begin(), ret->end());
         } else {
-          LOG_ALWAYS_FATAL("Error reading public native library list from \"%s\": %s",
-                           config_file_path.c_str(), ret.error().message().c_str());
+          LOG_ALWAYS_FATAL("Error reading extension library list: %s",
+                           ret.error().message().c_str());
         }
       }
     }
@@ -149,7 +149,7 @@ void ReadExtensionLibraries(const char* dirname, std::vector<std::string>* sonam
 
 static std::string InitDefaultPublicLibraries(bool for_preload) {
   std::string config_file = root_dir() + kDefaultPublicLibrariesFile;
-  auto sonames =
+  Result<std::vector<std::string>> sonames =
       ReadConfig(config_file, [&for_preload](const struct ConfigEntry& entry) -> Result<bool> {
         if (for_preload) {
           return !entry.nopreload;
@@ -158,8 +158,7 @@ static std::string InitDefaultPublicLibraries(bool for_preload) {
         }
       });
   if (!sonames.ok()) {
-    LOG_ALWAYS_FATAL("Error reading public native library list from \"%s\": %s",
-                     config_file.c_str(), sonames.error().message().c_str());
+    LOG_ALWAYS_FATAL("%s", sonames.error().message().c_str());
     return "";
   }
 
@@ -167,8 +166,8 @@ static std::string InitDefaultPublicLibraries(bool for_preload) {
   if (!for_preload) {
     // Remove the public libs provided by apexes because these libs are available
     // from apex namespaces.
-    for (const auto& p : apex_public_libraries()) {
-      auto public_libs = base::Split(p.second, ":");
+    for (const std::pair<std::string, std::string>& p : apex_public_libraries()) {
+      std::vector<std::string> public_libs = base::Split(p.second, ":");
       sonames->erase(std::remove_if(sonames->begin(),
                                     sonames->end(),
                                     [&public_libs](const std::string& v) {
@@ -186,7 +185,7 @@ static std::string InitDefaultPublicLibraries(bool for_preload) {
 
 static std::string InitVendorPublicLibraries() {
   // This file is optional, quietly ignore if the file does not exist.
-  auto sonames = ReadConfig(kVendorPublicLibrariesFile, always_true);
+  Result<std::vector<std::string>> sonames = ReadConfig(kVendorPublicLibrariesFile, always_true);
   if (!sonames.ok()) {
     ALOGI("InitVendorPublicLibraries skipped: %s", sonames.error().message().c_str());
     return "";
@@ -250,9 +249,9 @@ static std::string InitLlndkLibrariesVendor() {
   } else {
     config_file = kLlndkLibrariesNoVndkFile;
   }
-  auto sonames = ReadConfig(config_file, always_true);
+  Result<std::vector<std::string>> sonames = ReadConfig(config_file, always_true);
   if (!sonames.ok()) {
-    LOG_ALWAYS_FATAL("%s: %s", config_file.c_str(), sonames.error().message().c_str());
+    LOG_ALWAYS_FATAL("%s", sonames.error().message().c_str());
     return "";
   }
   std::string libs = android::base::Join(*sonames, ':');
@@ -272,9 +271,9 @@ static std::string InitLlndkLibrariesProduct() {
   } else {
     config_file = kLlndkLibrariesNoVndkFile;
   }
-  auto sonames = ReadConfig(config_file, always_true);
+  Result<std::vector<std::string>> sonames = ReadConfig(config_file, always_true);
   if (!sonames.ok()) {
-    LOG_ALWAYS_FATAL("%s: %s", config_file.c_str(), sonames.error().message().c_str());
+    LOG_ALWAYS_FATAL("%s", sonames.error().message().c_str());
     return "";
   }
   std::string libs = android::base::Join(*sonames, ':');
@@ -290,7 +289,7 @@ static std::string InitVndkspLibrariesVendor() {
 
   std::string config_file = kVndkLibrariesFile;
   InsertVndkVersionStr(&config_file, false);
-  auto sonames = ReadConfig(config_file, always_true);
+  Result<std::vector<std::string>> sonames = ReadConfig(config_file, always_true);
   if (!sonames.ok()) {
     LOG_ALWAYS_FATAL("%s", sonames.error().message().c_str());
     return "";
@@ -307,7 +306,7 @@ static std::string InitVndkspLibrariesProduct() {
   }
   std::string config_file = kVndkLibrariesFile;
   InsertVndkVersionStr(&config_file, true);
-  auto sonames = ReadConfig(config_file, always_true);
+  Result<std::vector<std::string>> sonames = ReadConfig(config_file, always_true);
   if (!sonames.ok()) {
     LOG_ALWAYS_FATAL("%s", sonames.error().message().c_str());
     return "";
@@ -460,8 +459,8 @@ Result<std::vector<std::string>> ParseConfig(
   std::vector<std::string> lines = base::Split(file_content, "\n");
 
   std::vector<std::string> sonames;
-  for (auto& line : lines) {
-    auto trimmed_line = base::Trim(line);
+  for (std::string& line : lines) {
+    std::string trimmed_line = base::Trim(line);
     if (trimmed_line[0] == '#' || trimmed_line.empty()) {
       continue;
     }
@@ -528,12 +527,12 @@ Result<std::vector<std::string>> ParseConfig(
 Result<std::map<std::string, std::string>> ParseApexLibrariesConfig(const std::string& file_content, const std::string& tag) {
   std::map<std::string, std::string> entries;
   std::vector<std::string> lines = base::Split(file_content, "\n");
-  for (auto& line : lines) {
-    auto trimmed_line = base::Trim(line);
+  for (std::string& line : lines) {
+    std::string trimmed_line = base::Trim(line);
     if (trimmed_line[0] == '#' || trimmed_line.empty()) {
       continue;
     }
-    auto config_line = ParseApexLibrariesConfigLine(trimmed_line);
+    Result<ApexLibrariesConfigLine> config_line = ParseApexLibrariesConfigLine(trimmed_line);
     if (!config_line.ok()) {
       return config_line.error();
     }
diff --git a/libnativeloader/test/Android.bp b/libnativeloader/test/Android.bp
index 68c9126407..6166725af7 100644
--- a/libnativeloader/test/Android.bp
+++ b/libnativeloader/test/Android.bp
@@ -66,16 +66,6 @@ android_test_helper_app {
     ],
 }
 
-java_library {
-    name: "loadlibrarytest_test_utils",
-    sdk_version: "31",
-    static_libs: [
-        "androidx.test.ext.junit",
-        "androidx.test.ext.truth",
-    ],
-    srcs: ["src/android/test/lib/TestUtils.java"],
-}
-
 // Test fixture that represents a shared library in /system/framework.
 java_library {
     name: "libnativeloader_system_shared_lib",
@@ -110,6 +100,24 @@ java_library {
     srcs: ["src/android/test/vendorsharedlib/VendorSharedLib.java"],
 }
 
+java_library {
+    name: "loadlibrarytest_testlib",
+    sdk_version: "system_31",
+    static_libs: [
+        "androidx.test.ext.junit",
+        "androidx.test.ext.truth",
+        "androidx.test.rules",
+        "modules-utils-build_system",
+    ],
+    libs: [
+        "libnativeloader_system_shared_lib",
+        "libnativeloader_system_ext_shared_lib",
+        "libnativeloader_product_shared_lib",
+        "libnativeloader_vendor_shared_lib",
+    ],
+    srcs: ["src/android/test/lib/*.java"],
+}
+
 java_defaults {
     name: "loadlibrarytest_app_defaults",
     defaults: ["art_module_source_build_java_defaults"],
@@ -120,9 +128,7 @@ java_defaults {
     // measures getting enabled in these tests, so set some high number.
     target_sdk_version: "9999",
     static_libs: [
-        "androidx.test.ext.junit",
-        "androidx.test.rules",
-        "loadlibrarytest_test_utils",
+        "loadlibrarytest_testlib",
     ],
     libs: [
         "libnativeloader_system_shared_lib",
diff --git a/libnativeloader/test/src/android/test/app/DataAppTest.java b/libnativeloader/test/src/android/test/app/DataAppTest.java
index 4ba8afcbd2..905b5bb0d1 100644
--- a/libnativeloader/test/src/android/test/app/DataAppTest.java
+++ b/libnativeloader/test/src/android/test/app/DataAppTest.java
@@ -16,25 +16,30 @@
 
 package android.test.app;
 
+import android.test.lib.AppTestCommon;
 import android.test.lib.TestUtils;
 import android.test.productsharedlib.ProductSharedLib;
 import android.test.systemextsharedlib.SystemExtSharedLib;
 import android.test.systemsharedlib.SystemSharedLib;
 import android.test.vendorsharedlib.VendorSharedLib;
+
 import androidx.test.filters.MediumTest;
-import androidx.test.runner.AndroidJUnit4;
+
 import org.junit.Test;
-import org.junit.runner.RunWith;
 
 @MediumTest
-@RunWith(AndroidJUnit4.class)
-public class DataAppTest {
+public class DataAppTest extends AppTestCommon {
+    @Override
+    public AppLocation getAppLocation() {
+        return AppLocation.DATA;
+    }
+
     @Test
     public void testLoadExtendedPublicLibraries() {
         System.loadLibrary("system_extpub.oem1");
         System.loadLibrary("system_extpub.oem2");
         System.loadLibrary("system_extpub1.oem1");
-        TestUtils.assertLinkerNamespaceError( // Missing <uses-native-library>.
+        TestUtils.assertLibraryInaccessible( // Missing <uses-native-library>.
                 () -> System.loadLibrary("system_extpub_nouses.oem2"));
         if (!TestUtils.skipPublicProductLibTests()) {
             System.loadLibrary("product_extpub.product1");
@@ -44,12 +49,12 @@ public class DataAppTest {
 
     @Test
     public void testLoadPrivateLibraries() {
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_private1"));
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("systemext_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("systemext_private1"));
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> System.loadLibrary("product_private1"));
+            TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("product_private1"));
         }
-        TestUtils.assertLibraryNotFound(() -> System.loadLibrary("vendor_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("vendor_private1"));
     }
 
     @Test
@@ -64,46 +69,60 @@ public class DataAppTest {
     public void testLoadPrivateLibrariesViaSystemSharedLib() {
         // TODO(b/237577392): Loading a private native system library via a shared system library
         // ought to work.
-        // SystemSharedLib.loadLibrary("system_private2");
-        // SystemSharedLib.loadLibrary("systemext_private2");
+        TestUtils.assertLibraryInaccessible(() -> SystemSharedLib.loadLibrary("system_private2"));
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemSharedLib.loadLibrary("systemext_private2"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("product_private2"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemSharedLib.loadLibrary("product_private2"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("vendor_private2"));
+
+        TestUtils.assertLibraryInaccessible(() -> SystemSharedLib.loadLibrary("vendor_private2"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaSystemExtSharedLib() {
         // TODO(b/237577392): Loading a private native system library via a shared system library
         // ought to work.
-        // SystemExtSharedLib.loadLibrary("system_private3");
-        // SystemExtSharedLib.loadLibrary("systemext_private3");
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("system_private3"));
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("systemext_private3"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(
+            TestUtils.assertLibraryInaccessible(
                     () -> SystemExtSharedLib.loadLibrary("product_private3"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemExtSharedLib.loadLibrary("vendor_private3"));
+
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("vendor_private3"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaProductSharedLib() {
-        TestUtils.assertLinkerNamespaceError(() -> ProductSharedLib.loadLibrary("system_private4"));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(() -> ProductSharedLib.loadLibrary("system_private4"));
+        TestUtils.assertLibraryInaccessible(
                 () -> ProductSharedLib.loadLibrary("systemext_private4"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
             ProductSharedLib.loadLibrary("product_private4");
         }
-        TestUtils.assertLibraryNotFound(() -> ProductSharedLib.loadLibrary("vendor_private4"));
+
+        TestUtils.assertLibraryInaccessible(() -> ProductSharedLib.loadLibrary("vendor_private4"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaVendorSharedLib() {
-        TestUtils.assertLinkerNamespaceError(() -> VendorSharedLib.loadLibrary("system_private5"));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(() -> VendorSharedLib.loadLibrary("system_private5"));
+        TestUtils.assertLibraryInaccessible(
                 () -> VendorSharedLib.loadLibrary("systemext_private5"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> VendorSharedLib.loadLibrary("product_private5"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> VendorSharedLib.loadLibrary("product_private5"));
         }
+
         VendorSharedLib.loadLibrary("vendor_private5");
     }
 
@@ -117,15 +136,15 @@ public class DataAppTest {
 
     @Test
     public void testLoadPrivateLibrariesWithAbsolutePaths() {
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/system", "system_private6")));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/system_ext", "systemext_private6")));
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLinkerNamespaceError(
+            TestUtils.assertLibraryInaccessible(
                     () -> System.load(TestUtils.libPath("/product", "product_private6")));
         }
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/vendor", "vendor_private6")));
     }
 }
diff --git a/libnativeloader/test/src/android/test/app/ProductAppTest.java b/libnativeloader/test/src/android/test/app/ProductAppTest.java
index 82d8b6e1f1..4cf379c9c6 100644
--- a/libnativeloader/test/src/android/test/app/ProductAppTest.java
+++ b/libnativeloader/test/src/android/test/app/ProductAppTest.java
@@ -16,8 +16,7 @@
 
 package android.test.app;
 
-import android.os.Build;
-import android.os.SystemProperties;
+import android.test.lib.AppTestCommon;
 import android.test.lib.TestUtils;
 import android.test.productsharedlib.ProductSharedLib;
 import android.test.systemextsharedlib.SystemExtSharedLib;
@@ -25,19 +24,19 @@ import android.test.systemsharedlib.SystemSharedLib;
 import android.test.vendorsharedlib.VendorSharedLib;
 
 import androidx.test.filters.MediumTest;
-import androidx.test.runner.AndroidJUnit4;
 
 import org.junit.Test;
-import org.junit.runner.RunWith;
 
 @MediumTest
-@RunWith(AndroidJUnit4.class)
-public class ProductAppTest {
-    // True if apps in product partitions get shared library namespaces, so we
-    // cannot test that libs in system and system_ext get blocked.
-    private static boolean productAppsAreShared() {
-        return Build.VERSION.SDK_INT <= 34 && // UPSIDE_DOWN_CAKE
-                SystemProperties.get("ro.product.vndk.version").isEmpty();
+public class ProductAppTest extends AppTestCommon {
+    @Override
+    public AppLocation getAppLocation() {
+        return AppLocation.PRODUCT;
+    }
+
+    @Test
+    public void testPrivateLibsExist() {
+        TestUtils.testPrivateLibsExist("/product", "product_private");
     }
 
     @Test
@@ -45,8 +44,8 @@ public class ProductAppTest {
         System.loadLibrary("system_extpub.oem1");
         System.loadLibrary("system_extpub.oem2");
         System.loadLibrary("system_extpub1.oem1");
-        if (!productAppsAreShared()) {
-            TestUtils.assertLinkerNamespaceError( // Missing <uses-native-library>.
+        if (!TestUtils.productAppsAreShared()) {
+            TestUtils.assertLibraryInaccessible( // Missing <uses-native-library>.
                     () -> System.loadLibrary("system_extpub_nouses.oem2"));
         }
         System.loadLibrary("product_extpub.product1");
@@ -55,12 +54,12 @@ public class ProductAppTest {
 
     @Test
     public void testLoadPrivateLibraries() {
-        if (!productAppsAreShared()) {
-            TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_private1"));
-            TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("systemext_private1"));
+        if (!TestUtils.productAppsAreShared()) {
+            TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_private1"));
+            TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("systemext_private1"));
         }
         System.loadLibrary("product_private1");
-        TestUtils.assertLibraryNotFound(() -> System.loadLibrary("vendor_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("vendor_private1"));
     }
 
     @Test
@@ -73,49 +72,65 @@ public class ProductAppTest {
 
     @Test
     public void testLoadPrivateLibrariesViaSystemSharedLib() {
-        // TODO(b/237577392): Loading a private native system library via a shared system library
-        // ought to work.
-        // SystemSharedLib.loadLibrary("system_private2");
-        // SystemSharedLib.loadLibrary("systemext_private2");
-        if (!productAppsAreShared()) {
-            TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("product_private2"));
+        if (!TestUtils.productAppsAreShared()) {
+            // TODO(b/237577392): Loading a private native system library via a shared system
+            // library ought to work.
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemSharedLib.loadLibrary("system_private2"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemSharedLib.loadLibrary("systemext_private2"));
+
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemSharedLib.loadLibrary("product_private2"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("vendor_private2"));
+
+        TestUtils.assertLibraryInaccessible(() -> SystemSharedLib.loadLibrary("vendor_private2"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaSystemExtSharedLib() {
-        // TODO(b/237577392): Loading a private native system library via a shared system library
-        // ought to work.
-        // SystemExtSharedLib.loadLibrary("system_private3");
-        // SystemExtSharedLib.loadLibrary("systemext_private3");
-        if (!productAppsAreShared()) {
-            TestUtils.assertLibraryNotFound(
+        if (!TestUtils.productAppsAreShared()) {
+            // TODO(b/237577392): Loading a private native system library via a shared system
+            // library ought to work.
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemExtSharedLib.loadLibrary("system_private3"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemExtSharedLib.loadLibrary("systemext_private3"));
+
+            TestUtils.assertLibraryInaccessible(
                     () -> SystemExtSharedLib.loadLibrary("product_private3"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemExtSharedLib.loadLibrary("vendor_private3"));
+
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("vendor_private3"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaProductSharedLib() {
-        if (!productAppsAreShared()) {
-            TestUtils.assertLinkerNamespaceError(
+        if (!TestUtils.productAppsAreShared()) {
+            TestUtils.assertLibraryInaccessible(
                     () -> ProductSharedLib.loadLibrary("system_private4"));
-            TestUtils.assertLinkerNamespaceError(
+            TestUtils.assertLibraryInaccessible(
                     () -> ProductSharedLib.loadLibrary("systemext_private4"));
         }
+
+        // Can load product_private4 by name only through the app classloader namespace.
         ProductSharedLib.loadLibrary("product_private4");
-        TestUtils.assertLibraryNotFound(() -> ProductSharedLib.loadLibrary("vendor_private4"));
+
+        TestUtils.assertLibraryInaccessible(() -> ProductSharedLib.loadLibrary("vendor_private4"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaVendorSharedLib() {
-        if (!productAppsAreShared()) {
-            TestUtils.assertLinkerNamespaceError(
+        if (!TestUtils.productAppsAreShared()) {
+            TestUtils.assertLibraryInaccessible(
                     () -> VendorSharedLib.loadLibrary("system_private5"));
-            TestUtils.assertLinkerNamespaceError(
+            TestUtils.assertLibraryInaccessible(
                     () -> VendorSharedLib.loadLibrary("systemext_private5"));
-            TestUtils.assertLibraryNotFound(() -> VendorSharedLib.loadLibrary("product_private5"));
+
+            TestUtils.assertLibraryInaccessible(
+                    () -> VendorSharedLib.loadLibrary("product_private5"));
+
             // When the app has a shared namespace, its libraries get loaded
             // with shared namespaces as well, inheriting the same paths. So
             // since the app wouldn't have access to /vendor/${LIB},
@@ -132,14 +147,14 @@ public class ProductAppTest {
 
     @Test
     public void testLoadPrivateLibrariesWithAbsolutePaths() {
-        if (!productAppsAreShared()) {
-            TestUtils.assertLinkerNamespaceError(
+        if (!TestUtils.productAppsAreShared()) {
+            TestUtils.assertLibraryInaccessible(
                     () -> System.load(TestUtils.libPath("/system", "system_private6")));
-            TestUtils.assertLinkerNamespaceError(
+            TestUtils.assertLibraryInaccessible(
                     () -> System.load(TestUtils.libPath("/system_ext", "systemext_private6")));
         }
         System.load(TestUtils.libPath("/product", "product_private6"));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/vendor", "vendor_private6")));
     }
 }
diff --git a/libnativeloader/test/src/android/test/app/SystemAppTest.java b/libnativeloader/test/src/android/test/app/SystemAppTest.java
index a909a4c49d..8d14753b18 100644
--- a/libnativeloader/test/src/android/test/app/SystemAppTest.java
+++ b/libnativeloader/test/src/android/test/app/SystemAppTest.java
@@ -16,20 +16,31 @@
 
 package android.test.app;
 
+import android.test.lib.AppTestCommon;
 import android.test.lib.TestUtils;
 import android.test.productsharedlib.ProductSharedLib;
 import android.test.systemextsharedlib.SystemExtSharedLib;
 import android.test.systemsharedlib.SystemSharedLib;
 import android.test.vendorsharedlib.VendorSharedLib;
+
 import androidx.test.filters.MediumTest;
-import androidx.test.runner.AndroidJUnit4;
+
 import org.junit.Test;
-import org.junit.runner.RunWith;
 
 // These tests are run from /system/app, /system/priv-app, and /system_ext/app.
 @MediumTest
-@RunWith(AndroidJUnit4.class)
-public class SystemAppTest {
+public class SystemAppTest extends AppTestCommon {
+    @Override
+    public AppLocation getAppLocation() {
+        return AppLocation.SYSTEM;
+    }
+
+    @Test
+    public void testPrivateLibsExist() {
+        TestUtils.testPrivateLibsExist("/system", "system_private");
+        TestUtils.testPrivateLibsExist("/system_ext", "systemext_private");
+    }
+
     @Test
     public void testLoadExtendedPublicLibraries() {
         System.loadLibrary("system_extpub.oem1");
@@ -49,9 +60,9 @@ public class SystemAppTest {
         System.loadLibrary("system_private1");
         System.loadLibrary("systemext_private1");
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> System.loadLibrary("product_private1"));
+            TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("product_private1"));
         }
-        TestUtils.assertLibraryNotFound(() -> System.loadLibrary("vendor_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("vendor_private1"));
     }
 
     @Test
@@ -66,41 +77,57 @@ public class SystemAppTest {
     public void testLoadPrivateLibrariesViaSystemSharedLib() {
         SystemSharedLib.loadLibrary("system_private2");
         SystemSharedLib.loadLibrary("systemext_private2");
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("product_private2"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemSharedLib.loadLibrary("product_private2"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("vendor_private2"));
+
+        TestUtils.assertLibraryInaccessible(() -> SystemSharedLib.loadLibrary("vendor_private2"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaSystemExtSharedLib() {
         SystemExtSharedLib.loadLibrary("system_private3");
         SystemExtSharedLib.loadLibrary("systemext_private3");
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(
+            TestUtils.assertLibraryInaccessible(
                     () -> SystemExtSharedLib.loadLibrary("product_private3"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemExtSharedLib.loadLibrary("vendor_private3"));
+
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("vendor_private3"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaProductSharedLib() {
+        // See AppTestCommon.isSharedSystemApp() for an explanation of these
+        // behaviours.
         ProductSharedLib.loadLibrary("system_private4");
         ProductSharedLib.loadLibrary("systemext_private4");
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> ProductSharedLib.loadLibrary("product_private4"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> ProductSharedLib.loadLibrary("product_private4"));
         }
-        TestUtils.assertLibraryNotFound(() -> ProductSharedLib.loadLibrary("vendor_private4"));
+
+        TestUtils.assertLibraryInaccessible(() -> ProductSharedLib.loadLibrary("vendor_private4"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaVendorSharedLib() {
+        // See AppTestCommon.isSharedSystemApp() for an explanation of these
+        // behaviours.
         VendorSharedLib.loadLibrary("system_private5");
         VendorSharedLib.loadLibrary("systemext_private5");
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> VendorSharedLib.loadLibrary("product_private5"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> VendorSharedLib.loadLibrary("product_private5"));
         }
-        TestUtils.assertLibraryNotFound(() -> VendorSharedLib.loadLibrary("vendor_private5"));
+
+        TestUtils.assertLibraryInaccessible(() -> VendorSharedLib.loadLibrary("vendor_private5"));
     }
 
     @Test
@@ -116,10 +143,10 @@ public class SystemAppTest {
         System.load(TestUtils.libPath("/system", "system_private6"));
         System.load(TestUtils.libPath("/system_ext", "systemext_private6"));
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLinkerNamespaceError(
+            TestUtils.assertLibraryInaccessible(
                     () -> System.load(TestUtils.libPath("/product", "product_private6")));
         }
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/vendor", "vendor_private6")));
     }
 }
diff --git a/libnativeloader/test/src/android/test/app/VendorAppTest.java b/libnativeloader/test/src/android/test/app/VendorAppTest.java
index 52688bbf35..377f670c74 100644
--- a/libnativeloader/test/src/android/test/app/VendorAppTest.java
+++ b/libnativeloader/test/src/android/test/app/VendorAppTest.java
@@ -16,25 +16,35 @@
 
 package android.test.app;
 
+import android.test.lib.AppTestCommon;
 import android.test.lib.TestUtils;
 import android.test.productsharedlib.ProductSharedLib;
 import android.test.systemextsharedlib.SystemExtSharedLib;
 import android.test.systemsharedlib.SystemSharedLib;
 import android.test.vendorsharedlib.VendorSharedLib;
+
 import androidx.test.filters.MediumTest;
-import androidx.test.runner.AndroidJUnit4;
+
 import org.junit.Test;
-import org.junit.runner.RunWith;
 
 @MediumTest
-@RunWith(AndroidJUnit4.class)
-public class VendorAppTest {
+public class VendorAppTest extends AppTestCommon {
+    @Override
+    public AppLocation getAppLocation() {
+        return AppLocation.VENDOR;
+    }
+
+    @Test
+    public void testPrivateLibsExist() {
+        TestUtils.testPrivateLibsExist("/vendor", "vendor_private");
+    }
+
     @Test
     public void testLoadExtendedPublicLibraries() {
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_extpub.oem1"));
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_extpub.oem2"));
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_extpub1.oem1"));
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_extpub_nouses.oem2"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_extpub.oem1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_extpub.oem2"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_extpub1.oem1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_extpub_nouses.oem2"));
         if (!TestUtils.skipPublicProductLibTests()) {
             System.loadLibrary("product_extpub.product1");
             System.loadLibrary("product_extpub1.product1");
@@ -43,10 +53,10 @@ public class VendorAppTest {
 
     @Test
     public void testLoadPrivateLibraries() {
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("system_private1"));
-        TestUtils.assertLinkerNamespaceError(() -> System.loadLibrary("systemext_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("system_private1"));
+        TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("systemext_private1"));
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> System.loadLibrary("product_private1"));
+            TestUtils.assertLibraryInaccessible(() -> System.loadLibrary("product_private1"));
         }
         System.loadLibrary("vendor_private1");
     }
@@ -63,52 +73,67 @@ public class VendorAppTest {
     public void testLoadPrivateLibrariesViaSystemSharedLib() {
         // TODO(b/237577392): Loading a private native system library via a shared system library
         // ought to work.
-        // SystemSharedLib.loadLibrary("system_private2");
-        // SystemSharedLib.loadLibrary("systemext_private2");
+        TestUtils.assertLibraryInaccessible(() -> SystemSharedLib.loadLibrary("system_private2"));
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemSharedLib.loadLibrary("systemext_private2"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("product_private2"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> SystemSharedLib.loadLibrary("product_private2"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemSharedLib.loadLibrary("vendor_private2"));
+
+        TestUtils.assertLibraryInaccessible(() -> SystemSharedLib.loadLibrary("vendor_private2"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaSystemExtSharedLib() {
         // TODO(b/237577392): Loading a private native system library via a shared system library
         // ought to work.
-        // SystemExtSharedLib.loadLibrary("system_private3");
-        // SystemExtSharedLib.loadLibrary("systemext_private3");
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("system_private3"));
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("systemext_private3"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(
+            TestUtils.assertLibraryInaccessible(
                     () -> SystemExtSharedLib.loadLibrary("product_private3"));
         }
-        TestUtils.assertLibraryNotFound(() -> SystemExtSharedLib.loadLibrary("vendor_private3"));
+
+        TestUtils.assertLibraryInaccessible(
+                () -> SystemExtSharedLib.loadLibrary("vendor_private3"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaProductSharedLib() {
-        TestUtils.assertLinkerNamespaceError(() -> ProductSharedLib.loadLibrary("system_private4"));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(() -> ProductSharedLib.loadLibrary("system_private4"));
+        TestUtils.assertLibraryInaccessible(
                 () -> ProductSharedLib.loadLibrary("systemext_private4"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
             ProductSharedLib.loadLibrary("product_private4");
         }
-        TestUtils.assertLibraryNotFound(() -> ProductSharedLib.loadLibrary("vendor_private4"));
+
+        TestUtils.assertLibraryInaccessible(() -> ProductSharedLib.loadLibrary("vendor_private4"));
     }
 
     @Test
     public void testLoadPrivateLibrariesViaVendorSharedLib() {
-        TestUtils.assertLinkerNamespaceError(() -> VendorSharedLib.loadLibrary("system_private5"));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(() -> VendorSharedLib.loadLibrary("system_private5"));
+        TestUtils.assertLibraryInaccessible(
                 () -> VendorSharedLib.loadLibrary("systemext_private5"));
+
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLibraryNotFound(() -> VendorSharedLib.loadLibrary("product_private5"));
+            TestUtils.assertLibraryInaccessible(
+                    () -> VendorSharedLib.loadLibrary("product_private5"));
         }
+
+        // Can load vendor_private5 by name only through the app classloader namespace.
         VendorSharedLib.loadLibrary("vendor_private5");
     }
 
     @Test
     public void testLoadExtendedPublicLibrariesWithAbsolutePaths() {
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/system", "system_extpub3.oem1")));
         if (!TestUtils.skipPublicProductLibTests()) {
             System.load(TestUtils.libPath("/product", "product_extpub3.product1"));
@@ -117,12 +142,12 @@ public class VendorAppTest {
 
     @Test
     public void testLoadPrivateLibrariesWithAbsolutePaths() {
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/system", "system_private6")));
-        TestUtils.assertLinkerNamespaceError(
+        TestUtils.assertLibraryInaccessible(
                 () -> System.load(TestUtils.libPath("/system_ext", "systemext_private6")));
         if (!TestUtils.skipPublicProductLibTests()) {
-            TestUtils.assertLinkerNamespaceError(
+            TestUtils.assertLibraryInaccessible(
                     () -> System.load(TestUtils.libPath("/product", "product_private6")));
         }
         System.load(TestUtils.libPath("/vendor", "vendor_private6"));
diff --git a/libnativeloader/test/src/android/test/hostside/LibnativeloaderTest.java b/libnativeloader/test/src/android/test/hostside/LibnativeloaderTest.java
index 55a6dd27b3..bcb4528ffe 100644
--- a/libnativeloader/test/src/android/test/hostside/LibnativeloaderTest.java
+++ b/libnativeloader/test/src/android/test/hostside/LibnativeloaderTest.java
@@ -278,7 +278,10 @@ public class LibnativeloaderTest extends BaseHostJUnit4Test {
         void pushPrivateLibs(ZipFile libApk) throws Exception {
             // Push the libraries once for each test. Since we cannot unload them, we need a fresh
             // never-before-loaded library in each loadLibrary call.
-            for (int i = 1; i <= 6; ++i) {
+            //
+            // Remember to update testPrivateLibsExist in TestUtils.java when
+            // the number of libraries changes.
+            for (int i = 1; i <= 10; ++i) {
                 pushNativeTestLib(libApk, "libsystem_testlib.so",
                         "/system/${LIB}/libsystem_private" + i + ".so");
                 pushNativeTestLib(libApk, "libsystem_testlib.so",
diff --git a/libnativeloader/test/src/android/test/lib/AppTestCommon.java b/libnativeloader/test/src/android/test/lib/AppTestCommon.java
new file mode 100644
index 0000000000..51f4655839
--- /dev/null
+++ b/libnativeloader/test/src/android/test/lib/AppTestCommon.java
@@ -0,0 +1,192 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package android.test.lib;
+
+import android.test.productsharedlib.ProductSharedLib;
+import android.test.systemextsharedlib.SystemExtSharedLib;
+import android.test.systemsharedlib.SystemSharedLib;
+import android.test.vendorsharedlib.VendorSharedLib;
+
+import androidx.test.runner.AndroidJUnit4;
+
+import org.junit.Test;
+import org.junit.runner.RunWith;
+
+@RunWith(AndroidJUnit4.class)
+public abstract class AppTestCommon {
+    public enum AppLocation { DATA, SYSTEM, PRODUCT, VENDOR }
+
+    public abstract AppLocation getAppLocation();
+
+    // Loading private libs using absolute paths through shared libs should
+    // normally only depend on the location of the shared lib, so these tests
+    // are shared for all apps, regardless of location.
+
+    // Returns true when system private native libs are accessible directly from
+    // the app classloader namespace.
+    private boolean systemPrivateLibsAccessibleFromAppNamespace() {
+        // Currently it only works from system apps. It also works from product
+        // apps on old versions where they were treated like system apps.
+        // TODO(b/237577392): Fix this to work from system shared libs.
+        return getAppLocation() == AppLocation.SYSTEM
+                || (getAppLocation() == AppLocation.PRODUCT && TestUtils.productAppsAreShared());
+    }
+
+    // Detect exception when product private libs are accessible directly from
+    // the app classloader namespace even when they shouldn't be.
+    private boolean productPrivateLibsAccessibleFromAppNamespace() {
+        // In old versions where product apps were treated like system apps, the
+        // product private libs were included in the system namespace, so
+        // they're accessible both from system and product apps.
+        // TODO(b/237577392): Fix this to work from product shared libs.
+        return (getAppLocation() == AppLocation.SYSTEM || getAppLocation() == AppLocation.PRODUCT)
+                && TestUtils.productAppsAreShared();
+    }
+
+    // Detect exception where we don't switch from a shared system namespace to
+    // a product or vendor "unbundled" namespace when calling into
+    // ProductSharedLib and VendorSharedLib. That means they still can load
+    // private system libs but not private libs in their own partition.
+    // TODO(mast): Stop propagating the shared property (isBundledApp in
+    // LoadedApk.java) down to public and vendor shared java libs?
+    private boolean noSwitchToVendorOrProductNamespace() {
+        // System apps get shared namespaces, and also product apps on old
+        // versions where they were treated like system apps.
+        return getAppLocation() == AppLocation.SYSTEM
+                || (getAppLocation() == AppLocation.PRODUCT && TestUtils.productAppsAreShared());
+    }
+
+    @Test
+    public void testLoadPrivateLibrariesViaSystemSharedLibWithAbsolutePaths() {
+        if (systemPrivateLibsAccessibleFromAppNamespace()) {
+            SystemSharedLib.load(TestUtils.libPath("/system", "system_private7"));
+            SystemSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private7"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                SystemSharedLib.load(TestUtils.libPath("/system", "system_private7"));
+            });
+            TestUtils.assertLibraryInaccessible(() -> {
+                SystemSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private7"));
+            });
+        }
+
+        if (productPrivateLibsAccessibleFromAppNamespace()) {
+            SystemSharedLib.load(TestUtils.libPath("/product", "product_private7"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                SystemSharedLib.load(TestUtils.libPath("/product", "product_private7"));
+            });
+        }
+
+        TestUtils.assertLibraryInaccessible(
+                () -> { SystemSharedLib.load(TestUtils.libPath("/vendor", "vendor_private7")); });
+    }
+
+    @Test
+    public void testLoadPrivateLibrariesViaSystemExtSharedLibWithAbsolutePaths() {
+        if (systemPrivateLibsAccessibleFromAppNamespace()) {
+            SystemExtSharedLib.load(TestUtils.libPath("/system", "system_private8"));
+            SystemExtSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private8"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                SystemExtSharedLib.load(TestUtils.libPath("/system", "system_private8"));
+            });
+            TestUtils.assertLibraryInaccessible(() -> {
+                SystemExtSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private8"));
+            });
+        }
+
+        if (productPrivateLibsAccessibleFromAppNamespace()) {
+            SystemExtSharedLib.load(TestUtils.libPath("/product", "product_private8"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                SystemExtSharedLib.load(TestUtils.libPath("/product", "product_private8"));
+            });
+        }
+
+        TestUtils.assertLibraryInaccessible(() -> {
+            SystemExtSharedLib.load(TestUtils.libPath("/vendor", "vendor_private8"));
+        });
+    }
+
+    @Test
+    public void testLoadPrivateLibrariesViaProductSharedLibWithAbsolutePaths() {
+        if (systemPrivateLibsAccessibleFromAppNamespace() || noSwitchToVendorOrProductNamespace()) {
+            ProductSharedLib.load(TestUtils.libPath("/system", "system_private9"));
+            ProductSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private9"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                ProductSharedLib.load(TestUtils.libPath("/system", "system_private9"));
+            });
+            TestUtils.assertLibraryInaccessible(() -> {
+                ProductSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private9"));
+            });
+        }
+
+        boolean loadPrivateProductLib;
+        if (TestUtils.productAppsAreShared()) {
+            // The library is accessible if the app is in either system or
+            // product, because both are loaded as system apps and private product
+            // libs are available for both.
+            loadPrivateProductLib = getAppLocation() == AppLocation.SYSTEM
+                    || getAppLocation() == AppLocation.PRODUCT;
+        } else {
+            loadPrivateProductLib = !noSwitchToVendorOrProductNamespace();
+        }
+        if (loadPrivateProductLib) {
+            ProductSharedLib.load(TestUtils.libPath("/product", "product_private9"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                ProductSharedLib.load(TestUtils.libPath("/product", "product_private9"));
+            });
+        }
+
+        TestUtils.assertLibraryInaccessible(
+                () -> { ProductSharedLib.load(TestUtils.libPath("/vendor", "vendor_private9")); });
+    }
+
+    @Test
+    public void testLoadPrivateLibrariesViaVendorSharedLibWithAbsolutePaths() {
+        if (systemPrivateLibsAccessibleFromAppNamespace() || noSwitchToVendorOrProductNamespace()) {
+            VendorSharedLib.load(TestUtils.libPath("/system", "system_private10"));
+            VendorSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private10"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                VendorSharedLib.load(TestUtils.libPath("/system", "system_private10"));
+            });
+            TestUtils.assertLibraryInaccessible(() -> {
+                VendorSharedLib.load(TestUtils.libPath("/system_ext", "systemext_private10"));
+            });
+        }
+
+        if (productPrivateLibsAccessibleFromAppNamespace()) {
+            VendorSharedLib.load(TestUtils.libPath("/product", "product_private10"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                VendorSharedLib.load(TestUtils.libPath("/product", "product_private10"));
+            });
+        }
+
+        if (!noSwitchToVendorOrProductNamespace()) {
+            VendorSharedLib.load(TestUtils.libPath("/vendor", "vendor_private10"));
+        } else {
+            TestUtils.assertLibraryInaccessible(() -> {
+                VendorSharedLib.load(TestUtils.libPath("/vendor", "vendor_private10"));
+            });
+        }
+    }
+}
diff --git a/libnativeloader/test/src/android/test/lib/TestUtils.java b/libnativeloader/test/src/android/test/lib/TestUtils.java
index ca64bad791..5f5cd911e8 100644
--- a/libnativeloader/test/src/android/test/lib/TestUtils.java
+++ b/libnativeloader/test/src/android/test/lib/TestUtils.java
@@ -17,25 +17,25 @@
 package android.test.lib;
 
 import static com.google.common.truth.Truth.assertThat;
+import static com.google.common.truth.Truth.assertWithMessage;
 
 import static org.junit.Assert.assertThrows;
 
-import android.os.Build;
+import android.os.SystemProperties;
 
 import androidx.test.platform.app.InstrumentationRegistry;
 
+import com.android.modules.utils.build.SdkLevel;
+
 import org.junit.function.ThrowingRunnable;
 
-public final class TestUtils {
-    public static void assertLibraryNotFound(ThrowingRunnable loadLibrary) {
-        Throwable t = assertThrows(UnsatisfiedLinkError.class, loadLibrary);
-        assertThat(t.getMessage()).containsMatch("dlopen failed: library .* not found");
-    }
+import java.io.File;
 
