[webkit-reviews] review requested: [Bug 136712] Support the type profiler in the DFG : [Attachment 239016] patch

[bugzilla-daemon at webkit.org]

Saam Barati <saambarati1 at gmail.com> has asked  for review:
Bug 136712: Support the type profiler in the DFG
https://bugs.webkit.org/show_bug.cgi?id=136712

Attachment 239016: patch
https://bugs.webkit.org/attachment.cgi?id=239016&action=review

------- Additional Comments from Saam Barati <saambarati1 at gmail.com>
- I left the additional class for slow path call w/ no arguments and no result
type. It doesn't work
  by just passing in NoResult to the existing class because then the operation
it calls is expected to 
  have a return type. I guess this can also be done by having a bogus return
type in the operation, 
  but that seems a bit hacky to me. What do you think?

- I made the necessary changes to reflect that TypeInteger is contained in
TypeNumber.
   This includes fixing the inlined type check when emitting code, and also the
convert to check
   for NumberUse in the Fixup phase. This NumberUse Check is now generated more
eagerly because
   I removed a few checks that were in place when TypeMachineInt wasn't
contained in TypeNumber.

- I also decided to use "isInt32Speculation(prediction)" instead of
"!(prediction & ~SpecInt32)" because it seems more consistent with the other
checks of this nature in the Fixup phase.