CODEWITHSUNDEEP
c++gpuopencl
MooMoo
MooMoo

Reputation: 1116

OpenCL get_global_id with wrong result

I have a problem with wrong global_id() result. I would like to convolute 3D voxel with dimension {35,35,35} with 3D kernel with dimension {5,5,5}. Therefore, I call "clEnqueueNDRangeKernel" with global_size = {35,35,35} and local size = { 5, 5, 5}

std::vector<size_t> local_nd  = { 5, 5, 5 };
std::vector<size_t> global_nd = { 35, 35, 35 };
err = clEnqueueNDRangeKernel( queue, hello_kernel, work_dim, NULL, global_nd.data(), local_nd.data(), 0, NULL, NULL);

What I expect when I call get_global_id() function is the global_id(0) should be between 0 to 34 global_id(1) should be between 0 to 34 and global_id(2) should be between 0 to 34.

However for global_id(0) and global_id(1) the results seem correct. However the global_id(2) the values range from 30 - 34 instead of 0 to 34 that I expect.

const int  ic0     =  get_global_id(0);  // icol
const int  ic1     =  get_global_id(1);  // irow  
const int  ic2     =  get_global_id(2);  // idep 


printf(" %d %d %d\n", ic0, ic1, ic2 ); 
// value of ic0 = [0  -> 34] correct!
// value of ic1 = [0  -> 34] correct!
// value of ic2 = [30 -> 34]  ( SHOULD IT BE [0->34] )?

my gpu is max-workgroup is max work-group item ND: { 1024, 1024, 64 }

Upvotes: 0

Views: 747

Answers (1)

MooMoo
MooMoo

Reputation: 1116

I found the problem as pmdj suggested. 

printf in kernels isn't always reliable - there's often a fixed-size buffer, and if you output too much, some messages may be dropped.

After I changed the OpenCL code with some conditions. ex: 

if( ic2< 10 )
    printf("ic2: %d ", ic2 );

The output ranges from [0 --> 34 as I expected]

Upvotes: 1

Related Questions