-    public static void assertLinkerNamespaceError(ThrowingRunnable loadLibrary) {
+public final class TestUtils {
+    public static void assertLibraryInaccessible(ThrowingRunnable loadLibrary) {
         Throwable t = assertThrows(UnsatisfiedLinkError.class, loadLibrary);
         assertThat(t.getMessage())
-                .containsMatch("dlopen failed: .* is not accessible for the namespace");
+                .containsMatch("dlopen failed: .* (not found|not accessible for the namespace)");
     }
 
     public static String libPath(String dir, String libName) {
@@ -46,6 +46,25 @@ public final class TestUtils {
     // True if we have to skip testing public libraries in the product
     // partition, which got supported in T.
     public static boolean skipPublicProductLibTests() {
-        return Build.VERSION.SDK_INT < 33; // TIRAMISU
+        return !SdkLevel.isAtLeastT();
+    }
+
+    // True if apps in product partitions get shared library namespaces, so we
+    // cannot test that libs in system and system_ext get blocked.
+    public static boolean productAppsAreShared() {
+        return !SdkLevel.isAtLeastU() && SystemProperties.get("ro.product.vndk.version").isEmpty();
+    }
+
+    // Test that private libs are present, as a safeguard so that the dlopen
+    // failures we expect in other tests aren't due to them not being there.
+    public static void testPrivateLibsExist(String libDir, String libStem) {
+        // Remember to update pushPrivateLibs in LibnativeloaderTest.java when
+        // the number of libraries changes.
+        for (int i = 1; i <= 10; ++i) {
+            String libPath = libPath(libDir, libStem + i);
+            assertWithMessage(libPath + " does not exist")
+                    .that(new File(libPath).exists())
+                    .isTrue();
+        }
     }
 }
diff --git a/libnativeloader/test/src/android/test/productsharedlib/ProductSharedLib.java b/libnativeloader/test/src/android/test/productsharedlib/ProductSharedLib.java
index a500d2a976..a8966d9825 100644
--- a/libnativeloader/test/src/android/test/productsharedlib/ProductSharedLib.java
+++ b/libnativeloader/test/src/android/test/productsharedlib/ProductSharedLib.java
@@ -17,5 +17,11 @@
 package android.test.productsharedlib;
 
 public final class ProductSharedLib {
-    public static void loadLibrary(String name) { System.loadLibrary(name); }
+    public static void loadLibrary(String name) {
+        System.loadLibrary(name);
+    }
+
+    public static void load(String path) {
+        System.load(path);
+    }
 }
diff --git a/libnativeloader/test/src/android/test/systemextsharedlib/SystemExtSharedLib.java b/libnativeloader/test/src/android/test/systemextsharedlib/SystemExtSharedLib.java
index 1240e12e55..ae9a6da053 100644
--- a/libnativeloader/test/src/android/test/systemextsharedlib/SystemExtSharedLib.java
+++ b/libnativeloader/test/src/android/test/systemextsharedlib/SystemExtSharedLib.java
@@ -17,5 +17,11 @@
 package android.test.systemextsharedlib;
 
 public final class SystemExtSharedLib {
-    public static void loadLibrary(String name) { System.loadLibrary(name); }
+    public static void loadLibrary(String name) {
+        System.loadLibrary(name);
+    }
+
+    public static void load(String path) {
+        System.load(path);
+    }
 }
diff --git a/libnativeloader/test/src/android/test/systemsharedlib/SystemSharedLib.java b/libnativeloader/test/src/android/test/systemsharedlib/SystemSharedLib.java
index 8e2af9f79c..ba8c9e2574 100644
--- a/libnativeloader/test/src/android/test/systemsharedlib/SystemSharedLib.java
+++ b/libnativeloader/test/src/android/test/systemsharedlib/SystemSharedLib.java
@@ -17,5 +17,11 @@
 package android.test.systemsharedlib;
 
 public final class SystemSharedLib {
-    public static void loadLibrary(String name) { System.loadLibrary(name); }
+    public static void loadLibrary(String name) {
+        System.loadLibrary(name);
+    }
+
+    public static void load(String path) {
+        System.load(path);
+    }
 }
diff --git a/libnativeloader/test/src/android/test/vendorsharedlib/VendorSharedLib.java b/libnativeloader/test/src/android/test/vendorsharedlib/VendorSharedLib.java
index 8859b63db4..14223d88f8 100644
--- a/libnativeloader/test/src/android/test/vendorsharedlib/VendorSharedLib.java
+++ b/libnativeloader/test/src/android/test/vendorsharedlib/VendorSharedLib.java
@@ -17,5 +17,11 @@
 package android.test.vendorsharedlib;
 
 public final class VendorSharedLib {
-    public static void loadLibrary(String name) { System.loadLibrary(name); }
+    public static void loadLibrary(String name) {
+        System.loadLibrary(name);
+    }
+
+    public static void load(String path) {
+        System.load(path);
+    }
 }
diff --git a/oatdump/oatdump.cc b/oatdump/oatdump.cc
index 69dc23ae3d..a93ad37e61 100644
--- a/oatdump/oatdump.cc
+++ b/oatdump/oatdump.cc
@@ -2109,7 +2109,7 @@ class ImageDumper {
     if (obj_class->IsArrayClass()) {
       os << StringPrintf("%p: %s length:%d\n", obj, obj_class->PrettyDescriptor().c_str(),
                          obj->AsArray()->GetLength());
-    } else if (obj->IsClass()) {
+    } else if (obj_class->IsClassClass()) {
       ObjPtr<mirror::Class> klass = obj->AsClass();
       os << StringPrintf("%p: java.lang.Class \"%s\" (",
                          obj,
@@ -2146,7 +2146,7 @@ class ImageDumper {
             (value == nullptr) ? obj_class->GetComponentType() : value->GetClass();
         PrettyObjectValue(os, value_class, value);
       }
-    } else if (obj->IsClass()) {
+    } else if (obj_class->IsClassClass()) {
       ObjPtr<mirror::Class> klass = obj->AsClass();
 
       if (kBitstringSubtypeCheckEnabled) {
@@ -2155,6 +2155,16 @@ class ImageDumper {
         os << "\n";
       }
 
+      if (klass->ShouldHaveEmbeddedVTable()) {
+        os << "EMBEDDED VTABLE:\n";
+        ScopedIndentation indent2(&vios_);
+        const PointerSize pointer_size = image_header_.GetPointerSize();
+        for (size_t i = 0, length = klass->GetEmbeddedVTableLength(); i != length; ++i) {
+          os << i << ": "
+             << ArtMethod::PrettyMethod(klass->GetEmbeddedVTableEntry(i, pointer_size)) << '\n';
+        }
+      }
+
       if (klass->NumStaticFields() != 0) {
         os << "STATICS:\n";
         ScopedIndentation indent2(&vios_);
diff --git a/oatdump/oatdump_app_test.cc b/oatdump/oatdump_app_test.cc
index 03b43cef11..d4dee4f2f6 100644
--- a/oatdump/oatdump_app_test.cc
+++ b/oatdump/oatdump_app_test.cc
@@ -20,7 +20,6 @@ namespace art {
 
 // Oat file compiled with a boot image. oatdump invoked with a boot image.
 TEST_P(OatDumpTest, TestDumpOatWithRuntimeWithBootImage) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam()));
   ASSERT_TRUE(Exec(GetParam(),
                    kArgOatApp | kArgBootImage | kArgBcp | kArgIsa,
@@ -30,7 +29,6 @@ TEST_P(OatDumpTest, TestDumpOatWithRuntimeWithBootImage) {
 
 // Oat file compiled without a boot image. oatdump invoked without a boot image.
 TEST_P(OatDumpTest, TestDumpOatWithRuntimeWithNoBootImage) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_FOR_DEBUG_BUILD();  // DCHECK failed.
   ASSERT_TRUE(GenerateAppOdexFile(GetParam(), {"--boot-image=/nonx/boot.art"}));
   ASSERT_TRUE(Exec(GetParam(),
@@ -41,7 +39,6 @@ TEST_P(OatDumpTest, TestDumpOatWithRuntimeWithNoBootImage) {
 
 // Dex code cannot be found in the vdex file, and no --dex-file is specified. Dump header only.
 TEST_P(OatDumpTest, TestDumpOatTryWithRuntimeDexNotFound) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(
       GenerateAppOdexFile(GetParam(), {"--dex-location=/nonx/app.jar", "--copy-dex-files=false"}));
   ASSERT_TRUE(Exec(GetParam(), kArgOatApp | kArgBootImage | kArgBcp | kArgIsa, {}, kExpectOat));
@@ -49,7 +46,6 @@ TEST_P(OatDumpTest, TestDumpOatTryWithRuntimeDexNotFound) {
 
 // Dex code cannot be found in the vdex file, but can be found in the specified dex file.
 TEST_P(OatDumpTest, TestDumpOatWithRuntimeDexSpecified) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(
       GenerateAppOdexFile(GetParam(), {"--dex-location=/nonx/app.jar", "--copy-dex-files=false"}));
   ASSERT_TRUE(Exec(GetParam(),
@@ -60,7 +56,6 @@ TEST_P(OatDumpTest, TestDumpOatWithRuntimeDexSpecified) {
 
 // Oat file compiled with a boot image. oatdump invoked without a boot image.
 TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeBcpMismatch) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam()));
   ASSERT_TRUE(Exec(GetParam(),
                    kArgOatApp | kArgBcp | kArgIsa,
@@ -70,14 +65,12 @@ TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeBcpMismatch) {
 
 // Bootclasspath not specified.
 TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeNoBcp) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam()));
   ASSERT_TRUE(Exec(GetParam(), kArgOatApp, {}, kExpectOat | kExpectCode | kExpectBssOffsetsForBcp));
 }
 
 // Dex code cannot be found in the vdex file, and no --dex-file is specified. Dump header only.
 TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeDexNotFound) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(
       GenerateAppOdexFile(GetParam(), {"--dex-location=/nonx/app.jar", "--copy-dex-files=false"}));
   ASSERT_TRUE(Exec(GetParam(), kArgOatApp, {}, kExpectOat));
@@ -85,7 +78,6 @@ TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeDexNotFound) {
 
 // Dex code cannot be found in the vdex file, but can be found in the specified dex file.
 TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeDexSpecified) {
-  TEST_DISABLED_FOR_RISCV64();
   ASSERT_TRUE(
       GenerateAppOdexFile(GetParam(), {"--dex-location=/nonx/app.jar", "--copy-dex-files=false"}));
   ASSERT_TRUE(Exec(
@@ -93,7 +85,6 @@ TEST_P(OatDumpTest, TestDumpOatWithoutRuntimeDexSpecified) {
 }
 
 TEST_P(OatDumpTest, TestDumpAppImageWithBootImage) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam(), {app_image_arg}));
@@ -105,7 +96,6 @@ TEST_P(OatDumpTest, TestDumpAppImageWithBootImage) {
 
 // Deprecated usage, but checked for compatibility.
 TEST_P(OatDumpTest, TestDumpAppImageWithBootImageLegacy) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam(), {app_image_arg}));
@@ -116,7 +106,6 @@ TEST_P(OatDumpTest, TestDumpAppImageWithBootImageLegacy) {
 }
 
 TEST_P(OatDumpTest, TestDumpAppImageInvalidPath) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam(), {app_image_arg}));
@@ -129,7 +118,6 @@ TEST_P(OatDumpTest, TestDumpAppImageInvalidPath) {
 
 // The runtime can start, but the boot image check should fail.
 TEST_P(OatDumpTest, TestDumpAppImageWithWrongBootImage) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam(), {app_image_arg}));
@@ -142,7 +130,6 @@ TEST_P(OatDumpTest, TestDumpAppImageWithWrongBootImage) {
 
 // Not possible.
 TEST_P(OatDumpTest, TestDumpAppImageWithoutRuntime) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(GetParam(), {app_image_arg}));
@@ -155,7 +142,6 @@ TEST_P(OatDumpTest, TestDumpAppImageWithoutRuntime) {
 
 // Dex code cannot be found in the vdex file, and no --dex-file is specified. Cannot dump app image.
 TEST_P(OatDumpTest, TestDumpAppImageDexNotFound) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(
@@ -169,7 +155,6 @@ TEST_P(OatDumpTest, TestDumpAppImageDexNotFound) {
 
 // Dex code cannot be found in the vdex file, but can be found in the specified dex file.
 TEST_P(OatDumpTest, TestDumpAppImageDexSpecified) {
-  TEST_DISABLED_FOR_RISCV64();
   TEST_DISABLED_WITHOUT_BAKER_READ_BARRIERS();  // GC bug, b/126305867
   const std::string app_image_arg = "--app-image-file=" + GetAppImageName();
   ASSERT_TRUE(GenerateAppOdexFile(
diff --git a/oatdump/oatdump_test.cc b/oatdump/oatdump_test.cc
index 88cd044a48..6cd3a9742a 100644
--- a/oatdump/oatdump_test.cc
+++ b/oatdump/oatdump_test.cc
@@ -66,10 +66,10 @@ TEST_P(OatDumpTest, TestListMethods) {
 }
 
 TEST_P(OatDumpTest, TestSymbolize) {
-  TEST_DISABLED_FOR_RISCV64();
   if (GetParam() == Flavor::kDynamic) {
     TEST_DISABLED_FOR_TARGET();  // Can not write files inside the apex directory.
   } else {
+    TEST_DISABLED_FOR_RISCV64();
     TEST_DISABLED_FOR_ARM_AND_ARM64();
   }
   std::string error_msg;
@@ -77,8 +77,8 @@ TEST_P(OatDumpTest, TestSymbolize) {
 }
 
 TEST_P(OatDumpTest, TestExportDex) {
-  TEST_DISABLED_FOR_RISCV64();
   if (GetParam() == Flavor::kStatic) {
+    TEST_DISABLED_FOR_RISCV64();
     TEST_DISABLED_FOR_ARM_AND_ARM64();
   }
   std::string error_msg;
diff --git a/odrefresh/Android.bp b/odrefresh/Android.bp
index 2a020da634..7b9aad7913 100644
--- a/odrefresh/Android.bp
+++ b/odrefresh/Android.bp
@@ -41,10 +41,10 @@ cc_defaults {
         "art-odrefresh-operator-srcs",
     ],
     shared_libs: [
-        "libartpalette",
         "libarttools", // Contains "libc++fs".
         "libbase",
         "liblog",
+        "libselinux",
     ],
     static_libs: [
         "libmodules-utils-build",
diff --git a/odrefresh/odrefresh.cc b/odrefresh/odrefresh.cc
index ea624b21bb..f515aefa10 100644
--- a/odrefresh/odrefresh.cc
+++ b/odrefresh/odrefresh.cc
@@ -85,8 +85,7 @@
 #include "odr_fs_utils.h"
 #include "odr_metrics.h"
 #include "odrefresh/odrefresh.h"
-#include "palette/palette.h"
-#include "palette/palette_types.h"
+#include "selinux/selinux.h"
 #include "tools/cmdline_builder.h"
 
 namespace art {
@@ -194,6 +193,28 @@ bool MoveOrEraseFiles(const std::vector<std::unique_ptr<File>>& files,
   return true;
 }
 
+Result<std::string> CreateStagingDirectory() {
+  std::string staging_dir = GetArtApexData() + "/staging";
+
+  std::error_code ec;
+  if (std::filesystem::exists(staging_dir, ec)) {
+    if (!std::filesystem::remove_all(staging_dir, ec)) {
+      return Errorf(
+          "Could not remove existing staging directory '{}': {}", staging_dir, ec.message());
+    }
+  }
+
+  if (mkdir(staging_dir.c_str(), S_IRWXU) != 0) {
+    return ErrnoErrorf("Could not create staging directory '{}'", staging_dir);
+  }
+
+  if (setfilecon(staging_dir.c_str(), "u:object_r:apex_art_staging_data_file:s0") != 0) {
+    return ErrnoErrorf("Could not set label on staging directory '{}'", staging_dir);
+  }
+
+  return staging_dir;
+}
+
 // Gets the `ApexInfo` associated with the currently active ART APEX.
 std::optional<apex::ApexInfo> GetArtApexInfo(const std::vector<apex::ApexInfo>& info_list) {
   auto it = std::find_if(info_list.begin(), info_list.end(), [](const apex::ApexInfo& info) {
@@ -2045,7 +2066,7 @@ OnDeviceRefresh::CompileSystemServer(const std::string& staging_dir,
 
 WARN_UNUSED ExitCode OnDeviceRefresh::Compile(OdrMetrics& metrics,
                                               CompilationOptions compilation_options) const {
-  const char* staging_dir = nullptr;
+  std::string staging_dir;
   metrics.SetStage(OdrMetrics::Stage::kPreparation);
 
   // If partial compilation is disabled, we should compile everything regardless of what's in
@@ -2083,13 +2104,16 @@ WARN_UNUSED ExitCode OnDeviceRefresh::Compile(OdrMetrics& metrics,
   }
 
   if (!config_.GetStagingDir().empty()) {
-    staging_dir = config_.GetStagingDir().c_str();
+    staging_dir = config_.GetStagingDir();
   } else {
     // Create staging area and assign label for generating compilation artifacts.
-    if (PaletteCreateOdrefreshStagingDirectory(&staging_dir) != PALETTE_STATUS_OK) {
+    Result<std::string> res = CreateStagingDirectory();
+    if (!res.ok()) {
+      LOG(ERROR) << res.error().message();
       metrics.SetStatus(OdrMetrics::Status::kStagingFailed);
       return ExitCode::kCleanupFailed;
     }
+    staging_dir = res.value();
   }
 
   std::string error_msg;
diff --git a/openjdkjvmti/ti_method.cc b/openjdkjvmti/ti_method.cc
index a460b39556..a1f25c466b 100644
--- a/openjdkjvmti/ti_method.cc
+++ b/openjdkjvmti/ti_method.cc
@@ -249,6 +249,7 @@ jvmtiError MethodUtil::GetLocalVariableTable(jvmtiEnv* env,
     return OK;
   };
 
+  // To avoid defining visitor in the same line as the `if`. We define the lambda and use std::move.
   auto visitor = [&](const art::DexFile::LocalInfo& entry) {
     if (err != OK) {
       return;
@@ -275,9 +276,8 @@ jvmtiError MethodUtil::GetLocalVariableTable(jvmtiEnv* env,
     });
   };
 
-  if (!accessor.DecodeDebugLocalInfo(art_method->IsStatic(),
-                                     art_method->GetDexMethodIndex(),
-                                     visitor)) {
+  if (!accessor.DecodeDebugLocalInfo(
+          art_method->IsStatic(), art_method->GetDexMethodIndex(), std::move(visitor))) {
     // Something went wrong with decoding the debug information. It might as well not be there.
     return ERR(ABSENT_INFORMATION);
   }
@@ -754,6 +754,7 @@ jvmtiError CommonLocalVariableClosure::GetSlotType(art::ArtMethod* method,
   bool found = false;
   *type = art::Primitive::kPrimVoid;
   descriptor->clear();
+  // To avoid defining visitor in the same line as the `if`. We define the lambda and use std::move.
   auto visitor = [&](const art::DexFile::LocalInfo& entry) {
     if (!found && entry.start_address_ <= dex_pc && entry.end_address_ > dex_pc &&
         entry.reg_ == slot_) {
@@ -762,7 +763,8 @@ jvmtiError CommonLocalVariableClosure::GetSlotType(art::ArtMethod* method,
       *descriptor = entry.descriptor_;
     }
   };
-  if (!accessor.DecodeDebugLocalInfo(method->IsStatic(), method->GetDexMethodIndex(), visitor) ||
+  if (!accessor.DecodeDebugLocalInfo(
+          method->IsStatic(), method->GetDexMethodIndex(), std::move(visitor)) ||
       !found) {
     // Something went wrong with decoding the debug information. It might as well not be there.
     // Try to find the type with the verifier.
diff --git a/runtime/Android.bp b/runtime/Android.bp
index d481d300a2..17f09cf814 100644
--- a/runtime/Android.bp
+++ b/runtime/Android.bp
@@ -310,6 +310,7 @@ cc_defaults {
         "jit/jit.cc",
         "jit/jit_code_cache.cc",
         "jit/jit_memory_region.cc",
+        "jit/jit_options.cc",
         "jit/profile_saver.cc",
         "jit/profiling_info.cc",
         "jit/small_pattern_matcher.cc",
diff --git a/runtime/art_method-inl.h b/runtime/art_method-inl.h
index cca20879c2..297b540ab2 100644
--- a/runtime/art_method-inl.h
+++ b/runtime/art_method-inl.h
@@ -19,7 +19,6 @@
 
 #include "art_method.h"
 
-#include "art_field.h"
 #include "base/callee_save_type.h"
 #include "class_linker-inl.h"
 #include "common_throws.h"
@@ -32,9 +31,7 @@
 #include "dex/signature.h"
 #include "gc_root-inl.h"
 #include "imtable-inl.h"
-#include "intrinsics_enum.h"
-#include "jit/jit.h"
-#include "jit/profiling_info.h"
+#include "jit/jit_options.h"
 #include "mirror/class-inl.h"
 #include "mirror/dex_cache-inl.h"
 #include "mirror/object-inl.h"
diff --git a/runtime/common_runtime_test.h b/runtime/common_runtime_test.h
index 0f164a7d07..8f9d7a80ff 100644
--- a/runtime/common_runtime_test.h
+++ b/runtime/common_runtime_test.h
@@ -283,6 +283,11 @@ class CheckJniAbortCatcher {
     GTEST_SKIP() << "WARNING: TEST DISABLED FOR ARM64"; \
   }
 
+#define TEST_DISABLED_FOR_RISCV64()                       \
+  if (kRuntimeISA == InstructionSet::kRiscv64) {          \
+    GTEST_SKIP() << "WARNING: TEST DISABLED FOR RISCV64"; \
+  }
+
 #define TEST_DISABLED_FOR_X86()                       \
   if (kRuntimeISA == InstructionSet::kX86) {          \
     GTEST_SKIP() << "WARNING: TEST DISABLED FOR X86"; \
diff --git a/runtime/fault_handler.cc b/runtime/fault_handler.cc
index ccf6fbd288..635de2af69 100644
--- a/runtime/fault_handler.cc
+++ b/runtime/fault_handler.cc
@@ -42,12 +42,16 @@ namespace art HIDDEN {
 // Static fault manger object accessed by signal handler.
 FaultManager fault_manager;
 
-// This needs to be NO_INLINE since some debuggers do not read the inline-info to set a breakpoint
-// if it isn't.
+// These need to be NO_INLINE since some debuggers do not read the inline-info to set a breakpoint
+// if they aren't.
 extern "C" NO_INLINE __attribute__((visibility("default"))) void art_sigsegv_fault() {
   // Set a breakpoint here to be informed when a SIGSEGV is unhandled by ART.
   VLOG(signals)<< "Caught unknown SIGSEGV in ART fault handler - chaining to next handler.";
 }
+extern "C" NO_INLINE __attribute__((visibility("default"))) void art_sigbus_fault() {
+  // Set a breakpoint here to be informed when a SIGBUS is unhandled by ART.
+  VLOG(signals) << "Caught unknown SIGBUS in ART fault handler - chaining to next handler.";
+}
 
 // Signal handler called on SIGSEGV.
 static bool art_sigsegv_handler(int sig, siginfo_t* info, void* context) {
@@ -232,7 +236,13 @@ bool FaultManager::HandleSigbusFault(int sig, siginfo_t* info, [[maybe_unused]]
   // Simulate a crash in a handler.
   raise(SIGBUS);
 #endif
-  return Runtime::Current()->GetHeap()->MarkCompactCollector()->SigbusHandler(info);
+  if (Runtime::Current()->GetHeap()->MarkCompactCollector()->SigbusHandler(info)) {
+    return true;
+  }
+
+  // Set a breakpoint in this function to catch unhandled signals.
+  art_sigbus_fault();
+  return false;
 }
 
 inline void FaultManager::CheckForUnrecognizedImplicitSuspendCheckInBootImage(
diff --git a/runtime/gc/space/image_space_test.cc b/runtime/gc/space/image_space_test.cc
index 4f6c4e9633..2979544d90 100644
--- a/runtime/gc/space/image_space_test.cc
+++ b/runtime/gc/space/image_space_test.cc
@@ -432,7 +432,6 @@ class ImageSpaceLoadingSingleComponentWithProfilesTest
 TEST_F(ImageSpaceLoadingSingleComponentWithProfilesTest, Test) {
   // Compiling the primary boot image into a single image is not allowed on host.
   TEST_DISABLED_FOR_HOST();
-  TEST_DISABLED_FOR_RISCV64();
 
   CheckImageSpaceAndOatFile(/*space_count=*/1);
 }
@@ -487,7 +486,6 @@ class ImageSpaceLoadingMultipleComponentsWithProfilesTest
 TEST_F(ImageSpaceLoadingMultipleComponentsWithProfilesTest, Test) {
   // Compiling the primary boot image into a single image is not allowed on host.
   TEST_DISABLED_FOR_HOST();
-  TEST_DISABLED_FOR_RISCV64();
 
   CheckImageSpaceAndOatFile(/*space_count=*/1);
 }
diff --git a/runtime/gc/verification.cc b/runtime/gc/verification.cc
index 926156fd8f..858c62a6e3 100644
--- a/runtime/gc/verification.cc
+++ b/runtime/gc/verification.cc
@@ -19,6 +19,8 @@
 #include <iomanip>
 #include <sstream>
 
+#include <android-base/unique_fd.h>
+
 #include "art_field-inl.h"
 #include "base/file_utils.h"
 #include "base/logging.h"
diff --git a/runtime/jit/jit.cc b/runtime/jit/jit.cc
index ea645a2688..54a56f2939 100644
--- a/runtime/jit/jit.cc
+++ b/runtime/jit/jit.cc
@@ -63,103 +63,9 @@ namespace jit {
 
 static constexpr bool kEnableOnStackReplacement = true;
 
-// Maximum permitted threshold value.
-static constexpr uint32_t kJitMaxThreshold = std::numeric_limits<uint16_t>::max();
-
-static constexpr uint32_t kJitDefaultOptimizeThreshold = 0xffff;
-// Different optimization threshold constants. These default to the equivalent optimization
-// thresholds divided by 2, but can be overridden at the command-line.
-static constexpr uint32_t kJitStressDefaultOptimizeThreshold = kJitDefaultOptimizeThreshold / 2;
-static constexpr uint32_t kJitSlowStressDefaultOptimizeThreshold =
-    kJitStressDefaultOptimizeThreshold / 2;
-
-static constexpr uint32_t kJitDefaultWarmupThreshold = 0x3fff;
-// Different warm-up threshold constants. These default to the equivalent warmup thresholds divided
-// by 2, but can be overridden at the command-line.
-static constexpr uint32_t kJitStressDefaultWarmupThreshold = kJitDefaultWarmupThreshold / 2;
-static constexpr uint32_t kJitSlowStressDefaultWarmupThreshold =
-    kJitStressDefaultWarmupThreshold / 2;
-
-DEFINE_RUNTIME_DEBUG_FLAG(Jit, kSlowMode);
-
 // JIT compiler
 JitCompilerInterface* Jit::jit_compiler_ = nullptr;
 
-JitOptions* JitOptions::CreateFromRuntimeArguments(const RuntimeArgumentMap& options) {
-  auto* jit_options = new JitOptions;
-  jit_options->use_jit_compilation_ = options.GetOrDefault(RuntimeArgumentMap::UseJitCompilation);
-  jit_options->use_profiled_jit_compilation_ =
-      options.GetOrDefault(RuntimeArgumentMap::UseProfiledJitCompilation);
-
-  jit_options->code_cache_initial_capacity_ =
-      options.GetOrDefault(RuntimeArgumentMap::JITCodeCacheInitialCapacity);
-  jit_options->code_cache_max_capacity_ =
-      options.GetOrDefault(RuntimeArgumentMap::JITCodeCacheMaxCapacity);
-  jit_options->dump_info_on_shutdown_ =
-      options.Exists(RuntimeArgumentMap::DumpJITInfoOnShutdown);
-  jit_options->profile_saver_options_ =
-      options.GetOrDefault(RuntimeArgumentMap::ProfileSaverOpts);
-  jit_options->thread_pool_pthread_priority_ =
-      options.GetOrDefault(RuntimeArgumentMap::JITPoolThreadPthreadPriority);
-  jit_options->zygote_thread_pool_pthread_priority_ =
-      options.GetOrDefault(RuntimeArgumentMap::JITZygotePoolThreadPthreadPriority);
-
-  // Set default optimize threshold to aid with checking defaults.
-  jit_options->optimize_threshold_ =
-      kIsDebugBuild
-      ? (Jit::kSlowMode
-         ? kJitSlowStressDefaultOptimizeThreshold
-         : kJitStressDefaultOptimizeThreshold)
-      : kJitDefaultOptimizeThreshold;
-
-  // Set default warm-up threshold to aid with checking defaults.
-  jit_options->warmup_threshold_ =
-      kIsDebugBuild ? (Jit::kSlowMode
-                       ? kJitSlowStressDefaultWarmupThreshold
-                       : kJitStressDefaultWarmupThreshold)
-      : kJitDefaultWarmupThreshold;
-
-  if (options.Exists(RuntimeArgumentMap::JITOptimizeThreshold)) {
-    jit_options->optimize_threshold_ = *options.Get(RuntimeArgumentMap::JITOptimizeThreshold);
-  }
-  DCHECK_LE(jit_options->optimize_threshold_, kJitMaxThreshold);
-
-  if (options.Exists(RuntimeArgumentMap::JITWarmupThreshold)) {
-    jit_options->warmup_threshold_ = *options.Get(RuntimeArgumentMap::JITWarmupThreshold);
-  }
-  DCHECK_LE(jit_options->warmup_threshold_, kJitMaxThreshold);
-
-  if (options.Exists(RuntimeArgumentMap::JITPriorityThreadWeight)) {
-    jit_options->priority_thread_weight_ =
-        *options.Get(RuntimeArgumentMap::JITPriorityThreadWeight);
-    if (jit_options->priority_thread_weight_ > jit_options->warmup_threshold_) {
-      LOG(FATAL) << "Priority thread weight is above the warmup threshold.";
-    } else if (jit_options->priority_thread_weight_ == 0) {
-      LOG(FATAL) << "Priority thread weight cannot be 0.";
-    }
-  } else {
-    jit_options->priority_thread_weight_ = std::max(
-        jit_options->warmup_threshold_ / Jit::kDefaultPriorityThreadWeightRatio,
-        static_cast<size_t>(1));
-  }
-
-  if (options.Exists(RuntimeArgumentMap::JITInvokeTransitionWeight)) {
-    jit_options->invoke_transition_weight_ =
-        *options.Get(RuntimeArgumentMap::JITInvokeTransitionWeight);
-    if (jit_options->invoke_transition_weight_ > jit_options->warmup_threshold_) {
-      LOG(FATAL) << "Invoke transition weight is above the warmup threshold.";
-    } else if (jit_options->invoke_transition_weight_  == 0) {
-      LOG(FATAL) << "Invoke transition weight cannot be 0.";
-    }
-  } else {
-    jit_options->invoke_transition_weight_ = std::max(
-        jit_options->warmup_threshold_ / Jit::kDefaultInvokeTransitionWeightRatio,
-        static_cast<size_t>(1));
-  }
-
-  return jit_options;
-}
-
 void Jit::DumpInfo(std::ostream& os) {
   code_cache_->Dump(os);
   cumulative_timings_.Dump(os);
diff --git a/runtime/jit/jit.h b/runtime/jit/jit.h
index e41675f1b4..64b522251d 100644
--- a/runtime/jit/jit.h
+++ b/runtime/jit/jit.h
@@ -24,14 +24,13 @@
 #include "base/histogram-inl.h"
 #include "base/macros.h"
 #include "base/mutex.h"
-#include "base/runtime_debug.h"
 #include "base/timing_logger.h"
 #include "compilation_kind.h"
 #include "handle.h"
 #include "offsets.h"
 #include "interpreter/mterp/nterp.h"
 #include "jit/debugger_interface.h"
-#include "jit/profile_saver_options.h"
+#include "jit_options.h"
 #include "obj_ptr.h"
 #include "thread_pool.h"
 
@@ -62,130 +61,6 @@ class JitOptions;
 
 static constexpr int16_t kJitCheckForOSR = -1;
 static constexpr int16_t kJitHotnessDisabled = -2;
-// At what priority to schedule jit threads. 9 is the lowest foreground priority on device.
-// See android/os/Process.java.
-static constexpr int kJitPoolThreadPthreadDefaultPriority = 9;
-// At what priority to schedule jit zygote threads compiling profiles in the background.
-// 19 is the lowest background priority on device.
-// See android/os/Process.java.
-static constexpr int kJitZygotePoolThreadPthreadDefaultPriority = 19;
-
-class JitOptions {
- public:
-  static JitOptions* CreateFromRuntimeArguments(const RuntimeArgumentMap& options);
-
-  uint16_t GetOptimizeThreshold() const {
-    return optimize_threshold_;
-  }
-
-  uint16_t GetWarmupThreshold() const {
-    return warmup_threshold_;
-  }
-
-  uint16_t GetPriorityThreadWeight() const {
-    return priority_thread_weight_;
-  }
-
-  uint16_t GetInvokeTransitionWeight() const {
-    return invoke_transition_weight_;
-  }
-
-  size_t GetCodeCacheInitialCapacity() const {
-    return code_cache_initial_capacity_;
-  }
-
-  size_t GetCodeCacheMaxCapacity() const {
-    return code_cache_max_capacity_;
-  }
-
-  bool DumpJitInfoOnShutdown() const {
-    return dump_info_on_shutdown_;
-  }
-
-  const ProfileSaverOptions& GetProfileSaverOptions() const {
-    return profile_saver_options_;
-  }
-
-  bool GetSaveProfilingInfo() const {
-    return profile_saver_options_.IsEnabled();
-  }
-
-  int GetThreadPoolPthreadPriority() const {
-    return thread_pool_pthread_priority_;
-  }
-
-  int GetZygoteThreadPoolPthreadPriority() const {
-    return zygote_thread_pool_pthread_priority_;
-  }
-
-  bool UseJitCompilation() const {
-    return use_jit_compilation_;
-  }
-
-  bool UseProfiledJitCompilation() const {
-    return use_profiled_jit_compilation_;
-  }
-
-  void SetUseJitCompilation(bool b) {
-    use_jit_compilation_ = b;
-  }
-
-  void SetSaveProfilingInfo(bool save_profiling_info) {
-    profile_saver_options_.SetEnabled(save_profiling_info);
-  }
-
-  void SetWaitForJitNotificationsToSaveProfile(bool value) {
-    profile_saver_options_.SetWaitForJitNotificationsToSave(value);
-  }
-
-  void SetJitAtFirstUse() {
-    use_jit_compilation_ = true;
-    optimize_threshold_ = 0;
-  }
-
-  void SetUseBaselineCompiler() {
-    use_baseline_compiler_ = true;
-  }
-
-  bool UseBaselineCompiler() const {
-    return use_baseline_compiler_;
-  }
-
- private:
-  // We add the sample in batches of size kJitSamplesBatchSize.
-  // This method rounds the threshold so that it is multiple of the batch size.
-  static uint32_t RoundUpThreshold(uint32_t threshold);
-
-  bool use_jit_compilation_;
-  bool use_profiled_jit_compilation_;
-  bool use_baseline_compiler_;
-  size_t code_cache_initial_capacity_;
-  size_t code_cache_max_capacity_;
-  uint32_t optimize_threshold_;
-  uint32_t warmup_threshold_;
-  uint16_t priority_thread_weight_;
-  uint16_t invoke_transition_weight_;
-  bool dump_info_on_shutdown_;
-  int thread_pool_pthread_priority_;
-  int zygote_thread_pool_pthread_priority_;
-  ProfileSaverOptions profile_saver_options_;
-
-  JitOptions()
-      : use_jit_compilation_(false),
-        use_profiled_jit_compilation_(false),
-        use_baseline_compiler_(false),
-        code_cache_initial_capacity_(0),
-        code_cache_max_capacity_(0),
-        optimize_threshold_(0),
-        warmup_threshold_(0),
-        priority_thread_weight_(0),
-        invoke_transition_weight_(0),
-        dump_info_on_shutdown_(false),
-        thread_pool_pthread_priority_(kJitPoolThreadPthreadDefaultPriority),
-        zygote_thread_pool_pthread_priority_(kJitZygotePoolThreadPthreadDefaultPriority) {}
-
-  DISALLOW_COPY_AND_ASSIGN(JitOptions);
-};
 
 // Implemented and provided by the compiler library.
 class JitCompilerInterface {
@@ -305,13 +180,9 @@ class JitThreadPool : public AbstractThreadPool {
 
 class Jit {
  public:
-  static constexpr size_t kDefaultPriorityThreadWeightRatio = 1000;
-  static constexpr size_t kDefaultInvokeTransitionWeightRatio = 500;
   // How frequently should the interpreter check to see if OSR compilation is ready.
   static constexpr int16_t kJitRecheckOSRThreshold = 101;  // Prime number to avoid patterns.
 
