1115886 - Store TypeObject::interpretedFunction in its addendum

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Brian Hackett [Laid off!]]

I think this was part of another bug a while ago that never landed, but now that TypeObject::addendum is back it makes a nice place to store the interpreted function for a TypeObject, which frees up a word (two words on 32 bit platforms).

Comment 1

[Brian Hackett [Laid off!]]

Attachment: patch

Comment 2

[Jan de Mooij [:jandem]]

Comment on attachment 8544282 [details] [diff] [review]
patch

Review of attachment 8544282 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good.

Comment 3

[Brian Hackett [Laid off!]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a27623547914

Comment 4

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/a27623547914

Comment 5

[Mayank Bansal]

Appears to have caused a 11.1% regression on AWFY Octane-Splay on machine 28.

http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=7de37d6b2f4c&tochange=a27623547914

Comment 6

[Brian Hackett [Laid off!]]

(In reply to mayankleoboy1 from comment #5)
> Appears to have caused a 11.1% regression on AWFY Octane-Splay on machine 28.
> 
> http://hg.mozilla.org/integration/mozilla-inbound/
> pushloghtml?fromchange=7de37d6b2f4c&tochange=a27623547914

Hmm, I'll look at this later but there isn't any reason this patch should have regressed splay (and the regression didn't happen on OS X x64 or on windows browser), though if our GC heuristics are especially brittle on this machine the reduction in size of TypeObject could have affected something.

Comment 7

[Brian Hackett [Laid off!]]

(In reply to Brian Hackett (:bhackett) from comment #6)
> (In reply to mayankleoboy1 from comment #5)
> > Appears to have caused a 11.1% regression on AWFY Octane-Splay on machine 28.
> > 
> > http://hg.mozilla.org/integration/mozilla-inbound/
> > pushloghtml?fromchange=7de37d6b2f4c&tochange=a27623547914
> 
> Hmm, I'll look at this later but there isn't any reason this patch should
> have regressed splay (and the regression didn't happen on OS X x64 or on
> windows browser), though if our GC heuristics are especially brittle on this
> machine the reduction in size of TypeObject could have affected something.

Running locally on 10.9 x86 I get similar scores before/after this patch when running splay by itself.  Similar to machine 28, if I run the entire suite the splay score drops a bit but the splay-latency score goes up a bit.  So I think shrinking sizeof(TypeObject) probably just changed where we do GCs on this benchmark when running in the shell.