Optimize FastDtoa, in particular Grisu3. In addition to making
a microbenchmark, there are a number of smaller and larger
changes here:
- Replace divisions by power-of-ten with multiplications by
their inverses, using an algorithm very similar to the one
in libdivide.
- For DiyFp::Times(), use 128-bit hardware multiplication
if available (which it generally is on 64-bit platforms).
- Where possible, send around a pointer to the end of the string,
instead of a pointer and a length, reducing register pressure
(especially for Intel). Where not (easily) possible, add
a local variable to make the compiler understand that length
and decimal_point cannot alias.
- Change some ints to unsigneds where it helps us avoid sign
extensions.
- Some minor changes to reduce instruction dependency chains.
- Inline BiggestPowerTen().
Actual performance gain is wildly different between platforms.
On my 3990X workstation (Zen 2), gains are about 21%. On a M1
Mac Mini, they are about 17%. But on my i7-10610U laptop
(Comet Lake, so Skylake microarchitecture), the function is
78% faster. This is probably because large divisions
(divisor over 255) seem to hurt a lot on Skylake, but I haven't
gone through it in detail.
Change-Id: I5b67c257d788a3f7d1be7065d055456852451d68
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/4110741
Commit-Queue: Steinar H Gunderson <sesse@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#84906}
The infrastructure runs everything already in Python3, so this is
mostly a clean-up.
For MB, a python2 holdover was removed and new lint errors were
fixed.
The renames were automated with:
git grep -e "/usr/bin/python$" |
cut -d':' -f1 |
xargs
sed -i 's/#!\/usr\/bin\/python$/#!\/usr\/bin\/python3/1'
and
git grep -e "/usr/bin/env python$" |
cut -d':' -f1 |
xargs
sed -i 's/#!\/usr\/bin\/env python$/#!\/usr\/bin\/env python3/1'
Bug: v8:13148
Change-Id: If4f3c7635e72fa134798d55314ac1aa92ddd01bf
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3811499
Reviewed-by: Liviu Rau <liviurau@google.com>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/main@{#82231}
- Removed duplication and unnecessary indirection from all suites testcfgs.
- Introduced a more comprehensive context to cover both command context and other platform specific concerns.
- Propagated above context to TestLoader to allow for test counting command execution on all platforms.
- Wrapped original pool with another class to give it a new interface and allow injecting different implementations in the future.
- Consolidated progress indicators under a single processor in the pipeline.
- Consolidated result retention requirements calculation outside of pipeline chain.
- Refactored LoaderProc and got it under tests.
- Added some more tests for the standard runner.
- Extracted BuildConfig class.
Bug: v8:12785
Change-Id: I87be040e91f792a983662bb5a10d55b36a14ea7f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3701595
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Liviu Rau <liviurau@chromium.org>
Cr-Commit-Position: refs/heads/main@{#81770}
To be consistent with the all the other tiers and avoid confusion, we
rename --opt to ---turbofan, and --always-opt to --always-turbofan.
Change-Id: Ie23dc8282b3fb4cf2fbf73b6c3d5264de5d09718
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3610431
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Jakob Linke <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/main@{#80336}
Otherwise, DCHECKs complain that the PageAllocator used for the
GlobalGCInfoTable may differ when setting up the platform repeatedly
through benchmarking SetUp().
Change-Id: I7e87e8c9d8c283105e1bd75a4cd176df7f304315
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3283075
Reviewed-by: Omer Katz <omerkatz@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77920}
This is a reland of 8f17052ce0
Changes can be seen in PS1..4 (re-marked kraken/audio-beat-detection)
Original change's description:
> [test] Unmark TSAN tests as SLOW/SKIP
>
> We allowed more time for tsan testing [1] so unmarking some tests.