-  DECLARE_RUNTIME_DEBUG_FLAG(kSlowMode);
-
   virtual ~Jit();
 
   // Create JIT itself.
diff --git a/runtime/jit/jit_code_cache.cc b/runtime/jit/jit_code_cache.cc
index 7bfbe15059..3560ac17ff 100644
--- a/runtime/jit/jit_code_cache.cc
+++ b/runtime/jit/jit_code_cache.cc
@@ -1429,18 +1429,20 @@ void JitCodeCache::GetProfiledMethods(const std::set<std::string>& dex_base_loca
   ScopedTrace trace(__FUNCTION__);
   Thread* self = Thread::Current();
   WaitUntilInlineCacheAccessible(self);
-  std::vector<ProfilingInfo*> copies;
+  SafeMap<ArtMethod*, ProfilingInfo*> profiling_infos;
+  std::vector<ArtMethod*> copies;
   // TODO: Avoid read barriers for potentially dead methods.
   // ScopedDebugDisallowReadBarriers sddrb(self);
   {
     MutexLock mu(self, *Locks::jit_lock_);
-    copies.reserve(profiling_infos_.size());
-    for (const auto& entry : profiling_infos_) {
+    profiling_infos = profiling_infos_;
+    for (const auto& entry : method_code_map_) {
       copies.push_back(entry.second);
     }
   }
-  for (ProfilingInfo* info : copies) {
-    ArtMethod* method = info->GetMethod();
+  for (ArtMethod* method : copies) {
+    auto it = profiling_infos.find(method);
+    ProfilingInfo* info = (it == profiling_infos.end()) ? nullptr : it->second;
     const DexFile* dex_file = method->GetDexFile();
     const std::string base_location = DexFileLoader::GetBaseLocation(dex_file->GetLocation());
     if (!ContainsElement(dex_base_locations, base_location)) {
@@ -1449,74 +1451,76 @@ void JitCodeCache::GetProfiledMethods(const std::set<std::string>& dex_base_loca
     }
     std::vector<ProfileMethodInfo::ProfileInlineCache> inline_caches;
 
-    // If the method is still baseline compiled and doesn't meet the inline cache threshold, don't
-    // save the inline caches because they might be incomplete.
-    // Although we don't deoptimize for incomplete inline caches in AOT-compiled code, inlining
-    // leads to larger generated code.
-    // If the inline cache is empty the compiler will generate a regular invoke virtual/interface.
-    const void* entry_point = method->GetEntryPointFromQuickCompiledCode();
-    if (ContainsPc(entry_point) &&
-        CodeInfo::IsBaseline(
-            OatQuickMethodHeader::FromEntryPoint(entry_point)->GetOptimizedCodeInfoPtr()) &&
-        (ProfilingInfo::GetOptimizeThreshold() - info->GetBaselineHotnessCount()) <
-            inline_cache_threshold) {
-      methods.emplace_back(/*ProfileMethodInfo*/
-          MethodReference(dex_file, method->GetDexMethodIndex()), inline_caches);
-      continue;
-    }
-
-    for (size_t i = 0; i < info->number_of_inline_caches_; ++i) {
-      std::vector<TypeReference> profile_classes;
-      const InlineCache& cache = info->GetInlineCaches()[i];
-      ArtMethod* caller = info->GetMethod();
-      bool is_missing_types = false;
-      for (size_t k = 0; k < InlineCache::kIndividualCacheSize; k++) {
-        mirror::Class* cls = cache.classes_[k].Read();
-        if (cls == nullptr) {
-          break;
-        }
+    if (info != nullptr) {
+      // If the method is still baseline compiled and doesn't meet the inline cache threshold, don't
+      // save the inline caches because they might be incomplete.
+      // Although we don't deoptimize for incomplete inline caches in AOT-compiled code, inlining
+      // leads to larger generated code.
+      // If the inline cache is empty the compiler will generate a regular invoke virtual/interface.
+      const void* entry_point = method->GetEntryPointFromQuickCompiledCode();
+      if (ContainsPc(entry_point) &&
+          CodeInfo::IsBaseline(
+              OatQuickMethodHeader::FromEntryPoint(entry_point)->GetOptimizedCodeInfoPtr()) &&
+          (ProfilingInfo::GetOptimizeThreshold() - info->GetBaselineHotnessCount()) <
+              inline_cache_threshold) {
+        methods.emplace_back(/*ProfileMethodInfo*/
+            MethodReference(dex_file, method->GetDexMethodIndex()), inline_caches);
+        continue;
+      }
 
-        // Check if the receiver is in the boot class path or if it's in the
-        // same class loader as the caller. If not, skip it, as there is not
-        // much we can do during AOT.
-        if (!cls->IsBootStrapClassLoaded() &&
-            caller->GetClassLoader() != cls->GetClassLoader()) {
-          is_missing_types = true;
-          continue;
-        }
+      for (size_t i = 0; i < info->number_of_inline_caches_; ++i) {
+        std::vector<TypeReference> profile_classes;
+        const InlineCache& cache = info->GetInlineCaches()[i];
+        ArtMethod* caller = info->GetMethod();
+        bool is_missing_types = false;
+        for (size_t k = 0; k < InlineCache::kIndividualCacheSize; k++) {
+          mirror::Class* cls = cache.classes_[k].Read();
+          if (cls == nullptr) {
+            break;
+          }
 
-        const DexFile* class_dex_file = nullptr;
-        dex::TypeIndex type_index;
+          // Check if the receiver is in the boot class path or if it's in the
+          // same class loader as the caller. If not, skip it, as there is not
+          // much we can do during AOT.
+          if (!cls->IsBootStrapClassLoaded() &&
+              caller->GetClassLoader() != cls->GetClassLoader()) {
+            is_missing_types = true;
+            continue;
+          }
 
-        if (cls->GetDexCache() == nullptr) {
-          DCHECK(cls->IsArrayClass()) << cls->PrettyClass();
-          // Make a best effort to find the type index in the method's dex file.
-          // We could search all open dex files but that might turn expensive
-          // and probably not worth it.
-          class_dex_file = dex_file;
-          type_index = cls->FindTypeIndexInOtherDexFile(*dex_file);
-        } else {
-          class_dex_file = &(cls->GetDexFile());
-          type_index = cls->GetDexTypeIndex();
-        }
-        if (!type_index.IsValid()) {
-          // Could be a proxy class or an array for which we couldn't find the type index.
-          is_missing_types = true;
-          continue;
+          const DexFile* class_dex_file = nullptr;
+          dex::TypeIndex type_index;
+
+          if (cls->GetDexCache() == nullptr) {
+            DCHECK(cls->IsArrayClass()) << cls->PrettyClass();
+            // Make a best effort to find the type index in the method's dex file.
+            // We could search all open dex files but that might turn expensive
+            // and probably not worth it.
+            class_dex_file = dex_file;
+            type_index = cls->FindTypeIndexInOtherDexFile(*dex_file);
+          } else {
+            class_dex_file = &(cls->GetDexFile());
+            type_index = cls->GetDexTypeIndex();
+          }
+          if (!type_index.IsValid()) {
+            // Could be a proxy class or an array for which we couldn't find the type index.
+            is_missing_types = true;
+            continue;
+          }
+          if (ContainsElement(dex_base_locations,
+                              DexFileLoader::GetBaseLocation(class_dex_file->GetLocation()))) {
+            // Only consider classes from the same apk (including multidex).
+            profile_classes.emplace_back(/*ProfileMethodInfo::ProfileClassReference*/
+                class_dex_file, type_index);
+          } else {
+            is_missing_types = true;
+          }
         }
-        if (ContainsElement(dex_base_locations,
-                            DexFileLoader::GetBaseLocation(class_dex_file->GetLocation()))) {
-          // Only consider classes from the same apk (including multidex).
-          profile_classes.emplace_back(/*ProfileMethodInfo::ProfileClassReference*/
-              class_dex_file, type_index);
-        } else {
-          is_missing_types = true;
+        if (!profile_classes.empty()) {
+          inline_caches.emplace_back(/*ProfileMethodInfo::ProfileInlineCache*/
+              cache.dex_pc_, is_missing_types, profile_classes);
         }
       }
-      if (!profile_classes.empty()) {
-        inline_caches.emplace_back(/*ProfileMethodInfo::ProfileInlineCache*/
-            cache.dex_pc_, is_missing_types, profile_classes);
-      }
     }
     methods.emplace_back(/*ProfileMethodInfo*/
         MethodReference(dex_file, method->GetDexMethodIndex()), inline_caches);
@@ -1535,17 +1539,13 @@ bool JitCodeCache::NotifyCompilationOf(ArtMethod* method,
                                        CompilationKind compilation_kind,
                                        bool prejit) {
   const void* existing_entry_point = method->GetEntryPointFromQuickCompiledCode();
-  if (compilation_kind != CompilationKind::kOsr && ContainsPc(existing_entry_point)) {
-    OatQuickMethodHeader* method_header =
-        OatQuickMethodHeader::FromEntryPoint(existing_entry_point);
-    bool is_baseline = (compilation_kind == CompilationKind::kBaseline);
-    if (CodeInfo::IsBaseline(method_header->GetOptimizedCodeInfoPtr()) == is_baseline) {
-      VLOG(jit) << "Not compiling "
-                << method->PrettyMethod()
-                << " because it has already been compiled"
-                << " kind=" << compilation_kind;
-      return false;
-    }
+  if (compilation_kind == CompilationKind::kBaseline && ContainsPc(existing_entry_point)) {
+    // The existing entry point is either already baseline, or optimized. No
+    // need to compile.
+    VLOG(jit) << "Not compiling "
+              << method->PrettyMethod()
+              << " baseline, because it has already been compiled";
+    return false;
   }
 
   if (method->NeedsClinitCheckBeforeCall() && !prejit) {
diff --git a/runtime/jit/jit_options.cc b/runtime/jit/jit_options.cc
new file mode 100644
index 0000000000..3c31d1dfcd
--- /dev/null
+++ b/runtime/jit/jit_options.cc
@@ -0,0 +1,117 @@
+/*
+ * Copyright 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "jit_options.h"
+
+#include "runtime_options.h"
+
+namespace art HIDDEN {
+namespace jit {
+
+// Maximum permitted threshold value.
+static constexpr uint32_t kJitMaxThreshold = std::numeric_limits<uint16_t>::max();
+
+static constexpr uint32_t kJitDefaultOptimizeThreshold = 0xffff;
+// Different optimization threshold constants. These default to the equivalent optimization
+// thresholds divided by 2, but can be overridden at the command-line.
+static constexpr uint32_t kJitStressDefaultOptimizeThreshold = kJitDefaultOptimizeThreshold / 2;
+static constexpr uint32_t kJitSlowStressDefaultOptimizeThreshold =
+    kJitStressDefaultOptimizeThreshold / 2;
+
+static constexpr uint32_t kJitDefaultWarmupThreshold = 0x3fff;
+// Different warm-up threshold constants. These default to the equivalent warmup thresholds divided
+// by 2, but can be overridden at the command-line.
+static constexpr uint32_t kJitStressDefaultWarmupThreshold = kJitDefaultWarmupThreshold / 2;
+static constexpr uint32_t kJitSlowStressDefaultWarmupThreshold =
+    kJitStressDefaultWarmupThreshold / 2;
+
+static constexpr size_t kDefaultPriorityThreadWeightRatio = 1000;
+static constexpr size_t kDefaultInvokeTransitionWeightRatio = 500;
+
+DEFINE_RUNTIME_DEBUG_FLAG(JitOptions, kSlowMode);
+
+JitOptions* JitOptions::CreateFromRuntimeArguments(const RuntimeArgumentMap& options) {
+  auto* jit_options = new JitOptions;
+  jit_options->use_jit_compilation_ = options.GetOrDefault(RuntimeArgumentMap::UseJitCompilation);
+  jit_options->use_profiled_jit_compilation_ =
+      options.GetOrDefault(RuntimeArgumentMap::UseProfiledJitCompilation);
+
+  jit_options->code_cache_initial_capacity_ =
+      options.GetOrDefault(RuntimeArgumentMap::JITCodeCacheInitialCapacity);
+  jit_options->code_cache_max_capacity_ =
+      options.GetOrDefault(RuntimeArgumentMap::JITCodeCacheMaxCapacity);
+  jit_options->dump_info_on_shutdown_ =
+      options.Exists(RuntimeArgumentMap::DumpJITInfoOnShutdown);
+  jit_options->profile_saver_options_ =
+      options.GetOrDefault(RuntimeArgumentMap::ProfileSaverOpts);
+  jit_options->thread_pool_pthread_priority_ =
+      options.GetOrDefault(RuntimeArgumentMap::JITPoolThreadPthreadPriority);
+  jit_options->zygote_thread_pool_pthread_priority_ =
+      options.GetOrDefault(RuntimeArgumentMap::JITZygotePoolThreadPthreadPriority);
+
+  // Set default optimize threshold to aid with checking defaults.
+  jit_options->optimize_threshold_ = kIsDebugBuild
+      ? (kSlowMode ? kJitSlowStressDefaultOptimizeThreshold : kJitStressDefaultOptimizeThreshold)
+      : kJitDefaultOptimizeThreshold;
+
+  // Set default warm-up threshold to aid with checking defaults.
+  jit_options->warmup_threshold_ = kIsDebugBuild
+      ? (kSlowMode ? kJitSlowStressDefaultWarmupThreshold : kJitStressDefaultWarmupThreshold)
+      : kJitDefaultWarmupThreshold;
+
+  if (options.Exists(RuntimeArgumentMap::JITOptimizeThreshold)) {
+    jit_options->optimize_threshold_ = *options.Get(RuntimeArgumentMap::JITOptimizeThreshold);
+  }
+  DCHECK_LE(jit_options->optimize_threshold_, kJitMaxThreshold);
+
+  if (options.Exists(RuntimeArgumentMap::JITWarmupThreshold)) {
+    jit_options->warmup_threshold_ = *options.Get(RuntimeArgumentMap::JITWarmupThreshold);
+  }
+  DCHECK_LE(jit_options->warmup_threshold_, kJitMaxThreshold);
+
+  if (options.Exists(RuntimeArgumentMap::JITPriorityThreadWeight)) {
+    jit_options->priority_thread_weight_ =
+        *options.Get(RuntimeArgumentMap::JITPriorityThreadWeight);
+    if (jit_options->priority_thread_weight_ > jit_options->warmup_threshold_) {
+      LOG(FATAL) << "Priority thread weight is above the warmup threshold.";
+    } else if (jit_options->priority_thread_weight_ == 0) {
+      LOG(FATAL) << "Priority thread weight cannot be 0.";
+    }
+  } else {
+    jit_options->priority_thread_weight_ = std::max(
+        jit_options->warmup_threshold_ / kDefaultPriorityThreadWeightRatio,
+        static_cast<size_t>(1));
+  }
+
+  if (options.Exists(RuntimeArgumentMap::JITInvokeTransitionWeight)) {
+    jit_options->invoke_transition_weight_ =
+        *options.Get(RuntimeArgumentMap::JITInvokeTransitionWeight);
+    if (jit_options->invoke_transition_weight_ > jit_options->warmup_threshold_) {
+      LOG(FATAL) << "Invoke transition weight is above the warmup threshold.";
+    } else if (jit_options->invoke_transition_weight_  == 0) {
+      LOG(FATAL) << "Invoke transition weight cannot be 0.";
+    }
+  } else {
+    jit_options->invoke_transition_weight_ = std::max(
+        jit_options->warmup_threshold_ / kDefaultInvokeTransitionWeightRatio,
+        static_cast<size_t>(1));
+  }
+
+  return jit_options;
+}
+
+}  // namespace jit
+}  // namespace art
diff --git a/runtime/jit/jit_options.h b/runtime/jit/jit_options.h
new file mode 100644
index 0000000000..be305b4d63
--- /dev/null
+++ b/runtime/jit/jit_options.h
@@ -0,0 +1,160 @@
+/*
+ * Copyright 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef ART_RUNTIME_JIT_JIT_OPTIONS_H_
+#define ART_RUNTIME_JIT_JIT_OPTIONS_H_
+
+#include "base/macros.h"
+#include "base/runtime_debug.h"
+#include "profile_saver_options.h"
+
+namespace art HIDDEN {
+
+struct RuntimeArgumentMap;
+
+namespace jit {
+
+// At what priority to schedule jit threads. 9 is the lowest foreground priority on device.
+// See android/os/Process.java.
+static constexpr int kJitPoolThreadPthreadDefaultPriority = 9;
+// At what priority to schedule jit zygote threads compiling profiles in the background.
+// 19 is the lowest background priority on device.
+// See android/os/Process.java.
+static constexpr int kJitZygotePoolThreadPthreadDefaultPriority = 19;
+
+class JitOptions {
+ public:
+  DECLARE_RUNTIME_DEBUG_FLAG(kSlowMode);
+
+  static JitOptions* CreateFromRuntimeArguments(const RuntimeArgumentMap& options);
+
+  uint16_t GetOptimizeThreshold() const {
+    return optimize_threshold_;
+  }
+
+  uint16_t GetWarmupThreshold() const {
+    return warmup_threshold_;
+  }
+
+  uint16_t GetPriorityThreadWeight() const {
+    return priority_thread_weight_;
+  }
+
+  uint16_t GetInvokeTransitionWeight() const {
+    return invoke_transition_weight_;
+  }
+
+  size_t GetCodeCacheInitialCapacity() const {
+    return code_cache_initial_capacity_;
+  }
+
+  size_t GetCodeCacheMaxCapacity() const {
+    return code_cache_max_capacity_;
+  }
+
+  bool DumpJitInfoOnShutdown() const {
+    return dump_info_on_shutdown_;
+  }
+
+  const ProfileSaverOptions& GetProfileSaverOptions() const {
+    return profile_saver_options_;
+  }
+
+  bool GetSaveProfilingInfo() const {
+    return profile_saver_options_.IsEnabled();
+  }
+
+  int GetThreadPoolPthreadPriority() const {
+    return thread_pool_pthread_priority_;
+  }
+
+  int GetZygoteThreadPoolPthreadPriority() const {
+    return zygote_thread_pool_pthread_priority_;
+  }
+
+  bool UseJitCompilation() const {
+    return use_jit_compilation_;
+  }
+
+  bool UseProfiledJitCompilation() const {
+    return use_profiled_jit_compilation_;
+  }
+
+  void SetUseJitCompilation(bool b) {
+    use_jit_compilation_ = b;
+  }
+
+  void SetSaveProfilingInfo(bool save_profiling_info) {
+    profile_saver_options_.SetEnabled(save_profiling_info);
+  }
+
+  void SetWaitForJitNotificationsToSaveProfile(bool value) {
+    profile_saver_options_.SetWaitForJitNotificationsToSave(value);
+  }
+
+  void SetJitAtFirstUse() {
+    use_jit_compilation_ = true;
+    optimize_threshold_ = 0;
+  }
+
+  void SetUseBaselineCompiler() {
+    use_baseline_compiler_ = true;
+  }
+
+  bool UseBaselineCompiler() const {
+    return use_baseline_compiler_;
+  }
+
+ private:
+  // We add the sample in batches of size kJitSamplesBatchSize.
+  // This method rounds the threshold so that it is multiple of the batch size.
+  static uint32_t RoundUpThreshold(uint32_t threshold);
+
+  bool use_jit_compilation_;
+  bool use_profiled_jit_compilation_;
+  bool use_baseline_compiler_;
+  size_t code_cache_initial_capacity_;
+  size_t code_cache_max_capacity_;
+  uint32_t optimize_threshold_;
+  uint32_t warmup_threshold_;
+  uint16_t priority_thread_weight_;
+  uint16_t invoke_transition_weight_;
+  bool dump_info_on_shutdown_;
+  int thread_pool_pthread_priority_;
+  int zygote_thread_pool_pthread_priority_;
+  ProfileSaverOptions profile_saver_options_;
+
+  JitOptions()
+      : use_jit_compilation_(false),
+        use_profiled_jit_compilation_(false),
+        use_baseline_compiler_(false),
+        code_cache_initial_capacity_(0),
+        code_cache_max_capacity_(0),
+        optimize_threshold_(0),
+        warmup_threshold_(0),
+        priority_thread_weight_(0),
+        invoke_transition_weight_(0),
+        dump_info_on_shutdown_(false),
+        thread_pool_pthread_priority_(kJitPoolThreadPthreadDefaultPriority),
+        zygote_thread_pool_pthread_priority_(kJitZygotePoolThreadPthreadDefaultPriority) {}
+
+  DISALLOW_COPY_AND_ASSIGN(JitOptions);
+};
+
+}  // namespace jit
+}  // namespace art
+
+#endif  // ART_RUNTIME_JIT_JIT_OPTIONS_H_
diff --git a/runtime/jit/profiling_info_test.cc b/runtime/jit/profiling_info_test.cc
index a8a95774d8..674cb73dcd 100644
--- a/runtime/jit/profiling_info_test.cc
+++ b/runtime/jit/profiling_info_test.cc
@@ -14,6 +14,8 @@
  * limitations under the License.
  */
 
+#include "profiling_info.h"
+
 #include <gtest/gtest.h>
 #include <stdio.h>
 
diff --git a/runtime/jni/jni_env_ext.cc b/runtime/jni/jni_env_ext.cc
index 26b154c89a..3b10ce29c9 100644
--- a/runtime/jni/jni_env_ext.cc
+++ b/runtime/jni/jni_env_ext.cc
@@ -142,27 +142,27 @@ void JNIEnvExt::PopFrame() {
 //       are tests in jni_internal_test to match the results against the actual values.
 
 // This is encoding the knowledge of the structure and layout of JNIEnv fields.
-static size_t JNIEnvSize(size_t pointer_size) {
+static size_t JNIEnvSize(PointerSize pointer_size) {
   // A single pointer.
-  return pointer_size;
+  return static_cast<size_t>(pointer_size);
 }
 
-inline MemberOffset JNIEnvExt::LocalReferenceTableOffset(size_t pointer_size) {
+inline MemberOffset JNIEnvExt::LocalReferenceTableOffset(PointerSize pointer_size) {
   return MemberOffset(JNIEnvSize(pointer_size) +
-                      2 * pointer_size);          // Thread* self + JavaVMExt* vm
+                      2 * static_cast<size_t>(pointer_size));  // Thread* self + JavaVMExt* vm
 }
 
-MemberOffset JNIEnvExt::SegmentStateOffset(size_t pointer_size) {
+MemberOffset JNIEnvExt::LrtSegmentStateOffset(PointerSize pointer_size) {
   return MemberOffset(LocalReferenceTableOffset(pointer_size).SizeValue() +
                       jni::LocalReferenceTable::SegmentStateOffset().SizeValue());
 }
 
-MemberOffset JNIEnvExt::LocalRefCookieOffset(size_t pointer_size) {
+MemberOffset JNIEnvExt::LrtPreviousStateOffset(PointerSize pointer_size) {
   return MemberOffset(LocalReferenceTableOffset(pointer_size).SizeValue() +
                       jni::LocalReferenceTable::PreviousStateOffset().SizeValue());
 }
 
-MemberOffset JNIEnvExt::SelfOffset(size_t pointer_size) {
+MemberOffset JNIEnvExt::SelfOffset(PointerSize pointer_size) {
   return MemberOffset(JNIEnvSize(pointer_size));
 }
 
diff --git a/runtime/jni/jni_env_ext.h b/runtime/jni/jni_env_ext.h
index 2e8b52e4cc..ca5cc0a7ba 100644
--- a/runtime/jni/jni_env_ext.h
+++ b/runtime/jni/jni_env_ext.h
@@ -42,9 +42,9 @@ class JNIEnvExt : public JNIEnv {
   // Creates a new JNIEnvExt. Returns null on error, in which case error_msg
   // will contain a description of the error.
   static JNIEnvExt* Create(Thread* self, JavaVMExt* vm, std::string* error_msg);
-  static MemberOffset SegmentStateOffset(size_t pointer_size);
-  static MemberOffset LocalRefCookieOffset(size_t pointer_size);
-  static MemberOffset SelfOffset(size_t pointer_size);
+  static MemberOffset LrtSegmentStateOffset(PointerSize pointer_size);
+  static MemberOffset LrtPreviousStateOffset(PointerSize pointer_size);
+  static MemberOffset SelfOffset(PointerSize pointer_size);
   static jint GetEnvHandler(JavaVMExt* vm, /*out*/void** out, jint version);
 
   ~JNIEnvExt();
@@ -147,7 +147,7 @@ class JNIEnvExt : public JNIEnv {
       REQUIRES(!Locks::thread_list_lock_, !Locks::jni_function_table_lock_);
 
  private:
-  static MemberOffset LocalReferenceTableOffset(size_t pointer_size);
+  static MemberOffset LocalReferenceTableOffset(PointerSize pointer_size);
 
   // Override of function tables. This applies to both default as well as instrumented (CheckJNI)
   // function tables.
diff --git a/runtime/jni/jni_internal_test.cc b/runtime/jni/jni_internal_test.cc
index 155048a30d..ed97e4d4c8 100644
--- a/runtime/jni/jni_internal_test.cc
+++ b/runtime/jni/jni_internal_test.cc
@@ -2634,19 +2634,22 @@ TEST_F(JniInternalTest, IndirectReferenceTableOffsets) {
 
 // Test the offset computation of JNIEnvExt offsets. b/26071368.
 TEST_F(JniInternalTest, JNIEnvExtOffsets) {
-  EXPECT_EQ(OFFSETOF_MEMBER(JNIEnvExt, self_), JNIEnvExt::SelfOffset(sizeof(void*)).Uint32Value());
+  EXPECT_EQ(OFFSETOF_MEMBER(JNIEnvExt, self_),
+            JNIEnvExt::SelfOffset(kRuntimePointerSize).Uint32Value());
 
   // `previous_state_` amd `segment_state_` are private in the IndirectReferenceTable.
   // So this test isn't as good as we'd hope it to be.
   uint32_t previous_state_now =
       OFFSETOF_MEMBER(JNIEnvExt, locals_) +
       jni::LocalReferenceTable::PreviousStateOffset().Uint32Value();
-  uint32_t previous_state_computed = JNIEnvExt::LocalRefCookieOffset(sizeof(void*)).Uint32Value();
+  uint32_t previous_state_computed =
+      JNIEnvExt::LrtPreviousStateOffset(kRuntimePointerSize).Uint32Value();
   EXPECT_EQ(previous_state_now, previous_state_computed);
   uint32_t segment_state_now =
       OFFSETOF_MEMBER(JNIEnvExt, locals_) +
       jni::LocalReferenceTable::SegmentStateOffset().Uint32Value();
-  uint32_t segment_state_computed = JNIEnvExt::SegmentStateOffset(sizeof(void*)).Uint32Value();
+  uint32_t segment_state_computed =
+      JNIEnvExt::LrtSegmentStateOffset(kRuntimePointerSize).Uint32Value();
   EXPECT_EQ(segment_state_now, segment_state_computed);
 }
 
diff --git a/runtime/metrics/statsd.cc b/runtime/metrics/statsd.cc
index e6d26d9746..67d677be9c 100644
--- a/runtime/metrics/statsd.cc
+++ b/runtime/metrics/statsd.cc
@@ -416,30 +416,18 @@ class StatsdBackend : public MetricsBackend {
 
 std::unique_ptr<MetricsBackend> CreateStatsdBackend() { return std::make_unique<StatsdBackend>(); }
 
-AStatsManager_PullAtomCallbackReturn DeviceStatusCallback(int32_t atom_tag,
-                                                          AStatsEventList* data,
-                                                          [[maybe_unused]] void* cookie) {
-  if (atom_tag == statsd::ART_DEVICE_STATUS) {
-    Runtime* runtime = Runtime::Current();
-    int32_t boot_image_status;
-    if (runtime->GetHeap()->HasBootImageSpace() && !runtime->HasImageWithProfile()) {
-      boot_image_status = statsd::ART_DEVICE_DATUM_REPORTED__BOOT_IMAGE_STATUS__STATUS_FULL;
-    } else if (runtime->GetHeap()->HasBootImageSpace() &&
-               runtime->GetHeap()->GetBootImageSpaces()[0]->GetProfileFiles().empty()) {
-      boot_image_status = statsd::ART_DEVICE_DATUM_REPORTED__BOOT_IMAGE_STATUS__STATUS_MINIMAL;
-    } else {
-      boot_image_status = statsd::ART_DEVICE_DATUM_REPORTED__BOOT_IMAGE_STATUS__STATUS_NONE;
-    }
-    statsd::addAStatsEvent(data, atom_tag, boot_image_status);
-    return AStatsManager_PULL_SUCCESS;
+void ReportDeviceMetrics() {
+  Runtime* runtime = Runtime::Current();
+  int32_t boot_image_status;
+  if (runtime->GetHeap()->HasBootImageSpace() && !runtime->HasImageWithProfile()) {
+    boot_image_status = statsd::ART_DEVICE_DATUM_REPORTED__BOOT_IMAGE_STATUS__STATUS_FULL;
+  } else if (runtime->GetHeap()->HasBootImageSpace() &&
+             runtime->GetHeap()->GetBootImageSpaces()[0]->GetProfileFiles().empty()) {
+    boot_image_status = statsd::ART_DEVICE_DATUM_REPORTED__BOOT_IMAGE_STATUS__STATUS_MINIMAL;
+  } else {
+    boot_image_status = statsd::ART_DEVICE_DATUM_REPORTED__BOOT_IMAGE_STATUS__STATUS_NONE;
   }
-
-  return AStatsManager_PULL_SKIP;
-}
-
-void SetupCallbackForDeviceStatus() {
-  AStatsManager_setPullAtomCallback(
-      statsd::ART_DEVICE_STATUS, /*metadata=*/nullptr, DeviceStatusCallback, /*cookie=*/nullptr);
+  statsd::stats_write(statsd::ART_DEVICE_DATUM_REPORTED, boot_image_status);
 }
 
 }  // namespace metrics
diff --git a/runtime/metrics/statsd.h b/runtime/metrics/statsd.h
index 00bd595ca1..ae53a22e85 100644
--- a/runtime/metrics/statsd.h
+++ b/runtime/metrics/statsd.h
@@ -29,10 +29,10 @@ class MetricsBackend;
 // Statsd is only supported on Android
 #ifdef __ANDROID__
 std::unique_ptr<MetricsBackend> CreateStatsdBackend();
-void SetupCallbackForDeviceStatus();
+void ReportDeviceMetrics();
 #else
 inline std::unique_ptr<MetricsBackend> CreateStatsdBackend() { return nullptr; }
-inline void SetupCallbackForDeviceStatus() {}
+inline void ReportDeviceMetrics() {}
 #endif
 
 }  // namespace metrics
diff --git a/runtime/mirror/class.h b/runtime/mirror/class.h
index 15f0fd306f..b69086c971 100644
--- a/runtime/mirror/class.h
+++ b/runtime/mirror/class.h
@@ -76,10 +76,7 @@ class DexCache;
 class Field;
 class IfTable;
 class Method;
-template <typename T> struct PACKED(8) DexCachePair;
-
-using StringDexCachePair = DexCachePair<String>;
-using StringDexCacheType = std::atomic<StringDexCachePair>;
+template <typename T> struct alignas(8) DexCachePair;
 
 // C++ mirror of java.lang.Class
 class EXPORT MANAGED Class final : public Object {
diff --git a/runtime/mirror/dex_cache.h b/runtime/mirror/dex_cache.h
index 921d8028e7..d93464759f 100644
--- a/runtime/mirror/dex_cache.h
+++ b/runtime/mirror/dex_cache.h
@@ -53,7 +53,7 @@ class DexCache;
 class MethodType;
 class String;
 
-template <typename T> struct PACKED(8) DexCachePair {
+template <typename T> struct alignas(8) DexCachePair {
   GcRoot<T> object;
   uint32_t index;
   // The array is initially [ {0,0}, {0,0}, {0,0} ... ]
@@ -90,7 +90,7 @@ template <typename T> struct PACKED(8) DexCachePair {
   T* GetObjectForIndex(uint32_t idx) REQUIRES_SHARED(Locks::mutator_lock_);
 };
 
-template <typename T> struct PACKED(2 * __SIZEOF_POINTER__) NativeDexCachePair {
+template <typename T> struct alignas(2 * __SIZEOF_POINTER__) NativeDexCachePair {
   T* object;
   size_t index;
   // This is similar to DexCachePair except that we're storing a native pointer
diff --git a/runtime/mirror/object-refvisitor-inl.h b/runtime/mirror/object-refvisitor-inl.h
index fa2fbaec27..d4f474974c 100644
--- a/runtime/mirror/object-refvisitor-inl.h
+++ b/runtime/mirror/object-refvisitor-inl.h
@@ -145,7 +145,6 @@ inline size_t Object::VisitRefsForCompaction(const Visitor& visitor,
     VisitInstanceFieldsReferences<kVerifyFlags, kReadBarrierOption>(klass, visitor);
     size = kFetchObjSize ? klass->GetObjectSize<kSizeOfFlags>() : 0;
   } else if ((class_flags & kClassFlagNoReferenceFields) != 0) {
-    CheckNoReferenceField<kVerifyFlags, kReadBarrierOption>(klass);
     if ((class_flags & kClassFlagString) != 0) {
       size = kFetchObjSize ? static_cast<String*>(this)->SizeOf<kSizeOfFlags>() : 0;
     } else if (klass->IsArrayClass<kVerifyFlags>()) {
@@ -168,7 +167,6 @@ inline size_t Object::VisitRefsForCompaction(const Visitor& visitor,
                                                                                    visitor);
     size = kFetchObjSize ? as_klass->SizeOf<kSizeOfFlags>() : 0;
   } else if (class_flags == kClassFlagObjectArray) {
-    DCHECK((klass->IsObjectArrayClass<kVerifyFlags, kReadBarrierOption>()));
     ObjPtr<ObjectArray<Object>> obj_arr = ObjPtr<ObjectArray<Object>>::DownCast(this);
     obj_arr->VisitReferences(visitor, begin, end);
     size = kFetchObjSize ?
diff --git a/runtime/monitor.cc b/runtime/monitor.cc
index cbd329f4f3..3458508306 100644
--- a/runtime/monitor.cc
+++ b/runtime/monitor.cc
@@ -1054,8 +1054,11 @@ void Monitor::Inflate(Thread* self, Thread* owner, ObjPtr<mirror::Object> obj, i
   }
 }
 
-void Monitor::InflateThinLocked(Thread* self, Handle<mirror::Object> obj, LockWord lock_word,
-                                uint32_t hash_code) {
+void Monitor::InflateThinLocked(Thread* self,
+                                Handle<mirror::Object> obj,
+                                LockWord lock_word,
+                                uint32_t hash_code,
+                                int attempt_of_4) {
   DCHECK_EQ(lock_word.GetState(), LockWord::kThinLocked);
   uint32_t owner_thread_id = lock_word.ThinLockOwner();
   if (owner_thread_id == self->GetThreadId()) {
@@ -1068,7 +1071,8 @@ void Monitor::InflateThinLocked(Thread* self, Handle<mirror::Object> obj, LockWo
     Thread* owner;
     {
       ScopedThreadSuspension sts(self, ThreadState::kWaitingForLockInflation);
-      owner = thread_list->SuspendThreadByThreadId(owner_thread_id, SuspendReason::kInternal);
+      owner = thread_list->SuspendThreadByThreadId(
+          owner_thread_id, SuspendReason::kInternal, attempt_of_4);
     }
     if (owner != nullptr) {
       // We succeeded in suspending the thread, check the lock's status didn't change.
@@ -1107,6 +1111,7 @@ ObjPtr<mirror::Object> Monitor::MonitorEnter(Thread* self,
   uint32_t thread_id = self->GetThreadId();
   size_t contention_count = 0;
   constexpr size_t kExtraSpinIters = 100;
+  int inflation_attempt = 1;
   StackHandleScope<1> hs(self);
   Handle<mirror::Object> h_obj(hs.NewHandle(obj));
   while (true) {
@@ -1153,7 +1158,7 @@ ObjPtr<mirror::Object> Monitor::MonitorEnter(Thread* self,
             continue;  // Go again.
           } else {
             // We'd overflow the recursion count, so inflate the monitor.
-            InflateThinLocked(self, h_obj, lock_word, 0);
+            InflateThinLocked(self, h_obj, lock_word, 0, inflation_attempt++);
           }
         } else {
           if (trylock) {
@@ -1174,7 +1179,7 @@ ObjPtr<mirror::Object> Monitor::MonitorEnter(Thread* self,
           } else {
             contention_count = 0;
             // No ordering required for initial lockword read. Install rereads it anyway.
-            InflateThinLocked(self, h_obj, lock_word, 0);
+            InflateThinLocked(self, h_obj, lock_word, 0, inflation_attempt++);
           }
         }
         continue;  // Start from the beginning.
diff --git a/runtime/monitor.h b/runtime/monitor.h
index d142212e18..0ee2dbb765 100644
--- a/runtime/monitor.h
+++ b/runtime/monitor.h
@@ -156,8 +156,14 @@ class Monitor {
   }
 
   // Inflate the lock on obj. May fail to inflate for spurious reasons, always re-check.
-  static void InflateThinLocked(Thread* self, Handle<mirror::Object> obj, LockWord lock_word,
-                                uint32_t hash_code) REQUIRES_SHARED(Locks::mutator_lock_);
+  // attempt_of_4 is in 1..4 inclusive or 0. A non-zero value indicates that we are retrying
+  // up to 4 times, and should only abort on 4. Zero means we are only trying once, with the
+  // full suspend timeout instead of a quarter.
+  static void InflateThinLocked(Thread* self,
+                                Handle<mirror::Object> obj,
+                                LockWord lock_word,
+                                uint32_t hash_code,
+                                int attempt_of_4 = 0) REQUIRES_SHARED(Locks::mutator_lock_);
 
   // Not exclusive because ImageWriter calls this during a Heap::VisitObjects() that
   // does not allow a thread suspension in the middle. TODO: maybe make this exclusive.
diff --git a/runtime/native/dalvik_system_VMRuntime.cc b/runtime/native/dalvik_system_VMRuntime.cc
index e774f07dc9..593b98dc80 100644
--- a/runtime/native/dalvik_system_VMRuntime.cc
+++ b/runtime/native/dalvik_system_VMRuntime.cc
@@ -66,9 +66,7 @@ extern "C" void android_set_application_target_sdk_version(uint32_t version);
 #include "thread-inl.h"
 #include "thread_list.h"
 
-// TODO(260881207): should be HIDDEN, but some apps fail to launch
-// (e.g. b/319255249)
-namespace art {
+namespace art HIDDEN {
 
 using android::base::StringPrintf;
 
@@ -592,6 +590,15 @@ static JNINativeMethod gMethods[] = {
 };
 
 void register_dalvik_system_VMRuntime(JNIEnv* env) {
+  if (Runtime::Current()->GetTargetSdkVersion() <= static_cast<uint32_t>(SdkVersion::kU)) {
+    real_register_dalvik_system_VMRuntime(env);
+  } else {
+    Runtime::Current()->Abort(
+        "Call to internal function 'register_dalvik_system_VMRuntime' is not allowed");
+  }
+}
+
+void real_register_dalvik_system_VMRuntime(JNIEnv* env) {
   REGISTER_NATIVE_METHODS("dalvik/system/VMRuntime");
 }
 
diff --git a/runtime/native/dalvik_system_VMRuntime.h b/runtime/native/dalvik_system_VMRuntime.h
index d66c107a30..3876a60753 100644
--- a/runtime/native/dalvik_system_VMRuntime.h
+++ b/runtime/native/dalvik_system_VMRuntime.h
@@ -21,11 +21,15 @@
 
 #include "base/macros.h"
 
-// TODO(260881207): should be HIDDEN, but some apps fail to launch
-// (e.g. b/319255249)
-namespace art {
+namespace art HIDDEN {
 
-void register_dalvik_system_VMRuntime(JNIEnv* env);
+// TODO(260881207): register_dalvik_system_VMRuntime should be HIDDEN,
+// but some apps fail to launch (e.g. b/319255249).
+// The function is still exported for now, but it does a targetSdk check
+// and aborts for SdkVersion after U. Libart code should use
+// `real_register...` until exported function is removed.
+EXPORT void register_dalvik_system_VMRuntime(JNIEnv* env);
+void real_register_dalvik_system_VMRuntime(JNIEnv* env);
 
