1094108 - 5% - 7% regression in octane-deltablue when inlining big functions with loops

Status: RESOLVED WONTFIX

(Core :: JavaScript Engine: JIT, defect)

Description

[Hannes Verschore [:h4writer]]

See bug 1087468 comment 14.

> This regresses deltablue with 5%-7%. So the big function
> with loop inlining also hits when having a constant/lambda :(.
> We definitely need to find out why "inlining big functions with loops"
> is cursed!
> 
> http://arewefastyet.com/#machine=11&view=single&suite=octane&subtest=DeltaBlue&start=1415069471&end=1415126055
> http://arewefastyet.com/#machine=12&view=single&suite=octane&subtest=DeltaBlue&start=1415069471&end=1415126055

Comment 1

[Hannes Verschore [:h4writer]]

From IRC:
> <jandem> h4writer: about the loop inlining, maybe we want something like: 
> inline functions with bigger loops 1 level deep (directly in the outer script)
> and bump the max inlining depth under that
> <jandem> h4writer: worth experimenting when we have time ;)

Comment 2

[Hannes Verschore [:h4writer]]

Loop inlining has been revamped. To workaround issues like these.