>
> [1]: https://chromium-review.googlesource.com/c/v8/v8/+/2979671
>
> Bug: v8:7790, v8:11600, v8:11906
> Change-Id: I026a32abe0380c2d32327aaa8b301ef51f22a5e2
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2979602
> Reviewed-by: Mythri Alle <mythria@chromium.org>
> Commit-Queue: Santiago Aboy Solanes <solanes@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#75357}
Bug: v8:7790, v8:11600, v8:11906
Change-Id: Ie5b8886820d8e7d18c8ceee612a5c4d9ef72bc49
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2983536
Reviewed-by: Mythri Alle <mythria@chromium.org>
Commit-Queue: Santiago Aboy Solanes <solanes@chromium.org>
Cr-Commit-Position: refs/heads/master@{#75408}
After the last CL on TSAN support for generated loads, we are seeing
timeouts in one of our TSAN bots.
Bug: v8:7790, v8:11600
Change-Id: I90924540c5ddcf9902f936849df28aff0f7bd3d5
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2972724
Reviewed-by: Mythri Alle <mythria@chromium.org>
Commit-Queue: Santiago Aboy Solanes <solanes@chromium.org>
Cr-Commit-Position: refs/heads/master@{#75241}
This finishes the TSAN support for loads as we do not use movb or movw
to load from memory
Bug: v8:7790, v8:11600
Change-Id: I3c319da95c24cfa03f4de2367e007fd4cf7dd355
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2953321
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Santiago Aboy Solanes <solanes@chromium.org>
Cr-Commit-Position: refs/heads/master@{#75204}
Expose third_party/google_benchmark:google_benchmark as a public dep
from third_party/google_benchmark:benchmark_main and a v8_libbase
dependency to empty_benchmark to expose base/macros.h.
Bug: v8:7330
Change-Id: I5205af365da2ac7c35647acff90ed65d831cfa07
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2851888
Commit-Queue: Dan Elphick <delphick@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Auto-Submit: Dan Elphick <delphick@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#74217}
Keep idiomatic state loop but rely on USE() to avoid
warning about unused variables.
Bug: v8:11687
Change-Id: Icde295723c5d389d827280f70a65776866d4e7a9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2850645
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#74200}
This missing include was undetected because trace_perf.cc is only
built if the checkout_google_benchmark custom gclient variable is
defined.
Bug: none
Change-Id: If2016edad4df382f14903593ea18066f7759c4d5
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2698387
Reviewed-by: Marja Hölttä <marja@chromium.org>
Commit-Queue: Chris Mumford <cmumford@google.com>
Cr-Commit-Position: refs/heads/master@{#72825}
The csuite.py script does not work correctly on Windows. It runs
correctly in baseline mode, but there are two problems when running in
compare mode:
1. In compare mode the output of benchmark.py is piped to the
compare-baseline.py script, but Windows only execute python files if
python.exe is the default program to open '.py' files, and this is
not the case, by default, when python is installed as part of the
depot_tools.
Fix: explicitly add the 'python' command before compare-baseline.py.
2. By default CSuite prints the results to stdout using escapes codes
that add color highlights. But this does not work on Windows when
compare-baseline.py is launched with a pipe:
python test/benchmarks/csuite/benchmark.py <...> |
python test/benchmarks/csuite/compare-baseline.py <baseline_results>
Fix: Do not use a pipe. Write the benchmark numbers for the
compare-run into a separate file, and pass the path to this file to
compare-baseline.py
Change-Id: Ic22d5bd4b47901f0ba0f35bc2496441346d21c6a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2656855
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Paolo Severini <paolosev@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#72807}
These tests are pretty slow with both TSAN and
--stress-incremental-marking enabled.
* Skip octane/raytrace for the variant stress when TSAN is enabled.
* Test wasm/shared-memory-worker-stress is slow with TSAN.
Change-Id: Ia4b8ce2986d04e6ef2fc7d28991cd5ce7faf31e5
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2571129
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#71598}
The CPPGC_BUILD_IN_V8 define (used for tracing) isn't propagated from
v8_base_without_compiler to cppgc_base, which breaks build with
perfetto. Instead use a gn args to specify standalone builds (defaulted
to false) and use that to choose the right tracing implementation.
Bug: chromium:1056170
Change-Id: I70bce819d45fb133b6f932a50a5d027e39f3e5b9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2555007
Auto-Submit: Omer Katz <omerkatz@chromium.org>
Reviewed-by: Hannes Payer <hpayer@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Commit-Queue: Omer Katz <omerkatz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#71356}
This skips the top outliers in debug mode and marks other tests as
slow that frequently flakily time out, see bug.
