Hello,
I am new to this forum, so please be gently. (Hopefully I didn’t miss any previous post regarding my problem.)
So here is my thing: I am writing an application for radiosity light calculation. For the formfactor calculation part: I am rendering many images on the GPU, get them via glGetTexImgae and evaluate them with OpenMP on the CPU. But, when I benchmarked the thing I realized glGetTexImage took nearly 70% of the time. So I wrote an OpenCL Kernel (in fact 2) to evaluate the rendered light ID image (from a skydome) on the GPU. I wrote this on my laptop with OpenCL 1.0 and Geforce 8600M GT. It worked just fine. In high resolutions I got a good speedup of 2 (for my old laptop).
Now, I wanted to test the application on my desktop with Geforce GTX570 (OpenCL 1.1, latest WHQL driver). It crashed. When I debugged the application, clWaitForEvents() returns CL_OUT_OF_RESOURCES after I launched my 2nd kernel. After reading a bit in this forum someone wrote, that this error might be a general error or that my kernel worksizes are set wrong. But it worked on my 8600M GT laptop.
- How do I figure out, what went wrong with CL_OUT_OF_RESOURCES?
- Reading out of memory bounds on the GPU in my kernel?
- Wrong worksizes in my Kernel?
- Not enough GPU memory? ;-)
Any hints and answers are very welcome. Thanks in advance! I can deliver code if you want, but I am afraid it would be a little overkill here.
Cheers, Christian