Chrome should be able to compile V8 with O3 in GN |
|||||||||
Issue descriptionChrome should be able to compile V8 with O3. Compile whole Chrome with O3 is not mandatory. Compiling V8 with O2 will result in a performance regression. This is the status quo when a GN build is done.
,
Jun 8 2016
,
Jun 8 2016
,
Jun 8 2016
Work in progress: https://codereview.chromium.org/2048163002/
,
Jun 8 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/96a6dfa2c30ab9b22abd20c87ed0e0d6ae41c40e commit 96a6dfa2c30ab9b22abd20c87ed0e0d6ae41c40e Author: dpranke <dpranke@chromium.org> Date: Wed Jun 08 22:28:05 2016 Change //build/config/compiler:optimize_max to use -O3. Certain components (e.g., v8) really want to be compiled with -O3, but the current ":optimize_max" setting just used -O2. Since "max" should theoretically mean "max", let's try making it be -O3 across the board and see what happens. R=brettw@chromium.org BUG= 616031 Review-Url: https://codereview.chromium.org/2048163002 Cr-Commit-Position: refs/heads/master@{#398704} [modify] https://crrev.com/96a6dfa2c30ab9b22abd20c87ed0e0d6ae41c40e/build/config/compiler/BUILD.gn
,
Jun 8 2016
Hopefully this is fixed now, and I sent a note to the perf sheriffs asking for any updates on perf regressions (or improvements). If there's more validation you want to do on the v8 side (or if you need to roll //build into v8), I think you should be able to do that now, so I'll punt this back to you?
,
Jun 9 2016
Thanks Dirk. Moving to machenbach@ who is owning this.
,
Jun 10 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/c4dc398543c8a7cbd4b5f8bfe3495d6182dc3f19 commit c4dc398543c8a7cbd4b5f8bfe3495d6182dc3f19 Author: dpranke <dpranke@chromium.org> Date: Fri Jun 10 21:41:19 2016 Revert of Change //build/config/compiler:optimize_max to use -O3. (patchset #1 id:1 of https://codereview.chromium.org/2048163002/ ) Reason for revert: Reverting, this caused a size regression for cronet that is more important to them than any likely perf gains. I'll re-land a different CL that's less broad. Original issue's description: > Change //build/config/compiler:optimize_max to use -O3. > > Certain components (e.g., v8) really want to be compiled with -O3, > but the current ":optimize_max" setting just used -O2. Since "max" > should theoretically mean "max", let's try making it be -O3 across > the board and see what happens. > > R=brettw@chromium.org > BUG= 616031 > > Committed: https://crrev.com/96a6dfa2c30ab9b22abd20c87ed0e0d6ae41c40e > Cr-Commit-Position: refs/heads/master@{#398704} TBR=brettw@chromium.org,machenbach@chromium.org # Not skipping CQ checks because original CL landed more than 1 days ago. BUG= 616031 Review-Url: https://codereview.chromium.org/2062433002 Cr-Commit-Position: refs/heads/master@{#399276} [modify] https://crrev.com/c4dc398543c8a7cbd4b5f8bfe3495d6182dc3f19/build/config/compiler/BUILD.gn
,
Jun 15 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/c4dc398543c8a7cbd4b5f8bfe3495d6182dc3f19 commit c4dc398543c8a7cbd4b5f8bfe3495d6182dc3f19 Author: dpranke <dpranke@chromium.org> Date: Fri Jun 10 21:41:19 2016 Revert of Change //build/config/compiler:optimize_max to use -O3. (patchset #1 id:1 of https://codereview.chromium.org/2048163002/ ) Reason for revert: Reverting, this caused a size regression for cronet that is more important to them than any likely perf gains. I'll re-land a different CL that's less broad. Original issue's description: > Change //build/config/compiler:optimize_max to use -O3. > > Certain components (e.g., v8) really want to be compiled with -O3, > but the current ":optimize_max" setting just used -O2. Since "max" > should theoretically mean "max", let's try making it be -O3 across > the board and see what happens. > > R=brettw@chromium.org > BUG= 616031 > > Committed: https://crrev.com/96a6dfa2c30ab9b22abd20c87ed0e0d6ae41c40e > Cr-Commit-Position: refs/heads/master@{#398704} TBR=brettw@chromium.org,machenbach@chromium.org # Not skipping CQ checks because original CL landed more than 1 days ago. BUG= 616031 Review-Url: https://codereview.chromium.org/2062433002 Cr-Commit-Position: refs/heads/master@{#399276} [modify] https://crrev.com/c4dc398543c8a7cbd4b5f8bfe3495d6182dc3f19/build/config/compiler/BUILD.gn
,
Jun 19 2016
,
Jun 20 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/c266bec7641b1926f65d1f48614aafb2008a7c57 commit c266bec7641b1926f65d1f48614aafb2008a7c57 Author: dpranke <dpranke@chromium.org> Date: Mon Jun 20 23:22:34 2016 Add a dedicated "optimize_speed" config to GN. The various GN configs to select different optimization levels are confusing, and sorting them out will take a decent amount of perf testing, but in order to achieve parity w/ GYP, for now we need a dedicated config to make sure some components (e.g., v8) are compiled w/ -O3 where appropriate. R=brettw@chromium.org, machenbach@chromium.org BUG= 616031 , 618678, 621335 Review-Url: https://codereview.chromium.org/2078223002 Cr-Commit-Position: refs/heads/master@{#400828} [modify] https://crrev.com/c266bec7641b1926f65d1f48614aafb2008a7c57/build/config/compiler/BUILD.gn [modify] https://crrev.com/c266bec7641b1926f65d1f48614aafb2008a7c57/third_party/opus/BUILD.gn
,
Jun 22 2016
just waiting for the fix to roll into Chromium now ...
,
Jun 22 2016
Rolled.
,
Aug 22 2016
|
|||||||||
►
Sign in to add a comment |
|||||||||
Comment 1 by machenb...@chromium.org
, May 31 2016