 }  // namespace art
 
diff --git a/runtime/native/jdk_internal_misc_Unsafe.cc b/runtime/native/jdk_internal_misc_Unsafe.cc
index 10c6b2da4f..ba64c818c6 100644
--- a/runtime/native/jdk_internal_misc_Unsafe.cc
+++ b/runtime/native/jdk_internal_misc_Unsafe.cc
@@ -491,8 +491,9 @@ static void Unsafe_unpark(JNIEnv* env, jobject, jobject jthread) {
     ThrowIllegalArgumentException("Argument to unpark() was not a Thread");
     return;
   }
-  art::MutexLock mu(soa.Self(), *art::Locks::thread_list_lock_);
-  art::Thread* thread = art::Thread::FromManagedThread(soa, mirror_thread);
+  Thread* self = soa.Self();
+  art::MutexLock mu(self, *art::Locks::thread_list_lock_);
+  art::Thread* thread = art::Thread::FromManagedThread(self, mirror_thread);
   if (thread != nullptr) {
     thread->Unpark();
   } else {
diff --git a/runtime/native/sun_misc_Unsafe.cc b/runtime/native/sun_misc_Unsafe.cc
index 573b5a9db2..38fe72555c 100644
--- a/runtime/native/sun_misc_Unsafe.cc
+++ b/runtime/native/sun_misc_Unsafe.cc
@@ -531,8 +531,9 @@ static void Unsafe_unpark(JNIEnv* env, jobject, jobject jthread) {
     ThrowIllegalArgumentException("Argument to unpark() was not a Thread");
     return;
   }
-  art::MutexLock mu(soa.Self(), *art::Locks::thread_list_lock_);
-  art::Thread* thread = art::Thread::FromManagedThread(soa, mirror_thread);
+  Thread* self = soa.Self();
+  art::MutexLock mu(self, *art::Locks::thread_list_lock_);
+  art::Thread* thread = art::Thread::FromManagedThread(self, mirror_thread);
   if (thread != nullptr) {
     thread->Unpark();
   } else {
diff --git a/runtime/nterp_helpers.cc b/runtime/nterp_helpers.cc
index cfe5610239..ba93df693d 100644
--- a/runtime/nterp_helpers.cc
+++ b/runtime/nterp_helpers.cc
@@ -234,244 +234,8 @@ bool CanMethodUseNterp(ArtMethod* method, InstructionSet isa) {
       method->IsProxyMethod()) {
     return false;
   }
-  if (isa == InstructionSet::kRiscv64) {
-    if (method->GetDexFile()->IsCompactDexFile()) {
-      return false;  // Riscv64 nterp does not support compact dex yet.
-    }
-    for (DexInstructionPcPair pair : method->DexInstructions()) {
-      // TODO(riscv64): Add support for more instructions.
-      // Remove the check when all instructions are supported.
-      // Cases are listed in opcode order (DEX_INSTRUCTION_LIST).
-      switch (pair->Opcode()) {
-        case Instruction::NOP:
-        case Instruction::MOVE:
-        case Instruction::MOVE_FROM16:
-        case Instruction::MOVE_16:
-        case Instruction::MOVE_WIDE:
-        case Instruction::MOVE_WIDE_FROM16:
-        case Instruction::MOVE_WIDE_16:
-        case Instruction::MOVE_OBJECT:
-        case Instruction::MOVE_OBJECT_FROM16:
-        case Instruction::MOVE_OBJECT_16:
-        case Instruction::MOVE_RESULT:
-        case Instruction::MOVE_RESULT_WIDE:
-        case Instruction::MOVE_RESULT_OBJECT:
-        case Instruction::MOVE_EXCEPTION:
-        case Instruction::RETURN_VOID:
-        case Instruction::RETURN:
-        case Instruction::RETURN_WIDE:
-        case Instruction::RETURN_OBJECT:
-        case Instruction::CONST_4:
-        case Instruction::CONST_16:
-        case Instruction::CONST:
-        case Instruction::CONST_HIGH16:
-        case Instruction::CONST_WIDE_16:
-        case Instruction::CONST_WIDE_32:
-        case Instruction::CONST_WIDE:
-        case Instruction::CONST_WIDE_HIGH16:
-        case Instruction::CONST_STRING:
-        case Instruction::CONST_STRING_JUMBO:
-        case Instruction::CONST_CLASS:
-        case Instruction::MONITOR_ENTER:
-        case Instruction::MONITOR_EXIT:
-        case Instruction::CHECK_CAST:
-        case Instruction::INSTANCE_OF:
-        case Instruction::ARRAY_LENGTH:
-        case Instruction::NEW_INSTANCE:
-        case Instruction::NEW_ARRAY:
-        case Instruction::FILLED_NEW_ARRAY:
-        case Instruction::FILLED_NEW_ARRAY_RANGE:
-        case Instruction::FILL_ARRAY_DATA:
-        case Instruction::THROW:
-        case Instruction::GOTO:
-        case Instruction::GOTO_16:
-        case Instruction::GOTO_32:
-        case Instruction::PACKED_SWITCH:
-        case Instruction::SPARSE_SWITCH:
-        case Instruction::CMPL_FLOAT:
-        case Instruction::CMPG_FLOAT:
-        case Instruction::CMPL_DOUBLE:
-        case Instruction::CMPG_DOUBLE:
-        case Instruction::CMP_LONG:
-        case Instruction::IF_EQ:
-        case Instruction::IF_NE:
-        case Instruction::IF_LT:
-        case Instruction::IF_GE:
-        case Instruction::IF_GT:
-        case Instruction::IF_LE:
-        case Instruction::IF_EQZ:
-        case Instruction::IF_NEZ:
-        case Instruction::IF_LTZ:
-        case Instruction::IF_GEZ:
-        case Instruction::IF_GTZ:
-        case Instruction::IF_LEZ:
-        case Instruction::AGET:
-        case Instruction::AGET_WIDE:
-        case Instruction::AGET_OBJECT:
-        case Instruction::AGET_BOOLEAN:
-        case Instruction::AGET_BYTE:
-        case Instruction::AGET_CHAR:
-        case Instruction::AGET_SHORT:
-        case Instruction::APUT:
-        case Instruction::APUT_WIDE:
-        case Instruction::APUT_OBJECT:
-        case Instruction::APUT_BOOLEAN:
-        case Instruction::APUT_BYTE:
-        case Instruction::APUT_CHAR:
-        case Instruction::APUT_SHORT:
-        case Instruction::IGET:
-        case Instruction::IGET_WIDE:
-        case Instruction::IGET_OBJECT:
-        case Instruction::IGET_BOOLEAN:
-        case Instruction::IGET_BYTE:
-        case Instruction::IGET_CHAR:
-        case Instruction::IGET_SHORT:
-        case Instruction::IPUT:
-        case Instruction::IPUT_WIDE:
-        case Instruction::IPUT_OBJECT:
-        case Instruction::IPUT_BOOLEAN:
-        case Instruction::IPUT_BYTE:
-        case Instruction::IPUT_CHAR:
-        case Instruction::IPUT_SHORT:
-        case Instruction::SGET:
-        case Instruction::SGET_WIDE:
-        case Instruction::SGET_OBJECT:
-        case Instruction::SGET_BOOLEAN:
-        case Instruction::SGET_BYTE:
-        case Instruction::SGET_CHAR:
-        case Instruction::SGET_SHORT:
-        case Instruction::SPUT:
-        case Instruction::SPUT_WIDE:
-        case Instruction::SPUT_OBJECT:
-        case Instruction::SPUT_BOOLEAN:
-        case Instruction::SPUT_BYTE:
-        case Instruction::SPUT_CHAR:
-        case Instruction::SPUT_SHORT:
-        case Instruction::INVOKE_VIRTUAL:
-        case Instruction::INVOKE_SUPER:
-        case Instruction::INVOKE_DIRECT:
-        case Instruction::INVOKE_STATIC:
-        case Instruction::INVOKE_INTERFACE:
-        case Instruction::INVOKE_VIRTUAL_RANGE:
-        case Instruction::INVOKE_SUPER_RANGE:
-        case Instruction::INVOKE_DIRECT_RANGE:
-        case Instruction::INVOKE_STATIC_RANGE:
-        case Instruction::INVOKE_INTERFACE_RANGE:
-        case Instruction::NEG_INT:
-        case Instruction::NOT_INT:
-        case Instruction::NEG_LONG:
-        case Instruction::NOT_LONG:
-        case Instruction::NEG_FLOAT:
-        case Instruction::NEG_DOUBLE:
-        case Instruction::INT_TO_LONG:
-        case Instruction::INT_TO_FLOAT:
-        case Instruction::INT_TO_DOUBLE:
-        case Instruction::LONG_TO_INT:
-        case Instruction::LONG_TO_FLOAT:
-        case Instruction::LONG_TO_DOUBLE:
-        case Instruction::FLOAT_TO_INT:
-        case Instruction::FLOAT_TO_LONG:
-        case Instruction::FLOAT_TO_DOUBLE:
-        case Instruction::DOUBLE_TO_INT:
-        case Instruction::DOUBLE_TO_LONG:
-        case Instruction::DOUBLE_TO_FLOAT:
-        case Instruction::INT_TO_BYTE:
-        case Instruction::INT_TO_CHAR:
-        case Instruction::INT_TO_SHORT:
-        case Instruction::ADD_INT:
-        case Instruction::SUB_INT:
-        case Instruction::MUL_INT:
-        case Instruction::DIV_INT:
-        case Instruction::REM_INT:
-        case Instruction::AND_INT:
-        case Instruction::OR_INT:
-        case Instruction::XOR_INT:
-        case Instruction::SHL_INT:
-        case Instruction::SHR_INT:
-        case Instruction::USHR_INT:
-        case Instruction::ADD_LONG:
-        case Instruction::SUB_LONG:
-        case Instruction::MUL_LONG:
-        case Instruction::DIV_LONG:
-        case Instruction::REM_LONG:
-        case Instruction::AND_LONG:
-        case Instruction::OR_LONG:
-        case Instruction::XOR_LONG:
-        case Instruction::SHL_LONG:
-        case Instruction::SHR_LONG:
-        case Instruction::USHR_LONG:
-        case Instruction::ADD_FLOAT:
-        case Instruction::SUB_FLOAT:
-        case Instruction::MUL_FLOAT:
-        case Instruction::DIV_FLOAT:
-        case Instruction::REM_FLOAT:
-        case Instruction::ADD_DOUBLE:
-        case Instruction::SUB_DOUBLE:
-        case Instruction::MUL_DOUBLE:
-        case Instruction::DIV_DOUBLE:
-        case Instruction::REM_DOUBLE:
-        case Instruction::ADD_INT_2ADDR:
-        case Instruction::SUB_INT_2ADDR:
-        case Instruction::MUL_INT_2ADDR:
-        case Instruction::DIV_INT_2ADDR:
-        case Instruction::REM_INT_2ADDR:
-        case Instruction::AND_INT_2ADDR:
-        case Instruction::OR_INT_2ADDR:
-        case Instruction::XOR_INT_2ADDR:
-        case Instruction::SHL_INT_2ADDR:
-        case Instruction::SHR_INT_2ADDR:
-        case Instruction::USHR_INT_2ADDR:
-        case Instruction::ADD_LONG_2ADDR:
-        case Instruction::SUB_LONG_2ADDR:
-        case Instruction::MUL_LONG_2ADDR:
-        case Instruction::DIV_LONG_2ADDR:
-        case Instruction::REM_LONG_2ADDR:
-        case Instruction::AND_LONG_2ADDR:
-        case Instruction::OR_LONG_2ADDR:
-        case Instruction::XOR_LONG_2ADDR:
-        case Instruction::SHL_LONG_2ADDR:
-        case Instruction::SHR_LONG_2ADDR:
-        case Instruction::USHR_LONG_2ADDR:
-        case Instruction::ADD_FLOAT_2ADDR:
-        case Instruction::SUB_FLOAT_2ADDR:
-        case Instruction::MUL_FLOAT_2ADDR:
-        case Instruction::DIV_FLOAT_2ADDR:
-        case Instruction::REM_FLOAT_2ADDR:
-        case Instruction::ADD_DOUBLE_2ADDR:
-        case Instruction::SUB_DOUBLE_2ADDR:
-        case Instruction::MUL_DOUBLE_2ADDR:
-        case Instruction::DIV_DOUBLE_2ADDR:
-        case Instruction::REM_DOUBLE_2ADDR:
-        case Instruction::ADD_INT_LIT16:
-        case Instruction::RSUB_INT:
-        case Instruction::MUL_INT_LIT16:
-        case Instruction::DIV_INT_LIT16:
-        case Instruction::REM_INT_LIT16:
-        case Instruction::AND_INT_LIT16:
-        case Instruction::OR_INT_LIT16:
-        case Instruction::XOR_INT_LIT16:
-        case Instruction::ADD_INT_LIT8:
-        case Instruction::RSUB_INT_LIT8:
-        case Instruction::MUL_INT_LIT8:
-        case Instruction::DIV_INT_LIT8:
-        case Instruction::REM_INT_LIT8:
-        case Instruction::AND_INT_LIT8:
-        case Instruction::OR_INT_LIT8:
-        case Instruction::XOR_INT_LIT8:
-        case Instruction::SHL_INT_LIT8:
-        case Instruction::SHR_INT_LIT8:
-        case Instruction::USHR_INT_LIT8:
-        case Instruction::INVOKE_POLYMORPHIC:
-        case Instruction::INVOKE_POLYMORPHIC_RANGE:
-        case Instruction::INVOKE_CUSTOM:
-        case Instruction::INVOKE_CUSTOM_RANGE:
-        case Instruction::CONST_METHOD_HANDLE:
-        case Instruction::CONST_METHOD_TYPE:
-          continue;
-        default:
-          return false;
-      }
-    }
+  if (isa == InstructionSet::kRiscv64 && method->GetDexFile()->IsCompactDexFile()) {
+    return false;  // Riscv64 nterp does not support compact dex yet.
   }
   // There is no need to add the alignment padding size for comparison with aligned limit.
   size_t frame_size_without_padding = NterpGetFrameSizeWithoutPadding(method, isa);
diff --git a/runtime/oat/elf_file.cc b/runtime/oat/elf_file.cc
index 96d7d0f4ac..30e0197470 100644
--- a/runtime/oat/elf_file.cc
+++ b/runtime/oat/elf_file.cc
@@ -1082,6 +1082,8 @@ static InstructionSet GetInstructionSetFromELF(uint16_t e_machine,
       return InstructionSet::kArm;
     case EM_AARCH64:
       return InstructionSet::kArm64;
+    case EM_RISCV:
+      return InstructionSet::kRiscv64;
     case EM_386:
       return InstructionSet::kX86;
     case EM_X86_64:
diff --git a/runtime/reflection.cc b/runtime/reflection.cc
index f368b5478a..c292f2a7f8 100644
--- a/runtime/reflection.cc
+++ b/runtime/reflection.cc
@@ -34,7 +34,7 @@
 #include "scoped_thread_state_change-inl.h"
 #include "stack_reference.h"
 #include "thread-inl.h"
-#include "well_known_classes.h"
+#include "well_known_classes-inl.h"
 
 namespace art HIDDEN {
 namespace {
@@ -259,15 +259,15 @@ class ArgArray {
         }
       }
 
-#define DO_FIRST_ARG(match_descriptor, get_fn, append) { \
+#define DO_FIRST_ARG(boxed, get_fn, append) { \
           if (LIKELY(arg != nullptr && \
-              arg->GetClass()->DescriptorEquals(match_descriptor))) { \
+                     arg->GetClass() == WellKnownClasses::java_lang_##boxed)) { \
             ArtField* primitive_field = arg->GetClass()->GetInstanceField(0); \
             append(primitive_field-> get_fn(arg.Get()));
 
-#define DO_ARG(match_descriptor, get_fn, append) \
+#define DO_ARG(boxed, get_fn, append) \
           } else if (LIKELY(arg != nullptr && \
-                            arg->GetClass<>()->DescriptorEquals(match_descriptor))) { \
+                            arg->GetClass() == WellKnownClasses::java_lang_##boxed)) { \
             ArtField* primitive_field = arg->GetClass()->GetInstanceField(0); \
             append(primitive_field-> get_fn(arg.Get()));
 
@@ -293,54 +293,54 @@ class ArgArray {
           Append(arg.Get());
           break;
         case 'Z':
-          DO_FIRST_ARG("Ljava/lang/Boolean;", GetBoolean, Append)
+          DO_FIRST_ARG(Boolean, GetBoolean, Append)
           DO_FAIL("boolean")
           break;
         case 'B':
-          DO_FIRST_ARG("Ljava/lang/Byte;", GetByte, Append)
+          DO_FIRST_ARG(Byte, GetByte, Append)
           DO_FAIL("byte")
           break;
         case 'C':
-          DO_FIRST_ARG("Ljava/lang/Character;", GetChar, Append)
+          DO_FIRST_ARG(Character, GetChar, Append)
           DO_FAIL("char")
           break;
         case 'S':
-          DO_FIRST_ARG("Ljava/lang/Short;", GetShort, Append)
-          DO_ARG("Ljava/lang/Byte;", GetByte, Append)
+          DO_FIRST_ARG(Short, GetShort, Append)
+          DO_ARG(Byte, GetByte, Append)
           DO_FAIL("short")
           break;
         case 'I':
-          DO_FIRST_ARG("Ljava/lang/Integer;", GetInt, Append)
-          DO_ARG("Ljava/lang/Character;", GetChar, Append)
-          DO_ARG("Ljava/lang/Short;", GetShort, Append)
-          DO_ARG("Ljava/lang/Byte;", GetByte, Append)
+          DO_FIRST_ARG(Integer, GetInt, Append)
+          DO_ARG(Character, GetChar, Append)
+          DO_ARG(Short, GetShort, Append)
+          DO_ARG(Byte, GetByte, Append)
           DO_FAIL("int")
           break;
         case 'J':
-          DO_FIRST_ARG("Ljava/lang/Long;", GetLong, AppendWide)
-          DO_ARG("Ljava/lang/Integer;", GetInt, AppendWide)
-          DO_ARG("Ljava/lang/Character;", GetChar, AppendWide)
-          DO_ARG("Ljava/lang/Short;", GetShort, AppendWide)
-          DO_ARG("Ljava/lang/Byte;", GetByte, AppendWide)
+          DO_FIRST_ARG(Long, GetLong, AppendWide)
+          DO_ARG(Integer, GetInt, AppendWide)
+          DO_ARG(Character, GetChar, AppendWide)
+          DO_ARG(Short, GetShort, AppendWide)
+          DO_ARG(Byte, GetByte, AppendWide)
           DO_FAIL("long")
           break;
         case 'F':
-          DO_FIRST_ARG("Ljava/lang/Float;", GetFloat, AppendFloat)
-          DO_ARG("Ljava/lang/Long;", GetLong, AppendFloat)
-          DO_ARG("Ljava/lang/Integer;", GetInt, AppendFloat)
-          DO_ARG("Ljava/lang/Character;", GetChar, AppendFloat)
-          DO_ARG("Ljava/lang/Short;", GetShort, AppendFloat)
-          DO_ARG("Ljava/lang/Byte;", GetByte, AppendFloat)
+          DO_FIRST_ARG(Float, GetFloat, AppendFloat)
+          DO_ARG(Long, GetLong, AppendFloat)
+          DO_ARG(Integer, GetInt, AppendFloat)
+          DO_ARG(Character, GetChar, AppendFloat)
+          DO_ARG(Short, GetShort, AppendFloat)
+          DO_ARG(Byte, GetByte, AppendFloat)
           DO_FAIL("float")
           break;
         case 'D':
-          DO_FIRST_ARG("Ljava/lang/Double;", GetDouble, AppendDouble)
-          DO_ARG("Ljava/lang/Float;", GetFloat, AppendDouble)
-          DO_ARG("Ljava/lang/Long;", GetLong, AppendDouble)
-          DO_ARG("Ljava/lang/Integer;", GetInt, AppendDouble)
-          DO_ARG("Ljava/lang/Character;", GetChar, AppendDouble)
-          DO_ARG("Ljava/lang/Short;", GetShort, AppendDouble)
-          DO_ARG("Ljava/lang/Byte;", GetByte, AppendDouble)
+          DO_FIRST_ARG(Double, GetDouble, AppendDouble)
+          DO_ARG(Float, GetFloat, AppendDouble)
+          DO_ARG(Long, GetLong, AppendDouble)
+          DO_ARG(Integer, GetInt, AppendDouble)
+          DO_ARG(Character, GetChar, AppendDouble)
+          DO_ARG(Short, GetShort, AppendDouble)
+          DO_ARG(Byte, GetByte, AppendDouble)
           DO_FAIL("double")
           break;
 #ifndef NDEBUG
@@ -952,28 +952,28 @@ static bool UnboxPrimitive(ObjPtr<mirror::Object> o,
   ObjPtr<mirror::Class> klass = o->GetClass();
   Primitive::Type primitive_type;
   ArtField* primitive_field = &klass->GetIFieldsPtr()->At(0);
-  if (klass->DescriptorEquals("Ljava/lang/Boolean;")) {
+  if (klass == WellKnownClasses::java_lang_Boolean) {
     primitive_type = Primitive::kPrimBoolean;
     boxed_value.SetZ(primitive_field->GetBoolean(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Byte;")) {
+  } else if (klass == WellKnownClasses::java_lang_Byte) {
     primitive_type = Primitive::kPrimByte;
     boxed_value.SetB(primitive_field->GetByte(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Character;")) {
+  } else if (klass == WellKnownClasses::java_lang_Character) {
     primitive_type = Primitive::kPrimChar;
     boxed_value.SetC(primitive_field->GetChar(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Float;")) {
+  } else if (klass == WellKnownClasses::java_lang_Float) {
     primitive_type = Primitive::kPrimFloat;
     boxed_value.SetF(primitive_field->GetFloat(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Double;")) {
+  } else if (klass == WellKnownClasses::java_lang_Double) {
     primitive_type = Primitive::kPrimDouble;
     boxed_value.SetD(primitive_field->GetDouble(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Integer;")) {
+  } else if (klass == WellKnownClasses::java_lang_Integer) {
     primitive_type = Primitive::kPrimInt;
     boxed_value.SetI(primitive_field->GetInt(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Long;")) {
+  } else if (klass == WellKnownClasses::java_lang_Long) {
     primitive_type = Primitive::kPrimLong;
     boxed_value.SetJ(primitive_field->GetLong(o));
-  } else if (klass->DescriptorEquals("Ljava/lang/Short;")) {
+  } else if (klass == WellKnownClasses::java_lang_Short) {
     primitive_type = Primitive::kPrimShort;
     boxed_value.SetS(primitive_field->GetShort(o));
   } else {
diff --git a/runtime/runtime.cc b/runtime/runtime.cc
index e6f8c576dc..b7587c0a91 100644
--- a/runtime/runtime.cc
+++ b/runtime/runtime.cc
@@ -1317,7 +1317,7 @@ void Runtime::InitNonZygoteOrPostFork(
     if (!odrefresh::UploadStatsIfAvailable(&err)) {
       LOG(WARNING) << "Failed to upload odrefresh metrics: " << err;
     }
-    metrics::SetupCallbackForDeviceStatus();
+    metrics::ReportDeviceMetrics();
   }
 
   if (LIKELY(automatically_set_jni_ids_indirection_) && CanSetJniIdType()) {
@@ -2381,7 +2381,7 @@ void Runtime::RegisterRuntimeNativeMethods(JNIEnv* env) {
   register_dalvik_system_DexFile(env);
   register_dalvik_system_BaseDexClassLoader(env);
   register_dalvik_system_VMDebug(env);
-  register_dalvik_system_VMRuntime(env);
+  real_register_dalvik_system_VMRuntime(env);
   register_dalvik_system_VMStack(env);
   register_dalvik_system_ZygoteHooks(env);
   register_java_lang_Class(env);
diff --git a/runtime/runtime_image.cc b/runtime/runtime_image.cc
index 0c85261829..5d304698bf 100644
--- a/runtime/runtime_image.cc
+++ b/runtime/runtime_image.cc
@@ -23,6 +23,7 @@
 #include "android-base/stringprintf.h"
 #include "android-base/strings.h"
 #include "arch/instruction_set.h"
+#include "arch/instruction_set_features.h"
 #include "base/arena_allocator.h"
 #include "base/arena_containers.h"
 #include "base/bit_utils.h"
@@ -603,8 +604,11 @@ class RuntimeImageHelper {
     }
 
     for (Handle<mirror::Class> cls : classes_to_write) {
-      ScopedAssertNoThreadSuspension sants("Writing class");
-      CopyClass(cls.Get());
+      {
+        ScopedAssertNoThreadSuspension sants("Writing class");
+        CopyClass(cls.Get());
+      }
+      self->AllowThreadSuspension();
     }
 
     // Relocate the type array entries. We do this now before creating image
diff --git a/runtime/runtime_options.h b/runtime/runtime_options.h
index 6ece459dad..3cadd09bf8 100644
--- a/runtime/runtime_options.h
+++ b/runtime/runtime_options.h
@@ -29,8 +29,8 @@
 #include "gc/collector_type.h"
 #include "gc/space/large_object_space.h"
 #include "hidden_api.h"
-#include "jit/jit.h"
 #include "jit/jit_code_cache.h"
+#include "jit/jit_options.h"
 #include "jit/profile_saver_options.h"
 #include "verifier/verifier_enums.h"
 
diff --git a/runtime/thread-inl.h b/runtime/thread-inl.h
index 83ab469c68..2fcc4b065b 100644
--- a/runtime/thread-inl.h
+++ b/runtime/thread-inl.h
@@ -284,10 +284,20 @@ inline void Thread::AddSuspend1Barrier(WrappedSuspend1Barrier* suspend1_barrier)
   tlsPtr_.active_suspend1_barriers = suspend1_barrier;
 }
 
-inline void Thread::RemoveFirstSuspend1Barrier() {
+inline void Thread::RemoveFirstSuspend1Barrier(WrappedSuspend1Barrier* suspend1_barrier) {
+  DCHECK_EQ(tlsPtr_.active_suspend1_barriers, suspend1_barrier);
   tlsPtr_.active_suspend1_barriers = tlsPtr_.active_suspend1_barriers->next_;
 }
 
+inline void Thread::RemoveSuspend1Barrier(WrappedSuspend1Barrier* barrier) {
+  // 'barrier' should be in the list. If not, we will get a SIGSEGV with fault address of 4 or 8.
+  WrappedSuspend1Barrier** last = &tlsPtr_.active_suspend1_barriers;
+  while (*last != barrier) {
+    last = &((*last)->next_);
+  }
+  *last = (*last)->next_;
+}
+
 inline bool Thread::HasActiveSuspendBarrier() {
   return tlsPtr_.active_suspend1_barriers != nullptr ||
          tlsPtr_.active_suspendall_barrier != nullptr;
diff --git a/runtime/thread.cc b/runtime/thread.cc
index e1ee9007bf..03632887d0 100644
--- a/runtime/thread.cc
+++ b/runtime/thread.cc
@@ -675,16 +675,15 @@ void* Thread::CreateCallback(void* arg) {
   return nullptr;
 }
 
-Thread* Thread::FromManagedThread(const ScopedObjectAccessAlreadyRunnable& soa,
-                                  ObjPtr<mirror::Object> thread_peer) {
+Thread* Thread::FromManagedThread(Thread* self, ObjPtr<mirror::Object> thread_peer) {
   ArtField* f = WellKnownClasses::java_lang_Thread_nativePeer;
   Thread* result = reinterpret_cast64<Thread*>(f->GetLong(thread_peer));
   // Check that if we have a result it is either suspended or we hold the thread_list_lock_
   // to stop it from going away.
   if (kIsDebugBuild) {
-    MutexLock mu(soa.Self(), *Locks::thread_suspend_count_lock_);
+    MutexLock mu(self, *Locks::thread_suspend_count_lock_);
     if (result != nullptr && !result->IsSuspended()) {
-      Locks::thread_list_lock_->AssertHeld(soa.Self());
+      Locks::thread_list_lock_->AssertHeld(self);
     }
   }
   return result;
@@ -692,7 +691,7 @@ Thread* Thread::FromManagedThread(const ScopedObjectAccessAlreadyRunnable& soa,
 
 Thread* Thread::FromManagedThread(const ScopedObjectAccessAlreadyRunnable& soa,
                                   jobject java_thread) {
-  return FromManagedThread(soa, soa.Decode<mirror::Object>(java_thread));
+  return FromManagedThread(soa.Self(), soa.Decode<mirror::Object>(java_thread));
 }
 
 static size_t FixStackSize(size_t stack_size) {
@@ -1518,20 +1517,27 @@ bool Thread::PassActiveSuspendBarriers() {
     }
     tlsPtr_.active_suspend1_barriers = nullptr;
     AtomicClearFlag(ThreadFlag::kActiveSuspendBarrier);
+    CHECK_GT(pass_barriers.size(), 0U);  // Since kActiveSuspendBarrier was set.
+    // Decrement suspend barrier(s) while we still hold the lock, since SuspendThread may
+    // remove and deallocate suspend barriers while holding suspend_count_lock_ .
+    // There will typically only be a single barrier to pass here.
+    for (AtomicInteger*& barrier : pass_barriers) {
+      int32_t old_val = barrier->fetch_sub(1, std::memory_order_release);
+      CHECK_GT(old_val, 0) << "Unexpected value for PassActiveSuspendBarriers(): " << old_val;
+      if (old_val != 1) {
+        // We're done with it.
+        barrier = nullptr;
+      }
+    }
   }
-
-  uint32_t barrier_count = 0;
+  // Finally do futex_wakes after releasing the lock.
   for (AtomicInteger* barrier : pass_barriers) {
-    ++barrier_count;
-    int32_t old_val = barrier->fetch_sub(1, std::memory_order_release);
-    CHECK_GT(old_val, 0) << "Unexpected value for PassActiveSuspendBarriers(): " << old_val;
 #if ART_USE_FUTEXES
-    if (old_val == 1) {
+    if (barrier != nullptr) {
       futex(barrier->Address(), FUTEX_WAKE_PRIVATE, INT_MAX, nullptr, nullptr, 0);
     }
 #endif
   }
-  CHECK_GT(barrier_count, 0U);
   return true;
 }
 
@@ -1721,7 +1727,7 @@ bool Thread::RequestSynchronousCheckpoint(Closure* function, ThreadState wait_st
       Locks::thread_list_lock_->ExclusiveUnlock(self);
       if (IsSuspended()) {
         // See the discussion in mutator_gc_coord.md and SuspendAllInternal for the race here.
-        RemoveFirstSuspend1Barrier();
+        RemoveFirstSuspend1Barrier(&wrapped_barrier);
         if (!HasActiveSuspendBarrier()) {
           AtomicClearFlag(ThreadFlag::kActiveSuspendBarrier);
         }
@@ -2613,10 +2619,6 @@ void Thread::Destroy(bool should_run_callbacks) {
       runtime->GetRuntimeCallbacks()->ThreadDeath(self);
     }
 
-    if (UNLIKELY(self->GetMethodTraceBuffer() != nullptr)) {
-      Trace::FlushThreadBuffer(self);
-    }
-
     // this.nativePeer = 0;
     SetNativePeer</*kSupportTransaction=*/ true>(tlsPtr_.opeer, nullptr);
 
@@ -2631,12 +2633,17 @@ void Thread::Destroy(bool should_run_callbacks) {
       ObjectLock<mirror::Object> locker(self, h_obj);
       locker.NotifyAll();
     }
+
     tlsPtr_.opeer = nullptr;
   }
 
   {
     ScopedObjectAccess soa(self);
     Runtime::Current()->GetHeap()->RevokeThreadLocalBuffers(this);
+
+    if (UNLIKELY(self->GetMethodTraceBuffer() != nullptr)) {
+      Trace::FlushThreadBuffer(self);
+    }
   }
   // Mark-stack revocation must be performed at the very end. No
   // checkpoint/flip-function or read-barrier should be called after this.
@@ -2685,9 +2692,7 @@ Thread::~Thread() {
   SetCachedThreadName(nullptr);  // Deallocate name.
   delete tlsPtr_.deps_or_stack_trace_sample.stack_trace_sample;
 
-  if (tlsPtr_.method_trace_buffer != nullptr) {
-    delete[] tlsPtr_.method_trace_buffer;
-  }
+  CHECK_EQ(tlsPtr_.method_trace_buffer, nullptr);
 
   Runtime::Current()->GetHeap()->AssertThreadLocalBuffersAreRevoked(this);
 
diff --git a/runtime/thread.h b/runtime/thread.h
index 5a9c55aa14..a59b10ae13 100644
--- a/runtime/thread.h
+++ b/runtime/thread.h
@@ -200,7 +200,7 @@ enum class WeakRefAccessState : int32_t {
 // See Thread.tlsPtr_.active_suspend1_barriers below for explanation.
 struct WrappedSuspend1Barrier {
   WrappedSuspend1Barrier() : barrier_(1), next_(nullptr) {}
-  AtomicInteger barrier_;
+  AtomicInteger barrier_;  // Only updated while holding thread_suspend_count_lock_ .
   struct WrappedSuspend1Barrier* next_ GUARDED_BY(Locks::thread_suspend_count_lock_);
 };
 
@@ -300,8 +300,7 @@ class EXPORT Thread {
   void CheckEmptyCheckpointFromWeakRefAccess(BaseMutex* cond_var_mutex);
   void CheckEmptyCheckpointFromMutex();
 
-  static Thread* FromManagedThread(const ScopedObjectAccessAlreadyRunnable& ts,
-                                   ObjPtr<mirror::Object> thread_peer)
+  static Thread* FromManagedThread(Thread* self, ObjPtr<mirror::Object> thread_peer)
       REQUIRES(Locks::thread_list_lock_, !Locks::thread_suspend_count_lock_)
       REQUIRES_SHARED(Locks::mutator_lock_);
   static Thread* FromManagedThread(const ScopedObjectAccessAlreadyRunnable& ts, jobject thread)
@@ -1762,7 +1761,14 @@ class EXPORT Thread {
 
   // Remove last-added entry from active_suspend1_barriers.
   // Only makes sense if we're still holding thread_suspend_count_lock_ since insertion.
-  ALWAYS_INLINE void RemoveFirstSuspend1Barrier() REQUIRES(Locks::thread_suspend_count_lock_);
+  // We redundantly pass in the barrier to be removed in order to enable a DCHECK.
+  ALWAYS_INLINE void RemoveFirstSuspend1Barrier(WrappedSuspend1Barrier* suspend1_barrier)
+      REQUIRES(Locks::thread_suspend_count_lock_);
+
+  // Remove the "barrier" from the list no matter where it appears. Called only under exceptional
+  // circumstances. The barrier must be in the list.
+  ALWAYS_INLINE void RemoveSuspend1Barrier(WrappedSuspend1Barrier* suspend1_barrier)
+      REQUIRES(Locks::thread_suspend_count_lock_);
 
   ALWAYS_INLINE bool HasActiveSuspendBarrier() REQUIRES(Locks::thread_suspend_count_lock_);
 
@@ -1946,7 +1952,7 @@ class EXPORT Thread {
   // first if possible.
   /***********************************************************************************************/
 
-  struct PACKED(4) tls_32bit_sized_values {
+  struct alignas(4) tls_32bit_sized_values {
     // We have no control over the size of 'bool', but want our boolean fields
     // to be 4-byte quantities.
     using bool32_t = uint32_t;
@@ -2074,7 +2080,7 @@ class EXPORT Thread {
     uint32_t shared_method_hotness;
   } tls32_;
 
-  struct PACKED(8) tls_64bit_sized_values {
+  struct alignas(8) tls_64bit_sized_values {
     tls_64bit_sized_values() : trace_clock_base(0) {
     }
 
@@ -2084,7 +2090,7 @@ class EXPORT Thread {
     RuntimeStats stats;
   } tls64_;
 
-  struct PACKED(sizeof(void*)) tls_ptr_sized_values {
+  struct alignas(sizeof(void*)) tls_ptr_sized_values {
       tls_ptr_sized_values() : card_table(nullptr),
                                exception(nullptr),
                                stack_end(nullptr),
diff --git a/runtime/thread_list.cc b/runtime/thread_list.cc
index d5face195d..5e63b27b20 100644
--- a/runtime/thread_list.cc
+++ b/runtime/thread_list.cc
@@ -659,8 +659,9 @@ void ThreadList::FlipThreadRoots(Closure* thread_flip_visitor,
 // failures are expected.
 static constexpr bool kShortSuspendTimeouts = false;
 
+static constexpr unsigned kSuspendBarrierIters = kShortSuspendTimeouts ? 5 : 20;
+
 #if ART_USE_FUTEXES
-static constexpr int kSuspendBarrierIters = 5;
 
 // Returns true if it timed out.
 static bool WaitOnceForSuspendBarrier(AtomicInteger* barrier,
@@ -669,8 +670,9 @@ static bool WaitOnceForSuspendBarrier(AtomicInteger* barrier,
   timespec wait_timeout;
   if (kShortSuspendTimeouts) {
     timeout_ns = MsToNs(kSuspendBarrierIters);
+    CHECK_GE(NsToMs(timeout_ns / kSuspendBarrierIters), 1ul);
   } else {
-    DCHECK_GE(NsToMs(timeout_ns / kSuspendBarrierIters), 100ul);
+    DCHECK_GE(NsToMs(timeout_ns / kSuspendBarrierIters), 10ul);
   }
   InitTimeSpec(false, CLOCK_MONOTONIC, NsToMs(timeout_ns / kSuspendBarrierIters), 0, &wait_timeout);
   if (futex(barrier->Address(), FUTEX_WAIT_PRIVATE, cur_val, &wait_timeout, nullptr, 0) != 0) {
@@ -684,16 +686,15 @@ static bool WaitOnceForSuspendBarrier(AtomicInteger* barrier,
 }
 
 #else
-static constexpr int kSuspendBarrierIters = 10;
 
 static bool WaitOnceForSuspendBarrier(AtomicInteger* barrier,
                                       int32_t cur_val,
                                       uint64_t timeout_ns) {
-  static constexpr int kIters = kShortSuspendTimeouts ? 10'000 : 1'000'000;
-  if (!kShortSuspendTimeouts) {
-    DCHECK_GE(NsToMs(timeout_ns / kSuspendBarrierIters), 100ul);
-  }
-  for (int i = 0; i < kIters; ++i) {
+  // In the normal case, aim for a couple of hundred milliseconds.
+  static constexpr unsigned kInnerIters =
+      kShortSuspendTimeouts ? 1'000 : (timeout_ns / 1000) / kSuspendBarrierIters;
+  DCHECK_GE(kInnerIters, 1'000u);
+  for (int i = 0; i < kInnerIters; ++i) {
     sched_yield();
     if (barrier->load(std::memory_order_acquire) == 0) {
       return false;
@@ -701,7 +702,8 @@ static bool WaitOnceForSuspendBarrier(AtomicInteger* barrier,
   }
   return true;
 }
-#endif
+
+#endif  // ART_USE_FUTEXES
 
 // Return a short string describing the scheduling state of the thread with the given tid.
 static std::string GetThreadState(pid_t t) {
@@ -726,18 +728,23 @@ static std::string GetThreadState(pid_t t) {
 #endif
 }
 
-std::optional<std::string> ThreadList::WaitForSuspendBarrier(AtomicInteger* barrier, pid_t t) {
+std::optional<std::string> ThreadList::WaitForSuspendBarrier(AtomicInteger* barrier,
+                                                             pid_t t,
+                                                             int attempt_of_4) {
   // Only fail after kIter timeouts, to make us robust against app freezing.
 #if ART_USE_FUTEXES
   const uint64_t start_time = NanoTime();
 #endif
+  uint64_t timeout_ns =
+      attempt_of_4 == 0 ? thread_suspend_timeout_ns_ : thread_suspend_timeout_ns_ / 4;
+  bool collect_state = (t != 0 && (attempt_of_4 == 0 || attempt_of_4 == 4));
   int32_t cur_val = barrier->load(std::memory_order_acquire);
   if (cur_val <= 0) {
     DCHECK_EQ(cur_val, 0);
     return std::nullopt;
   }
-  int i = 0;
-  if (WaitOnceForSuspendBarrier(barrier, cur_val, thread_suspend_timeout_ns_)) {
+  unsigned i = 0;
+  if (WaitOnceForSuspendBarrier(barrier, cur_val, timeout_ns)) {
     i = 1;
   }
   cur_val = barrier->load(std::memory_order_acquire);
@@ -747,14 +754,13 @@ std::optional<std::string> ThreadList::WaitForSuspendBarrier(AtomicInteger* barr
   }
 
