Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(36)

Issue 2831213006: Enable ThinLTO and LLD for POSIX LTO by default on Linux. (Closed)

Created:
3 years, 8 months ago by krasin1
Modified:
3 years, 8 months ago
Reviewers:
Dirk Pranke, Nico
CC:
chromium-reviews, kcc2, pcc1
Target Ref:
refs/heads/master
Project:
chromium
Visibility:
Public.

Description

Enable ThinLTO and LLD for POSIX LTO by default on Linux. Eventually, we will enable LLD on Linux even for regular builds, but it's natural to make incremental steps here. ThinLTO brings multi-threaded linking for LinkTimeOptimization builds, which allows to speed up codegen considerably (up to 4x). ThinLTO also manages a cache inside out/<gn-config>/thinlto-cache directory, that improves incremental linking as well. There's a cache pruning policy that will prevent the cache from growing indefinitely. The policy is not yet finalized, and complaints / suggestions are welcome. BUG=660216, 607968 Review-Url: https://codereview.chromium.org/2831213006 Cr-Commit-Position: refs/heads/master@{#466732} Committed: https://chromium.googlesource.com/chromium/src/+/7997bbe88aef7eeca37111dcb8426f0430347fb5

Patch Set 1 #

Patch Set 2 : fmt #

Patch Set 3 : Update a comment #

Patch Set 4 : Update another comment #

Unified diffs Side-by-side diffs Delta from patch set Stats (+12 lines, -12 lines) Patch
M build/toolchain/toolchain.gni View 1 2 3 2 chunks +12 lines, -12 lines 0 comments Download

Messages

Total messages: 34 (18 generated)
krasin1
3 years, 8 months ago (2017-04-21 18:00:32 UTC) #7
Nico
lgtm, but please say "64-bit" in CL description. What's the story for 32-bit?
3 years, 8 months ago (2017-04-21 18:07:04 UTC) #13
Nico
(also, please watch bot cycle time on bots that do full builds, since parallel links ...
3 years, 8 months ago (2017-04-21 18:07:36 UTC) #14
krasin1
On 2017/04/21 18:07:04, Nico wrote: > lgtm, but please say "64-bit" in CL description. What's ...
3 years, 8 months ago (2017-04-21 18:08:07 UTC) #15
Nico
Ah right, we don't ship 32-bit linux, you're right. (We do have a 32-bit chromeos ...
3 years, 8 months ago (2017-04-21 18:10:06 UTC) #18
krasin1
On 2017/04/21 18:07:36, Nico wrote: > (also, please watch bot cycle time on bots that ...
3 years, 8 months ago (2017-04-21 18:10:14 UTC) #19
Nico
..so what's the plan for arm? :-) My point is "why do we need the ...
3 years, 8 months ago (2017-04-21 18:10:39 UTC) #20
krasin1
The plan is to win on Linux x86-64, and then expand to more platforms and ...
3 years, 8 months ago (2017-04-21 18:12:09 UTC) #21
Dirk Pranke
lgtm
3 years, 8 months ago (2017-04-21 21:18:34 UTC) #22
commit-bot: I haz the power
CQ is trying da patch. Follow status at: https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.org/2831213006/60001
3 years, 8 months ago (2017-04-24 16:15:24 UTC) #24
Nico
On 2017/04/24 16:15:24, commit-bot: I haz the power wrote: > CQ is trying da patch. ...
3 years, 8 months ago (2017-04-24 16:39:14 UTC) #26
krasin1
On 2017/04/24 16:39:14, Nico wrote: > On 2017/04/24 16:15:24, commit-bot: I haz the power wrote: ...
3 years, 8 months ago (2017-04-24 16:51:47 UTC) #27
krasin1
given the new information (that it has been broken for a month now), I think ...
3 years, 8 months ago (2017-04-24 20:10:35 UTC) #28
commit-bot: I haz the power
CQ is trying da patch. Follow status at: https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.org/2831213006/60001
3 years, 8 months ago (2017-04-24 20:11:48 UTC) #30
commit-bot: I haz the power
Committed patchset #4 (id:60001) as https://chromium.googlesource.com/chromium/src/+/7997bbe88aef7eeca37111dcb8426f0430347fb5
3 years, 8 months ago (2017-04-24 20:19:32 UTC) #33
krasin1
3 years, 8 months ago (2017-04-25 00:41:17 UTC) #34
Message was sent while issue was closed.
A revert of this CL (patchset #4 id:60001) has been created in
https://codereview.chromium.org/2842683002/ by krasin@chromium.org.

The reason for reverting is: Code size increase:
https://chromeperf.appspot.com/report?sid=7826328365affe1842778994c934cbe08c7...

Postmortem is coming. TL;DR: worse dead code elimination in ThinLTO.

Powered by Google App Engine
This is Rietveld 408576698