[webkit-reviews] review granted: [Bug 174750] B3 should do LICM : [Attachment 316241] the patch

[bugzilla-daemon at webkit.org]

Keith Miller <keith_miller at apple.com> has granted Filip Pizlo
<fpizlo at apple.com>'s request for review:
Bug 174750: B3 should do LICM
https://bugs.webkit.org/show_bug.cgi?id=174750

Attachment 316241: the patch

https://bugs.webkit.org/attachment.cgi?id=316241&action=review




--- Comment #12 from Keith Miller <keith_miller at apple.com> ---
Comment on attachment 316241
  --> https://bugs.webkit.org/attachment.cgi?id=316241
the patch

View in context: https://bugs.webkit.org/attachment.cgi?id=316241&action=review

r=me with some nits.

> Source/JavaScriptCore/ChangeLog:8
> +	   Added a LICM phase to B3. This phase is called
hoistLoopInvariantValues, to conform to the B3 naming

I dunno HLIV doesn't roll off the tongue the same way LICM does...

> Source/JavaScriptCore/b3/B3BasicBlock.cpp:60
> -    m_values[m_values.size() - 1] = value;
> +    m_values[m_values.size() - 2] = value;

legit...

> Source/JavaScriptCore/b3/B3CFG.h:62
> +    

revert please.

> Source/JavaScriptCore/b3/B3EnsureLoopPreHeaders.cpp:71
> +	   loop.header()->addPredecessor(preHeader);

Nit: It would probably be a little clearer if you moved this line to where you
set the successor of the preHeader.