No=Try: true
Bug: v8:10909
Change-Id: I26b22cceba4a93496f340fe25af0685391fa762b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2407897
Commit-Queue: Maya Lekova <mslekova@chromium.org>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Auto-Submit: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#69865}
csuite.py compare command currently throws exception
"_csv.Error: iterator should return strings, not bytes
(did you open the file in text mode?)"
This should fix it.
Change-Id: I69c0ec43575a8c3627dac81dc99e47ba6adf6f61
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2316833
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Z Nguyen-Huu <duongn@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#69050}
This CL removes the GetTraceDescriptor virtual call from garbage
collected mixins and replaces it with querying the object start
bitmap.
The CL also removes the mixin macros which are now no longer needed.
Bug: chromium:1056170
Change-Id: I27ed299f93025d09a3bb3f0d17b14bed3c200565
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2287508
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Anton Bikineev <bikineev@chromium.org>
Commit-Queue: Omer Katz <omerkatz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#68787}
This adds Google benchmark for microbenchmarking C++ code as an
optional dependency.
To enable, add the following to the .gclient before syncing
"custom_vars": {
"checkout_google_benchmark": True
}
Change-Id: Id0eab772dd71558906658ef4bb60e31acd665948
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2275964
Reviewed-by: Hannes Payer <hpayer@chromium.org>
Reviewed-by: Anton Bikineev <bikineev@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#68663}
The serializer currently cannot handle a heap state containing
arbitrary compiled Code objects. As a quick fix for the
--stress-snapshot d8 flag, we clear compiled data from the isolate
prior to the serialize-deserialize-verify pass.
With this change, mjsunit tests pass on x64.
The %SerializeDeserializeNow() runtime function would require more
work, since it is not possible to mutate the heap to this extent while
still preserving a runnable host context and isolate. We will need
another solution there.
Drive-by: Skip the stress_snapshot variant except for the mjsunit
suite.
Tbr: machenbach@chromium.org
Bug: v8:10493,v8:10416
Change-Id: Ie110da8b51613fcd69c7f391d3cf8589d6b04dd8
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2182429
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67585}
Before: running different suites is removing non-related master data
After: different suites have separate master data files
Change-Id: Iab46d663fdf2a1d709aeb7da77a443c2ed1bdcc2
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2071870
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Marcin Wiącek <marcin@mwiacek.com>
Cr-Commit-Position: refs/heads/master@{#66453}
Embedded builtins are now unconditionally enabled, which removes the
need to differentiate between enabled/disabled embedded builtins.
This Cl removes the 'embedded_builtins' variant and related
*.status entries.
R=machenbach@chromium.org
Bug: v8:8519
Change-Id: I55d0dd54735b7cc437832af6fa2836fd6c14a317
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1864936
Commit-Queue: Simon Zünd <szuend@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64464}
This is a reland of 9efe315ee2 after marking
box2d slow.
Original change's description:
> Mark functions for optimization only on bytecode budget interrupts
>
> We used to mark functions for optimization on any interrupt. This sometimes
> causes functions to OSR when not needed. The implementation was such because
> we didn't have a different runtime function to distinguish bytecode budget
> interrupts from other interrupts. For lazy feedback allocation we added a
> new runtime function for bytecode budget interrupts so it makes it easier
> to actually mark functions only when needed.
>
> This also includes a fix to reduce the stack limits for interrupts when
> entering a scope that allows interrupts from a postponed interrupt scope.
>
> Bug: chromium:993061
> Change-Id: Iaf7b4dccb7a503e5b6bfcbb993bc7482aa593955
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1829218
> Reviewed-by: Leszek Swirski <leszeks@chromium.org>
> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
> Commit-Queue: Mythri Alle <mythria@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#64048}
Bug: chromium:993061
Change-Id: I24dae03357d6c368e4173db3f071e8ab09e9d6dc
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1832173
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Mythri Alle <mythria@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64082}
