Bug 38154 – REGRESSION: after 56028 FastMalloc calls madvise too often

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Stephanie Lewis]

<rdar://problem/7834433> REGRESSSION: 1.5% PLT regression due to 56028 (return memory quicker).  
        To save on madvise calls when recommitting memory recommit the entire chosen span and then carve the span
        instead of carving the span up and only committing the part that will be used immediately.

[Stephanie Lewis]

Created attachment 54354 [details]
Change the recommit policy to commit entire spans at one time.

Change the recommit policy to recommit entire spans at one time.

[WebKit Review Bot]

Attachment 54354 [details] did not pass style-queue:

Failed to run "['WebKitTools/Scripts/check-webkit-style', '--no-squash']" exit_code: 1
JavaScriptCore/wtf/FastMalloc.cpp:1688:  Weird number of spaces at line-start.  Are you using a 4-space indent?  [whitespace/indent] [3]
JavaScriptCore/wtf/FastMalloc.cpp:1694:  Weird number of spaces at line-start.  Are you using a 4-space indent?  [whitespace/indent] [3]
Total errors found: 2 in 2 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Geoffrey Garen]

Comment on attachment 54354 [details]
Change the recommit policy to commit entire spans at one time.

r=me

Any data on how much of an improvement this is?

[Stephanie Lewis]

Committed http://trac.webkit.org/changeset/58346.

This is about a .5% improvement.  The FastMalloc file has historically not conformed to style guidelines, and my changes matched the file's style.