[Webkit-unassigned] [Bug 160588] Refactor MathIC compilation process in Baseline and DFG to turn temporary registers usage more flexible

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=160588

Saam Barati <sbarati at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #288391|review?                     |review-
              Flags|                            |

--- Comment #4 from Saam Barati <sbarati at apple.com> ---
Comment on attachment 288391
  --> https://bugs.webkit.org/attachment.cgi?id=288391
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=288391&action=review

> Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp:3457
> +    for (int i = 0; i < MATH_IC_MAX_GPR; i++)

should be uint32_t or size_t. (For all loops below, too.)

> Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp:3482
> +    if (scratchGPRRegCount == 1)

Hmm, this doesn't like it'd scale well.

> Source/JavaScriptCore/ftl/FTLLowerDFGToB3.cpp:1583
> +        patchpoint->numGPScratchRegisters = numGPScratchRegisters;
> +        patchpoint->numFPScratchRegisters = 2 + numFPScratchRegisters;

The idea of doing this is so that we don't require the 2 extra FP registers when not needed. So the caller should be responsible of providing these numbers.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20160920/dcaf210e/attachment.html>