I committed the first tweak you made so you can 'git pull' (you might need `git fetch` `git reset --hard origin/gpu_integration`)
I'm not sure why you get the error for
Code:
CUDA_CHECK(cudaSetDeviceFlags (cudaDeviceScheduleBlockingSync));
Google suggests this might happen if I've already started run things but the code shouldn't have at that point.
You could try replacing line 135 in cudakernel.cu (while leaving line 591 commented out)
Code:
- errCheck (cudaSetDeviceFlags (cudaDeviceScheduleYield));
+ errCheck (cudaSetDeviceFlags (cudaDeviceScheduleBlockingSync));
If anyone else experiences this let me know so I can try to fix but I'm going to ignore for now.