What's the state of play now with >8bit colour modes in snow leopard? I've been building an application in leopard that uses QC extensively. I'm using a lot of render-in-image patches set to 32bit mode to do some high quality processing. In Leopard, I get good performance, and an unfiltered (i.e. pixelated) image rendered at the end (this is expected).
>From a quick test under QC4 last night, it looks like performance is actually lower, CPU usage quite a bit higher, and the end result is filtered. I've not had time to do any thorough testing yet (the iphone project I'm working on just plain doesn't work under SL somehow, and the deadline is close so I'm staying with 10.5) but perhaps somebody could enlighten me as to what's going on? First, colour depth: is filtering now possible with 32bit mode? I thought it wasn't possible in hardware (I'm using an ATI 2600). Or has it dropped back to 8bit for some reason. And is 16bit fixed so it works on nvidia cards now? Next, core image performance. I've read somewhere that CI is now implemented using OpenCL, and as OpenCL isn't supported on my GPU that would imply CI is now running on the CPU (this would be a disaster for me, and I'd have to drop SL support until the situation changes or I can re-write the whole thing in GLSL :( ). If it IS running on CPU, that would explain why performance is down and CPU usage is up. Can anyone confirm what's going on here? Last quick question: I can't access any of the openCL patches in QC4 it seems. It's supposed to run on the CPU if the GPU isn't supported, so is there any way to tell QC to let me play with them even if performance is bad? Thanks Chris
_______________________________________________ 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]
