clBuildProgram() hangs with 99% CPU-consumption

[ebfe]

Hi,

I’m the maintainer of Pyrit and currently try to get my OpenCL-code working with NVIDIA’s CUDA 3.0-beta related drivers. While the code has worked with previous versions of NVIDIA’s OpenCL-implementation (and works with OpenCL on ATI through Stream 2.0), the currently available beta-versions related to CUDA 3.0 run into problems.

Tracking down the problem, I can see that clBuildProgram() hangs indefinitely (the call never returns and consumes 99% CPU-time) and the whole process has to be SIGKILLed. As it is, I can’t really provide any more debug-information. I invite everyone to take a look at the code and maybe provide some insight.

[TRT_Sam]

Hi,

I’m the maintainer of Pyrit and currently try to get my OpenCL-code working with NVIDIA’s CUDA 3.0-beta related drivers. While the code has worked with previous versions of NVIDIA’s OpenCL-implementation (and works with OpenCL on ATI through Stream 2.0), the currently available beta-versions related to CUDA 3.0 run into problems.

Tracking down the problem, I can see that clBuildProgram() hangs indefinitely (the call never returns and consumes 99% CPU-time) and the whole process has to be SIGKILLed. As it is, I can’t really provide any more debug-information. I invite everyone to take a look at the code and maybe provide some insight.

Hi,

I push up an old thread cause I got exactly the same problem. I’m trying to adapt Apple’s FFT sample and got the clBuildProgram consuming almost all CPU-time! Did you find out what was wrong?

Thanks,

Sam

[TRT_Sam]

Hi,

I’m the maintainer of Pyrit and currently try to get my OpenCL-code working with NVIDIA’s CUDA 3.0-beta related drivers. While the code has worked with previous versions of NVIDIA’s OpenCL-implementation (and works with OpenCL on ATI through Stream 2.0), the currently available beta-versions related to CUDA 3.0 run into problems.

Tracking down the problem, I can see that clBuildProgram() hangs indefinitely (the call never returns and consumes 99% CPU-time) and the whole process has to be SIGKILLed. As it is, I can’t really provide any more debug-information. I invite everyone to take a look at the code and maybe provide some insight.

Hi,

I push up an old thread cause I got exactly the same problem. I’m trying to adapt Apple’s FFT sample and got the clBuildProgram consuming almost all CPU-time! Did you find out what was wrong?

Thanks,

Sam

[Yuan]

Sounds like a known problem in 195/3.0. Try a later release. It should have been fixed.