[webkit-reviews] review granted: [Bug 112839] fourthTier: DFG should be able to run on a separate thread : [Attachment 202380] the patch

[bugzilla-daemon at webkit.org]

Geoffrey Garen <ggaren at apple.com> has granted Filip Pizlo <fpizlo at apple.com>'s
request for review:
Bug 112839: fourthTier: DFG should be able to run on a separate thread
https://bugs.webkit.org/show_bug.cgi?id=112839

Attachment 202380: the patch
https://bugs.webkit.org/attachment.cgi?id=202380&action=review

------- Additional Comments from Geoffrey Garen <ggaren at apple.com>
View in context: https://bugs.webkit.org/attachment.cgi?id=202380&action=review


r=me

> Source/JavaScriptCore/jit/JITStubs.cpp:959
> +	   worklistState =
> +	      
stackFrame.vm->worklist->completeAllReadyPlansForVM(*stackFrame.vm, codeBlock);


This seems OK for now, but it might become a problem if LLVM compilation sticks
at ~4ms / function. If you had 4 functions queued up, one call to cti_optimize
would cause you to drop an animation frame. If you had 16 functions queued up,
you'd drop 4 frames. Basically, there's a tradeoff here between startup time
and responsiveness, and the nob is turned all the way toward startup time right
now, which is probably the wrong balance.

> Source/JavaScriptCore/runtime/JSSegmentedVariableObject.h:97
> +    typedef ConcurrentJITLock Lock;
> +    typedef ConcurrentJITLocker Locker;

Actually, I think this reads even better without the typedef, since it calls
out the other client you're locking against.