A problem occur when I called the DLL built from CUDA

Hi all,

I built my CUDA code into a DLL and call it from LabVIEW, the strange thing I found is once I run my EXE, the CUDA would append the LabVIEW.exe to one of the core of my dual core machine… See the attach picture (the ‘CPU 1’ is unchecked):

This would make the LabVIEW.exe only runs on one of the cores letting the second core idling, I am not sure whether this is the expected behaviour or do you have any suggestion to make the second core run in parallel with GPU.

Thanks a lot.
CPU1.bmp (297 KB)

I saw the processing power of using the combination of 1 GPU + 1 CPU, just wonder whether is it possible to use 1 GPU + 2 CPUs in the CUDA architecture.

are you using the profiler? the profiler sets CPU affinity automatically

I didn’t use the profiler, I just called the DLL and measured the time it uses.