The work is pretty self-contained: custom quickstubs, typedarrays, and canvas 2d bits. The impact is much faster canvas imagedata ops; speedups are 30% to 50% depending on testcase we've looked at, probably mostly from the better cache locality that we get with the 4x smaller imagedata. This moves the needle from being at about safari-parity on most of these testcases (they've been slowly creeping faster in their dot-releases) to a lot faster.
Added the relevant dependent bugs; we wouldn't need all of them, just fixes/changes to qsgen.py, and then the new code for typed arrays + the canvas 2d quickstubs. We'd want to make sure that we don't expose the typed array constructors to web content, but that shouldn't be too hard.
Do we still want to do this?
There's no point; this would have helped to avoid being branded as slow 18 months ago, but now it's not like anyone cares the performance of 3.6.
Status: NEW → RESOLVED
Last Resolved: 7 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.