A few hundred thousand imacs and a lot of mac pros would make a reasonable economic case I suspect, but if the things I've heard about GPGPU on ATI 2/3xxx cards are right the rewards might not be worth the effort.
What's always (for small values of "always" -- since wwdc 08, when I learned of cl's existence ;) interested me would be having the cl compiler be smart enough to know what classes of kernels can be gpu accelerated, and target them appropriately (similar to how CoreImage will run kernels on the GPU when it can, and will fall back to CPU if it can't) -- even my pathetic GMA950 can surely work on _some_ subset of CL problems (even if it's just the lame "add a bunch of numbers" kernels), and I'm guessing the 2/3xxx series are even more capable, even if not completely CL capable; the docs say otherwise, but I don't know one way or the other. Performance would be all over the place, of course, and it would take a lot of work to get it up, running, and debugged.
Purely academic, I know, but it's hard to not consider the possibilities sometimes, no matter how small :)
-- [ christopher wright ] [email protected] http://kineme.net/
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Do not post admin requests to the list. They will be ignored. Quartzcomposer-dev mailing list ([email protected]) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com This email sent to [email protected]
