[Public WebGL] Re: webgl/swiftshader

Florian Bösch [email protected]
Wed Feb 22 06:05:54 PST 2012


Maybe we're approaching the issue of wildly different speeds from a wrong
angle. In essence app-devs just want to know how much performance they can
assume. Trying to figure that out from flags or gpu model or anything would
be both tedious and impossible because of the increase in user-identifyable
bits. The other way we can figure that out would be by offering some sort
of test-runner and performance settings to the user, which may be what's
sometimes desired, but is always a lot of work to do.

Maybe we can come up with some meaningful performance "profiles" that are
tied to a set of authored benchmarks, and they'd roughly indicate the level
of performance in a simple grading (say a scale between 0 and 1, 0 being
the slowest implementation one can find that still works, and 1 being the
fastest). And app-devs could use this as a hint to configure their app and
settings (or set reasonable defaults for the user to change).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://khronos.org/pipermail/public_webgl_khronos.org/attachments/20120222/b1a216cc/attachment.html>


More information about the public_webgl mailing list