   // Long wait; gather information in case of timeout.
-  std::string sampled_state = t == 0 ? "" : GetThreadState(t);
+  std::string sampled_state = collect_state ? GetThreadState(t) : "";
   while (i < kSuspendBarrierIters) {
-    if (WaitOnceForSuspendBarrier(barrier, cur_val, thread_suspend_timeout_ns_)) {
+    if (WaitOnceForSuspendBarrier(barrier, cur_val, timeout_ns)) {
       ++i;
 #if ART_USE_FUTEXES
       if (!kShortSuspendTimeouts) {
-        CHECK_GE(NanoTime() - start_time,
-                 i * thread_suspend_timeout_ns_ / kSuspendBarrierIters - 1'000'000);
+        CHECK_GE(NanoTime() - start_time, i * timeout_ns / kSuspendBarrierIters - 1'000'000);
       }
 #endif
     }
@@ -764,11 +770,10 @@ std::optional<std::string> ThreadList::WaitForSuspendBarrier(AtomicInteger* barr
       return std::nullopt;
     }
   }
-  std::string result = t == 0 ? "" :
-                                "Target states: [" + sampled_state + ", " + GetThreadState(t) +
-                                    "]" + std::to_string(cur_val) + "@" +
-                                    std::to_string((uintptr_t)barrier) + "->";
-  return result + std::to_string(barrier->load(std::memory_order_acquire));
+  return collect_state ? "Target states: [" + sampled_state + ", " + GetThreadState(t) + "]" +
+                             std::to_string(cur_val) + "@" + std::to_string((uintptr_t)barrier) +
+                             " Final wait time: " + PrettyDuration(NanoTime() - start_time) :
+                         "";
 }
 
 void ThreadList::SuspendAll(const char* cause, bool long_suspend) {
@@ -834,7 +839,6 @@ void ThreadList::SuspendAll(const char* cause, bool long_suspend) {
 // Ensures all threads running Java suspend and that those not running Java don't start.
 void ThreadList::SuspendAllInternal(Thread* self, SuspendReason reason) {
   // self can be nullptr if this is an unregistered thread.
-  const uint64_t start_time = NanoTime();
   Locks::mutator_lock_->AssertNotExclusiveHeld(self);
   Locks::thread_list_lock_->AssertNotHeld(self);
   Locks::thread_suspend_count_lock_->AssertNotHeld(self);
@@ -914,24 +918,47 @@ void ThreadList::SuspendAllInternal(Thread* self, SuspendReason reason) {
     // We're already not runnable, so an attempt to suspend us should succeed.
   }
 
-  if (WaitForSuspendBarrier(&pending_threads).has_value()) {
-    const uint64_t wait_time = NanoTime() - start_time;
-    MutexLock mu(self, *Locks::thread_list_lock_);
-    MutexLock mu2(self, *Locks::thread_suspend_count_lock_);
-    std::ostringstream oss;
-    oss << "Unsuspended threads: ";
-    Thread* culprit = nullptr;
-    for (const auto& thread : list_) {
-      if (thread != self && !thread->IsSuspended()) {
-        culprit = thread;
-        oss << *thread << ", ";
-      }
+  Thread* culprit = nullptr;
+  pid_t tid = 0;
+  std::ostringstream oss;
+  for (int attempt_of_4 = 1; attempt_of_4 <= 4; ++attempt_of_4) {
+    auto result = WaitForSuspendBarrier(&pending_threads, tid, attempt_of_4);
+    if (!result.has_value()) {
+      // Wait succeeded.
+      break;
     }
-    oss << "waited for " << PrettyDuration(wait_time);
-    if (culprit == nullptr) {
-      LOG(FATAL) << "SuspendAll timeout. " << oss.str();
-    } else {
-      culprit->AbortInThis("SuspendAll timeout. " + oss.str());
+    if (attempt_of_4 == 3) {
+      // Second to the last attempt; Try to gather more information in case we time out.
+      MutexLock mu(self, *Locks::thread_list_lock_);
+      MutexLock mu2(self, *Locks::thread_suspend_count_lock_);
+      oss << "Unsuspended threads: ";
+      for (const auto& thread : list_) {
+        if (thread != self && !thread->IsSuspended()) {
+          culprit = thread;
+          oss << *thread << ", ";
+        }
+      }
+      if (culprit != nullptr) {
+        tid = culprit->GetTid();
+      }
+    } else if (attempt_of_4 == 4) {
+      // Final attempt still timed out.
+      if (culprit == nullptr) {
+        LOG(FATAL) << "SuspendAll timeout. Couldn't find holdouts.";
+      } else {
+        std::string name;
+        culprit->GetThreadName(name);
+        oss << "Info for " << *culprit << ":";
+        std::string thr_descr =
+            StringPrintf("%s tid: %d, state&flags: 0x%x, priority: %d,  barrier value: %d, ",
+                         name.c_str(),
+                         tid,
+                         culprit->GetStateAndFlags(std::memory_order_relaxed).GetValue(),
+                         culprit->GetNativePriority(),
+                         pending_threads.load());
+        oss << thr_descr << result.value();
+        culprit->AbortInThis("SuspendAll timeout: " + oss.str());
+      }
     }
   }
 }
@@ -1007,15 +1034,15 @@ bool ThreadList::Resume(Thread* thread, SuspendReason reason) {
     // To check IsSuspended.
     MutexLock mu2(self, *Locks::thread_suspend_count_lock_);
     if (UNLIKELY(!thread->IsSuspended())) {
-      LOG(ERROR) << "Resume(" << reinterpret_cast<void*>(thread)
-          << ") thread not suspended";
+      LOG(reason == SuspendReason::kForUserCode ? ERROR : FATAL)
+          << "Resume(" << reinterpret_cast<void*>(thread) << ") thread not suspended";
       return false;
     }
     if (!Contains(thread)) {
       // We only expect threads within the thread-list to have been suspended otherwise we can't
       // stop such threads from delete-ing themselves.
-      LOG(ERROR) << "Resume(" << reinterpret_cast<void*>(thread)
-          << ") thread not within thread list";
+      LOG(reason == SuspendReason::kForUserCode ? ERROR : FATAL)
+          << "Resume(" << reinterpret_cast<void*>(thread) << ") thread not within thread list";
       return false;
     }
     thread->DecrementSuspendCount(self, /*for_user_code=*/(reason == SuspendReason::kForUserCode));
@@ -1026,135 +1053,42 @@ bool ThreadList::Resume(Thread* thread, SuspendReason reason) {
   return true;
 }
 
-static void ThreadSuspendByPeerWarning(ScopedObjectAccess& soa,
-                                       LogSeverity severity,
-                                       const char* message,
-                                       jobject peer) REQUIRES_SHARED(Locks::mutator_lock_) {
-  ObjPtr<mirror::Object> name =
-      WellKnownClasses::java_lang_Thread_name->GetObject(soa.Decode<mirror::Object>(peer));
-  if (name == nullptr) {
-    LOG(severity) << message << ": " << peer;
-  } else {
-    LOG(severity) << message << ": " << peer << ":" << name->AsString()->ToModifiedUtf8();
-  }
-}
-
-Thread* ThreadList::SuspendThreadByPeer(jobject peer, SuspendReason reason) {
-  bool is_suspended = false;
-  Thread* const self = Thread::Current();
-  VLOG(threads) << "SuspendThreadByPeer starting";
-  Thread* thread;
-  WrappedSuspend1Barrier wrapped_barrier{};
-  for (int iter_count = 1;; ++iter_count) {
-    {
-      // Note: this will transition to runnable and potentially suspend.
-      ScopedObjectAccess soa(self);
-      MutexLock thread_list_mu(self, *Locks::thread_list_lock_);
-      thread = Thread::FromManagedThread(soa, peer);
-      if (thread == nullptr) {
-        ThreadSuspendByPeerWarning(soa,
-                                   ::android::base::WARNING,
-                                    "No such thread for suspend",
-                                    peer);
-        return nullptr;
-      }
-      if (!Contains(thread)) {
-        VLOG(threads) << "SuspendThreadByPeer failed for unattached thread: "
-            << reinterpret_cast<void*>(thread);
-        return nullptr;
-      }
-      // IsSuspended on the current thread will fail as the current thread is changed into
-      // Runnable above. As the suspend count is now raised if this is the current thread
-      // it will self suspend on transition to Runnable, making it hard to work with. It's simpler
-      // to just explicitly handle the current thread in the callers to this code.
-      CHECK_NE(thread, self) << "Attempt to suspend the current thread for the debugger";
-      VLOG(threads) << "SuspendThreadByPeer found thread: " << *thread;
-      {
-        MutexLock suspend_count_mu(self, *Locks::thread_suspend_count_lock_);
-        if (LIKELY(self->GetSuspendCount() == 0)) {
-          thread->IncrementSuspendCount(self, nullptr, &wrapped_barrier, reason);
-          if (thread->IsSuspended()) {
-            // See the discussion in mutator_gc_coord.md and SuspendAllInternal for the race here.
-            thread->RemoveFirstSuspend1Barrier();
-            if (!thread->HasActiveSuspendBarrier()) {
-              thread->AtomicClearFlag(ThreadFlag::kActiveSuspendBarrier);
-            }
-            is_suspended = true;
-          }
-          DCHECK_GT(thread->GetSuspendCount(), 0);
-          break;
-        }
-        // Else we hold the suspend count lock but another thread is trying to suspend us,
-        // making it unsafe to try to suspend another thread in case we get a cycle.
-        // We start the loop again, which will allow this thread to be suspended.
-      }
-    }
-    // All locks are released, and we should quickly exit the suspend-unfriendly state. Retry.
-    if (iter_count >= kMaxSuspendRetries) {
-      LOG(FATAL) << "Too many suspend retries";
-    }
-    usleep(kThreadSuspendSleepUs);
-  }
-  // Now wait for target to decrement suspend barrier.
-  if (is_suspended || !WaitForSuspendBarrier(&wrapped_barrier.barrier_).has_value()) {
-    // wrapped_barrier.barrier_ has been decremented and will no longer be accessed.
-    VLOG(threads) << "SuspendThreadByPeer thread suspended: " << *thread;
-    if (ATraceEnabled()) {
-      std::string name;
-      thread->GetThreadName(name);
-      ATraceBegin(
-          StringPrintf("SuspendThreadByPeer suspended %s for peer=%p", name.c_str(), peer).c_str());
-    }
-    DCHECK(thread->IsSuspended());
-    return thread;
-  } else {
-    LOG(WARNING) << "Suspended thread state_and_flags: " << thread->StateAndFlagsAsHexString();
-    // thread still has a pointer to wrapped_barrier. Returning and continuing would be unsafe
-    // without additional cleanup.
-    {
-      ScopedObjectAccess soa(self);
-      ThreadSuspendByPeerWarning(
-          soa, ::android::base::FATAL, "SuspendThreadByPeer timed out", peer);
-    }
-    UNREACHABLE();
-  }
-}
-
-Thread* ThreadList::SuspendThreadByThreadId(uint32_t thread_id, SuspendReason reason) {
+bool ThreadList::SuspendThread(Thread* self,
+                               Thread* thread,
+                               SuspendReason reason,
+                               ThreadState self_state,
+                               const char* func_name,
+                               int attempt_of_4) {
   bool is_suspended = false;
-  Thread* const self = Thread::Current();
-  CHECK_NE(thread_id, kInvalidThreadId);
-  VLOG(threads) << "SuspendThreadByThreadId starting";
-  Thread* thread;
-  pid_t tid;
+  VLOG(threads) << func_name << "starting";
+  pid_t tid = thread->GetTid();
   uint8_t suspended_count;
   uint8_t checkpoint_count;
   WrappedSuspend1Barrier wrapped_barrier{};
   static_assert(sizeof wrapped_barrier.barrier_ == sizeof(uint32_t));
-  for (int iter_count = 1;; ++iter_count) {
+  ThreadExitFlag tef;
+  bool exited = false;
+  thread->NotifyOnThreadExit(&tef);
+  int iter_count = 1;
+  do {
     {
+      Locks::mutator_lock_->AssertSharedHeld(self);
+      Locks::thread_list_lock_->AssertHeld(self);
       // Note: this will transition to runnable and potentially suspend.
-      ScopedObjectAccess soa(self);
-      MutexLock thread_list_mu(self, *Locks::thread_list_lock_);
-      thread = FindThreadByThreadId(thread_id);
-      if (thread == nullptr) {
-        // There's a race in inflating a lock and the owner giving up ownership and then dying.
-        LOG(WARNING) << StringPrintf("No such thread id %d for suspend", thread_id);
-        return nullptr;
-      }
       DCHECK(Contains(thread));
-      CHECK_NE(thread, self) << "Attempt to suspend the current thread for the debugger";
-      VLOG(threads) << "SuspendThreadByThreadId found thread: " << *thread;
+      // This implementation fails if thread == self. Let the clients handle that case
+      // appropriately.
+      CHECK_NE(thread, self) << func_name << "(self)";
+      VLOG(threads) << func_name << " suspending: " << *thread;
       {
         MutexLock suspend_count_mu(self, *Locks::thread_suspend_count_lock_);
         if (LIKELY(self->GetSuspendCount() == 0)) {
-          tid = thread->GetTid();
           suspended_count = thread->suspended_count_;
           checkpoint_count = thread->checkpoint_count_;
           thread->IncrementSuspendCount(self, nullptr, &wrapped_barrier, reason);
           if (thread->IsSuspended()) {
             // See the discussion in mutator_gc_coord.md and SuspendAllInternal for the race here.
-            thread->RemoveFirstSuspend1Barrier();
+            thread->RemoveFirstSuspend1Barrier(&wrapped_barrier);
             if (!thread->HasActiveSuspendBarrier()) {
               thread->AtomicClearFlag(ThreadFlag::kActiveSuspendBarrier);
             }
@@ -1172,7 +1106,23 @@ Thread* ThreadList::SuspendThreadByThreadId(uint32_t thread_id, SuspendReason re
     if (iter_count >= kMaxSuspendRetries) {
       LOG(FATAL) << "Too many suspend retries";
     }
-    usleep(kThreadSuspendSleepUs);
+    Locks::thread_list_lock_->ExclusiveUnlock(self);
+    {
+      ScopedThreadSuspension sts(self, ThreadState::kSuspended);
+      usleep(kThreadSuspendSleepUs);
+      ++iter_count;
+    }
+    Locks::thread_list_lock_->ExclusiveLock(self);
+    exited = tef.HasExited();
+  } while (!exited);
+  thread->UnregisterThreadExitFlag(&tef);
+  Locks::thread_list_lock_->ExclusiveUnlock(self);
+  self->TransitionFromRunnableToSuspended(self_state);
+  if (exited) {
+    // This is OK: There's a race in inflating a lock and the owner giving up ownership and then
+    // dying.
+    LOG(WARNING) << StringPrintf("Thread with tid %d exited before suspending", tid);
+    return false;
   }
   // Now wait for target to decrement suspend barrier.
   std::optional<std::string> failure_info;
@@ -1180,26 +1130,31 @@ Thread* ThreadList::SuspendThreadByThreadId(uint32_t thread_id, SuspendReason re
     // As an experiment, redundantly trigger suspension. TODO: Remove this.
     std::atomic_thread_fence(std::memory_order_seq_cst);
     thread->TriggerSuspend();
-    failure_info = WaitForSuspendBarrier(&wrapped_barrier.barrier_, tid);
+    failure_info = WaitForSuspendBarrier(&wrapped_barrier.barrier_, tid, attempt_of_4);
     if (!failure_info.has_value()) {
       is_suspended = true;
     }
   }
-  if (is_suspended) {
-    // wrapped_barrier.barrier_ has been decremented and will no longer be accessed.
-    VLOG(threads) << "SuspendThreadByThreadId thread suspended: " << *thread;
-    if (ATraceEnabled()) {
-      std::string name;
-      thread->GetThreadName(name);
-      ATraceBegin(
-          StringPrintf("SuspendThreadByPeer suspended %s for id=%d", name.c_str(), thread_id)
-              .c_str());
+  while (!is_suspended) {
+    if (attempt_of_4 > 0 && attempt_of_4 < 4) {
+      // Caller will try again. Give up and resume the thread for now.  We need to make sure
+      // that wrapped_barrier is removed from the list before we deallocate it.
+      MutexLock suspend_count_mu(self, *Locks::thread_suspend_count_lock_);
+      if (wrapped_barrier.barrier_.load() == 0) {
+        // Succeeded in the meantime.
+        is_suspended = true;
+        continue;
+      }
+      thread->RemoveSuspend1Barrier(&wrapped_barrier);
+      if (!thread->HasActiveSuspendBarrier()) {
+        thread->AtomicClearFlag(ThreadFlag::kActiveSuspendBarrier);
+      }
+      // Do not call Resume(), since we are probably not fully suspended.
+      thread->DecrementSuspendCount(self,
+                                    /*for_user_code=*/(reason == SuspendReason::kForUserCode));
+      Thread::resume_cond_->Broadcast(self);
+      return false;
     }
-    DCHECK(thread->IsSuspended());
-    return thread;
-  } else {
-    // thread still has a pointer to wrapped_barrier. Returning and continuing would be unsafe
-    // without additional cleanup.
     std::string name;
     thread->GetThreadName(name);
     WrappedSuspend1Barrier* first_barrier;
@@ -1207,12 +1162,13 @@ Thread* ThreadList::SuspendThreadByThreadId(uint32_t thread_id, SuspendReason re
       MutexLock suspend_count_mu(self, *Locks::thread_suspend_count_lock_);
       first_barrier = thread->tlsPtr_.active_suspend1_barriers;
     }
-    // 'thread' should still be suspended, and hence stick around. Try to abort there, since its
-    // stack trace is much more interesting than ours.
-    thread->AbortInThis(StringPrintf(
-        "Caused SuspendThreadByThreadId to time out: %d (%s), state&flags: 0x%x, priority: %d,"
+    // 'thread' should still have a suspend request pending, and hence stick around. Try to abort
+    // there, since its stack trace is much more interesting than ours.
+    std::string message = StringPrintf(
+        "%s timed out: %d (%s), state&flags: 0x%x, priority: %d,"
         " barriers: %p, ours: %p, barrier value: %d, nsusps: %d, ncheckpts: %d, thread_info: %s",
-        thread_id,
+        func_name,
+        thread->GetTid(),
         name.c_str(),
         thread->GetStateAndFlags(std::memory_order_relaxed).GetValue(),
         thread->GetNativePriority(),
@@ -1221,9 +1177,81 @@ Thread* ThreadList::SuspendThreadByThreadId(uint32_t thread_id, SuspendReason re
         wrapped_barrier.barrier_.load(),
         thread->suspended_count_ - suspended_count,
         thread->checkpoint_count_ - checkpoint_count,
-        failure_info.value().c_str()));
-    UNREACHABLE();
+        failure_info.value().c_str());
+    // Check one last time whether thread passed the suspend barrier. Empirically this seems to
+    // happen maybe between 1 and 5% of the time.
+    if (wrapped_barrier.barrier_.load() != 0) {
+      // thread still has a pointer to wrapped_barrier. Returning and continuing would be unsafe
+      // without additional cleanup.
+      thread->AbortInThis(message);
+      UNREACHABLE();
+    }
+    is_suspended = true;
   }
+  // wrapped_barrier.barrier_ has been decremented and will no longer be accessed.
+  VLOG(threads) << func_name << " suspended: " << *thread;
+  if (ATraceEnabled()) {
+    std::string name;
+    thread->GetThreadName(name);
+    ATraceBegin(
+        StringPrintf("%s suspended %s for tid=%d", func_name, name.c_str(), thread->GetTid())
+            .c_str());
+  }
+  DCHECK(thread->IsSuspended());
+  return true;
+}
+
+Thread* ThreadList::SuspendThreadByPeer(jobject peer, SuspendReason reason) {
+  Thread* const self = Thread::Current();
+  ThreadState old_self_state = self->GetState();
+  self->TransitionFromSuspendedToRunnable();
+  Locks::thread_list_lock_->ExclusiveLock(self);
+  ObjPtr<mirror::Object> thread_ptr = self->DecodeJObject(peer);
+  Thread* thread = Thread::FromManagedThread(self, thread_ptr);
+  if (thread == nullptr || !Contains(thread)) {
+    if (thread == nullptr) {
+      ObjPtr<mirror::Object> name = WellKnownClasses::java_lang_Thread_name->GetObject(thread_ptr);
+      std::string thr_name = (name == nullptr ? "<unknown>" : name->AsString()->ToModifiedUtf8());
+      LOG(WARNING) << "No such thread for suspend"
+                   << ": " << peer << ":" << thr_name;
+    } else {
+      LOG(WARNING) << "SuspendThreadByPeer failed for unattached thread: "
+                   << reinterpret_cast<void*>(thread);
+    }
+    Locks::thread_list_lock_->ExclusiveUnlock(self);
+    self->TransitionFromRunnableToSuspended(old_self_state);
+    return nullptr;
+  }
+  VLOG(threads) << "SuspendThreadByPeer found thread: " << *thread;
+  // Releases thread_list_lock_ and mutator lock.
+  bool success = SuspendThread(self, thread, reason, old_self_state, __func__, 0);
+  Locks::thread_list_lock_->AssertNotHeld(self);
+  return success ? thread : nullptr;
+}
+
+Thread* ThreadList::SuspendThreadByThreadId(uint32_t thread_id,
+                                            SuspendReason reason,
+                                            int attempt_of_4) {
+  Thread* const self = Thread::Current();
+  ThreadState old_self_state = self->GetState();
+  CHECK_NE(thread_id, kInvalidThreadId);
+  VLOG(threads) << "SuspendThreadByThreadId starting";
+  self->TransitionFromSuspendedToRunnable();
+  Locks::thread_list_lock_->ExclusiveLock(self);
+  Thread* thread = FindThreadByThreadId(thread_id);
+  if (thread == nullptr) {
+    // There's a race in inflating a lock and the owner giving up ownership and then dying.
+    LOG(WARNING) << StringPrintf("No such thread id %d for suspend", thread_id);
+    Locks::thread_list_lock_->ExclusiveUnlock(self);
+    self->TransitionFromRunnableToSuspended(old_self_state);
+    return nullptr;
+  }
+  DCHECK(Contains(thread));
+  VLOG(threads) << "SuspendThreadByThreadId found thread: " << *thread;
+  // Releases thread_list_lock_ and mutator lock.
+  bool success = SuspendThread(self, thread, reason, old_self_state, __func__, attempt_of_4);
+  Locks::thread_list_lock_->AssertNotHeld(self);
+  return success ? thread : nullptr;
 }
 
 Thread* ThreadList::FindThreadByThreadId(uint32_t thread_id) {
diff --git a/runtime/thread_list.h b/runtime/thread_list.h
index cb7744056f..8be2d4e02f 100644
--- a/runtime/thread_list.h
+++ b/runtime/thread_list.h
@@ -17,6 +17,10 @@
 #ifndef ART_RUNTIME_THREAD_LIST_H_
 #define ART_RUNTIME_THREAD_LIST_H_
 
+#include <bitset>
+#include <list>
+#include <vector>
+
 #include "barrier.h"
 #include "base/histogram.h"
 #include "base/mutex.h"
@@ -25,10 +29,7 @@
 #include "jni.h"
 #include "reflective_handle_scope.h"
 #include "suspend_reason.h"
-
-#include <bitset>
-#include <list>
-#include <vector>
+#include "thread_state.h"
 
 namespace art HIDDEN {
 namespace gc {
@@ -94,8 +95,10 @@ class ThreadList {
   // Suspend a thread using its thread id, typically used by lock/monitor inflation. Returns the
   // thread on success else null. The thread id is used to identify the thread to avoid races with
   // the thread terminating. Note that as thread ids are recycled this may not suspend the expected
-  // thread, that may be terminating.
-  Thread* SuspendThreadByThreadId(uint32_t thread_id, SuspendReason reason)
+  // thread, that may be terminating. 'attempt_of_4' is zero if this is the only
+  // attempt, or 1..4 to try 4 times with fractional timeouts.
+  // TODO: Reconsider the use of thread_id, now that we have ThreadExitFlag.
+  Thread* SuspendThreadByThreadId(uint32_t thread_id, SuspendReason reason, int attempt_of_4 = 0)
       REQUIRES(!Locks::mutator_lock_,
                !Locks::thread_list_lock_,
                !Locks::thread_suspend_count_lock_);
@@ -217,7 +220,9 @@ class ThreadList {
   // the diagnostic information. If 0 is passed, we return an empty string on timeout.  Normally
   // the caller does not hold the mutator lock. See the comment at the call in
   // RequestSynchronousCheckpoint for the only exception.
-  std::optional<std::string> WaitForSuspendBarrier(AtomicInteger* barrier, pid_t t = 0)
+  std::optional<std::string> WaitForSuspendBarrier(AtomicInteger* barrier,
+                                                   pid_t t = 0,
+                                                   int attempt_of_4 = 0)
       REQUIRES(!Locks::thread_list_lock_, !Locks::thread_suspend_count_lock_);
 
  private:
@@ -234,6 +239,19 @@ class ThreadList {
       REQUIRES(Locks::thread_list_lock_, Locks::thread_suspend_count_lock_)
           UNLOCK_FUNCTION(Locks::mutator_lock_);
 
+  // Helper to actually suspend a single thread. This is called with thread_list_lock_ held and
+  // the caller guarantees that *thread is valid until that is released.  We "release the mutator
+  // lock", by switching to self_state.  'attempt_of_4' is 0 if we only attempt once, and 1..4 if
+  // we are going to try 4 times with a quarter of the full timeout. 'func_name' is used only to
+  // identify ourselves for logging.
+  bool SuspendThread(Thread* self,
+                     Thread* thread,
+                     SuspendReason reason,
+                     ThreadState self_state,
+                     const char* func_name,
+                     int attempt_of_4) RELEASE(Locks::thread_list_lock_)
+      RELEASE_SHARED(Locks::mutator_lock_);
+
   void SuspendAllInternal(Thread* self, SuspendReason reason = SuspendReason::kInternal)
       REQUIRES(!Locks::thread_list_lock_,
                !Locks::thread_suspend_count_lock_,
diff --git a/runtime/trace.cc b/runtime/trace.cc
index 3a410ef9d4..04f4de6c89 100644
--- a/runtime/trace.cc
+++ b/runtime/trace.cc
@@ -21,10 +21,10 @@
 
 #include "android-base/macros.h"
 #include "android-base/stringprintf.h"
-
 #include "art_method-inl.h"
 #include "base/casts.h"
 #include "base/enums.h"
+#include "base/leb128.h"
 #include "base/os.h"
 #include "base/stl_util.h"
 #include "base/systrace.h"
@@ -53,6 +53,13 @@
 
 namespace art HIDDEN {
 
+struct MethodTraceRecord {
+  ArtMethod* method;
+  TraceAction action;
+  uint32_t wall_clock_time;
+  uint32_t thread_cpu_time;
+};
+
 using android::base::StringPrintf;
 
 static constexpr size_t TraceActionBits = MinimumBitsToStore(
@@ -70,6 +77,22 @@ static const uint16_t kTraceRecordSizeSingleClock = 10;  // using v2
 static const uint16_t kTraceRecordSizeDualClock   = 14;  // using v3 with two timestamps
 static const size_t kNumTracePoolBuffers = 32;
 
+// Packet type encoding for the new method tracing format.
+static const int kThreadInfoHeaderV2 = 0;
+static const int kMethodInfoHeaderV2 = 1;
+static const int kEntryHeaderV2 = 2;
+static const int kSummaryHeaderV2 = 3;
+
+// Packet sizes for the new method trace format.
+static const uint16_t kTraceHeaderLengthV2 = 32;
+static const uint16_t kTraceRecordSizeSingleClockV2 = 6;
+static const uint16_t kTraceRecordSizeDualClockV2 = kTraceRecordSizeSingleClockV2 + 2;
+static const uint16_t kEntryHeaderSizeSingleClockV2 = 17;
+static const uint16_t kEntryHeaderSizeDualClockV2 = kEntryHeaderSizeSingleClockV2 + 4;
+
+static const uint16_t kTraceVersionSingleClockV2 = 4;
+static const uint16_t kTraceVersionDualClockV2 = 5;
+
 TraceClockSource Trace::default_clock_source_ = kDefaultTraceClockSource;
 
 Trace* volatile Trace::the_trace_ = nullptr;
@@ -310,14 +333,24 @@ void Trace::SetDefaultClockSource(TraceClockSource clock_source) {
 #endif
 }
 
-static uint16_t GetTraceVersion(TraceClockSource clock_source) {
-  return (clock_source == TraceClockSource::kDual) ? kTraceVersionDualClock
-                                                   : kTraceVersionSingleClock;
+static uint16_t GetTraceVersion(TraceClockSource clock_source, int version) {
+  if (version == Trace::kFormatV1) {
+    return (clock_source == TraceClockSource::kDual) ? kTraceVersionDualClock :
+                                                       kTraceVersionSingleClock;
+  } else {
+    return (clock_source == TraceClockSource::kDual) ? kTraceVersionDualClockV2 :
+                                                       kTraceVersionSingleClockV2;
+  }
 }
 
-static uint16_t GetRecordSize(TraceClockSource clock_source) {
-  return (clock_source == TraceClockSource::kDual) ? kTraceRecordSizeDualClock
-                                                   : kTraceRecordSizeSingleClock;
+static uint16_t GetRecordSize(TraceClockSource clock_source, int version) {
+  if (version == Trace::kFormatV1) {
+    return (clock_source == TraceClockSource::kDual) ? kTraceRecordSizeDualClock :
+                                                       kTraceRecordSizeSingleClock;
+  } else {
+    return (clock_source == TraceClockSource::kDual) ? kTraceRecordSizeDualClockV2 :
+                                                       kTraceRecordSizeSingleClockV2;
+  }
 }
 
 static uint16_t GetNumEntries(TraceClockSource clock_source) {
@@ -699,7 +732,12 @@ void Trace::StopTracing(bool flush_entries) {
     // We also flush the buffer when destroying a thread which expects the_trace_ to be valid so
     // make sure that the per-thread buffer is reset before resetting the_trace_.
     {
+      MutexLock mu(self, *Locks::trace_lock_);
       MutexLock tl_lock(Thread::Current(), *Locks::thread_list_lock_);
+      // Flush the per-thread buffers and reset the trace inside the trace_lock_ to avoid any
+      // race if the thread is detaching and trying to flush the buffer too. Since we hold the
+      // trace_lock_ both here and when flushing on a thread detach only one of them will succeed
+      // in actually flushing the buffer.
       for (Thread* thread : Runtime::Current()->GetThreadList()->GetList()) {
         if (thread->GetMethodTraceBuffer() != nullptr) {
           // We may have pending requests to flush the data. So just enqueue a
@@ -709,12 +747,9 @@ void Trace::StopTracing(bool flush_entries) {
               thread, /* is_sync= */ false, /* free_buffer= */ true);
         }
       }
+      the_trace_ = nullptr;
+      sampling_pthread_ = 0U;
     }
-
-    // Reset the_trace_ by taking a trace_lock
-    MutexLock mu(self, *Locks::trace_lock_);
-    the_trace_ = nullptr;
-    sampling_pthread_ = 0U;
   }
 
   // At this point, code may read buf_ as its writers are shutdown
@@ -731,6 +766,12 @@ void Trace::StopTracing(bool flush_entries) {
 
 void Trace::FlushThreadBuffer(Thread* self) {
   MutexLock mu(self, *Locks::trace_lock_);
+  // Check if we still need to flush inside the trace_lock_. If we are stopping tracing it is
+  // possible we already deleted the trace and flushed the buffer too.
+  if (the_trace_ == nullptr) {
+    DCHECK_EQ(self->GetMethodTraceBuffer(), nullptr);
+    return;
+  }
   the_trace_->trace_writer_->FlushBuffer(self, /* is_sync= */ false, /* free_buffer= */ true);
 }
 
@@ -791,6 +832,11 @@ TraceClockSource GetClockSourceFromFlags(int flags) {
   }
 }
 
+int GetTraceFormatVersionFromFlags(int flags) {
+  int version = (flags & Trace::kTraceFormatVersionFlagMask) >> Trace::kTraceFormatVersionShift;
+  return version;
+}
+
 }  // namespace
 
 TraceWriter::TraceWriter(File* trace_file,
@@ -798,34 +844,46 @@ TraceWriter::TraceWriter(File* trace_file,
                          TraceClockSource clock_source,
                          size_t buffer_size,
                          int num_trace_buffers,
+                         int trace_format_version,
                          uint32_t clock_overhead_ns)
     : trace_file_(trace_file),
       trace_output_mode_(output_mode),
       clock_source_(clock_source),
       buf_(new uint8_t[std::max(kMinBufSize, buffer_size)]()),
       buffer_size_(std::max(kMinBufSize, buffer_size)),
+      trace_format_version_(trace_format_version),
       start_time_(GetMicroTime(GetTimestamp())),
       overflow_(false),
+      num_records_(0),
       clock_overhead_ns_(clock_overhead_ns),
       owner_tids_(num_trace_buffers),
       tracing_lock_("tracing lock", LockLevel::kTracingStreamingLock) {
-  uint16_t trace_version = GetTraceVersion(clock_source_);
+  uint16_t trace_version = GetTraceVersion(clock_source_, trace_format_version_);
   if (output_mode == TraceOutputMode::kStreaming) {
     trace_version |= 0xF0U;
   }
+
   // Set up the beginning of the trace.
-  memset(buf_.get(), 0, kTraceHeaderLength);
-  Append4LE(buf_.get(), kTraceMagicValue);
-  Append2LE(buf_.get() + 4, trace_version);
-  Append2LE(buf_.get() + 6, kTraceHeaderLength);
-  Append8LE(buf_.get() + 8, start_time_);
-  if (trace_version >= kTraceVersionDualClock) {
-    uint16_t record_size = GetRecordSize(clock_source_);
-    Append2LE(buf_.get() + 16, record_size);
-  }
-  static_assert(18 <= kMinBufSize, "Minimum buffer size not large enough for trace header");
+  if (trace_format_version_ == Trace::kFormatV1) {
+    memset(buf_.get(), 0, kTraceHeaderLength);
+    Append4LE(buf_.get(), kTraceMagicValue);
+    Append2LE(buf_.get() + 4, trace_version);
+    Append2LE(buf_.get() + 6, kTraceHeaderLength);
+    Append8LE(buf_.get() + 8, start_time_);
+    if (trace_version >= kTraceVersionDualClock) {
+      uint16_t record_size = GetRecordSize(clock_source_, trace_format_version_);
+      Append2LE(buf_.get() + 16, record_size);
+    }
+    static_assert(18 <= kMinBufSize, "Minimum buffer size not large enough for trace header");
 
-  cur_offset_ = kTraceHeaderLength;
+    cur_offset_ = kTraceHeaderLength;
+  } else {
+    memset(buf_.get(), 0, kTraceHeaderLengthV2);
+    Append4LE(buf_.get(), kTraceMagicValue);
+    Append2LE(buf_.get() + 4, trace_version);
+    Append8LE(buf_.get() + 6, start_time_);
+    cur_offset_ = kTraceHeaderLengthV2;
+  }
 
   if (output_mode == TraceOutputMode::kStreaming) {
     // Flush the header information to the file. We use a per thread buffer, so
@@ -863,6 +921,7 @@ Trace::Trace(File* trace_file,
       stop_tracing_(false) {
   CHECK_IMPLIES(trace_file == nullptr, output_mode == TraceOutputMode::kDDMS);
 
+  int trace_format_version = GetTraceFormatVersionFromFlags(flags_);
   // In streaming mode, we only need a buffer big enough to store data per each
   // thread buffer. In non-streaming mode this is specified by the user and we
   // stop tracing when the buffer is full.
@@ -874,6 +933,7 @@ Trace::Trace(File* trace_file,
                                       clock_source_,
                                       buf_size,
                                       kNumTracePoolBuffers,
+                                      trace_format_version,
                                       GetClockOverheadNanoSeconds()));
 }
 
@@ -903,7 +963,7 @@ void TraceWriter::FinishTracing(int flags, bool flush_entries) {
     std::ostringstream os;
 
     os << StringPrintf("%cversion\n", kTraceTokenChar);
-    os << StringPrintf("%d\n", GetTraceVersion(clock_source_));
+    os << StringPrintf("%d\n", GetTraceVersion(clock_source_, trace_format_version_));
     os << StringPrintf("data-file-overflow=%s\n", overflow_ ? "true" : "false");
     if (UseThreadCpuClock(clock_source_)) {
       if (UseWallClock(clock_source_)) {
@@ -916,8 +976,7 @@ void TraceWriter::FinishTracing(int flags, bool flush_entries) {
     }
     os << StringPrintf("elapsed-time-usec=%" PRIu64 "\n", elapsed);
     if (trace_output_mode_ != TraceOutputMode::kStreaming) {
-      size_t num_records = (final_offset - kTraceHeaderLength) / GetRecordSize(clock_source_);
-      os << StringPrintf("num-method-calls=%zd\n", num_records);
+      os << StringPrintf("num-method-calls=%zd\n", num_records_);
     }
     os << StringPrintf("clock-call-overhead-nsec=%d\n", clock_overhead_ns_);
     os << StringPrintf("vm=art\n");
@@ -927,17 +986,13 @@ void TraceWriter::FinishTracing(int flags, bool flush_entries) {
       os << "alloc-size=" << Runtime::Current()->GetStat(KIND_ALLOCATED_BYTES) << "\n";
       os << "gc-count=" <<  Runtime::Current()->GetStat(KIND_GC_INVOCATIONS) << "\n";
     }
-    os << StringPrintf("%cthreads\n", kTraceTokenChar);
-    {
-      // TODO(b/280558212): Moving the Mutexlock out of DumpThreadList to try and
-      // narrow down where seg fault is happening. Change this after the bug is
-      // fixed.
-      CHECK_NE(self, nullptr);
-      MutexLock mu(self, tracing_lock_);
+
+    if (trace_format_version_ == Trace::kFormatV1) {
+      os << StringPrintf("%cthreads\n", kTraceTokenChar);
       DumpThreadList(os);
+      os << StringPrintf("%cmethods\n", kTraceTokenChar);
+      DumpMethodList(os);
     }
-    os << StringPrintf("%cmethods\n", kTraceTokenChar);
-    DumpMethodList(os);
     os << StringPrintf("%cend\n", kTraceTokenChar);
     std::string header(os.str());
 
@@ -947,15 +1002,27 @@ void TraceWriter::FinishTracing(int flags, bool flush_entries) {
       // cannot be any writes to trace_file_ after finish tracing.
       // Write a special token to mark the end of trace records and the start of
       // trace summary.
-      uint8_t buf[7];
-      Append2LE(buf, 0);
-      buf[2] = kOpTraceSummary;
-      Append4LE(buf + 3, static_cast<uint32_t>(header.length()));
-      // Write the trace summary. The summary is identical to the file header when
-      // the output mode is not streaming (except for methods).
-      if (!trace_file_->WriteFully(buf, sizeof(buf)) ||
-          !trace_file_->WriteFully(header.c_str(), header.length())) {
-        PLOG(WARNING) << "Failed streaming a tracing event.";
+      if (trace_format_version_ == Trace::kFormatV1) {
+        uint8_t buf[7];
+        Append2LE(buf, 0);
+        buf[2] = kOpTraceSummary;
+        Append4LE(buf + 3, static_cast<uint32_t>(header.length()));
+        // Write the trace summary. The summary is identical to the file header when
+        // the output mode is not streaming (except for methods).
+        if (!trace_file_->WriteFully(buf, sizeof(buf)) ||
+            !trace_file_->WriteFully(header.c_str(), header.length())) {
+          PLOG(WARNING) << "Failed streaming a tracing event.";
+        }
+      } else {
+        uint8_t buf[3];
+        buf[0] = kSummaryHeaderV2;
+        Append2LE(buf + 1, static_cast<uint32_t>(header.length()));
+        // Write the trace summary. Reports information about tracing mode, number of records and
+        // clock overhead in plain text format.
+        if (!trace_file_->WriteFully(buf, sizeof(buf)) ||
+            !trace_file_->WriteFully(header.c_str(), header.length())) {
+          PLOG(WARNING) << "Failed streaming a tracing event.";
+        }
       }
     } else {
       if (trace_file_.get() == nullptr) {
@@ -1140,9 +1207,14 @@ void TraceWriter::RecordThreadInfo(Thread* thread) {
 
   static constexpr size_t kThreadNameHeaderSize = 7;
   uint8_t header[kThreadNameHeaderSize];
-  Append2LE(header, 0);
-  header[2] = kOpNewThread;
-  Append2LE(header + 3, GetThreadEncoding(thread->GetTid()));
+  if (trace_format_version_ == Trace::kFormatV1) {
+    Append2LE(header, 0);
+    header[2] = kOpNewThread;
+    Append2LE(header + 3, GetThreadEncoding(thread->GetTid()));
+  } else {
+    header[0] = kThreadInfoHeaderV2;
+    Append4LE(header + 1, thread->GetTid());
+  }
   DCHECK(thread_name.length() < (1 << 16));
   Append2LE(header + 5, static_cast<uint16_t>(thread_name.length()));
 
