CODEWITHSUNDEEP
cmultithreadingopenclparallel-processinggpu
Tudor
Tudor

Reputation: 62439

OpenCL Kernel fails to compile

I keep reading and reading this matrix multiplication kernel code and I just don't get why the call to clBuildProgram returns CL_BUILD_PROGRAM_FAILURE. Here is my kernel code:

__kernel void MatMulKernel(__global const float* A, 
                           __global const float* B, 
                                          float* C, 
                                    const int size1, 
                                    const int size2, 
                                    const int size3) 
{   
    int k = get_global_id(0);
    int i;

    int line = k / size3;
    int column = k % size3;

    float partial = 0;

    for(i = 0; i < size2; i++)
    {
        partial += A[line * size2 + i] * B[i * size3 + column];
    }

    C[k] = partial; 
}

Can anyone spot the problem? Thank you.

Note: The code that does the initializations is correct, as I have tested with other kernels and they compile correctly.

Edit: Ok prunge's answer did the trick, but now I'm running into a different problem. The kernel execution actually causes a crash. Here is the code:

err = clSetKernelArg(hKernel, 0, sizeof(cl_mem), (void *)&hDeviceMemA);
err = clSetKernelArg(hKernel, 1, sizeof(cl_mem), (void *)&hDeviceMemB);
err = clSetKernelArg(hKernel, 2, sizeof(cl_mem), (void *)&hDeviceMemC);
err = clSetKernelArg(hKernel, 3, sizeof(cl_int), (void *)&s1);
err = clSetKernelArg(hKernel, 4, sizeof(cl_int), (void *)&s2);      
err = clSetKernelArg(hKernel, 5, sizeof(cl_int), (void *)&s3);  

cl_event events[1];
// execute kernel
start = clock();
err = clEnqueueNDRangeKernel(hCmdQueue, hKernel, 1, 0, (const size_t *)BENCH_SIZE_COMP, 0, 0, 0, &events[0]);       
clWaitForEvents(1, events);

All err values for the calls to clSetKernelArg are CL_SUCCESS. When the program reaches clEnqueueNDRangeKernel it crashes.

Upvotes: 0

Views: 1205

Answers (1)

prunge
prunge

Reputation: 23238

This is the error I got:

error: kernel pointer arguments must point to addrSpace global, local, or constant

The float* C parameter should probably be __global. All kernel pointer arguments need an address space qualifier.

Upvotes: 2

Related Questions