Hi. I'm trying to learn OpenCL using my Cypress graphics card, and I've got a few questions.
1. Where can i find information on the characteristics (ULPS etc.) of the native math functions in OpenCL? Specifically on ATI GPUs (Cypress).
2. I've seen words like "wavefronts" used, where should I go to learn more about what is actually going on (wavefronts, coalesced memory access etc). Is there any good source for a beginner to learn these things? The OpenCL specification document is great for learning OpenCL, but it doesn't help with the specifics of GPGPU or AMD GPUs.
3. Is there any indication when/if double precision fp operations and datatypes will be fully supported in OpenCL. The preview DP support available is not complete if I'm not mistaken.
4. Is AMD (doing now or planning to) artificially reducing double precision performance on consumer products, like a certain competitor seems to be doing with their upcomming flagship product?
Thanks a lot for any help