Bug 202324 - OSR exit shouldn't bother updating get_by_id array profiles that have changed modes
Summary: OSR exit shouldn't bother updating get_by_id array profiles that have changed...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Keith Miller
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2019-09-27 11:10 PDT by Keith Miller
Modified: 2019-10-02 14:10 PDT (History)
8 users (show)

See Also:

Attachments
Patch (3.94 KB, patch)
2019-09-27 11:11 PDT, Keith Miller 		no flags Details | Formatted Diff | Diff
Patch for landing (4.08 KB, patch)
2019-09-27 11:44 PDT, Keith Miller 		no flags Details | Formatted Diff | Diff
Show Obsolete (1) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Keith Miller 2019-09-27 11:10:32 PDT 
OSR exit shouldn't bother updating get_by_id array profiles that have changed modes
Comment 1 Keith Miller 2019-09-27 11:11:50 PDT 
Created attachment 379744 [details]
Patch
Comment 2 Keith Miller 2019-09-27 11:11:52 PDT 
<rdar://problem/52669110>
Comment 3 Yusuke Suzuki 2019-09-27 11:33:29 PDT 
Comment on attachment 379744 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=379744&action=review

r=me

> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:523
> +            bool doProfile = instruction->opcodeID() != op_get_by_id || instruction->as<OpGetById>().metadata(profiledCodeBlock).m_modeMetadata.mode == GetByIdMode::ArrayLength;

You can use `!instruction->is<OpGetById>()`.

> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:1176
> +                if (instruction->opcodeID() == op_get_by_id) {

Ditto. You can use `instruction->is<OpGetById>()`.
Comment 4 Keith Miller 2019-09-27 11:34:16 PDT 
Comment on attachment 379744 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=379744&action=review

>> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:523
>> +            bool doProfile = instruction->opcodeID() != op_get_by_id || instruction->as<OpGetById>().metadata(profiledCodeBlock).m_modeMetadata.mode == GetByIdMode::ArrayLength;
> 
> You can use `!instruction->is<OpGetById>()`.

Nice! Fixed, thanks!

>> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:1176
>> +                if (instruction->opcodeID() == op_get_by_id) {
> 
> Ditto. You can use `instruction->is<OpGetById>()`.

ditto.
Comment 5 Mark Lam 2019-09-27 11:34:48 PDT 
Comment on attachment 379744 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=379744&action=review

> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:1174
>              if (ArrayProfile* arrayProfile = jit.baselineCodeBlockFor(codeOrigin)->getArrayProfile(codeOrigin.bytecodeIndex())) {
> +                const Instruction* instruction = jit.baselineCodeBlockFor(codeOrigin)->instructions().at(codeOrigin.bytecodeIndex()).ptr();

Can you just compute jit.baselineCodeBlockFor(codeOrigin) once and pre-cache it?
Comment 6 Keith Miller 2019-09-27 11:42:09 PDT 
Comment on attachment 379744 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=379744&action=review

>> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:1174
>> +                const Instruction* instruction = jit.baselineCodeBlockFor(codeOrigin)->instructions().at(codeOrigin.bytecodeIndex()).ptr();
> 
> Can you just compute jit.baselineCodeBlockFor(codeOrigin) once and pre-cache it?

Sure, done.
Comment 7 Keith Miller 2019-09-27 11:44:43 PDT 
Created attachment 379749 [details]
Patch for landing
Comment 8 WebKit Commit Bot 2019-09-27 12:30:14 PDT 
Comment on attachment 379749 [details]
Patch for landing

Clearing flags on attachment: 379749

Committed r250440: <https://trac.webkit.org/changeset/250440>
Comment 9 WebKit Commit Bot 2019-09-27 12:30:16 PDT 
All reviewed patches have been landed.  Closing bug.
Comment 10 Saam Barati 2019-10-01 17:33:30 PDT 
You missed the FTL
Comment 11 Keith Miller 2019-10-02 14:02:35 PDT 
(In reply to Saam Barati from comment #10)
> You missed the FTL

Whoops, will fix.
Comment 12 Keith Miller 2019-10-02 14:10:23 PDT 
See: https://bugs.webkit.org/show_bug.cgi?id=202493 for FTL version.