Bug 187448 – [JSC] Optimize padding of UnlinkedCodeBlock to shrink

• Status: REOPENED
• Component: New Bugs

[Yusuke Suzuki]

[JSC] Optimize padding of UnlinkedCodeBlock to shrink

[Yusuke Suzuki]

Created attachment 344548 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 344548 [details]
Patch

Clearing flags on attachment: 344548

Committed r233629: <https://trac.webkit.org/changeset/233629>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/41950243>

[Keith Miller]

Comment on attachment 344548 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=344548&action=review

> Source/JavaScriptCore/bytecode/UnlinkedCodeBlock.h:465
> +    unsigned m_codeType : 2; // CodeType
> +    unsigned m_didOptimize : 2; // TriState

Nit: I think these can still be the enum types btw.

[Yusuke Suzuki]

(In reply to Keith Miller from comment #5)
> Comment on attachment 344548 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=344548&action=review
> 
> > Source/JavaScriptCore/bytecode/UnlinkedCodeBlock.h:465
> > +    unsigned m_codeType : 2; // CodeType
> > +    unsigned m_didOptimize : 2; // TriState
> 
> Nit: I think these can still be the enum types btw.

IIRC, using different types for bit fields adds unnecessary paddings in MSVC.
So, conservatively, I choose `unsigned` here.

[Keith Miller]

I'm reverting this with https://bugs.webkit.org/show_bug.cgi?id=187746 since it seems to have broken internal wasm benchmarks. I think this patch just happened to exemplify an existing issue though. I'll investigate...