[Webkit-unassigned] [Bug 43987] Downloading using XHR is much slower than before

Mon Aug 16 10:57:18 PDT 2010

https://bugs.webkit.org/show_bug.cgi?id=43987

--- Comment #1 from Geoffrey Garen <ggaren at apple.com>  2010-08-16 10:57:18 PST ---
(In reply to comment #0)

Thanks for spotting this.

> I think, this issue is introduced since the introduction of the new JS parser. Specifically, JavaScriptCore/wtf/Vector.h has templates for AlignedBuffer with the size of 1 byte to 64 bytes. 64 bytes is too small for downloading big chunks of data. Basically, what is happening is that we keep expanding the vector by only 64 bytes many times and thats why the ScriptString::operator+ is so slow.

I don't think Vector is the problem here. 64 bytes is the initial, inline capacity of the vector, not its growth policy. Vectors use a 1.25X growth policy. See Vector<T, inlineCapacity>::expandCapacity().

The problem here is that ScriptString creates a new vector for each append operation, instead of using a persistent vector a rope.

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