diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2023-10-25 16:52:00 +0100 |
|---|---|---|
| committer | Tobias Hieta <tobias@hieta.se> | 2023-11-13 11:05:25 +0100 |
| commit | 69b3baf9b87eabfa307d8f74f7d0eea9df28cd5a (patch) | |
| tree | 6cdc5aca9e3f96f0e972cf6b7782a80f10b17a08 | |
| parent | e7dc53b94212b2a65f9fb995fd3e0f9a2d192fdc (diff) | |
| download | llvm-69b3baf9b87eabfa307d8f74f7d0eea9df28cd5a.tar.gz | |
[DAG] WidenVectorOperand - add basic handling for *_EXTEND_VECTOR_INREG nodes
Fixes Issue #70208
(cherry picked from commit c9c9bf0f20fd820b86fac35113bbd8049ff4e72a)
| -rw-r--r-- | llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h | 1 | ||||
| -rw-r--r-- | llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp | 10 | ||||
| -rw-r--r-- | llvm/test/CodeGen/PowerPC/pr70208.ll | 25 |
3 files changed, 36 insertions, 0 deletions
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h index db8f61eee606..ad70655de349 100644 --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h @@ -1000,6 +1000,7 @@ private: SDValue WidenVecOp_EXTRACT_VECTOR_ELT(SDNode *N); SDValue WidenVecOp_INSERT_SUBVECTOR(SDNode *N); SDValue WidenVecOp_EXTRACT_SUBVECTOR(SDNode *N); + SDValue WidenVecOp_EXTEND_VECTOR_INREG(SDNode *N); SDValue WidenVecOp_STORE(SDNode* N); SDValue WidenVecOp_VP_STORE(SDNode *N, unsigned OpNo); SDValue WidenVecOp_VP_STRIDED_STORE(SDNode *N, unsigned OpNo); diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp index 8c117c1c74dc..9c1839f2576e 100644 --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp @@ -5946,6 +5946,11 @@ bool DAGTypeLegalizer::WidenVectorOperand(SDNode *N, unsigned OpNo) { case ISD::EXPERIMENTAL_VP_STRIDED_STORE: Res = WidenVecOp_VP_STRIDED_STORE(N, OpNo); break; + case ISD::ANY_EXTEND_VECTOR_INREG: + case ISD::SIGN_EXTEND_VECTOR_INREG: + case ISD::ZERO_EXTEND_VECTOR_INREG: + Res = WidenVecOp_EXTEND_VECTOR_INREG(N); + break; case ISD::MSTORE: Res = WidenVecOp_MSTORE(N, OpNo); break; case ISD::MGATHER: Res = WidenVecOp_MGATHER(N, OpNo); break; case ISD::MSCATTER: Res = WidenVecOp_MSCATTER(N, OpNo); break; @@ -6338,6 +6343,11 @@ SDValue DAGTypeLegalizer::WidenVecOp_EXTRACT_VECTOR_ELT(SDNode *N) { N->getValueType(0), InOp, N->getOperand(1)); } +SDValue DAGTypeLegalizer::WidenVecOp_EXTEND_VECTOR_INREG(SDNode *N) { + SDValue InOp = GetWidenedVector(N->getOperand(0)); + return DAG.getNode(N->getOpcode(), SDLoc(N), N->getValueType(0), InOp); +} + SDValue DAGTypeLegalizer::WidenVecOp_STORE(SDNode *N) { // We have to widen the value, but we want only to store the original // vector type. diff --git a/llvm/test/CodeGen/PowerPC/pr70208.ll b/llvm/test/CodeGen/PowerPC/pr70208.ll new file mode 100644 index 000000000000..3a8d0eb00061 --- /dev/null +++ b/llvm/test/CodeGen/PowerPC/pr70208.ll @@ -0,0 +1,25 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3 +; RUN: llc -verify-machineinstrs -mtriple=ppc64le-unknown-linux-gnu < %s | FileCheck %s + +define <4 x i64> @widget(<8 x i16> %call) { +; CHECK-LABEL: widget: +; CHECK: # %bb.0: +; CHECK-NEXT: addis 3, 2, .LCPI0_0@toc@ha +; CHECK-NEXT: addis 4, 2, .LCPI0_1@toc@ha +; CHECK-NEXT: xxlxor 37, 37, 37 +; CHECK-NEXT: addi 3, 3, .LCPI0_0@toc@l +; CHECK-NEXT: lxvd2x 0, 0, 3 +; CHECK-NEXT: addi 3, 4, .LCPI0_1@toc@l +; CHECK-NEXT: lxvd2x 1, 0, 3 +; CHECK-NEXT: xxswapd 35, 0 +; CHECK-NEXT: xxswapd 32, 1 +; CHECK-NEXT: vperm 4, 5, 2, 3 +; CHECK-NEXT: vperm 3, 5, 2, 0 +; CHECK-NEXT: vmr 2, 4 +; CHECK-NEXT: blr + %bitcast = bitcast <8 x i16> %call to <16 x i8> + %shufflevector = shufflevector <16 x i8> <i8 poison, i8 poison, i8 0, i8 0, i8 poison, i8 poison, i8 0, i8 0, i8 poison, i8 poison, i8 0, i8 0, i8 poison, i8 poison, i8 0, i8 0>, <16 x i8> %bitcast, <16 x i32> <i32 16, i32 17, i32 2, i32 3, i32 18, i32 19, i32 6, i32 7, i32 20, i32 21, i32 10, i32 11, i32 22, i32 23, i32 14, i32 15> + %bitcast1 = bitcast <16 x i8> %shufflevector to <4 x i32> + %zext = zext <4 x i32> %bitcast1 to <4 x i64> + ret <4 x i64> %zext +} |