Age | Commit message (Collapse) | Author |
|
Change-Id: Ie5ef0a122dee6dabe62077f340f9a0844bb4219a
|
|
Remove dependencies on iteration number for micro/ShifterOperand.java
methods.
Test: ./test/test.py
Change-Id: Ic34435bd04e9b547fb3a3c144411d0e57717a877
|
|
Change-Id: I97c6ecd27c74325cc495c33638fb926bded876b1
|
|
Change-Id: I11932a8362d8e8700ff9a2eb03ee47ac22bccaf2
|
|
When we have 2 sets of statistics that are not equal
Change-Id: I3811b216f6dc8503df0ba22bb8cdf2bacfbe9009
|
|
|
|
|
|
|
|
Change-Id: Iebb3359fa7efe9bbd60fac5b4471e1a73d520c00
|
|
The internal number of iterations wasn't defined by the original
benchmark and was set to some arbitrary number. So they were
adjusted to decrease running time.
Change-Id: I7ea9b6b24fd600c1bf9462160a1a8fc0a146d3d4
|
|
Get rid of unnecessary string allocations in this benchmark. This
decreases execution time by 30s.
Change-Id: If0b2ca9ff81bdde6f63477db64c0052196f0bb10
|
|
Change-Id: I46b441c6f65e84240cb037a1adf150baf8a8b7f9
|
|
|
|
|
|
Change-Id: I0a03784e055f01132382ab5da8ccea76988c8a3d
|
|
Make the benchmarks single iteration time not dependent on the
total iteration number.
Change-Id: I47561be6c3d240317755cec264af63dde4c33e4c
|
|
This change is to fix the Exception we get when we attempt to sort a
list using both floats and strings.
Change-Id: I57fad4bccfa86f3321773c20772df33663b9279e
|
|
After removing GC effects the running time has decreased for
revcomp, that's why the number of iterations for
dont-auto-calibrate mode has been changed.
Change-Id: I82693023d95b78d95b262495c3223c55e858e49f
|
|
Change-Id: I1e95b1c27d6f1b6b1c49c1f3aafc8b88018e0b39
|
|
Manually configure the number of iterations to have each benchmark
running for 1.0-1.5 sec on Cortex-A53 in dont-auto-calibrate mode.
This is required by perf scripts.
Change-Id: I0ce5d224af0a6876fb4f5a5668cffa197b54a9ae
|
|
Change encoding from UTF-8 with CRLF to ascii.
Change-Id: I68af43a0546ebec2a255e43712055029f6bd47ad
|
|
Change-Id: I3abe07acb4d6aa463a89469fc74931d1140158d1
|
|
|
|
|
|
|
|
|
|
Change-Id: Id6c2215fcd604194c580c06f5846b15f15aefe52
|
|
A new folder benchmarks/jit_aot/ is added to compare and benchmark
ART JIT and AOT performance.
Currently two micro benchmarks are added:
- benchmarks/jit_aot/Invoke.java
- benchmarks/jit_aot/LoadCheck.java
benchmarks/jit_aot/Invoke.java:
This benchmark compares the performance of ART JIT and AOT,
on their abilities to optimize and inline invoke-virtual and invoke-static.
benchmarks/jit_aot/LoadCheck.java:
This benchmark compares the performance of ART JIT and AOT,
on their abilities to optimize invokes to other user defined class (LoadClassCheck).
Change-Id: I3ae388abe37f376f959983c4a4c6344a769482f7
|
|
|
|
Change-Id: Ibf5b1e2de30b84c7ff92b79a1bb414c42ce76065
|
|
This change fixes the script execution when the directory separators
on the host and on the device differ.
Change-Id: Ib02fec966e989a83598a5c762e7983a5b7cd6534
|
|
Change-Id: I43b48fdcd805821d06737dfdf83207614d9a21f4
|
|
|
|
|
|
This is a hack. Ideally we want to fix the root cause, by providing more
samples or better handling cases when we have few samples.
Change-Id: I3cfcc200aae036b35b71589969cd9e597dcff341
|
|
|
|
|
|
Change-Id: I1a91e8867407eeae5ff623a0c0ff83527c39169b
|
|
This also fixes a bug for `--order-by-diff` for nested results.
Change-Id: I42462ad78297b05d16d05bda5040c82a7fe8b938
|
|
When running test cases with ART JIT mode with default auto calibration,
the performance actually replies on whether performance critical functions
are JIT compiled during the auto calibration phase (warm up phase).
Currently, the default thresholds that impacts benchmark JIT mode:
- In art-testing framework, default minimum calibration time is 50ms.
- In ART JIT, default threshold to trigger jit is 10000 (hotness count).
For some of the test cases, for example benchmarks/stanford/Oscar,
above thresholds are not enough to trigger JIT compilation for some critical functions,
which results in much slower scores than AOT mode.
This patch proposes to change jit threshold when running jit mode, to solve this problem.
Change-Id: I8178969b6fc48524fa47682ffdf3e9921d637b12
|
|
The option now works with results obtained by running the top-level `run.py`.
Change-Id: I5eae94f91ebc674a6a92d2cf0d3fbb25b4b3053a
|
|
Change-Id: I39b3744a7f29c5db522f703d43d4614feb984cba
|
|
Change-Id: I5b562fcdd843d264dc42098a516b98efa33a5c12
|
|
|
|
Change-Id: Ibea9621dfeafa87cff9fa40d9215aaa18615607f
|
|
Change-Id: I0f42b3242d163eb2b57a8dd7b4b085b8d45f779f
|
|
It is now a 'cumulative' option. Allowing multiple levels of information
details.
This removes the 'mean' information from the default output, simplifying it and
making it more readable.
Change-Id: I5a6bad275c55ed450ab4dc3166f70da8c2bc545e
|
|
Change-Id: Ie11fef7f2caf0f0499d72785e9e58b838ad3f3a5
|
|
Change-Id: I9defa1fb28e51bae99e9c37120bf7d1eb66540f6
|
|
Added requirements.txt for easier installation inside virtualenv using
pip packaging.
This change is needed for a smoother integration with art-reports.
Change-Id: I580f6f80942355787231f5b2e6a7e223f543c2f6
Signed-off-by: Milosz Wasilewski <milosz.wasilewski@linaro.org>
|