[Webkit-unassigned] [Bug 43742] REGRESSION(r58469): Math.pow() always returns double-backed JSValue which is extremely slow as array subscript

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=43742


Geoffrey Garen <ggaren at apple.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ggaren at apple.com




--- Comment #11 from Geoffrey Garen <ggaren at apple.com>  2010-08-18 14:56:44 PST ---
> I am unsure what the best fix is, move forward with the patch I've posted here, modify the JIT's storeDouble() to emulate JSValue(double), or something else entirely. Please advise :-)

I don't think you should modify storeDouble. storeDouble is used after an operation on doubles -- something that's unlikely to produce an int.

I think the two options are:

1. Add code to the specialized assembly math functions to try to convert to int if possible.
2. Move forward with a faster case for double indexing into an array.

I'd say try #1 and see if it's a performance regression. If not, it's pure win.

#2 also seems reasonable, but it won't pay off as much as #1 would, since it still involves a call out to a helper function.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.