Bug 184316 – Add pointer profiling support to the DFG and supporting files.

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Mark Lam]

Patch coming.

[Radar WebKit Bug Importer]

<rdar://problem/39188524>

[Mark Lam]

Created attachment 337245 [details]
proposed patch.

Let's get some baking time on the EWS.

[EWS Watchlist]

Attachment 337245 [details] did not pass style-queue:


ERROR: Source/JavaScriptCore/dfg/DFGSlowPathGenerator.h:170:  Wrong number of spaces before statement. (expected: 12)  [whitespace/indent] [4]
Total errors found: 1 in 30 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Filip Pizlo]

Comment on attachment 337245 [details]
proposed patch.

View in context: https://bugs.webkit.org/attachment.cgi?id=337245&action=review

> Source/JavaScriptCore/b3/air/AirCCallSpecial.cpp:137
> +        jit.call(scratchRegister, B3CCallPtrTag);

This might be OK, but can you file a bug for passing the ptrtag to the CCallValue, so that the client of B3 can select what tag is used for each call they emit?

> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:755
> +#if USE(POINTER_PROFILING)
> +            void* oldEntrySP = cpu.fp<uint8_t*>() + sizeof(CallerFrameAndPC);
> +            void* newEntrySP = cpu.fp<uint8_t*>() + inlineCallFrame->returnPCOffset() + sizeof(void*);
> +            returnPC = retagCodePtr(returnPC, bitwise_cast<PtrTag>(oldEntrySP), bitwise_cast<PtrTag>(newEntrySP));
> +#endif

Wait... won't this fail miserably if the OSR exit is reused at a different stack height?

[Filip Pizlo]

Comment on attachment 337245 [details]
proposed patch.

View in context: https://bugs.webkit.org/attachment.cgi?id=337245&action=review

OK, now that Mark explained this to me, I get it.  r=me

>> Source/JavaScriptCore/dfg/DFGOSRExit.cpp:755
>> +#endif
> 
> Wait... won't this fail miserably if the OSR exit is reused at a different stack height?

Never mind, this is the probe code.  It's right.

[Mark Lam]

Created attachment 337254 [details]
patch for landing.

[EWS Watchlist]

Attachment 337254 [details] did not pass style-queue:


ERROR: Source/JavaScriptCore/dfg/DFGSlowPathGenerator.h:170:  Wrong number of spaces before statement. (expected: 12)  [whitespace/indent] [4]
Total errors found: 1 in 30 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Mark Lam]

Thanks for the review.  Landed in r230294: <http://trac.webkit.org/r230294>.