[webkit-reviews] review canceled: [Bug 136936] CSS JIT: Don't add backtracking register count to minimum register count if it's not necessary : [Attachment 238633] Patch

[bugzilla-daemon at webkit.org]

Benjamin Poulain <benjamin at webkit.org> has canceled Yusuke Suzuki
<utatane.tea at gmail.com>'s request for review:
Bug 136936: CSS JIT: Don't add backtracking register count to minimum register
count if it's not necessary
https://bugs.webkit.org/show_bug.cgi?id=136936

Attachment 238633: Patch
https://bugs.webkit.org/attachment.cgi?id=238633&action=review

------- Additional Comments from Benjamin Poulain <benjamin at webkit.org>
Clearing the review flag.

The patch is likely correct, but I don't think the gain/risk is worth it at
this very moment.

My rationale:

All the code has been verified for 6 registers for the rightmost fragment and 5
for backtracking. Verifying all the generators for 5 registers would be very
time consuming, it would probably take me an entire day of review.

We are on the verge of adding complex new capabilities to the engine. There are
at least two features that will need great attention: 1) matching multiple
pseudo elements with :matches() 2) Multilevel backtracking for :nth-child(),
:matches() and :not().

The minus side of not including this patch is some performance impact on ARMv7.
But even today without it we are several times faster than any competing
engine.