596457 - JM: Turn off methodjit if SSE2 is not supported

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[David Mandelin [:dmandelin]]

We think lack of SSE2 is the main cause of the topcrash in bug 595351. 

The assembler should already have an SSE2 detector, so for now, we can just check for SSE2 before calling into the methodjit.

Comment 1

[David Anderson [:dvander] - inactive, e-mail if emergency]

Attachment: fix

Tested on a 633MHz Celeron to have fixed the problem.

Comment 2

[David Mandelin [:dmandelin]]

Comment on attachment 475655 [details] [diff] [review]
fix

I didn't even know we had that updateJITEnabled thing. It is good.

Comment 3

[David Anderson [:dvander] - inactive, e-mail if emergency]

http://hg.mozilla.org/tracemonkey/rev/89283cd55a98

This should probably land on m-c early - can someone set a blocking flag?

Comment 4

[thewolf86]

According to AWFY Tracemonkey seems borked with this.

Comment 5

[David Anderson [:dvander] - inactive, e-mail if emergency]

Thanks, good catch, I accidentally inverted a test.

http://hg.mozilla.org/tracemonkey/rev/b9113b21b88c

Comment 6

[chris hofmann]

sayrer, can you mark blocking?

Comment 7

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/89283cd55a98

Comment 8

[Brendan Eich [:brendan]]

(In reply to comment #7)
> http://hg.mozilla.org/mozilla-central/rev/89283cd55a98

This does not including David's followup linked in comment 5.

BTW, I think if (!(a || b)) is always less clear (more cluttered, harder to read, requiring mental application of De Morgan's Theorem) than if (!a && !b). rs=me on a fix to do that.

/be

Comment 9

[Brendan Eich [:brendan]]

Sayrer got the followup:

http://hg.mozilla.org/mozilla-central/rev/b9113b21b88c

/be