645889 - JM: Add RegExp.exec -> RegExp.test optimization

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Brian Hackett [Laid off!]]

The optimization which the tracer does to transform re.exec into re.test when possible is currently responsible for most or all of the difference between JM+TI and JM+TM on awfy.  This optimization should also be added to JM (type inference not needed, it's very simple to swap natives when generating native call stubs) to avoid the clutter.

Comment 1

[Brian Hackett [Laid off!]]

Patch on the JM tree.  Wrong author field on the commit, sorry!

http://hg.mozilla.org/projects/jaegermonkey/rev/a3632d306b13

Comment 2

[David Mandelin [:dmandelin]]

Would this land cleanly on the TM tree?

Comment 3

[Brian Hackett [Laid off!]]

Almost.  I'll put together a patch for review.

Comment 4

[Brian Hackett [Laid off!]]

Attachment: patch against TM

Comment 5

[David Mandelin [:dmandelin]]

Comment on attachment 522564 [details] [diff] [review]
patch against TM

Nice factoring. Only nit is that the name "IsCallResultUnusedOrTested" is kind of weird, but the comment at the top makes it pretty obvious anyway, and it's only used in these places. "IsCallResultNotStored"? "IsCallResultNullTestedOnly"? "IsCallResultObjectNotUsed"?

Comment 6

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

CallResultEscapes, and invert the test?

Comment 7

[Brian Hackett [Laid off!]]

I like CallResultEscapes.

http://hg.mozilla.org/tracemonkey/rev/ba8cc57e0848

Comment 8

[Chris Leary [:cdleary] (not checking bugmail)]

cdleary-bot mozilla-central merge info:
http://hg.mozilla.org/mozilla-central/rev/ba8cc57e0848