Bug 19732
| Summary: | Fairly epic regression in insane mandelbrot test | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Oliver Hunt <oliver> |
| Component: | JavaScriptCore | Assignee: | Cameron Zwarich (cpst) <zwarich> |
| Status: | RESOLVED FIXED | ||
| Severity: | Normal | CC: | ggaren, mjs, zwarich |
| Priority: | P1 | Keywords: | Regression |
| Version: | 528+ (Nightly build) | ||
| Hardware: | Mac | ||
| OS: | OS X 10.5 | ||
| URL: | REGRESSION: http://nerget.com/jstests/dom-mandelbrot.html | ||
Oliver Hunt
I see a change from 2s with a build from a few weeks ago, to 10s with a nightly from a few days ago.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Geoffrey Garen
Somebody want to see if codegen for the test changed?
Maciej Stachowiak
Step 1 would be to narrow the point of regression.
Cameron Zwarich (cpst)
I made some codegen changes recently. They mostly affect code that needs a full scope chain, so it doesn't seem like they are at fault. Regardless, I'll assign this to myself.
Cameron Zwarich (cpst)
I don't see a regression. It is in the same range for all of the nightlies I have on my machine going back to the beginning of June. The only way I can get that kind of a slowdown is if I use a debug build instead of a release build.
Do you remember which nightlies two nightlies demonstrated a regression for you?
Cameron Zwarich (cpst)
This was something random that is no longer a problem.