[webkit-reviews] review requested: [Bug 174818] WebAssembly: generate smaller binaries : [Attachment 316386] patch

[bugzilla-daemon at webkit.org]

JF Bastien <jfbastien at apple.com> has asked  for review:
Bug 174818: WebAssembly: generate smaller binaries
https://bugs.webkit.org/show_bug.cgi?id=174818

Attachment 316386: patch

https://bugs.webkit.org/attachment.cgi?id=316386&action=review




--- Comment #8 from JF Bastien <jfbastien at apple.com> ---
Created attachment 316386

  --> https://bugs.webkit.org/attachment.cgi?id=316386&action=review

patch

> I think it's still worth landing the change even without 1.  I think we
> should figure out how to make the tiering checks less expensive somehow.

Sure, here's an updated patch. I agree we want to make it cheaper. An
interpreter would be a good fix, or an intermediate helper on all non-tiered
function entry for tier-up check (which would handle non-loop recording and
tier-up). I agree with either of these the no-tierup approach isn't as useful.

> I don't think it's OK to say that some functions never tier.	That's just
> too dangerous and breaks too many of our assumptions.

Eh, I disagree but it's not worth arguing. So much of the code is small and
trivial, straight-line. If BBQ messes it up worst than a few extra registers
and spills then we should fix that bug. I'd have liked to have a small gain,
but we can get back to it in a few months.