[webkit-reviews] review denied: [Bug 170215] WebAssembly: Air::Inst::generate crashes on large binary on A64 : [Attachment 308255] quick hack

[bugzilla-daemon at webkit.org]

JF Bastien <jfbastien at apple.com> has denied  review:
Bug 170215: WebAssembly: Air::Inst::generate crashes on large binary on A64
https://bugs.webkit.org/show_bug.cgi?id=170215

Attachment 308255: quick hack

https://bugs.webkit.org/attachment.cgi?id=308255&action=review




--- Comment #7 from JF Bastien <jfbastien at apple.com> ---
Created attachment 308255

  --> https://bugs.webkit.org/attachment.cgi?id=308255&action=review

quick hack

Here's a hack which fixes part of the issue by reserving x15 on ARM64. I'm not
super confident about:
 - The reservation occurs in reservedHardwareRegisters. Is that weird?
 - Not giving it another name when I use it for this purpose.
 - How far away stackAddrImpl is from AirLowerStackArgs.cpp, yet they're now
intertwined (if the former fails, the later does similar work on top).
 - I don't think the #else clause needs to be there anymore? x86 simply can't
fail, right?

Note that x15 is only explicitly used in the ExecutableAllocator's
jitWriteThunkGenerator, and this is a fine place to use it AFAICT.

There's another bug at O1, very similar but this time in add64. It's trying to
materialize the frame size, it's too big, and the same assert about scratch
register fires. I assume it's in the same post-RA code area. I'll fix in this
bug as well.