Open
Bug 1169950
Opened 9 years ago
Updated 2 years ago
Firefox is slower than Chrome on Massive-poppler-cold-preparation
Categories
(Core :: JavaScript Engine, defect)
Core
JavaScript Engine
Tracking
()
NEW
Tracking | Status | |
---|---|---|
firefox41 | --- | affected |
People
(Reporter: guijoselito, Unassigned)
References
Details
My scores (win7): Firefox 38 - 223 Nightly 41 - 220 Chrome 43 - 628 I'm only filling this bug because this is the only test where Chrome beats Firefox on Massive Benchmark.
Comment 1•9 years ago
|
||
This is expected. cold-preparation tests measure startup times without anything in the cache. Firefox does AOT compilation of asm.js code, which lets it run faster later, but it does have a startup cost. Chrome JITs the code with far fewer optimizations, but it does so more quickly. It might still be worth looking at the AOT times, though, for anything to optimize.
Comment 2•9 years ago
|
||
I imagine bug 1169167 could help a lot here.
Comment 3•9 years ago
|
||
Yes, I'm not totally familiar with the TurboAsm compilation strategy, but iirc, it doesn't compile *fully* AOT, just AOT-er. Does cold-prep run *any* code or purely just load the script?
Comment 4•9 years ago
|
||
main() is not run, so essentially no code executes, just parsing/compilation and emscripten setup until main() is ready to be called (including global constructors, but no big ones there).
Updated•2 years ago
|
Severity: normal → S3
You need to log in
before you can comment on or make changes to this bug.
Description
•