Bug 202324 – OSR exit shouldn't bother updating get_by_id array profiles that have changed modes

• Component: New Bugs
• Status: RESOLVED FIXED

[Keith Miller]

OSR exit shouldn't bother updating get_by_id array profiles that have changed modes

[Keith Miller]

Created attachment 379744 [details]
Patch

[Keith Miller]

<rdar://problem/52669110>

[Yusuke Suzuki]

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>()`.

[Keith Miller]

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.

[Mark Lam]

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?

[Keith Miller]

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.

[Keith Miller]

Created attachment 379749 [details]
Patch for landing

[WebKit Commit Bot]

Comment on attachment 379749 [details]
Patch for landing

Clearing flags on attachment: 379749

Committed r250440: <https://trac.webkit.org/changeset/250440>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Saam Barati]

You missed the FTL

[Keith Miller]

(In reply to Saam Barati from comment #10)
> You missed the FTL

Whoops, will fix.

[Keith Miller]

See: https://bugs.webkit.org/show_bug.cgi?id=202493 for FTL version.