Bug 116700 - Inline TCMalloc_Central_FreeList's RemoveRange and FetchFromSpansSafe
Summary: Inline TCMalloc_Central_FreeList's RemoveRange and FetchFromSpansSafe
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Benjamin Poulain
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-05-23 20:56 PDT by Benjamin Poulain
Modified: 2013-05-24 18:12 PDT (History)
5 users (show)

See Also:


Attachments
Patch (1.71 KB, patch)
2013-05-23 20:58 PDT, Benjamin Poulain
no flags Details | Formatted Diff | Diff
Archive of layout-test-results from webkit-ews-16 for mac-mountainlion-wk2 (1.34 MB, application/zip)
2013-05-24 15:52 PDT, Build Bot
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Benjamin Poulain 2013-05-23 20:56:03 PDT
Inline TCMalloc_Central_FreeList's RemoveRange and FetchFromSpansSafe
Comment 1 Benjamin Poulain 2013-05-23 20:58:59 PDT
Created attachment 202761 [details]
Patch
Comment 2 Oliver Hunt 2013-05-23 22:32:02 PDT
Comment on attachment 202761 [details]
Patch

do you have evidence that it's a win? ALWAYS_INLINE is a stick we use to beat the compiler on the head when it's making the wrong choice
Comment 3 WebKit Commit Bot 2013-05-23 22:49:19 PDT
Attachment 202761 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/FastMalloc.cpp']" exit_code: 1
Source/WTF/wtf/FastMalloc.cpp:3054:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
Source/WTF/wtf/FastMalloc.cpp:3092:  Place brace on its own line for function definitions.  [whitespace/braces] [4]
Total errors found: 2 in 2 files


If any of these errors are false positives, please file a bug against check-webkit-style.
Comment 4 Benjamin Poulain 2013-05-24 00:07:12 PDT
(In reply to comment #2)
> (From update of attachment 202761 [details])
> do you have evidence that it's a win? ALWAYS_INLINE is a stick we use to beat the compiler on the head when it's making the wrong choice

Yeah, I was surprised clang did not inline those already. Usually clang is really good at figuring out this kind of stuff.

On the (very long) benchmark I ran, fastMalloc takes ~2%, and that was removing 1% of those 2%, so it helped very very little.

You know this code quite well. Do you know of any opportunity to make fastMalloc faster on ARM?
Comment 5 Build Bot 2013-05-24 15:52:00 PDT
Comment on attachment 202761 [details]
Patch

Attachment 202761 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/720033

New failing tests:
editing/undo/undo-after-setting-value.html
Comment 6 Build Bot 2013-05-24 15:52:02 PDT
Created attachment 202859 [details]
Archive of layout-test-results from webkit-ews-16 for mac-mountainlion-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: webkit-ews-16  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.3
Comment 7 Benjamin Poulain 2013-05-24 18:11:58 PDT
Comment on attachment 202761 [details]
Patch

Clearing flags on attachment: 202761

Committed r150676: <http://trac.webkit.org/changeset/150676>
Comment 8 Benjamin Poulain 2013-05-24 18:12:01 PDT
All reviewed patches have been landed.  Closing bug.