need help with gpu code

1 view (last 30 days)
m4 Chrennikov
m4 Chrennikov on 4 Feb 2012
Im trying to use mutex in my code with no luck. Matlab says that kernel reaches timeout.
kernel:
__global__ void getGrid( int* mask,
int* mutex ) {
unsigned int idx = blockIdx.x * blockDim.x + threadIdx.x;
while( idx < 100 ){
while( atomicCAS( mutex, 0, 1 ) != 0 );
mask[ idx ] = idx;
atomicExch( mutex, 0 );
//*mutex = 0;
//while( atomicCAS( mutex, idx, idx + 1 ) != idx );
idx += blockDim.x * gridDim.x;
}
}
script:
mutex = parallel.gpu.GPUArray.zeros('int32');
kern0 = parallel.gpu.CUDAKernel('testkern.ptx', 'testkern.cu');
kern0.ThreadBlockSize = [256 1 1];
kern0.GridSize = [7*20 1];
mask = parallel.gpu.GPUArray.zeros(int32(256*7*20),int32(1),'int32');
tic
[mask mutex] = feval(kern0, mask, mutex);
toc
Commented atomicCAS works properly, but first one don't. Any mistakes or.. ?
  3 Comments
m4 Chrennikov
m4 Chrennikov on 4 Feb 2012
It is shared for graphics but this is not the cause.
Konrad Malkowski
Konrad Malkowski on 15 Feb 2012
Does this work outside of MATLAB?

Sign in to comment.

Answers (0)

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!