@@ -1179,20 +1251,34 @@ void TraceWriter::PreProcessTraceForMethodInfos(
 }
 
 void TraceWriter::RecordMethodInfo(const std::string& method_info_line, uint32_t method_id) {
-  std::string method_line(GetMethodLine(method_info_line, method_id));
   // Write a special block with the name.
-  static constexpr size_t kMethodNameHeaderSize = 5;
-  uint8_t method_header[kMethodNameHeaderSize];
-  DCHECK_LT(kMethodNameHeaderSize, kPerThreadBufSize);
-  Append2LE(method_header, 0);
-  method_header[2] = kOpNewMethod;
-
+  std::string method_line;
+  size_t header_size;
+  static constexpr size_t kMaxMethodNameHeaderSize = 7;
+  uint8_t method_header[kMaxMethodNameHeaderSize];
   uint16_t method_line_length = static_cast<uint16_t>(method_line.length());
   DCHECK(method_line.length() < (1 << 16));
-  Append2LE(method_header + 3, method_line_length);
+  if (trace_format_version_ == Trace::kFormatV1) {
+    // Write a special block with the name.
+    static constexpr size_t kMethodNameHeaderSize = 5;
+    DCHECK_LT(kMethodNameHeaderSize, kPerThreadBufSize);
+    Append2LE(method_header, 0);
+    method_header[2] = kOpNewMethod;
+    method_line = GetMethodLine(method_info_line, method_id);
+    method_line_length = static_cast<uint16_t>(method_line.length());
+    Append2LE(method_header + 3, method_line_length);
+    header_size = kMethodNameHeaderSize;
+  } else {
+    method_line = method_info_line;
+    method_line_length = static_cast<uint16_t>(method_line.length());
+    method_header[0] = kMethodInfoHeaderV2;
+    Append4LE(method_header + 1, method_id);
+    Append2LE(method_header + 5, method_line_length);
+    header_size = 7;
+  }
 
   const uint8_t* ptr = reinterpret_cast<const uint8_t*>(method_line.c_str());
-  if (!trace_file_->WriteFully(method_header, kMethodNameHeaderSize) ||
+  if (!trace_file_->WriteFully(method_header, header_size) ||
       !trace_file_->WriteFully(ptr, method_line_length)) {
     PLOG(WARNING) << "Failed streaming a tracing event.";
   }
@@ -1289,9 +1375,11 @@ void TraceWriter::FlushBuffer(Thread* thread, bool is_sync, bool release) {
     // This is a synchronous flush, so no need to allocate a new buffer. This is used either
     // when the tracing has finished or in non-streaming mode.
     // Just reset the buffer pointer to the initial value, so we can reuse the same buffer.
-    *current_offset = kPerThreadBufSize;
     if (release) {
       thread->SetMethodTraceBuffer(nullptr);
+      *current_offset = 0;
+    } else {
+      *current_offset = kPerThreadBufSize;
     }
   } else {
     int old_index = GetMethodTraceIndex(method_trace_entries);
@@ -1300,17 +1388,156 @@ void TraceWriter::FlushBuffer(Thread* thread, bool is_sync, bool release) {
     thread_pool_->AddTask(
         Thread::Current(),
         new TraceWriterTask(this, old_index, method_trace_entries, *current_offset, tid));
-    *current_offset = kPerThreadBufSize;
     if (release) {
       thread->SetMethodTraceBuffer(nullptr);
+      *current_offset = 0;
     } else {
       thread->SetMethodTraceBuffer(AcquireTraceBuffer(tid));
+      *current_offset = kPerThreadBufSize;
     }
   }
 
   return;
 }
 
+void TraceWriter::ReadValuesFromRecord(uintptr_t* method_trace_entries,
+                                       size_t record_index,
+                                       MethodTraceRecord& record,
+                                       bool has_thread_cpu_clock,
+                                       bool has_wall_clock) {
+  uintptr_t method_and_action = method_trace_entries[record_index++];
+  record.method = reinterpret_cast<ArtMethod*>(method_and_action & kMaskTraceAction);
+  CHECK(record.method != nullptr);
+  record.action = DecodeTraceAction(method_and_action);
+
+  record.thread_cpu_time = 0;
+  record.wall_clock_time = 0;
+  if (has_thread_cpu_clock) {
+    record.thread_cpu_time = method_trace_entries[record_index++];
+  }
+  if (has_wall_clock) {
+    uint64_t timestamp = method_trace_entries[record_index++];
+    if (art::kRuntimePointerSize == PointerSize::k32) {
+      // On 32-bit architectures timestamp is stored as two 32-bit values.
+      uint64_t high_timestamp = method_trace_entries[record_index++];
+      timestamp = (high_timestamp << 32 | timestamp);
+    }
+    record.wall_clock_time = GetMicroTime(timestamp) - start_time_;
+  }
+}
+
+void TraceWriter::FlushEntriesFormatV1(
+    uintptr_t* method_trace_entries,
+    size_t tid,
+    const std::unordered_map<ArtMethod*, std::string>& method_infos,
+    size_t end_offset,
+    size_t* current_index,
+    uint8_t* buffer_ptr) {
+  uint16_t thread_id = GetThreadEncoding(tid);
+  bool has_thread_cpu_clock = UseThreadCpuClock(clock_source_);
+  bool has_wall_clock = UseWallClock(clock_source_);
+  size_t buffer_index = *current_index;
+  size_t num_entries = GetNumEntries(clock_source_);
+  const size_t record_size = GetRecordSize(clock_source_, trace_format_version_);
+
+  for (size_t entry_index = kPerThreadBufSize; entry_index != end_offset;) {
+    entry_index -= num_entries;
+
+    MethodTraceRecord record;
+    ReadValuesFromRecord(
+        method_trace_entries, entry_index, record, has_thread_cpu_clock, has_wall_clock);
+
+    auto [method_id, is_new_method] = GetMethodEncoding(record.method);
+    if (is_new_method && trace_output_mode_ == TraceOutputMode::kStreaming) {
+      RecordMethodInfo(method_infos.find(record.method)->second, method_id);
+    }
+
+    DCHECK_LT(buffer_index + record_size, buffer_size_);
+    EncodeEventEntry(buffer_ptr + buffer_index,
+                     thread_id,
+                     method_id,
+                     record.action,
+                     record.thread_cpu_time,
+                     record.wall_clock_time);
+    buffer_index += record_size;
+  }
+  *current_index = buffer_index;
+}
+
+void TraceWriter::FlushEntriesFormatV2(
+    uintptr_t* method_trace_entries,
+    size_t tid,
+    const std::unordered_map<ArtMethod*, std::string>& method_infos,
+    size_t num_records,
+    size_t* current_index,
+    uint8_t* init_buffer_ptr) {
+  bool has_thread_cpu_clock = UseThreadCpuClock(clock_source_);
+  bool has_wall_clock = UseWallClock(clock_source_);
+  size_t num_entries = GetNumEntries(clock_source_);
+  uint32_t prev_wall_timestamp = 0;
+  uint32_t prev_thread_timestamp = 0;
+  int32_t init_method_action_encoding = 0;
+  bool is_first_entry = true;
+  uint8_t* current_buffer_ptr = init_buffer_ptr;
+  uint32_t header_size = (clock_source_ == TraceClockSource::kDual) ? kEntryHeaderSizeDualClockV2 :
+                                                                      kEntryHeaderSizeSingleClockV2;
+
+  size_t entry_index = kPerThreadBufSize;
+  for (size_t i = 0; i < num_records; i++) {
+    entry_index -= num_entries;
+
+    MethodTraceRecord record;
+    ReadValuesFromRecord(
+        method_trace_entries, entry_index, record, has_thread_cpu_clock, has_wall_clock);
+
+    // TODO(mythria): Explore the possibility of using method pointer instead of having an encoding.
+    // On 64-bit this means method ids would use 8 bytes but that is okay since we only encode the
+    // full method id in the header and then encode the diff against the method id in the header.
+    // The diff is usually expected to be small.
+    auto [method_id, is_new_method] = GetMethodEncoding(record.method);
+    if (is_new_method && trace_output_mode_ == TraceOutputMode::kStreaming) {
+      RecordMethodInfo(method_infos.find(record.method)->second, method_id);
+    }
+    DCHECK(method_id < (1 << (31 - TraceActionBits)));
+    uint32_t method_action_encoding = (method_id << TraceActionBits) | record.action;
+
+    if (is_first_entry) {
+      prev_wall_timestamp = record.wall_clock_time;
+      prev_thread_timestamp = record.thread_cpu_time;
+      init_method_action_encoding = method_action_encoding;
+      is_first_entry = false;
+
+      EncodeEventBlockHeader(init_buffer_ptr,
+                             tid,
+                             method_action_encoding,
+                             prev_thread_timestamp,
+                             prev_wall_timestamp,
+                             num_records);
+      current_buffer_ptr += header_size;
+    } else {
+      current_buffer_ptr = EncodeSignedLeb128(current_buffer_ptr,
+                                           (method_action_encoding - init_method_action_encoding));
+
+      if (has_wall_clock) {
+        current_buffer_ptr =
+            EncodeUnsignedLeb128(current_buffer_ptr, (record.wall_clock_time - prev_wall_timestamp));
+        prev_wall_timestamp = record.wall_clock_time;
+      }
+
+      if (has_thread_cpu_clock) {
+        current_buffer_ptr =
+            EncodeUnsignedLeb128(current_buffer_ptr, (record.thread_cpu_time - prev_thread_timestamp));
+        prev_thread_timestamp = record.thread_cpu_time;
+      }
+    }
+  }
+
+  // Update the total size of the block excluding header size.
+  uint8_t* total_size_loc = init_buffer_ptr + header_size - 2;
+  Append2LE(total_size_loc, current_buffer_ptr - (init_buffer_ptr + header_size));
+  *current_index += current_buffer_ptr - init_buffer_ptr;
+}
+
 void TraceWriter::FlushBuffer(uintptr_t* method_trace_entries,
                               size_t current_offset,
                               size_t tid,
@@ -1319,67 +1546,36 @@ void TraceWriter::FlushBuffer(uintptr_t* method_trace_entries,
   // method id for each method. We do that by maintaining a map from id to method for each newly
   // seen method. tracing_lock_ is required to serialize these.
   MutexLock mu(Thread::Current(), tracing_lock_);
-  size_t current_index;
+  size_t current_index = 0;
   uint8_t* buffer_ptr = buf_.get();
   size_t buffer_size = buffer_size_;
-  if (trace_output_mode_ == TraceOutputMode::kStreaming) {
-    // In streaming mode, we flush the data to file each time we flush the per-thread buffer.
-    // Just reuse the entire buffer.
-    current_index = 0;
-  } else {
-    // In non-streaming mode we only flush at the end, so retain the earlier data. If the buffer
-    // is full we don't process any more entries.
-    current_index = cur_offset_;
-  }
-  uint16_t thread_id = GetThreadEncoding(tid);
 
-  bool has_thread_cpu_clock = UseThreadCpuClock(clock_source_);
-  bool has_wall_clock = UseWallClock(clock_source_);
-  const size_t record_size = GetRecordSize(clock_source_);
-  DCHECK_LT(record_size, kPerThreadBufSize);
   size_t num_entries = GetNumEntries(clock_source_);
   size_t num_records = (kPerThreadBufSize - current_offset) / num_entries;
   DCHECK_EQ((kPerThreadBufSize - current_offset) % num_entries, 0u);
+  const size_t record_size = GetRecordSize(clock_source_, trace_format_version_);
+  DCHECK_LT(record_size, kPerThreadBufSize);
 
-  // Check if there is sufficient place in the buffer for non-streaming case. If not return early.
-  if (cur_offset_ + record_size * num_records >= buffer_size &&
-      trace_output_mode_ != TraceOutputMode::kStreaming) {
-    overflow_ = true;
-    return;
-  }
-
-  DCHECK_GT(buffer_size_, record_size * num_entries);
-  for (size_t entry_index = kPerThreadBufSize; entry_index != current_offset;) {
-    entry_index -= num_entries;
-    size_t record_index = entry_index;
-    uintptr_t method_and_action = method_trace_entries[record_index++];
-    ArtMethod* method = reinterpret_cast<ArtMethod*>(method_and_action & kMaskTraceAction);
-    CHECK(method != nullptr);
-    TraceAction action = DecodeTraceAction(method_and_action);
-    uint32_t thread_time = 0;
-    uint32_t wall_time = 0;
-    if (has_thread_cpu_clock) {
-      thread_time = method_trace_entries[record_index++];
-    }
-    if (has_wall_clock) {
-      uint64_t timestamp = method_trace_entries[record_index++];
-      if (art::kRuntimePointerSize == PointerSize::k32) {
-        // On 32-bit architectures timestamp is stored as two 32-bit values.
-        uint64_t high_timestamp = method_trace_entries[record_index++];
-        timestamp = (high_timestamp << 32 | timestamp);
-      }
-      wall_time = GetMicroTime(timestamp) - start_time_;
-    }
+  if (trace_output_mode_ != TraceOutputMode::kStreaming) {
+    // In non-streaming mode we only flush to file at the end, so retain the earlier data. If the
+    // buffer is full we don't process any more entries.
+    current_index = cur_offset_;
 
-    auto [method_id, is_new_method] = GetMethodEncoding(method);
-    if (is_new_method && trace_output_mode_ == TraceOutputMode::kStreaming) {
-      RecordMethodInfo(method_infos.find(method)->second, method_id);
+    // Check if there is sufficient place in the buffer for non-streaming case. If not return early.
+    if (cur_offset_ + record_size * num_records >= buffer_size) {
+      overflow_ = true;
+      return;
     }
+  }
+  num_records_ += num_records;
 
-    DCHECK_LT(current_index + record_size,  buffer_size);
-    EncodeEventEntry(
-        buffer_ptr + current_index, thread_id, method_id, action, thread_time, wall_time);
-    current_index += record_size;
+  DCHECK_GT(buffer_size_, record_size * num_entries);
+  if (trace_format_version_ == Trace::kFormatV1) {
+    FlushEntriesFormatV1(
+        method_trace_entries, tid, method_infos, current_offset, &current_index, buffer_ptr);
+  } else {
+    FlushEntriesFormatV2(
+        method_trace_entries, tid, method_infos, num_records, &current_index, buffer_ptr);
   }
 
   if (trace_output_mode_ == TraceOutputMode::kStreaming) {
@@ -1476,6 +1672,30 @@ void TraceWriter::EncodeEventEntry(uint8_t* ptr,
   static_assert(kPacketSize == 2 + 4 + 4 + 4, "Packet size incorrect.");
 }
 
+void TraceWriter::EncodeEventBlockHeader(uint8_t* ptr,
+                                         uint32_t thread_id,
+                                         uint32_t init_method_index,
+                                         uint32_t init_thread_clock,
+                                         uint32_t init_wall_clock,
+                                         uint16_t num_records) {
+  ptr[0] = kEntryHeaderV2;
+  Append4LE(ptr + 1, thread_id);
+  Append4LE(ptr + 5, init_method_index);
+  ptr += 9;
+
+  if (UseThreadCpuClock(clock_source_)) {
+    Append4LE(ptr, init_thread_clock);
+    ptr += 4;
+  }
+  if (UseWallClock(clock_source_)) {
+    Append4LE(ptr, init_wall_clock);
+    ptr += 4;
+  }
+  // This specifies the total number of records encoded in the block using lebs. We encode the first
+  // entry in the header, so the block contains one less than num_records.
+  Append2LE(ptr, num_records - 1);
+}
+
 void TraceWriter::EnsureSpace(uint8_t* buffer,
                               size_t* current_index,
                               size_t buffer_size,
@@ -1498,6 +1718,7 @@ void TraceWriter::DumpMethodList(std::ostream& os) {
 }
 
 void TraceWriter::DumpThreadList(std::ostream& os) {
+  MutexLock mu(Thread::Current(), tracing_lock_);
   for (const auto& it : threads_list_) {
     os << it.first << "\t" << it.second << "\n";
   }
diff --git a/runtime/trace.h b/runtime/trace.h
index 7ff68132fc..227955ff2d 100644
--- a/runtime/trace.h
+++ b/runtime/trace.h
@@ -48,6 +48,8 @@ class DexFile;
 class ShadowFrame;
 class Thread;
 
+struct MethodTraceRecord;
+
 using DexIndexBitSet = std::bitset<65536>;
 
 enum TracingMode {
@@ -146,6 +148,7 @@ class TraceWriter {
               TraceClockSource clock_source,
               size_t buffer_size,
               int num_trace_buffers,
+              int trace_format_version,
               uint32_t clock_overhead_ns);
 
   // This encodes all the events in the per-thread trace buffer and writes it to the trace file /
@@ -210,6 +213,25 @@ class TraceWriter {
   int GetMethodTraceIndex(uintptr_t* current_buffer);
 
  private:
+  void ReadValuesFromRecord(uintptr_t* method_trace_entries,
+                            size_t record_index,
+                            MethodTraceRecord& record,
+                            bool has_thread_cpu_clock,
+                            bool has_wall_clock);
+
+  void FlushEntriesFormatV2(uintptr_t* method_trace_entries,
+                            size_t tid,
+                            const std::unordered_map<ArtMethod*, std::string>& method_infos,
+                            size_t num_records,
+                            size_t* current_index,
+                            uint8_t* init_buffer_ptr) REQUIRES(tracing_lock_);
+
+  void FlushEntriesFormatV1(uintptr_t* method_trace_entries,
+                            size_t tid,
+                            const std::unordered_map<ArtMethod*, std::string>& method_infos,
+                            size_t end_offset,
+                            size_t* current_index,
+                            uint8_t* buffer_ptr) REQUIRES(tracing_lock_);
   // Get a 32-bit id for the method and specify if the method hasn't been seen before. If this is
   // the first time we see this method record information (like method name, declaring class etc.,)
   // about the method.
@@ -237,6 +259,15 @@ class TraceWriter {
                         uint32_t thread_clock_diff,
                         uint32_t wall_clock_diff) REQUIRES(tracing_lock_);
 
+  // Encodes the header for the events block. This assumes that there is enough space reserved to
+  // encode the entry.
+  void EncodeEventBlockHeader(uint8_t* ptr,
+                              uint32_t thread_id,
+                              uint32_t method_index,
+                              uint32_t init_thread_clock_time,
+                              uint32_t init_wall_clock_time,
+                              uint16_t num_records) REQUIRES(tracing_lock_);
+
   // Ensures there is sufficient space in the buffer to record the requested_size. If there is not
   // enough sufficient space the current contents of the buffer are written to the file and
   // current_index is reset to 0. This doesn't check if buffer_size is big enough to hold the
@@ -253,7 +284,7 @@ class TraceWriter {
   // Methods to output traced methods and threads.
   void DumpMethodList(std::ostream& os) REQUIRES_SHARED(Locks::mutator_lock_)
       REQUIRES(!tracing_lock_);
-  void DumpThreadList(std::ostream& os) REQUIRES(!Locks::thread_list_lock_) REQUIRES(tracing_lock_);
+  void DumpThreadList(std::ostream& os) REQUIRES(!Locks::thread_list_lock_, !tracing_lock_);
 
   // File to write trace data out to, null if direct to ddms.
   std::unique_ptr<File> trace_file_;
@@ -291,12 +322,18 @@ class TraceWriter {
   // Size of buf_.
   const size_t buffer_size_;
 
+  // Version of trace output
+  const int trace_format_version_;
+
   // Time trace was created.
   const uint64_t start_time_;
 
   // Did we overflow the buffer recording traces?
   bool overflow_;
 
+  // Total number of records flushed to file.
+  size_t num_records_;
+
   // Clock overhead.
   const uint32_t clock_overhead_ns_;
 
@@ -322,6 +359,11 @@ class Trace final : public instrumentation::InstrumentationListener {
     kTraceClockSourceThreadCpu = 0x100,
   };
 
+  static const int kFormatV1 = 0;
+  static const int kFormatV2 = 1;
+  static const int kTraceFormatVersionFlagMask = 0b110;
+  static const int kTraceFormatVersionShift = 1;
+
   enum class TraceMode {
     kMethodTracing,
     kSampling
diff --git a/runtime/well_known_classes.h b/runtime/well_known_classes.h
index 75be6747e8..bd8bbe0108 100644
--- a/runtime/well_known_classes.h
+++ b/runtime/well_known_classes.h
@@ -268,11 +268,14 @@ struct EXPORT WellKnownClasses {
       java_lang_Integer_IntegerCache;
   static constexpr ClassFromField<&java_lang_Long_LongCache_cache> java_lang_Long_LongCache;
 
+  static constexpr ClassFromMethod<&java_lang_Boolean_valueOf> java_lang_Boolean;
   static constexpr ClassFromMethod<&java_lang_Byte_valueOf> java_lang_Byte;
   static constexpr ClassFromMethod<&java_lang_Character_valueOf> java_lang_Character;
   static constexpr ClassFromMethod<&java_lang_Short_valueOf> java_lang_Short;
   static constexpr ClassFromMethod<&java_lang_Integer_valueOf> java_lang_Integer;
+  static constexpr ClassFromMethod<&java_lang_Float_valueOf> java_lang_Float;
   static constexpr ClassFromMethod<&java_lang_Long_valueOf> java_lang_Long;
+  static constexpr ClassFromMethod<&java_lang_Double_valueOf> java_lang_Double;
 };
 
 }  // namespace art
diff --git a/test/115-native-bridge/nativebridge.cc b/test/115-native-bridge/nativebridge.cc
index 3dcc1fcc4a..be3a2011d9 100644
--- a/test/115-native-bridge/nativebridge.cc
+++ b/test/115-native-bridge/nativebridge.cc
@@ -30,6 +30,7 @@
 #include <jni.h>
 #include <nativebridge/native_bridge.h>
 
+#include "base/casts.h"
 #include "base/macros.h"
 
 struct NativeBridgeMethod {
@@ -309,7 +310,9 @@ class SignalHandlerTestStatus {
 
   void AssertState(TestStatus expected) {
     if (state_ != expected) {
-      printf("ERROR: unexpected state, was %d, expected %d\n", state_, expected);
+      printf("ERROR: unexpected state, was %d, expected %d\n",
+             art::enum_cast<int>(state_),
+             art::enum_cast<int>(expected));
     }
   }
 };
diff --git a/test/2246-trace-v2/Android.bp b/test/2246-trace-v2/Android.bp
new file mode 100644
index 0000000000..6655926552
--- /dev/null
+++ b/test/2246-trace-v2/Android.bp
@@ -0,0 +1,40 @@
+// Generated by `regen-test-files`. Do not edit manually.
+
+// Build rules for ART run-test `2246-trace-v2`.
+
+package {
+    // See: http://go/android-license-faq
+    // A large-scale-change added 'default_applicable_licenses' to import
+    // all of the 'license_kinds' from "art_license"
+    // to get the below license kinds:
+    //   SPDX-license-identifier-Apache-2.0
+    default_applicable_licenses: ["art_license"],
+}
+
+// Test's Dex code.
+java_test {
+    name: "art-run-test-2246-trace-v2",
+    defaults: ["art-run-test-defaults"],
+    test_config_template: ":art-run-test-target-no-test-suite-tag-template",
+    srcs: ["src/**/*.java"],
+    data: [
+        ":art-run-test-2246-trace-v2-expected-stdout",
+        ":art-run-test-2246-trace-v2-expected-stderr",
+    ],
+}
+
+// Test's expected standard output.
+genrule {
+    name: "art-run-test-2246-trace-v2-expected-stdout",
+    out: ["art-run-test-2246-trace-v2-expected-stdout.txt"],
+    srcs: ["expected-stdout.txt"],
+    cmd: "cp -f $(in) $(out)",
+}
+
+// Test's expected standard error.
+genrule {
+    name: "art-run-test-2246-trace-v2-expected-stderr",
+    out: ["art-run-test-2246-trace-v2-expected-stderr.txt"],
+    srcs: ["expected-stderr.txt"],
+    cmd: "cp -f $(in) $(out)",
+}
diff --git a/test/2246-trace-v2/dump_trace.cc b/test/2246-trace-v2/dump_trace.cc
new file mode 100644
index 0000000000..6a094921fe
--- /dev/null
+++ b/test/2246-trace-v2/dump_trace.cc
@@ -0,0 +1,233 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <stdio.h>
+
+#include <map>
+#include <memory>
+
+#include "base/leb128.h"
+#include "base/os.h"
+#include "base/unix_file/fd_file.h"
+#include "jni.h"
+
+namespace art {
+namespace {
+
+static const int kMagicValue = 0x574f4c53;
+static const int kVersionDualClock = 0xf5;
+static const int kThreadInfo = 0;
+static const int kMethodInfo = 1;
+static const int kTraceEntries = 2;
+static const int kTraceActionBits = 2;
+static const int kSummary = 3;
+
+int ReadNumber(int num_bytes, uint8_t* header) {
+  int number = 0;
+  for (int i = 0; i < num_bytes; i++) {
+    number += header[i] << (i * 8);
+  }
+  return number;
+}
+
+bool ProcessThreadOrMethodInfo(std::unique_ptr<File>& file, std::map<int, std::string>& name_map) {
+  uint8_t header[6];
+  if (!file->ReadFully(&header, sizeof(header))) {
+    printf("Couldn't read header\n");
+    return false;
+  }
+  int id = ReadNumber(4, header);
+  int length = ReadNumber(2, header + 4);
+
+  char* name = new char[length];
+  if (!file->ReadFully(name, length)) {
+    delete[] name;
+    return false;
+  }
+  std::string str(name, length);
+  std::replace(str.begin(), str.end(), '\t', ' ');
+  name_map.emplace(id, str);
+  delete[] name;
+  return true;
+}
+
+void print_trace_entry(const std::string& thread_name,
+                       const std::string& method_name,
+                       int* current_depth,
+                       int event_type) {
+  std::string entry;
+  for (int i = 0; i < *current_depth; i++) {
+    entry.push_back('.');
+  }
+  if (event_type == 0) {
+    *current_depth += 1;
+    entry.append(".>> ");
+  } else if (event_type == 1) {
+    *current_depth -= 1;
+    entry.append("<< ");
+  } else if (event_type == 2) {
+    *current_depth -= 1;
+    entry.append("<<E ");
+  } else {
+    entry.append("?? ");
+  }
+  entry.append(thread_name);
+  entry.append(" ");
+  entry.append(method_name);
+  printf("%s", entry.c_str());
+}
+
+bool ProcessTraceEntries(std::unique_ptr<File>& file,
+                         std::map<int, int>& current_depth_map,
+                         std::map<int, std::string>& thread_map,
+                         std::map<int, std::string>& method_map,
+                         bool is_dual_clock,
+                         const char* thread_name_filter) {
+  uint8_t header[20];
+  int header_size = is_dual_clock ? 20 : 16;
+  if (!file->ReadFully(header, header_size)) {
+    return false;
+  }
+
+  uint32_t thread_id = ReadNumber(4, header);
+  uint32_t method_value = ReadNumber(4, header + 4);
+  int offset = 8;
+  if (is_dual_clock) {
+    // Read timestamp.
+    ReadNumber(4, header + offset);
+    offset += 4;
+  }
+  // Read timestamp.
+  ReadNumber(4, header + offset);
+  offset += 4;
+  int num_records = ReadNumber(2, header + offset);
+  offset += 2;
+  int total_size = ReadNumber(2, header + offset);
+  uint8_t* buffer = new uint8_t[total_size];
+  if (!file->ReadFully(buffer, total_size)) {
+    delete[] buffer;
+    return false;
+  }
+
+  const uint8_t* current_buffer_ptr = buffer;
+  int32_t method_id = method_value >> kTraceActionBits;
+  uint8_t event_type = method_value & 0x3;
+  int current_depth = 0;
+  if (current_depth_map.find(thread_id) != current_depth_map.end()) {
+    // Get the current call stack depth. If it is the first method we are seeing on this thread
+    // then this map wouldn't haven an entry we start with the depth of 0.
+    current_depth = current_depth_map[thread_id];
+  }
+  std::string thread_name = thread_map[thread_id];
+  bool print_thread_events = (thread_name.compare(thread_name_filter) == 0);
+  if (method_map.find(method_id) == method_map.end()) {
+    LOG(FATAL) << "No entry for init method " << method_id;
+  }
+  if (print_thread_events) {
+    print_trace_entry(thread_name, method_map[method_id], &current_depth, event_type);
+  }
+  for (int i = 0; i < num_records; i++) {
+    int32_t diff = 0;
+    bool success = DecodeSignedLeb128Checked(&current_buffer_ptr, buffer + total_size - 1, &diff);
+    if (!success) {
+      LOG(FATAL) << "Reading past the buffer???";
+    }
+    int32_t curr_method_value = method_value + diff;
+    method_id = curr_method_value >> kTraceActionBits;
+    event_type = curr_method_value & 0x3;
+    if (print_thread_events) {
+      print_trace_entry(thread_name, method_map[method_id], &current_depth, event_type);
+    }
+    // Read timestamps
+    DecodeUnsignedLeb128(&current_buffer_ptr);
+    if (is_dual_clock) {
+      DecodeUnsignedLeb128(&current_buffer_ptr);
+    }
+  }
+  current_depth_map[thread_id] = current_depth;
+  return true;
+}
+
+extern "C" JNIEXPORT void JNICALL Java_Main_dumpTrace(JNIEnv* env,
+                                                      jclass,
+                                                      jstring fileName,
+                                                      jstring threadName) {
+  const char* file_name = env->GetStringUTFChars(fileName, nullptr);
+  const char* thread_name = env->GetStringUTFChars(threadName, nullptr);
+  std::map<int, std::string> thread_map;
+  std::map<int, std::string> method_map;
+  std::map<int, int> current_depth_map;
+
+  std::unique_ptr<File> file(OS::OpenFileForReading(file_name));
+  if (file == nullptr) {
+    printf("Couldn't open file\n");
+    return;
+  }
+
+  uint8_t header[32];
+  const bool success = file->ReadFully(&header, sizeof(header));
+  if (!success) {
+    printf("Couldn't read header\n");
+    return;
+  }
+  int magic_value = ReadNumber(4, header);
+  if (magic_value != kMagicValue) {
+    printf("Incorrect magic value\n");
+    return;
+  }
+  int version = ReadNumber(2, header + 4);
+  if (success) {
+    printf("version=%0x\n", version);
+  }
+
+  bool is_dual_clock = (version == kVersionDualClock);
+  bool has_entries = true;
+  while (has_entries) {
+    uint8_t entry_header;
+    if (!file->ReadFully(&entry_header, sizeof(entry_header))) {
+      break;
+    }
+    switch (entry_header) {
+      case kThreadInfo:
+        if (!ProcessThreadOrMethodInfo(file, thread_map)) {
+          has_entries = false;
+        }
+        break;
+      case kMethodInfo:
+        if (!ProcessThreadOrMethodInfo(file, method_map)) {
+          has_entries = false;
+        }
+        break;
+      case kTraceEntries:
+        ProcessTraceEntries(
+            file, current_depth_map, thread_map, method_map, is_dual_clock, thread_name);
+        break;
+      case kSummary:
+        has_entries = false;
+        break;
+      default:
+        printf("Invalid Header %d\n", entry_header);
+        has_entries = false;
+        break;
+    }
+  }
+
+  env->ReleaseStringUTFChars(fileName, file_name);
+  env->ReleaseStringUTFChars(threadName, thread_name);
+}
+
+}  // namespace
+}  // namespace art
diff --git a/test/2246-trace-v2/expected-stderr.txt b/test/2246-trace-v2/expected-stderr.txt
new file mode 100644
index 0000000000..e69de29bb2
--- /dev/null
+++ b/test/2246-trace-v2/expected-stderr.txt
diff --git a/test/2246-trace-v2/expected-stdout.txt b/test/2246-trace-v2/expected-stdout.txt
new file mode 100644
index 0000000000..53430a6195
--- /dev/null
+++ b/test/2246-trace-v2/expected-stdout.txt
@@ -0,0 +1,773 @@
+JNI_OnLoad called
+***** streaming test - dual clock *******
+version=f5
+.>> TestThread2246 java.lang.Thread run ()V Thread.java
+..>> TestThread2246 Main$$ExternalSyntheticLambda0 run ()V D8$$SyntheticClass
+...>> TestThread2246 Main lambda$testTracing$0 (IZLMain;)V Main.java
+....>> TestThread2246 Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+.....>> TestThread2246 java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+......>> TestThread2246 dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+.......>> TestThread2246 dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+.......<< TestThread2246 dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+......<< TestThread2246 dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+.....<< TestThread2246 java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+....<< TestThread2246 Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+....>> TestThread2246 Main <init> ()V Main.java
+.....>> TestThread2246 java.lang.Object <init> ()V Object.java
+.....<< TestThread2246 java.lang.Object <init> ()V Object.java
+....<< TestThread2246 Main <init> ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWork ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWork ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main$VMDebug $noinline$stopMethodTracing ()V Main.java
+.....>> TestThread2246 java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+......>> TestThread2246 dalvik.system.VMDebug stopMethodTracing ()V VMDebug.java
+version=f5
+.>> main Main main ([Ljava/lang/String;)V Main.java
+..>> main Main testTracing (ZII)V Main.java
+...>> main Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+....>> main java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+.....>> main dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+......>> main dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+......<< main dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+.....<< main dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+....<< main java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+...<< main Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+...>> main Main $noinline$doSomeWork ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWork ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main doSomeWorkThrow ()V Main.java
+....>> main Main callThrowFunction ()V Main.java
+.....>> main java.lang.Exception <init> (Ljava/lang/String;)V Exception.java
+......>> main java.lang.Throwable <init> (Ljava/lang/String;)V Throwable.java
+.......>> main java.lang.Object <init> ()V Object.java
+.......<< main java.lang.Object <init> ()V Object.java
+.......>> main java.util.Collections emptyList ()Ljava/util/List; Collections.java
+.......<< main java.util.Collections emptyList ()Ljava/util/List; Collections.java
+.......>> main java.lang.Throwable fillInStackTrace ()Ljava/lang/Throwable; Throwable.java
+........>> main java.lang.Throwable nativeFillInStackTrace ()Ljava/lang/Object; Throwable.java
+........<< main java.lang.Throwable nativeFillInStackTrace ()Ljava/lang/Object; Throwable.java
+.......<< main java.lang.Throwable fillInStackTrace ()Ljava/lang/Throwable; Throwable.java
+......<< main java.lang.Throwable <init> (Ljava/lang/String;)V Throwable.java
+.....<< main java.lang.Exception <init> (Ljava/lang/String;)V Exception.java
+....<<E main Main callThrowFunction ()V Main.java
+...<< main Main doSomeWorkThrow ()V Main.java
+...>> main Main$VMDebug $noinline$stopMethodTracing ()V Main.java
+....>> main java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+.....>> main dalvik.system.VMDebug stopMethodTracing ()V VMDebug.java
+***** streaming test - wall clock *******
+version=f4
+.>> TestThread2246 java.lang.Thread run ()V Thread.java
+..>> TestThread2246 Main$$ExternalSyntheticLambda0 run ()V D8$$SyntheticClass
+...>> TestThread2246 Main lambda$testTracing$0 (IZLMain;)V Main.java
+....>> TestThread2246 Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+.....>> TestThread2246 java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+......>> TestThread2246 dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+.......>> TestThread2246 dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+.......<< TestThread2246 dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+......<< TestThread2246 dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+.....<< TestThread2246 java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+....<< TestThread2246 Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+....>> TestThread2246 Main <init> ()V Main.java
+.....>> TestThread2246 java.lang.Object <init> ()V Object.java
+.....<< TestThread2246 java.lang.Object <init> ()V Object.java
+....<< TestThread2246 Main <init> ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWork ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWork ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+.....>> TestThread2246 Main callOuterFunction ()V Main.java
+......>> TestThread2246 Main callLeafFunction ()V Main.java
+......<< TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callOuterFunction ()V Main.java
+.....>> TestThread2246 Main callLeafFunction ()V Main.java
+.....<< TestThread2246 Main callLeafFunction ()V Main.java
+....<< TestThread2246 Main $noinline$doSomeWorkJIT ()V Main.java
+....>> TestThread2246 Main$VMDebug $noinline$stopMethodTracing ()V Main.java
+.....>> TestThread2246 java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+......>> TestThread2246 dalvik.system.VMDebug stopMethodTracing ()V VMDebug.java
+version=f4
+.>> main Main main ([Ljava/lang/String;)V Main.java
+..>> main Main testTracing (ZII)V Main.java
+...>> main Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+....>> main java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+.....>> main dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+......>> main dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+......<< main dalvik.system.VMDebug startMethodTracingFd (Ljava/lang/String;IIIZIZ)V VMDebug.java
+.....<< main dalvik.system.VMDebug startMethodTracing (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V VMDebug.java
+....<< main java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+...<< main Main$VMDebug startMethodTracingV2 (Ljava/lang/String;Ljava/io/FileDescriptor;IIZIZ)V Main.java
+...>> main Main $noinline$doSomeWork ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWork ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main $noinline$doSomeWorkJIT ()V Main.java
+....>> main Main callOuterFunction ()V Main.java
+.....>> main Main callLeafFunction ()V Main.java
+.....<< main Main callLeafFunction ()V Main.java
+....<< main Main callOuterFunction ()V Main.java
+....>> main Main callLeafFunction ()V Main.java
+....<< main Main callLeafFunction ()V Main.java
+...<< main Main $noinline$doSomeWorkJIT ()V Main.java
+...>> main Main doSomeWorkThrow ()V Main.java
+....>> main Main callThrowFunction ()V Main.java
+.....>> main java.lang.Exception <init> (Ljava/lang/String;)V Exception.java
+......>> main java.lang.Throwable <init> (Ljava/lang/String;)V Throwable.java
+.......>> main java.lang.Object <init> ()V Object.java
+.......<< main java.lang.Object <init> ()V Object.java
+.......>> main java.util.Collections emptyList ()Ljava/util/List; Collections.java
+.......<< main java.util.Collections emptyList ()Ljava/util/List; Collections.java
+.......>> main java.lang.Throwable fillInStackTrace ()Ljava/lang/Throwable; Throwable.java
+........>> main java.lang.Throwable nativeFillInStackTrace ()Ljava/lang/Object; Throwable.java
+........<< main java.lang.Throwable nativeFillInStackTrace ()Ljava/lang/Object; Throwable.java
+.......<< main java.lang.Throwable fillInStackTrace ()Ljava/lang/Throwable; Throwable.java
+......<< main java.lang.Throwable <init> (Ljava/lang/String;)V Throwable.java
+.....<< main java.lang.Exception <init> (Ljava/lang/String;)V Exception.java
+....<<E main Main callThrowFunction ()V Main.java
+...<< main Main doSomeWorkThrow ()V Main.java
+...>> main Main$VMDebug $noinline$stopMethodTracing ()V Main.java
+....>> main java.lang.reflect.Method invoke (Ljava/lang/Object;[Ljava/lang/Object;)Ljava/lang/Object; Method.java
+.....>> main dalvik.system.VMDebug stopMethodTracing ()V VMDebug.java
diff --git a/test/2246-trace-v2/info.txt b/test/2246-trace-v2/info.txt
new file mode 100644
index 0000000000..fa93a971b4
--- /dev/null
+++ b/test/2246-trace-v2/info.txt
@@ -0,0 +1,2 @@
+Tests streaming method tracing. It verifies the format of the generated file is
+as expected.
diff --git a/test/2246-trace-v2/run.py b/test/2246-trace-v2/run.py
new file mode 100644
index 0000000000..4c0d8584f7
--- /dev/null
+++ b/test/2246-trace-v2/run.py
@@ -0,0 +1,22 @@
+#!/bin/bash
+#
+# Copyright (C) 2024 The Android Open Source Project
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+def run(ctx, args):
+  # The expected output on non debuggable configurations isn't consistent.
+  # TODO(mythria): Investigate why the output is different and update the test to work for
+  # non debuggable runtimes too.
+  ctx.default_run(args,  Xcompiler_option=["--debuggable"])
diff --git a/test/2246-trace-v2/src/Main.java b/test/2246-trace-v2/src/Main.java
new file mode 100644
index 0000000000..b5ccb9d0ec
--- /dev/null
+++ b/test/2246-trace-v2/src/Main.java
@@ -0,0 +1,178 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import java.io.File;
+import java.io.FileDescriptor;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.lang.reflect.Method;
+
+public class Main {
+    private static final String TEMP_FILE_NAME_PREFIX = "test";
+    private static final String TEMP_FILE_NAME_SUFFIX = ".trace";
+    private static final int WALL_CLOCK_FLAG = 0x010;
+    private static final int TRACE_OUTPUT_V2_FLAG = 0b010;
+    private static final int STREAMING_DUAL_CLOCK_VERSION = 1;
+    private static final int STREAMING_WALL_CLOCK_VERSION = 1;
+    private static File file;
+
+    public static void main(String[] args) throws Exception {
+        System.loadLibrary(args[0]);
+        String name = System.getProperty("java.vm.name");
+        if (!"Dalvik".equals(name)) {
+            System.out.println("This test is not supported on " + name);
+            return;
+        }
+
+        ensureJitCompiled(Main.class, "$noinline$doSomeWorkJIT");
+
+        System.out.println("***** streaming test - dual clock *******");
+        testTracing(
+                /* streaming=*/true, /* flags= */ 0, STREAMING_DUAL_CLOCK_VERSION);
+
+        System.out.println("***** streaming test - wall clock *******");
+        testTracing(
+                /* streaming=*/true, /* flags= */ WALL_CLOCK_FLAG, STREAMING_WALL_CLOCK_VERSION);
+    }
+
+    public static void testTracing(boolean streaming, int flags, int expected_version)
+            throws Exception {
+        Main m = new Main();
+        Thread t = new Thread(() -> {
+            try {
+                file = createTempFile();
+                FileOutputStream out_file = new FileOutputStream(file);
+                VMDebug.startMethodTracingV2(
+                        file.getPath(), out_file.getFD(), 0, flags, false, 0, streaming);
+                Main m1 = new Main();
+                m1.$noinline$doSomeWork();
+                // Call JITed code multiple times to flush out any issues with timestamps.
+                for (int i = 0; i < 20; i++) {
+                    m.$noinline$doSomeWorkJIT();
+                }
+                VMDebug.$noinline$stopMethodTracing();
+                out_file.close();
+                dumpTrace(file.getAbsolutePath(), "TestThread2246");
+                file.delete();
+            } catch (Exception e) {
+                System.out.println("Exception in thread " + e);
+                e.printStackTrace();
+            } finally {
+                file.delete();
+            }
+        }, "TestThread2246");
+        try {
+            if (VMDebug.getMethodTracingMode() != 0) {
+                VMDebug.$noinline$stopMethodTracing();
+            }
+
+            t.start();
+            t.join();
+
+            file = createTempFile();
+            FileOutputStream main_out_file = new FileOutputStream(file);
+            VMDebug.startMethodTracingV2(
+                    file.getPath(), main_out_file.getFD(), 0, flags, false, 0, streaming);
+            m.$noinline$doSomeWork();
+            // Call JITed code multiple times to flush out any issues with timestamps.
+            for (int i = 0; i < 20; i++) {
+                m.$noinline$doSomeWorkJIT();
+            }
+            m.doSomeWorkThrow();
+            VMDebug.$noinline$stopMethodTracing();
+            main_out_file.close();
+            dumpTrace(file.getAbsolutePath(), "main");
+            file.delete();
+        } finally {
+            file.delete();
+        }
+    }
+
+    private static File createTempFile() throws Exception {
+        try {
+            return File.createTempFile(TEMP_FILE_NAME_PREFIX, TEMP_FILE_NAME_SUFFIX);
+        } catch (IOException e) {
+            System.setProperty("java.io.tmpdir", "/data/local/tmp");
+            try {
+                return File.createTempFile(TEMP_FILE_NAME_PREFIX, TEMP_FILE_NAME_SUFFIX);
+            } catch (IOException e2) {
+                System.setProperty("java.io.tmpdir", "/sdcard");
+                return File.createTempFile(TEMP_FILE_NAME_PREFIX, TEMP_FILE_NAME_SUFFIX);
+            }
+        }
+    }
+
+    public void callOuterFunction() {
+        callLeafFunction();
+    }
+
+    public void callLeafFunction() {}
+
+    public void $noinline$doSomeWork() {
+        callOuterFunction();
+        callLeafFunction();
+    }
+
+    public void $noinline$doSomeWorkJIT() {
+        callOuterFunction();
+        callLeafFunction();
+    }
+
+    public void callThrowFunction() throws Exception {
+        throw new Exception("test");
+    }
+
+    public void doSomeWorkThrow() {
+        try {
+            callThrowFunction();
+        } catch (Exception e) {
+        }
+    }
+
+    private static class VMDebug {
+        private static final Method startMethodTracingMethod;
+        private static final Method stopMethodTracingMethod;
+        private static final Method getMethodTracingModeMethod;
+        static {
+            try {
+                Class<?> c = Class.forName("dalvik.system.VMDebug");
+                startMethodTracingMethod = c.getDeclaredMethod("startMethodTracing", String.class,
+                        FileDescriptor.class, Integer.TYPE, Integer.TYPE, Boolean.TYPE,
+                        Integer.TYPE, Boolean.TYPE);
+                stopMethodTracingMethod = c.getDeclaredMethod("stopMethodTracing");
+                getMethodTracingModeMethod = c.getDeclaredMethod("getMethodTracingMode");
+            } catch (Exception e) {
+                throw new RuntimeException(e);
+            }
+        }
+
+        public static void startMethodTracingV2(String filename, FileDescriptor fd, int bufferSize,
+                int flags, boolean samplingEnabled, int intervalUs, boolean streaming)
+                throws Exception {
+            startMethodTracingMethod.invoke(null, filename, fd, bufferSize,
+                    flags | TRACE_OUTPUT_V2_FLAG, samplingEnabled, intervalUs, streaming);
+        }
+        public static void $noinline$stopMethodTracing() throws Exception {
+            stopMethodTracingMethod.invoke(null);
+        }
+        public static int getMethodTracingMode() throws Exception {
+            return (int) getMethodTracingModeMethod.invoke(null);
+        }
+    }
+
+    private static native void ensureJitCompiled(Class<?> cls, String methodName);
+    private static native void dumpTrace(String fileName, String threadName);
+}
diff --git a/test/2247-checker-write-barrier-elimination/Android.bp b/test/2247-checker-write-barrier-elimination/Android.bp
index 5848cb496e..c9744e9b00 100644
--- a/test/2247-checker-write-barrier-elimination/Android.bp
+++ b/test/2247-checker-write-barrier-elimination/Android.bp
@@ -15,7 +15,7 @@ package {
 java_test {
     name: "art-run-test-2247-checker-write-barrier-elimination",
     defaults: ["art-run-test-defaults"],
-    test_config_template: ":art-run-test-target-no-test-suite-tag-template",
+    test_config_template: ":art-run-test-target-template",
     srcs: ["src/**/*.java"],
     data: [
         ":art-run-test-2247-checker-write-barrier-elimination-expected-stdout",
diff --git a/test/2272-checker-codegen-honor-write-barrier-kind/Android.bp b/test/2272-checker-codegen-honor-write-barrier-kind/Android.bp
new file mode 100644
index 0000000000..dc4292b4d8
--- /dev/null
+++ b/test/2272-checker-codegen-honor-write-barrier-kind/Android.bp
@@ -0,0 +1,43 @@
+// Generated by `regen-test-files`. Do not edit manually.
+
+// Build rules for ART run-test `2272-checker-codegen-honor-write-barrier-kind`.
+
+package {
+    // See: http://go/android-license-faq
+    // A large-scale-change added 'default_applicable_licenses' to import
+    // all of the 'license_kinds' from "art_license"
+    // to get the below license kinds:
+    //   SPDX-license-identifier-Apache-2.0
+    default_applicable_licenses: ["art_license"],
+}
+
+// Test's Dex code.
+java_test {
+    name: "art-run-test-2272-checker-codegen-honor-write-barrier-kind",
+    defaults: ["art-run-test-defaults"],
+    test_config_template: ":art-run-test-target-no-test-suite-tag-template",
+    srcs: ["src/**/*.java"],
+    data: [
+        ":art-run-test-2272-checker-codegen-honor-write-barrier-kind-expected-stdout",
+        ":art-run-test-2272-checker-codegen-honor-write-barrier-kind-expected-stderr",
+    ],
+    // Include the Java source files in the test's artifacts, to make Checker assertions
+    // available to the TradeFed test runner.
+    include_srcs: true,
+}
+
+// Test's expected standard output.
+genrule {
+    name: "art-run-test-2272-checker-codegen-honor-write-barrier-kind-expected-stdout",
+    out: ["art-run-test-2272-checker-codegen-honor-write-barrier-kind-expected-stdout.txt"],
+    srcs: ["expected-stdout.txt"],
+    cmd: "cp -f $(in) $(out)",
+}
+
+// Test's expected standard error.
+genrule {
+    name: "art-run-test-2272-checker-codegen-honor-write-barrier-kind-expected-stderr",
+    out: ["art-run-test-2272-checker-codegen-honor-write-barrier-kind-expected-stderr.txt"],
+    srcs: ["expected-stderr.txt"],
+    cmd: "cp -f $(in) $(out)",
+}
diff --git a/test/2272-checker-codegen-honor-write-barrier-kind/src/Main.java b/test/2272-checker-codegen-honor-write-barrier-kind/src/Main.java
index f07286b5b9..7d67249c6d 100644
--- a/test/2272-checker-codegen-honor-write-barrier-kind/src/Main.java
+++ b/test/2272-checker-codegen-honor-write-barrier-kind/src/Main.java
@@ -42,21 +42,20 @@ public class Main {
     /// CHECK-START: java.lang.String[] Main.$noinline$testArraySetsHonorWriteBarrier(java.lang.String[], java.lang.String) prepare_for_register_allocation (before)
     /// CHECK: <<Null:l\d+>>   NullConstant
     /// CHECK: <<BT:l\d+>>     BoundType [<<Null>>]
-    /// CHECK: ArraySet [<<arr:l\d+>>,<<index:i\d+>>,<<BT>>] value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:EmitBeingReliedOn
-    /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:DontEmit
+    /// CHECK: ArraySet [<<arr:l\d+>>,<<index:i\d+>>,<<BT>>] value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:DontEmit
+    /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:EmitNotBeingReliedOn
 
