[webkit-reviews] review granted: [Bug 190906] Try ripping out inferred types because it might be a performance improvement : [Attachment 359187] patch

[bugzilla-daemon at webkit.org]

Yusuke Suzuki <yusukesuzuki at slowstart.org> has granted Saam Barati
<sbarati at apple.com>'s request for review:
Bug 190906: Try ripping out inferred types because it might be a performance
improvement
https://bugs.webkit.org/show_bug.cgi?id=190906

Attachment 359187: patch

https://bugs.webkit.org/attachment.cgi?id=359187&action=review




--- Comment #11 from Yusuke Suzuki <yusukesuzuki at slowstart.org> ---
Comment on attachment 359187
  --> https://bugs.webkit.org/attachment.cgi?id=359187
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=359187&action=review

r=me with additional cleanup suggestions.

> Source/JavaScriptCore/bytecode/PutByIdFlags.h:41
>  inline PutByIdFlags encodeStructureID(StructureID id)

This function is no longer used.

> Source/JavaScriptCore/bytecode/PutByIdFlags.h:54
>  #if USE(JSVALUE64)

Ditto.

> Source/JavaScriptCore/dfg/DFGAbstractInterpreterInlines.h:-2921
> -		       result.merge(
> -			   m_graph.inferredValueForProperty(
> -			       value, uid, status[i].offset(),
m_state.structureClobberState()));

This would perform tryGetConstantProperty, don't we need to do this?

> Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:-1354
> -    loadi OpPutById::Metadata::flags[t5], t1

Now, I think that this metadata.flag is no longer necessary. All the
information can be fit into byte code.flag.