     /// CHECK-START: java.lang.String[] Main.$noinline$testArraySetsHonorWriteBarrier(java.lang.String[], java.lang.String) prepare_for_register_allocation (after)
     /// CHECK: <<Null:l\d+>>   NullConstant
-    /// CHECK: ArraySet [<<arr:l\d+>>,<<index:i\d+>>,<<Null>>] value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:EmitBeingReliedOn
-    /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:DontEmit
+    /// CHECK: ArraySet [<<arr:l\d+>>,<<index:i\d+>>,<<Null>>] value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:DontEmit
+    /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:EmitNotBeingReliedOn
 
     /// CHECK-START: java.lang.String[] Main.$noinline$testArraySetsHonorWriteBarrier(java.lang.String[], java.lang.String) prepare_for_register_allocation (after)
     /// CHECK-NOT: BoundType
 
     /// CHECK-START: java.lang.String[] Main.$noinline$testArraySetsHonorWriteBarrier(java.lang.String[], java.lang.String) disassembly (after)
-    /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:EmitBeingReliedOn
-    // / CHECK: ; card_table
     /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:DontEmit
+    /// CHECK: ArraySet value_can_be_null:true needs_type_check:false can_trigger_gc:false write_barrier_kind:EmitNotBeingReliedOn
     private static java.lang.String[] $noinline$testArraySetsHonorWriteBarrier(
             String[] arr, String o2) {
         Object o = null;
diff --git a/test/2273-checker-unreachable-intrinsics/Android.bp b/test/2273-checker-unreachable-intrinsics/Android.bp
new file mode 100644
index 0000000000..be4dd29437
--- /dev/null
+++ b/test/2273-checker-unreachable-intrinsics/Android.bp
@@ -0,0 +1,43 @@
+// Generated by `regen-test-files`. Do not edit manually.
+
+// Build rules for ART run-test `2273-checker-unreachable-intrinsics`.
+
+package {
+    // See: http://go/android-license-faq
+    // A large-scale-change added 'default_applicable_licenses' to import
+    // all of the 'license_kinds' from "art_license"
+    // to get the below license kinds:
+    //   SPDX-license-identifier-Apache-2.0
+    default_applicable_licenses: ["art_license"],
+}
+
+// Test's Dex code.
+java_test {
+    name: "art-run-test-2273-checker-unreachable-intrinsics",
+    defaults: ["art-run-test-defaults"],
+    test_config_template: ":art-run-test-target-template",
+    srcs: ["src/**/*.java"],
+    data: [
+        ":art-run-test-2273-checker-unreachable-intrinsics-expected-stdout",
+        ":art-run-test-2273-checker-unreachable-intrinsics-expected-stderr",
+    ],
+    // Include the Java source files in the test's artifacts, to make Checker assertions
+    // available to the TradeFed test runner.
+    include_srcs: true,
+}
+
+// Test's expected standard output.
+genrule {
+    name: "art-run-test-2273-checker-unreachable-intrinsics-expected-stdout",
+    out: ["art-run-test-2273-checker-unreachable-intrinsics-expected-stdout.txt"],
+    srcs: ["expected-stdout.txt"],
+    cmd: "cp -f $(in) $(out)",
+}
+
+// Test's expected standard error.
+genrule {
+    name: "art-run-test-2273-checker-unreachable-intrinsics-expected-stderr",
+    out: ["art-run-test-2273-checker-unreachable-intrinsics-expected-stderr.txt"],
+    srcs: ["expected-stderr.txt"],
+    cmd: "cp -f $(in) $(out)",
+}
diff --git a/test/2274-checker-bitwise-gvn/expected-stderr.txt b/test/2274-checker-bitwise-gvn/expected-stderr.txt
new file mode 100644
index 0000000000..e69de29bb2
--- /dev/null
+++ b/test/2274-checker-bitwise-gvn/expected-stderr.txt
diff --git a/test/2274-checker-bitwise-gvn/expected-stdout.txt b/test/2274-checker-bitwise-gvn/expected-stdout.txt
new file mode 100644
index 0000000000..e69de29bb2
--- /dev/null
+++ b/test/2274-checker-bitwise-gvn/expected-stdout.txt
diff --git a/test/2274-checker-bitwise-gvn/info.txt b/test/2274-checker-bitwise-gvn/info.txt
new file mode 100644
index 0000000000..1874dd6109
--- /dev/null
+++ b/test/2274-checker-bitwise-gvn/info.txt
@@ -0,0 +1,2 @@
+Tests that GVN doesn't deduplicate HBitwiseNegatedRight
+instructions with different kind.
diff --git a/test/2274-checker-bitwise-gvn/src/Main.java b/test/2274-checker-bitwise-gvn/src/Main.java
new file mode 100644
index 0000000000..57cdf15c65
--- /dev/null
+++ b/test/2274-checker-bitwise-gvn/src/Main.java
@@ -0,0 +1,54 @@
+/*
+ * Copyright (C) 2024 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+public class Main {
+    public static void main(String[] args) {
+        // Test with even/odd input, and even/odd amount of loop iterations.
+        assertEquals(-33, $noinline$TwoBitwiseOperations(32, 4));
+        assertEquals(1, $noinline$TwoBitwiseOperations(32, 5));
+        assertEquals(-31, $noinline$TwoBitwiseOperations(31, 4));
+        assertEquals(0, $noinline$TwoBitwiseOperations(31, 5));
+    }
+
+    /// CHECK-START-ARM: int Main.$noinline$TwoBitwiseOperations(int, int) instruction_simplifier_arm (after)
+    /// CHECK:       BitwiseNegatedRight kind:And
+    /// CHECK:       BitwiseNegatedRight kind:Or
+
+    /// CHECK-START-ARM64: int Main.$noinline$TwoBitwiseOperations(int, int) instruction_simplifier_arm64 (after)
+    /// CHECK:       BitwiseNegatedRight kind:And
+    /// CHECK:       BitwiseNegatedRight kind:Or
+
+    /// CHECK-START-{ARM,ARM64}: int Main.$noinline$TwoBitwiseOperations(int, int) disassembly (after)
+    /// CHECK:       BitwiseNegatedRight kind:And
+    /// CHECK:       BitwiseNegatedRight kind:Or
+    private static int $noinline$TwoBitwiseOperations(int a, int n) {
+        int result = 0;
+        for (int i = 0; i < n; ++i) {
+            if (i % 2 == 0) {
+                result = (~a) & 1;
+            } else {
+                result = (~a) | 1;
+            }
+        }
+        return result;
+    }
+
+    public static void assertEquals(int expected, int actual) {
+        if (expected != actual) {
+            throw new Error("Expected: " + expected + ", found: " + actual);
+        }
+    }
+}
diff --git a/test/458-checker-instruct-simplification/src/Main.java b/test/458-checker-instruct-simplification/src/Main.java
index 76239be741..784acf7949 100644
--- a/test/458-checker-instruct-simplification/src/Main.java
+++ b/test/458-checker-instruct-simplification/src/Main.java
@@ -3692,6 +3692,206 @@ public class Main {
     return arg << 25 >> 26;
   }
 
+  // Check that we don't introduce new implicit type conversions so the following pattern
+  // does not occur in the graph:
+  //
+  // <<ImplicitConv>> TypeConversion
+  // <<ExplicitConv>> TypeConversonn [<<ImplicitConv>>]
+  //
+  // That will lead to a crash because InstructionSimplifier removes implicit type conversions
+  // and during visiting TypeConversion instruction expects that its inputs have been already
+  // simplified.
+  //
+  // The structure of the following tests is
+  //
+  //   (T) ((x << N) >> N) or (T) ((x << N) >>> N)
+  //
+  // where
+  //   * K is a type of x
+  //   * Shifts correspond to implicit type conversion K -> M
+  //   * M -> T conversion is explicit
+  //
+  // T itself doesn't matter, the only important thing is that M -> T is explicit.
+  //
+  // We check cases when shifts correspond to the following implicit type conversions:
+  //   byte -> byte
+  //   byte -> short
+  //   unsigned byte -> unsigned byte
+  //   unsigned byte -> short
+  //   unsigned byte -> char
+  //   short -> short
+  //   char -> char
+  //
+  // To produce unsigned byte bitwise AND with 0xFF is used.
+
+  /// CHECK-START: int Main.$noinline$testByteToByteToChar(byte) instruction_simplifier (before)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Const24:i\d+>> IntConstant 24
+  /// CHECK:     <<Shl:i\d+>>     Shl [<<Param>>,<<Const24>>]
+  /// CHECK:     <<Shr:i\d+>>     Shr [<<Shl>>,<<Const24>>]
+  /// CHECK:     <<Conv:c\d+>>    TypeConversion [<<Shr>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testByteToByteToChar(byte) instruction_simplifier (after)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Conv:c\d+>>    TypeConversion [<<Param>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testByteToByteToChar(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testByteToByteToChar(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+  private static int $noinline$testByteToByteToChar(byte arg) {
+    return (char) ((arg << 24) >> 24);
+  }
+
+  /// CHECK-START: int Main.$noinline$testByteToShortToByte(byte) instruction_simplifier (before)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Const16:i\d+>> IntConstant 16
+  /// CHECK:     <<Shl:i\d+>>     Shl [<<Param>>,<<Const16>>]
+  /// CHECK:     <<Shr:i\d+>>     Shr [<<Shl>>,<<Const16>>]
+  /// CHECK:     <<Conv:b\d+>>    TypeConversion [<<Shr>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Return:v\d+>>  Return [<<Param>>]
+
+  /// CHECK-START: int Main.$noinline$testByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+
+  /// CHECK-START: int Main.$noinline$testByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  TypeConversion
+  private static int $noinline$testByteToShortToByte(byte arg) {
+    return (byte) ((arg << 16) >> 16);
+  }
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToUnsignedByteToByte(byte) instruction_simplifier (before)
+  /// CHECK:     <<Param:b\d+>>    ParameterValue
+  /// CHECK:     <<Const255:i\d+>> IntConstant 255
+  /// CHECK:     <<Const24:i\d+>>  IntConstant 24
+  /// CHECK:     <<And:i\d+>>      And [<<Param>>,<<Const255>>]
+  /// CHECK:     <<Shl:i\d+>>      Shl [<<And>>,<<Const24>>]
+  /// CHECK:     <<UShr:i\d+>>     UShr [<<Shl>>,<<Const24>>]
+  /// CHECK:     <<Conv:b\d+>>     TypeConversion [<<UShr>>]
+  /// CHECK:     <<Return:v\d+>>   Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToUnsignedByteToByte(byte) instruction_simplifier (after)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Return:v\d+>>  Return [<<Param>>]
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToUnsignedByteToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToUnsignedByteToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToUnsignedByteToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  TypeConversion
+  private static int $noinline$testUnsignedByteToUnsignedByteToByte(byte arg) {
+    return (byte) (((arg & 0xFF) << 24) >>> 24);
+  }
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToShortToByte(byte) instruction_simplifier (before)
+  /// CHECK:     <<Param:b\d+>>    ParameterValue
+  /// CHECK:     <<Const255:i\d+>> IntConstant 255
+  /// CHECK:     <<Const16:i\d+>>  IntConstant 16
+  /// CHECK:     <<And:i\d+>>      And [<<Param>>,<<Const255>>]
+  /// CHECK:     <<Shl:i\d+>>      Shl [<<And>>,<<Const16>>]
+  /// CHECK:     <<Shr:i\d+>>      Shr [<<Shl>>,<<Const16>>]
+  /// CHECK:     <<Conv:b\d+>>     TypeConversion [<<Shr>>]
+  /// CHECK:     <<Return:v\d+>>   Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Return:v\d+>>  Return [<<Param>>]
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToShortToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  TypeConversion
+  private static int $noinline$testUnsignedByteToShortToByte(byte arg) {
+    return (byte) (((arg & 0xFF) << 16) >> 16);
+  }
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToCharToByte(byte) instruction_simplifier (before)
+  /// CHECK:     <<Param:b\d+>>    ParameterValue
+  /// CHECK:     <<Const255:i\d+>> IntConstant 255
+  /// CHECK:     <<Const16:i\d+>>  IntConstant 16
+  /// CHECK:     <<And:i\d+>>      And [<<Param>>,<<Const255>>]
+  /// CHECK:     <<Shl:i\d+>>      Shl [<<And>>,<<Const16>>]
+  /// CHECK:     <<UShr:i\d+>>     UShr [<<Shl>>,<<Const16>>]
+  /// CHECK:     <<Conv:b\d+>>     TypeConversion [<<UShr>>]
+  /// CHECK:     <<Return:v\d+>>   Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToCharToByte(byte) instruction_simplifier (after)
+  /// CHECK:     <<Param:b\d+>>   ParameterValue
+  /// CHECK:     <<Return:v\d+>>  Return [<<Param>>]
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToCharToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToCharToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+
+  /// CHECK-START: int Main.$noinline$testUnsignedByteToCharToByte(byte) instruction_simplifier (after)
+  /// CHECK-NOT:                  TypeConversion
+  private static int $noinline$testUnsignedByteToCharToByte(byte arg) {
+    return (byte) (((arg & 0xFF) << 16) >>> 16);
+  }
+
+  /// CHECK-START: int Main.$noinline$testShortToShortToByte(short) instruction_simplifier (before)
+  /// CHECK:     <<Param:s\d+>>   ParameterValue
+  /// CHECK:     <<Const16:i\d+>> IntConstant 16
+  /// CHECK:     <<Shl:i\d+>>     Shl [<<Param>>,<<Const16>>]
+  /// CHECK:     <<Shr:i\d+>>     Shr [<<Shl>>,<<Const16>>]
+  /// CHECK:     <<Conv:b\d+>>    TypeConversion [<<Shr>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testShortToShortToByte(short) instruction_simplifier (after)
+  /// CHECK:     <<Param:s\d+>>   ParameterValue
+  /// CHECK:     <<Conv:b\d+>>    TypeConversion [<<Param>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testShortToShortToByte(short) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testShortToShortToByte(short) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+  private static int $noinline$testShortToShortToByte(short arg) {
+    return (byte) ((arg << 16) >> 16);
+  }
+
+  /// CHECK-START: int Main.$noinline$testCharToCharToByte(char) instruction_simplifier (before)
+  /// CHECK:     <<Param:c\d+>>   ParameterValue
+  /// CHECK:     <<Const16:i\d+>> IntConstant 16
+  /// CHECK:     <<Shl:i\d+>>     Shl [<<Param>>,<<Const16>>]
+  /// CHECK:     <<UShr:i\d+>>    UShr [<<Shl>>,<<Const16>>]
+  /// CHECK:     <<Conv:b\d+>>    TypeConversion [<<UShr>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testCharToCharToByte(char) instruction_simplifier (after)
+  /// CHECK:     <<Param:c\d+>>   ParameterValue
+  /// CHECK:     <<Conv:b\d+>>    TypeConversion [<<Param>>]
+  /// CHECK:     <<Return:v\d+>>  Return [<<Conv>>]
+
+  /// CHECK-START: int Main.$noinline$testCharToCharToByte(char) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shl
+
+  /// CHECK-START: int Main.$noinline$testCharToCharToByte(char) instruction_simplifier (after)
+  /// CHECK-NOT:                  Shr
+  private static int $noinline$testCharToCharToByte(char arg) {
+    return (byte) ((arg << 16) >>> 16);
+  }
+
   public static void main(String[] args) throws Exception {
     Class smaliTests2 = Class.forName("SmaliTests2");
     Method $noinline$XorAllOnes = smaliTests2.getMethod("$noinline$XorAllOnes", int.class);
@@ -4107,6 +4307,27 @@ public class Main {
         $noinline$testUnsignedPromotionPatternWithDifferentShiftAmountConstants(0xaabbccdd));
     assertIntEquals(0xffffffee,
         $noinline$testSignedPromotionPatternWithDifferentShiftAmountConstants(0xaabbccdd));
+
+    assertIntEquals(0xffaa, $noinline$testByteToByteToChar((byte) 0xaa));
+    assertIntEquals(0x0a, $noinline$testByteToByteToChar((byte) 0x0a));
+
+    assertIntEquals(0x0a, $noinline$testByteToShortToByte((byte) 0x0a));
+    assertIntEquals(0xffffffaa, $noinline$testByteToShortToByte((byte) 0xaa));
+
+    assertIntEquals(0x0a, $noinline$testUnsignedByteToUnsignedByteToByte((byte) 0x0a));
+    assertIntEquals(0xffffffaa, $noinline$testUnsignedByteToUnsignedByteToByte((byte) 0xaa));
+
+    assertIntEquals(0x0a, $noinline$testUnsignedByteToShortToByte((byte) 0x0a));
+    assertIntEquals(0xffffffaa, $noinline$testUnsignedByteToShortToByte((byte) 0xaa));
+
+    assertIntEquals(0x0a, $noinline$testUnsignedByteToCharToByte((byte) 0x0a));
+    assertIntEquals(0xffffffaa, $noinline$testUnsignedByteToCharToByte((byte) 0xaa));
+
+    assertIntEquals(0x0b, $noinline$testShortToShortToByte((short) 0xaa0b));
+    assertIntEquals(0xffffffbb, $noinline$testShortToShortToByte((short) 0xaabb));
+
+    assertIntEquals(0x0b, $noinline$testCharToCharToByte((char) 0xaa0b));
+    assertIntEquals(0xffffffbb, $noinline$testCharToCharToByte((char) 0xaabb));
   }
 
   private static boolean $inline$true() { return true; }
diff --git a/test/557-checker-instruct-simplifier-ror/src/Main.java b/test/557-checker-instruct-simplifier-ror/src/Main.java
index 5d4bb7ab33..667b35f5d9 100644
--- a/test/557-checker-instruct-simplifier-ror/src/Main.java
+++ b/test/557-checker-instruct-simplifier-ror/src/Main.java
@@ -503,6 +503,7 @@ public class Main {
   }
 
   //  (j << distance) + (j >>> -distance)
+  // We can't perform the optimization as distance might be `0`, resulting in the wrong value.
 
   /// CHECK-START: long Main.rol_long_reg_v_negv_add(long, int) instruction_simplifier (before)
   /// CHECK:          <<ArgValue:j\d+>>     ParameterValue
@@ -516,19 +517,17 @@ public class Main {
   /// CHECK-START: long Main.rol_long_reg_v_negv_add(long, int) instruction_simplifier (after)
   /// CHECK:          <<ArgValue:j\d+>>     ParameterValue
   /// CHECK:          <<ArgDistance:i\d+>>  ParameterValue
-  /// CHECK:          <<Neg:i\d+>>          Neg [<<ArgDistance>>]
-  /// CHECK:          <<Ror:j\d+>>          Ror [<<ArgValue>>,<<Neg>>]
-  /// CHECK:                                Return [<<Ror>>]
-
-  /// CHECK-START: long Main.rol_long_reg_v_negv_add(long, int) instruction_simplifier (after)
-  /// CHECK-NOT:  Add
-  /// CHECK-NOT:  Shl
-  /// CHECK-NOT:  UShr
+  /// CHECK-DAG:      <<Neg:i\d+>>          Neg [<<ArgDistance>>]
+  /// CHECK-DAG:      <<UShr:j\d+>>         UShr [<<ArgValue>>,<<Neg>>]
+  /// CHECK-DAG:      <<Shl:j\d+>>          Shl [<<ArgValue>>,<<ArgDistance>>]
+  /// CHECK:          <<Add:j\d+>>          Add [<<Shl>>,<<UShr>>]
+  /// CHECK:                                Return [<<Add>>]
   public static long rol_long_reg_v_negv_add(long value, int distance) {
     return (value << distance) + (value >>> -distance);
   }
 
   //  (j << distance) ^ (j >>> -distance)
+  // We can't perform the optimization as distance might be `0`, resulting in the wrong value.
 
   /// CHECK-START: long Main.rol_long_reg_v_negv_xor(long, int) instruction_simplifier (before)
   /// CHECK:          <<ArgValue:j\d+>>     ParameterValue
@@ -542,18 +541,60 @@ public class Main {
   /// CHECK-START: long Main.rol_long_reg_v_negv_xor(long, int) instruction_simplifier (after)
   /// CHECK:          <<ArgValue:j\d+>>     ParameterValue
   /// CHECK:          <<ArgDistance:i\d+>>  ParameterValue
-  /// CHECK:          <<Neg:i\d+>>          Neg [<<ArgDistance>>]
-  /// CHECK:          <<Ror:j\d+>>          Ror [<<ArgValue>>,<<Neg>>]
-  /// CHECK:                                Return [<<Ror>>]
+  /// CHECK-DAG:      <<Neg:i\d+>>          Neg [<<ArgDistance>>]
+  /// CHECK-DAG:      <<UShr:j\d+>>         UShr [<<ArgValue>>,<<Neg>>]
+  /// CHECK-DAG:      <<Shl:j\d+>>          Shl [<<ArgValue>>,<<ArgDistance>>]
+  /// CHECK:          <<Xor:j\d+>>          Xor [<<Shl>>,<<UShr>>]
+  /// CHECK:                                Return [<<Xor>>]
 
-  /// CHECK-START: long Main.rol_long_reg_v_negv_xor(long, int) instruction_simplifier (after)
-  /// CHECK-NOT:  Xor
-  /// CHECK-NOT:  Shl
-  /// CHECK-NOT:  UShr
   public static long rol_long_reg_v_negv_xor(long value, int distance) {
     return (value << distance) ^ (value >>> -distance);
   }
 
+  /// CHECK-START: void Main.$noinline$testDontOptimizeAddIntoRotate_Int() disassembly (after)
+  /// CHECK-NOT: Ror
+  public static void $noinline$testDontOptimizeAddIntoRotate_Int() {
+      int distance = returnFalse() ? 1 : 0;
+      int value = -512667375;
+      int expected_result = 2 * value;
+      int result = (value >>> distance) + (value << -distance);
+      assertIntEquals(expected_result, result);
+  }
+
+  /// CHECK-START: void Main.$noinline$testDontOptimizeAddIntoRotate_Long() disassembly (after)
+  /// CHECK-NOT: Ror
+  public static void $noinline$testDontOptimizeAddIntoRotate_Long() {
+      int distance = returnFalse() ? 1 : 0;
+      long value = -512667375L;
+      long expected_result = 2L * value;
+      long result = (value >>> distance) + (value << -distance);
+      assertLongEquals(expected_result, result);
+  }
+
+  /// CHECK-START: void Main.$noinline$testDontOptimizeXorIntoRotate_Int() disassembly (after)
+  /// CHECK-NOT: Ror
+  public static void $noinline$testDontOptimizeXorIntoRotate_Int() {
+      int distance = returnFalse() ? 1 : 0;
+      int value = -512667375;
+      int expected_result = 0;
+      int result = (value >>> distance) ^ (value << -distance);
+      assertIntEquals(expected_result, result);
+  }
+
+  /// CHECK-START: void Main.$noinline$testDontOptimizeXorIntoRotate_Long() disassembly (after)
+  /// CHECK-NOT: Ror
+  public static void $noinline$testDontOptimizeXorIntoRotate_Long() {
+      int distance = returnFalse() ? 1 : 0;
+      long value = -512667375L;
+      long expected_result = 0;
+      long result = (value >>> distance) ^ (value << -distance);
+      assertLongEquals(expected_result, result);
+  }
+
+  static boolean returnFalse() {
+    return false;
+  }
+
   public static void main(String[] args) {
     assertIntEquals(2, ror_int_constant_c_c(8));
     assertIntEquals(2, ror_int_constant_c_c_0(8));
@@ -581,5 +622,11 @@ public class Main {
 
     assertLongEquals(32L, rol_long_reg_v_negv_add(8L, 2));
     assertLongEquals(32L, rol_long_reg_v_negv_xor(8L, 2));
+
+    $noinline$testDontOptimizeAddIntoRotate_Int();
+    $noinline$testDontOptimizeAddIntoRotate_Long();
+
+    $noinline$testDontOptimizeXorIntoRotate_Int();
+    $noinline$testDontOptimizeXorIntoRotate_Long();
   }
 }
diff --git a/test/638-checker-inline-cache-intrinsic/src/Main.java b/test/638-checker-inline-cache-intrinsic/src/Main.java
index f25d03a894..86430e6e19 100644
--- a/test/638-checker-inline-cache-intrinsic/src/Main.java
+++ b/test/638-checker-inline-cache-intrinsic/src/Main.java
@@ -36,8 +36,8 @@ public class Main {
   /// CHECK:       InvokeInterface method_name:java.lang.CharSequence.charAt
 
   /// CHECK-START: char Main.$noinline$inlinePolymorphic(java.lang.CharSequence) inliner (after)
-  /// CHECK:       InvokeVirtual method_name:java.lang.String.charAt intrinsic:StringCharAt
-  /// CHECK:       Deoptimize
+  /// CHECK-DAG:   InvokeVirtual method_name:java.lang.String.charAt intrinsic:StringCharAt
+  /// CHECK-DAG:   Deoptimize
 
   /// CHECK-START: char Main.$noinline$inlinePolymorphic(java.lang.CharSequence) instruction_simplifier$after_inlining (after)
   /// CHECK:       Deoptimize
diff --git a/test/928-jni-table/jni_table.cc b/test/928-jni-table/jni_table.cc
index 1dfe34bd6b..52cd374f8e 100644
--- a/test/928-jni-table/jni_table.cc
+++ b/test/928-jni-table/jni_table.cc
@@ -44,7 +44,6 @@ static void DoDeleteGlobalRef(JNIEnv* env, jobject o) {
   CHECK(thr != nullptr);
   if (env->IsInstanceOf(o, thr)) {
     jvmtiThreadInfo jti;
-    // b/146170834: This could cause DCHECK failures.
     CHECK_EQ(jvmti_env->GetThreadInfo(reinterpret_cast<jthread>(o), &jti), JVMTI_ERROR_NONE);
   }
   gOriginalEnv->DeleteGlobalRef(env, o);
diff --git a/test/941-recursive-obsolete-jit/src/Main.java b/test/941-recursive-obsolete-jit/src/Main.java
index e3065a7117..f2a397e6bb 100644
--- a/test/941-recursive-obsolete-jit/src/Main.java
+++ b/test/941-recursive-obsolete-jit/src/Main.java
@@ -135,6 +135,12 @@ public class Main {
     do {
       // Run ensureJitCompiled here since it might get GCd
       ensureJitCompiled(Transform.class, "sayHi");
+      // We want to make sure sayHi method gets deoptimized. So we cannot allow any runtime frames
+      // between sayHi and the run method where the transformation is happening. If the run method
+      // is interpreted there will be a runtime frame to transition from JIT to interpreted code.
+      // So ensure the run method is JITed too, so we don't loop for a long time in the hope of
+      // getting the run method JITed.
+      ensureJitCompiled(do_redefinition.getClass(), "run");
       // Clear output.
       reporter.clear();
       t.sayHi(2, reporter, () -> { reporter.accept("Not doing anything here"); });
diff --git a/test/943-private-recursive-jit/src/Main.java b/test/943-private-recursive-jit/src/Main.java
index 09337bae26..9fa6607e85 100644
--- a/test/943-private-recursive-jit/src/Main.java
+++ b/test/943-private-recursive-jit/src/Main.java
@@ -151,6 +151,12 @@ public class Main {
       // Run ensureJitCompiled here since it might get GCd
       ensureJitCompiled(Transform.class, "sayHi");
       ensureJitCompiled(Transform.class, "privateSayHi");
+      // We want to make sure sayHi method gets deoptimized. So we cannot allow any runtime frames
+      // between sayHi and the run method where the transformation is happening. If the run method
+      // is interpreted there will be a runtime frame to transition from JIT to interpreted code.
+      // So ensure the run method is JITed too, so we don't loop for a long time in the hope of
+      // getting the run method JITed.
+      ensureJitCompiled(do_redefinition.getClass(), "run");
       // Clear output.
       reporter.clear();
       t.sayHi(2, reporter, () -> { reporter.accept("Not doing anything here"); });
diff --git a/test/Android.bp b/test/Android.bp
index 5406dfd976..1768340530 100644
--- a/test/Android.bp
+++ b/test/Android.bp
@@ -924,6 +924,7 @@ cc_defaults {
         "2040-huge-native-alloc/huge_native_buf.cc",
         "2048-bad-native-registry/native_finalizer.cc",
         "2235-JdkUnsafeTest/unsafe_test.cc",
+        "2246-trace-v2/dump_trace.cc",
         "2262-miranda-methods/jni_invoke.cc",
         "2270-mh-internal-hiddenapi-use/mh-internal-hidden-api.cc",
         "common/runtime_state.cc",
diff --git a/test/common/stack_inspect.cc b/test/common/stack_inspect.cc
index 1dd5f5a79d..8be355005a 100644
--- a/test/common/stack_inspect.cc
+++ b/test/common/stack_inspect.cc
@@ -82,6 +82,8 @@ static bool IsMethodInterpreted(Thread* self,
   StackVisitor::WalkStack(
       [&](const art::StackVisitor* stack_visitor) REQUIRES_SHARED(Locks::mutator_lock_) {
         if (goal == stack_visitor->GetMethod()) {
+          // We don't deoptimize beyond a runtime frame. So if we need the method to be
+          // deoptimizeable we cannot allow the previous frame to be a runtime frame.
           *method_is_interpreted =
               (require_deoptable && prev_was_runtime) || stack_visitor->IsShadowFrame();
           method_found = true;
diff --git a/test/knownfailures.json b/test/knownfailures.json
index a0b7dcd760..b47a1a1b35 100644
--- a/test/knownfailures.json
+++ b/test/knownfailures.json
@@ -1238,6 +1238,7 @@
                   "2238-checker-polymorphic-recursive-inlining",
                   "2240-tracing-non-invokable-method",
                   "2246-trace-stream",
+                  "2246-trace-v2",
                   "2254-class-value-before-and-after-u",
                   "2261-badcleaner-in-systemcleaner",
                   "2263-method-trace-jit",
@@ -1571,7 +1572,7 @@
     },
     {
       "tests": ["2271-profile-inline-cache"],
-      "variant": "jit-on-first-use",
+      "variant": "jit-on-first-use | debuggable | trace |  stream",
       "description": ["Test relies on profiling done by baseline compiled code. Meanwhile, it",
                       "can't use --baseline because it has a test case checking the behavior when",
                       "a method is optimized compiled."]
diff --git a/test/run-test b/test/run-test
index de676a0eaa..17d5227816 100755
--- a/test/run-test
+++ b/test/run-test
@@ -113,14 +113,14 @@ if True:
     tmp_dir = f"{TMPDIR}/{test_dir}"
   checker = f"{progdir}/../tools/checker/checker.py"
 
-  ON_VM = os.environ.get("ART_TEST_ON_VM")
-  SSH_USER = os.environ.get("ART_TEST_SSH_USER")
-  SSH_HOST = os.environ.get("ART_TEST_SSH_HOST")
-  SSH_PORT = os.environ.get("ART_TEST_SSH_PORT")
-  SSH_CMD = os.environ.get("ART_SSH_CMD")
-  SCP_CMD = os.environ.get("ART_SCP_CMD")
-  CHROOT = os.environ.get("ART_TEST_CHROOT")
-  CHROOT_CMD = os.environ.get("ART_CHROOT_CMD")
+  ON_VM = env.ART_TEST_ON_VM
+  SSH_USER = env.ART_TEST_SSH_USER
+  SSH_HOST = env.ART_TEST_SSH_HOST
+  SSH_PORT = env.ART_TEST_SSH_PORT
+  SSH_CMD = env.ART_SSH_CMD
+  SCP_CMD = env.ART_SCP_CMD
+  CHROOT = env.ART_TEST_CHROOT
+  CHROOT_CMD = env.ART_CHROOT_CMD
 
   def fail(message: str, caller:Optional[FrameInfo]=None):
     caller = caller or getframeinfo(currentframe().f_back)  # type: ignore
diff --git a/test/testrunner/env.py b/test/testrunner/env.py
index 200de4a97d..8313756995 100644
--- a/test/testrunner/env.py
+++ b/test/testrunner/env.py
@@ -148,4 +148,17 @@ SOONG_OUT_DIR = _get_build_var('SOONG_OUT_DIR')
 ART_TEST_RUN_ON_ARM_FVP = _getEnvBoolean('ART_TEST_RUN_ON_ARM_FVP', False)
 
 ART_TEST_ON_VM = _env.get('ART_TEST_ON_VM')
-ART_SSH_CMD = _env.get('ART_SSH_CMD')
+
+ART_TEST_SSH_PORT = _env.get('ART_TEST_SSH_PORT', 10001)
+ART_TEST_SSH_USER = _env.get('ART_TEST_SSH_USER', 'ubuntu')
+ART_TEST_SSH_HOST = _env.get('ART_TEST_SSH_HOST', 'localhost')
+ART_SSH_CMD = _env.get('ART_SSH_CMD', f"ssh -q -i ~/.ssh/ubuntu -p {ART_TEST_SSH_PORT} "
+                                      f"-o StrictHostKeyChecking=no "
+                                      f"{ART_TEST_SSH_USER}@{ART_TEST_SSH_HOST}")
+ART_SCP_CMD = _env.get('ART_SCP_CMD', f"scp -i ~/.ssh/ubuntu  -P {ART_TEST_SSH_PORT} "
+                                      f"-o StrictHostKeyChecking=no -p -r")
+ART_CHROOT_CMD = _env.get('ART_CHROOT_CMD', "unshare --user --map-root-user chroot art-test-chroot")
+if ART_TEST_ON_VM:
+  ART_TEST_CHROOT = _env.get('ART_TEST_CHROOT', f"/home/{ART_TEST_SSH_USER}/art-test-chroot")
+else:
+  ART_TEST_CHROOT = _env.get('ART_TEST_CHROOT', "/data/local/art-test-chroot")
diff --git a/test/utils/regen-test-files b/test/utils/regen-test-files
index 36ac9ef9b1..5da8549135 100755
--- a/test/utils/regen-test-files
+++ b/test/utils/regen-test-files
@@ -216,19 +216,15 @@ known_failing_tests = frozenset([
   "2243-single-step-default",
   "2262-miranda-methods",
   "2262-default-conflict-methods",
-  # 2247-checker-write-barrier-elimination: Disabled while we investigate failures
-  "2247-checker-write-barrier-elimination"
 ])
 
-# These tests are new and have not had enough post-submit runs to meet
-# pre-submit SLOs. Monitor their post-submit runs before removing them
-# from this set (in order to promote them to presubmits).
+# These ART run-tests are new and have not had enough post-submit runs
+# to meet pre-submit SLOs. Monitor their post-submit runs before
+# removing them from this set (in order to promote them to
+# presubmits).
 postsubmit_only_tests = frozenset([
-  "851-null-instanceof",
-  "853-checker-inlining",
-  "2266-checker-remove-empty-ifs",
-  "2268-checker-remove-dead-phis",
-  "2269-checker-constant-folding-instrinsics",
+  "2247-checker-write-barrier-elimination",
+  "2273-checker-unreachable-intrinsics",
 ])
 
 known_failing_on_hwasan_tests = frozenset([
@@ -269,11 +265,33 @@ art_gtest_eng_only_module_names = [
 # All supported ART gtests.
 art_gtest_module_names = sorted(art_gtest_user_module_names + art_gtest_eng_only_module_names)
 
+# These ART gtests are new and have not had enough post-submit runs
+# to meet pre-submit SLOs. Monitor their post-submit runs before
+# removing them from this set (in order to promote them to
+# presubmits).
+art_gtest_postsubmit_only_module_names = [
+  "art_standalone_dexopt_chroot_setup_tests",
+]
+
 # ART gtests supported in MTS that do not need root access to the device.
 art_gtest_mts_user_module_names = copy.copy(art_gtest_user_module_names)
 
+# ART gtests supported in presubmits.
+art_gtest_presubmit_module_names = [t for t in art_gtest_module_names
+                                    if t not in art_gtest_postsubmit_only_module_names]
+
 # ART gtests supported in Mainline presubmits.
-art_gtests_mainline_presubmit_module_names = copy.copy(art_gtest_module_names)
+art_gtest_mainline_presubmit_module_names = copy.copy(art_gtest_presubmit_module_names)
+
+# ART gtests supported in postsubmits.
+unknown_art_gtest_postsubmit_only_module_names = [t for t in art_gtest_postsubmit_only_module_names
+                                                  if t not in art_gtest_module_names]
+if unknown_art_gtest_postsubmit_only_module_names:
+  logging.error(textwrap.dedent("""\
+  The following `art_gtest_postsubmit_only_module_names` elements are not part of
+  `art_gtest_module_names`: """) + str(unknown_art_gtest_postsubmit_only_module_names))
+  sys.exit(1)
+art_gtest_postsubmit_module_names = copy.copy(art_gtest_postsubmit_only_module_names)
 
 # Tests exhibiting a flaky behavior, currently exluded from MTS for
 # the stake of stability / confidence (b/209958457).
@@ -309,10 +327,24 @@ flaky_tests_excluded_from_mts = {
     ]
 }
 
+# Tests excluded from all test mapping test groups.
+#
+# Example of admissible values in this dictionary:
+#
+#   "art_standalone_cmdline_tests": ["CmdlineParserTest#TestCompilerOption"],
+#   "art_standalone_dexopt_chroot_setup_tests": ["DexoptChrootSetupTest#HelloWorld"],
+#
+failing_tests_excluded_from_test_mapping = {
+  # Empty.
+}
+
 # Tests failing because of linking issues, currently exluded from MTS
 # and Mainline Presubmits to minimize noise in continuous runs while
 # we investigate.
 #
+# Example of admissible values in this dictionary: same as for
+# `failing_tests_excluded_from_test_mapping` (see above).
+#
 # TODO(b/247108425): Address the linking issues and re-enable these
 # tests.
 failing_tests_excluded_from_mts_and_mainline_presubmits = {
@@ -320,6 +352,11 @@ failing_tests_excluded_from_mts_and_mainline_presubmits = {
     "art_standalone_libartpalette_tests": ["PaletteClientJniTest*"],
 }
 
+failing_tests_excluded_from_mainline_presubmits = (
+  failing_tests_excluded_from_test_mapping |
+  failing_tests_excluded_from_mts_and_mainline_presubmits
+)
+
 # Is `run_test` a Checker test (i.e. a test containing Checker
 # assertions)?
 def is_checker_test(run_test):
@@ -634,25 +671,34 @@ class Generator:
                                         for t in art_run_tests
                                         if t in postsubmit_only_tests]
 
+    def gen_tests_dict(tests, excluded_test_cases = {}, excluded_test_modules = [], suffix = ""):
+      return [
+          ({"name": t + suffix,
+            "options": [
+                {"exclude-filter": e}
+                for e in excluded_test_cases[t]
+            ]}
+           if t in excluded_test_cases
+           else {"name": t + suffix})
+          for t in tests
+          if t not in excluded_test_modules
+      ]
+
     # Mainline presubmits.
     mainline_presubmit_apex_suffix = "[com.google.android.art.apex]"
     mainline_other_presubmit_tests = []
     mainline_presubmit_tests = (mainline_other_presubmit_tests + presubmit_run_test_module_names +
-                                art_gtests_mainline_presubmit_module_names)
-    mainline_presubmit_tests_dict = [
-        ({"name": t + mainline_presubmit_apex_suffix,
-          "options": [
-              {"exclude-filter": e}
-              for e in failing_tests_excluded_from_mts_and_mainline_presubmits[t]
-          ]}
-         if t in failing_tests_excluded_from_mts_and_mainline_presubmits
-         else {"name": t + mainline_presubmit_apex_suffix})
-        for t in mainline_presubmit_tests
-    ]
+                                art_gtest_mainline_presubmit_module_names)
+    mainline_presubmit_tests_dict = \
+      gen_tests_dict(mainline_presubmit_tests,
+                     failing_tests_excluded_from_mainline_presubmits,
+                     [],
+                     mainline_presubmit_apex_suffix)
 
     # Android Virtualization Framework presubmits
     avf_presubmit_tests = ["ComposHostTestCases"]
-    avf_presubmit_tests_dict = [{"name": t} for t in avf_presubmit_tests]
+    avf_presubmit_tests_dict = gen_tests_dict(avf_presubmit_tests,
+                                              failing_tests_excluded_from_test_mapping)
 
     # Presubmits.
     other_presubmit_tests = [
@@ -663,14 +709,18 @@ class Generator:
         "art_standalone_dexpreopt_tests",
     ]
     presubmit_tests = (other_presubmit_tests + presubmit_run_test_module_names +
-                       art_gtest_module_names)
-    presubmit_tests_dict = [{"name": t} for t in presubmit_tests]
-    hwasan_presubmit_tests_dict = [{"name": t} for t in presubmit_tests
-                                   if t not in known_failing_on_hwasan_tests]
+                       art_gtest_presubmit_module_names)
+    presubmit_tests_dict = gen_tests_dict(presubmit_tests,
+                                          failing_tests_excluded_from_test_mapping)
+    hwasan_presubmit_tests_dict = gen_tests_dict(presubmit_tests,
+                                                 failing_tests_excluded_from_test_mapping,
+                                                 known_failing_on_hwasan_tests)
 
     # Postsubmits.
-    postsubmit_tests = postsubmit_run_test_module_names
+    postsubmit_tests = postsubmit_run_test_module_names + art_gtest_postsubmit_module_names
     postsubmit_tests_dict = [{"name": t} for t in postsubmit_tests]
+    postsubmit_tests_dict = gen_tests_dict(postsubmit_tests,
+                                           failing_tests_excluded_from_test_mapping)
 
     # Use an `OrderedDict` container to preserve the order in which items are inserted.
     # Do not produce an entry for a test group if it is empty.
@@ -986,19 +1036,28 @@ class Generator:
     expected_succeeding_tests_percentage = int(
         num_expected_succeeding_tests * 100 / len(run_tests))
 
-    mainline_presubmit_gtests_percentage = int(
-        len(art_gtests_mainline_presubmit_module_names) * 100 / len(art_gtest_module_names))
+    num_gtests = len(art_gtest_module_names)
+
+    num_presubmit_gtests = len(art_gtest_presubmit_module_names)
+    presubmit_gtests_percentage = int(num_presubmit_gtests * 100 / num_gtests)
+
+    num_mainline_presubmit_gtests = len(art_gtest_mainline_presubmit_module_names)
+    mainline_presubmit_gtests_percentage = int(num_mainline_presubmit_gtests * 100 / num_gtests)
+
+    num_postsubmit_gtests = len(art_gtest_postsubmit_module_names)
+    postsubmit_gtests_percentage = int(num_postsubmit_gtests * 100 / num_gtests)
 
     print(f"Generated TEST_MAPPING entries for {num_expected_succeeding_tests} ART run-tests out"
           f" of {len(run_tests)} ({expected_succeeding_tests_percentage}%):")
     for (num_tests, test_kind, tests_percentage, test_group_name) in [
         (num_mainline_presubmit_run_tests, "ART run-tests", mainline_presubmit_run_tests_percentage,
          "mainline-presubmit"),
-        (len(art_gtests_mainline_presubmit_module_names), "ART gtests",
-         mainline_presubmit_gtests_percentage, "mainline-presubmit"),
         (num_presubmit_run_tests, "ART run-tests", presubmit_run_tests_percentage, "presubmit"),
         (num_postsubmit_run_tests, "ART run-tests", postsubmit_run_tests_percentage, "postsubmit"),
-        (len(art_gtest_module_names), "ART gtests", 100, "presubmit"),
+        (num_mainline_presubmit_gtests, "ART gtests", mainline_presubmit_gtests_percentage,
+         "mainline-presubmit"),
+        (num_presubmit_gtests, "ART gtests", presubmit_gtests_percentage, "presubmit"),
+        (num_postsubmit_gtests, "ART gtests", postsubmit_gtests_percentage, "presubmit"),
     ]:
       print(
           f"  {num_tests:3d} {test_kind} ({tests_percentage}%) in `{test_group_name}` test group.")
diff --git a/tools/buildbot-build.sh b/tools/buildbot-build.sh
index ca36813840..2bcb4e6a05 100755
--- a/tools/buildbot-build.sh
+++ b/tools/buildbot-build.sh
@@ -175,6 +175,8 @@ if [[ $build_target == "yes" ]]; then
   make_command+=" deapexer"
   # Needed to generate the primary boot image for testing.
   make_command+=" generate-boot-image"
+  # Data file needed by the `ArtExecTest.SetTaskProfiles` test.
+  make_command+=" task_profiles.json"
   # Build/install the required APEXes.
   make_command+=" ${apexes[*]}"
   make_command+=" ${specific_targets}"
diff --git a/tools/buildbot-utils.sh b/tools/buildbot-utils.sh
index 70fa01a895..1cdd275b98 100755
--- a/tools/buildbot-utils.sh
+++ b/tools/buildbot-utils.sh
@@ -87,7 +87,7 @@ if [[ -n "$ART_TEST_ON_VM" ]]; then
   export RSYNC_RSH="ssh -i ~/.ssh/ubuntu -p $ART_TEST_SSH_PORT -o StrictHostKeyChecking=no" # don't prefix with "ART_", rsync expects this name
 
   if [[ "$TARGET_ARCH" =~ ^(arm64|riscv64)$ ]]; then
-    export ART_TEST_VM_IMG="ubuntu-22.04-server-cloudimg-$TARGET_ARCH.img"
+    export ART_TEST_VM_IMG="ubuntu-23.10-server-cloudimg-$TARGET_ARCH.img"
     export ART_TEST_VM_DIR="$ANDROID_BUILD_TOP/vm/$TARGET_ARCH"
     export ART_TEST_VM="$ART_TEST_VM_DIR/$ART_TEST_VM_IMG"
   else
diff --git a/tools/buildbot-vm.sh b/tools/buildbot-vm.sh
index 6c324c3bba..a15f9e8cbe 100755
--- a/tools/buildbot-vm.sh
+++ b/tools/buildbot-vm.sh
@@ -45,24 +45,24 @@ if [[ $action = create ]]; then
 
     # sudo apt install qemu-system-<arch> qemu-efi cloud-image-utils
 
-    # Get the cloud image for Ubunty 22.04 (Jammy)
-    wget "http://cloud-images.ubuntu.com/releases/22.04/release/$ART_TEST_VM_IMG"
+    # Get the cloud image for Ubunty 23.10 (Mantic Minotaur)
+    wget "http://cloud-images.ubuntu.com/releases/23.10/release/$ART_TEST_VM_IMG"
 
     if [[ "$TARGET_ARCH" = "riscv64" ]]; then
         # Get U-Boot for Ubuntu 22.04 (Jammy)
         get_stable_binary \
-            u/u-boot/u-boot-qemu_2022.01+dfsg-2ubuntu2.3_all.deb \
+            u/u-boot/u-boot-qemu_2023.07+dfsg-1ubuntu2_all.deb \
             usr/lib/u-boot/qemu-riscv64_smode/uboot.elf
 
         # Get OpenSBI for Ubuntu 22.04 (Jammy)
         get_stable_binary \
-            o/opensbi/opensbi_1.3-1ubuntu0.22.04.2_all.deb \
+            o/opensbi/opensbi_1.3-1ubuntu0.23.04.2_all.deb \
             usr/lib/riscv64-linux-gnu/opensbi/generic/fw_jump.elf
 
     elif [[ "$TARGET_ARCH" = "arm64" ]]; then
-        # Get EFI (ARM64) for Ubuntu 22.04 (Jammy)
+        # Get EFI (ARM64)
         get_stable_binary \
-            e/edk2/qemu-efi-aarch64_2022.02-3ubuntu0.22.04.1_all.deb \
+            e/edk2/qemu-efi-aarch64_2023.05-2ubuntu0.1_all.deb \
             usr/share/qemu-efi-aarch64/QEMU_EFI.fd
 
         dd if=/dev/zero of=flash0.img bs=1M count=64
@@ -111,6 +111,7 @@ elif [[ $action = boot ]]; then
             -nographic \
             -bios fw_jump.elf \
             -kernel uboot.elf \
+            -cpu rv64,v=true,vlen=256,vext_spec=v1.0 \
             -drive file="$ART_TEST_VM_IMG",if=virtio \
             -drive file=user-data.img,format=raw,if=virtio \
             -device virtio-net-device,netdev=usernet \
@@ -134,7 +135,7 @@ elif [[ $action = boot ]]; then
         (qemu-system-aarch64 \
             -m 16G \
             -smp 8 \
-            -cpu cortex-a57 \
+            -cpu cortex-a710,sve=on \
             -M virt \
             -nographic \
             -drive if=none,file="$ART_TEST_VM_IMG",id=hd0 \
@@ -145,9 +146,10 @@ elif [[ $action = boot ]]; then
             -device virtio-net-device,netdev=usernet \
             -netdev user,id=usernet,hostfwd=tcp::$ART_TEST_SSH_PORT-:22 > $SCRIPT_DIR/boot.out &)
         echo "Now listening for successful boot"
+        finish_str='.*finished at.*'
         while IFS= read -d $'\0' -n 1 a ; do
             line+="${a}"
-            if [[ "$line" =~ '.*finished.*' ]] ; then
+            if [[ "$line" =~ $finish_str ]] ; then
                 echo $line
                 echo "VM Successfully booted!"
                 exit 0
diff --git a/tools/create_minidebuginfo/create_minidebuginfo.cc b/tools/create_minidebuginfo/create_minidebuginfo.cc
index 506661a377..f13b62579e 100644
--- a/tools/create_minidebuginfo/create_minidebuginfo.cc
+++ b/tools/create_minidebuginfo/create_minidebuginfo.cc
@@ -94,22 +94,20 @@ static void WriteMinidebugInfo(const std::vector<uint8_t>& input, std::vector<ui
   auto* debug_frame = builder->GetDebugFrame();
   debug_frame->Start();
   {
-    std::map<std::basic_string_view<uint8_t>, Elf_Addr> cie_dedup;
+    std::map<std::string_view, Elf_Addr> cie_dedup;
     std::unordered_map<const CIE*, Elf_Addr> new_cie_offset;
     std::deque<std::pair<const FDE*, const CIE*>> entries;
     // Read, de-duplicate and write CIE entries.  Read FDE entries.
     reader.VisitDebugFrame(
         [&](const CIE* cie) {
-          std::basic_string_view<uint8_t> key(cie->data(), cie->size());
+          std::string_view key(reinterpret_cast<const char*>(cie->data()), cie->size());
           auto it = cie_dedup.emplace(key, debug_frame->GetPosition());
           if (/* inserted */ it.second) {
             debug_frame->WriteFully(cie->data(), cie->size());
           }
           new_cie_offset[cie] = it.first->second;
         },
-        [&](const FDE* fde, const CIE* cie) {
-          entries.emplace_back(std::make_pair(fde, cie));
-        });
+        [&](const FDE* fde, const CIE* cie) { entries.emplace_back(std::make_pair(fde, cie)); });
     // Sort FDE entries by opcodes to improve locality for compression (saves ~25%).
     std::stable_sort(entries.begin(), entries.end(), [](const auto& lhs, const auto& rhs) {
       constexpr size_t opcode_offset = sizeof(FDE);
diff --git a/tools/fuzzer/corpus/recursive_encoded_array.dex b/tools/fuzzer/corpus/recursive_encoded_array.dex
new file mode 100644
index 0000000000..c775e744cd
--- /dev/null
+++ b/tools/fuzzer/corpus/recursive_encoded_array.dex
diff --git a/tools/libcore_failures.txt b/tools/libcore_failures.txt
index fc28acd22b..a3c07ccdcf 100644
--- a/tools/libcore_failures.txt
+++ b/tools/libcore_failures.txt
@@ -290,8 +290,7 @@
   bug: 228441328,
   names: ["test.java.lang.Math.CeilAndFloorTests#nearIntegerTests",
           "test.java.time.chrono.TestEraDisplayName",
-          "test.java.time.format.TestDateTimeFormatterBuilderWithLocale",
-          "test.java.util.TestFormatter"]
+          "test.java.time.format.TestDateTimeFormatterBuilderWithLocale"]
 },
 {
   description: "Fails on armv8 device",
diff --git a/tools/libcore_gcstress_debug_failures.txt b/tools/libcore_gcstress_debug_failures.txt
index 168b0ad207..a1c771107c 100644
--- a/tools/libcore_gcstress_debug_failures.txt
+++ b/tools/libcore_gcstress_debug_failures.txt
@@ -47,6 +47,9 @@
           "org.apache.harmony.luni.tests.internal.net.www.protocol.http.HttpURLConnectionTest#testUsingProxy",
           "org.apache.harmony.luni.tests.internal.net.www.protocol.http.HttpURLConnectionTest#testUsingProxySelector",
           "org.apache.harmony.luni.tests.internal.net.www.protocol.https.HttpsURLConnectionTest#testConsequentProxyConnection",
+          "org.apache.harmony.luni.tests.internal.net.www.protocol.https.HttpsURLConnectionTest#testHttpsConnection",
+          "org.apache.harmony.luni.tests.internal.net.www.protocol.https.HttpsURLConnectionTest#testProxyAuthConnection",
+          "org.apache.harmony.luni.tests.internal.net.www.protocol.https.HttpsURLConnectionTest#testProxyAuthConnection_doOutput",
           "org.apache.harmony.tests.java.lang.ref.ReferenceQueueTest#test_removeJ",
           "org.apache.harmony.tests.java.lang.ProcessManagerTest#testSleep",
           "org.apache.harmony.tests.java.util.TimerTest#testOverdueTaskExecutesImmediately",
@@ -63,6 +66,16 @@
   ]
 },
 {
+  description: "Failure with gcstress and debug.",
+  bug: 313922528,
+  result: ERROR,
+  modes: [host],
+  names: ["libcore.java.util.jar.OldManifestTest#test_equals",
+          "test.java.util.Collection",
+          "test.java.util.TestFormatter"
+  ]
+},
+{
   description: "Time-sensitive test fails check of elapsed time with gcstress",
   result: EXEC_FAILED,
   bug: 205007075,
diff --git a/tools/luci/config/generated/cr-buildbucket.cfg b/tools/luci/config/generated/cr-buildbucket.cfg
index 6708ed1374..c6cdc023f8 100644
--- a/tools/luci/config/generated/cr-buildbucket.cfg
+++ b/tools/luci/config/generated/cr-buildbucket.cfg
@@ -2,7 +2,7 @@
 # Do not modify manually.
 #
 # For the schema of this file, see BuildbucketCfg message:
-#   https://luci-config.appspot.com/schemas/projects:buildbucket.cfg
+#   https://config.luci.app/schemas/projects:buildbucket.cfg
 
 buckets {
   name: "ci"
@@ -23,11 +23,13 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:32"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:true"
         properties_j: "device:\"angler-armv7\""
         properties_j: "generational_cc:true"
+        properties_j: "product:\"arm_krait\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -51,11 +53,13 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:32"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:false"
         properties_j: "device:\"angler-armv7\""
         properties_j: "generational_cc:true"
+        properties_j: "product:\"arm_krait\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -80,11 +84,13 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:32"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:false"
         properties_j: "debug:true"
         properties_j: "device:\"angler-armv7\""
         properties_j: "generational_cc:false"
+        properties_j: "product:\"arm_krait\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -108,11 +114,13 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:true"
         properties_j: "device:\"angler-armv8\""
         properties_j: "generational_cc:true"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -136,11 +144,13 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:false"
         properties_j: "device:\"angler-armv8\""
         properties_j: "generational_cc:true"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -165,11 +175,13 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:false"
         properties_j: "debug:true"
         properties_j: "device:\"angler-armv8\""
         properties_j: "generational_cc:false"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -193,12 +205,14 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:32"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:false"
         properties_j: "device:\"bullhead-armv7\""
         properties_j: "gcstress:true"
         properties_j: "generational_cc:true"
+        properties_j: "product:\"arm_krait\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -222,12 +236,14 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:true"
         properties_j: "device:\"bullhead-armv8\""
         properties_j: "gcstress:true"
         properties_j: "generational_cc:true"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -251,12 +267,14 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:false"
         properties_j: "device:\"bullhead-armv8\""
         properties_j: "gcstress:true"
         properties_j: "generational_cc:true"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -585,6 +603,36 @@ buckets {
       }
     }
     builders {
+      name: "qemu-armv8-ndebug"
+      swarming_host: "chromium-swarm.appspot.com"
+      dimensions: "os:Linux"
+      dimensions: "pool:luci.art.ci"
+      recipe {
+        name: "art"
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
+        properties_j: "builder_group:\"client.art\""
+        properties_j: "concurrent_collector:true"
+        properties_j: "debug:false"
+        properties_j: "device:\"qemu-armv8\""
+        properties_j: "generational_cc:true"
+        properties_j: "on_virtual_machine:true"
+      }
+      execution_timeout_secs: 108000
+      expiration_secs: 61200
+      caches {
+        name: "art"
+        path: "art"
+      }
+      build_numbers: YES
+      service_account: "art-ci-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
       name: "qemu-riscv64-ndebug"
       swarming_host: "chromium-swarm.appspot.com"
       dimensions: "os:Linux"
@@ -593,6 +641,7 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:false"
@@ -622,6 +671,7 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "build_only:true"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
@@ -652,12 +702,14 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:32"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:true"
         properties_j: "device:\"walleye-armv7\""
         properties_j: "generational_cc:true"
         properties_j: "heap_poisoning:true"
+        properties_j: "product:\"arm_krait\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -681,12 +733,14 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:true"
         properties_j: "device:\"walleye-armv8\""
         properties_j: "generational_cc:true"
         properties_j: "heap_poisoning:true"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
@@ -710,12 +764,14 @@ buckets {
         name: "art"
         cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
         cipd_version: "refs/heads/main"
+        properties_j: "bitness:64"
         properties_j: "builder_group:\"client.art\""
         properties_j: "concurrent_collector:true"
         properties_j: "debug:false"
         properties_j: "device:\"walleye-armv8\""
         properties_j: "generational_cc:true"
         properties_j: "heap_poisoning:true"
+        properties_j: "product:\"armv8\""
       }
       execution_timeout_secs: 108000
       expiration_secs: 61200
diff --git a/tools/luci/config/generated/luci-logdog.cfg b/tools/luci/config/generated/luci-logdog.cfg
index adc75bef49..01a391261d 100644
--- a/tools/luci/config/generated/luci-logdog.cfg
+++ b/tools/luci/config/generated/luci-logdog.cfg
@@ -2,7 +2,7 @@
 # Do not modify manually.
 #
 # For the schema of this file, see ProjectConfig message:
-#   https://luci-config.appspot.com/schemas/projects:luci-logdog.cfg
+#   https://config.luci.app/schemas/projects:luci-logdog.cfg
 
 reader_auth_groups: "all"
 writer_auth_groups: "luci-logdog-chromium-writers"
diff --git a/tools/luci/config/generated/luci-milo.cfg b/tools/luci/config/generated/luci-milo.cfg
index 8863af65ce..979a5d1fb9 100644
--- a/tools/luci/config/generated/luci-milo.cfg
+++ b/tools/luci/config/generated/luci-milo.cfg
@@ -2,7 +2,7 @@
 # Do not modify manually.
 #
 # For the schema of this file, see Project message:
-#   https://luci-config.appspot.com/schemas/projects:luci-milo.cfg
+#   https://config.luci.app/schemas/projects:luci-milo.cfg
 
 consoles {
   id: "luci"
diff --git a/tools/luci/config/generated/luci-notify.cfg b/tools/luci/config/generated/luci-notify.cfg
index 4456fa3778..a281a827ec 100644
--- a/tools/luci/config/generated/luci-notify.cfg
+++ b/tools/luci/config/generated/luci-notify.cfg
@@ -2,7 +2,7 @@
 # Do not modify manually.
 #
 # For the schema of this file, see ProjectConfig message:
-#   https://luci-config.appspot.com/schemas/projects:luci-notify.cfg
+#   https://config.luci.app/schemas/projects:luci-notify.cfg
 
 notifiers {
   notifications {
@@ -274,6 +274,19 @@ notifiers {
   }
   builders {
     bucket: "ci"
+    name: "qemu-armv8-ndebug"
+  }
+}
+notifiers {
+  notifications {
+    on_new_status: FAILURE
+    on_new_status: INFRA_FAILURE
+    email {
+      recipients: "art-team+chromium-buildbot@google.com"
+    }
+  }
+  builders {
+    bucket: "ci"
     name: "qemu-riscv64-ndebug"
   }
 }
diff --git a/tools/luci/config/generated/luci-scheduler.cfg b/tools/luci/config/generated/luci-scheduler.cfg
index 291ed09677..479447321d 100644
--- a/tools/luci/config/generated/luci-scheduler.cfg
+++ b/tools/luci/config/generated/luci-scheduler.cfg
@@ -2,7 +2,7 @@
 # Do not modify manually.
 #
 # For the schema of this file, see ProjectConfig message:
-#   https://luci-config.appspot.com/schemas/projects:luci-scheduler.cfg
+#   https://config.luci.app/schemas/projects:luci-scheduler.cfg
 
 job {
   id: "angler-armv7-debug"
@@ -205,6 +205,16 @@ job {
   }
 }
 job {
+  id: "qemu-armv8-ndebug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "qemu-armv8-ndebug"
+  }
+}
+job {
   id: "qemu-riscv64-ndebug"
   realm: "ci"
   acl_sets: "ci"
@@ -278,6 +288,7 @@ trigger {
   triggers: "host-x86_64-ndebug"
   triggers: "host-x86_64-non-gen-cc"
   triggers: "host-x86_64-poison-debug"
+  triggers: "qemu-armv8-ndebug"
   triggers: "qemu-riscv64-ndebug"
   triggers: "qemu-riscv64-ndebug-build_only"
   triggers: "walleye-armv7-poison-debug"
@@ -312,6 +323,7 @@ trigger {
   triggers: "host-x86_64-ndebug"
   triggers: "host-x86_64-non-gen-cc"
   triggers: "host-x86_64-poison-debug"
+  triggers: "qemu-armv8-ndebug"
   triggers: "qemu-riscv64-ndebug"
   triggers: "qemu-riscv64-ndebug-build_only"
   triggers: "walleye-armv7-poison-debug"
@@ -346,6 +358,7 @@ trigger {
   triggers: "host-x86_64-ndebug"
   triggers: "host-x86_64-non-gen-cc"
   triggers: "host-x86_64-poison-debug"
+  triggers: "qemu-armv8-ndebug"
   triggers: "qemu-riscv64-ndebug"
   triggers: "qemu-riscv64-ndebug-build_only"
   triggers: "walleye-armv7-poison-debug"
@@ -380,6 +393,7 @@ trigger {
   triggers: "host-x86_64-ndebug"
   triggers: "host-x86_64-non-gen-cc"
   triggers: "host-x86_64-poison-debug"
+  triggers: "qemu-armv8-ndebug"
   triggers: "qemu-riscv64-ndebug"
   triggers: "qemu-riscv64-ndebug-build_only"
   triggers: "walleye-armv7-poison-debug"
diff --git a/tools/luci/config/generated/project.cfg b/tools/luci/config/generated/project.cfg
index 584e84f46a..d517637b4b 100644
--- a/tools/luci/config/generated/project.cfg
+++ b/tools/luci/config/generated/project.cfg
@@ -2,12 +2,12 @@
 # Do not modify manually.
 #
 # For the schema of this file, see ProjectCfg message:
-#   https://luci-config.appspot.com/schemas/projects:project.cfg
+#   https://config.luci.app/schemas/projects:project.cfg
 
 name: "art"
 access: "group:all"
 lucicfg {
-  version: "1.38.1"
+  version: "1.43.4"
   package_dir: ".."
   config_dir: "generated"
   entry_point: "main.star"
diff --git a/tools/luci/config/generated/realms.cfg b/tools/luci/config/generated/realms.cfg
index 777f8685e7..c45317659a 100644
--- a/tools/luci/config/generated/realms.cfg
+++ b/tools/luci/config/generated/realms.cfg
@@ -2,7 +2,7 @@
 # Do not modify manually.
 #
 # For the schema of this file, see RealmsCfg message:
-#   https://luci-config.appspot.com/schemas/projects:realms.cfg
+#   https://config.luci.app/schemas/projects:realms.cfg
 
 realms {
   name: "@root"
diff --git a/tools/luci/config/main.star b/tools/luci/config/main.star
index af746d6735..0320ce68cf 100755
--- a/tools/luci/config/main.star
+++ b/tools/luci/config/main.star
@@ -217,8 +217,10 @@ def target_builders():
         short_name="dbg",
         dimensions=target_dims,
         properties={
+            "bitness": 32,
             "device": "angler-armv7",
             "debug": True,
+            "product": "arm_krait",
         }
     )
     ci_builder(
@@ -227,10 +229,12 @@ def target_builders():
         short_name="ngen",
         dimensions=userfault_gc_target_dims,
         properties={
+            "bitness": 32,
             "device": "angler-armv7",
             "debug": True,
             "concurrent_collector": False,
             "generational_cc": False,
+            "product": "arm_krait",
         }
     )
     ci_builder(
@@ -239,8 +243,10 @@ def target_builders():
         short_name="ndbg",
         dimensions=target_dims,
         properties={
+            "bitness": 32,
             "device": "angler-armv7",
             "debug": False,
+            "product": "arm_krait",
         }
     )
     ci_builder(
@@ -249,8 +255,10 @@ def target_builders():
         short_name="dbg",
         dimensions=target_dims,
         properties={
+            "bitness": 64,
             "device": "angler-armv8",
             "debug": True,
+            "product": "armv8",
         }
     )
     ci_builder(
@@ -259,10 +267,12 @@ def target_builders():
         short_name="ngen",
         dimensions=userfault_gc_target_dims,
         properties={
+            "bitness": 64,
             "device": "angler-armv8",
             "debug": True,
             "concurrent_collector": False,
             "generational_cc": False,
+            "product": "armv8",
         }
     )
     ci_builder(
@@ -271,8 +281,10 @@ def target_builders():
         short_name="ndbg",
         dimensions=target_dims,
         properties={
+            "bitness": 64,
             "device": "angler-armv8",
             "debug": False,
+            "product": "armv8",
         }
     )
     ci_builder(
@@ -281,9 +293,11 @@ def target_builders():
         short_name="dbg",
         dimensions=target_dims,
         properties={
+            "bitness": 32,
             "device": "bullhead-armv7",
             "debug": False,
             "gcstress": True,
+            "product": "arm_krait",
         }
     )
     ci_builder(
@@ -292,9 +306,11 @@ def target_builders():
         short_name="dbg",
         dimensions=target_dims,
         properties={
+            "bitness": 64,
             "device": "bullhead-armv8",
             "debug": True,
             "gcstress": True,
+            "product": "armv8",
         }
     )
     ci_builder(
@@ -303,9 +319,11 @@ def target_builders():
         short_name="ndbg",
         dimensions=target_dims,
         properties={
+            "bitness": 64,
             "device": "bullhead-armv8",
             "debug": False,
             "gcstress": True,
+            "product": "armv8",
         }
     )
     ci_builder(
@@ -314,9 +332,11 @@ def target_builders():
         short_name="dbg",
         dimensions=target_dims,
         properties={
+            "bitness": 32,
             "device": "walleye-armv7",
             "debug": True,
             "heap_poisoning": True,
+            "product": "arm_krait",
         }
     )
     ci_builder(
@@ -325,9 +345,11 @@ def target_builders():
         short_name="dbg",
         dimensions=target_dims,
         properties={
+            "bitness": 64,
             "device": "walleye-armv8",
             "debug": True,
             "heap_poisoning": True,
+            "product": "armv8",
         }
     )
     ci_builder(
@@ -336,9 +358,11 @@ def target_builders():
         short_name="ndbg",
         dimensions=target_dims,
         properties={
+            "bitness": 64,
             "device": "walleye-armv8",
             "debug": False,
             "heap_poisoning": True,
+            "product": "armv8",
         }
     )
 
@@ -465,12 +489,26 @@ def host_builders():
         }
     )
     ci_builder(
+        name="qemu-armv8-ndebug",
+        category="qemu|armv8",
+        short_name="ndbg",
+        dimensions=host_dims,
+        is_fyi=True,
+        properties={
+            "bitness": 64,
+            "debug": False,
+            "device": "qemu-armv8",
+            "on_virtual_machine": True,
+        }
+    )
+    ci_builder(
         name="qemu-riscv64-ndebug",
         category="qemu|riscv64",
         short_name="ndbg",
         dimensions=host_dims,
         is_fyi=True,
         properties={
+            "bitness": 64,
             "debug": False,
             "device": "qemu-riscv64",
             "on_virtual_machine": True,
@@ -482,6 +520,7 @@ def host_builders():
         short_name="bo",
         dimensions=host_dims,
         properties={
+            "bitness": 64,
             "build_only": True,
             "debug": False,
             "device": "qemu-riscv64",