OpenCL Kernel code issue

OpenCL
1

Hi everyone.
I am pretty new to OpenCL. I am having some problems with the clBuildProgram function. In my case, the program is not getting built and the error code is giving the error CL_BUILD_PROGRAM_FAILURE.

I am trying to parallelise the following code snippet:

for (k = kernel.width-1 ; k >= 0 ; k–)
sum += *ppp++ * kernel.data[k];

This is the kernel code that I have written for parallelizing the same:

__kernel void convolveImageHoriz(__global float *ppp, __global float *kernel, float sum, int width)
{
int k = get_global_id(0);

if (k >= 0 && k < width)
    {
        sum = ppp[width-k-1] * kernel[k];
    }

}

I know for a fact that there is an issue with my Kernel code because when I load some other working kernel, then the program builds successfully.

I am under a deadline so I would request someone to respond quickly.

2

mohitu

  1. Try to print the buld log to see the build error:

#include <cstdio>
#include <vector>

//....

size_t n = 0;
cl_int ret = clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, n, NULL, &n);
if (n > 1) {
	std::vector<char> buildLog;
	buildLog.resize(n);
	ret = clGetProgramBuildInfo(program, device, CL_PROGRAM_BUILD_LOG, n, &buildLog[0], &n);
	fprintf(stderr, "
OpenCL Build log: %s
", &buildLog[0]);
}

I have result of compilation of your kernel:

OpenCL Build log: :1:71: error: expected ‘)’
__kernel void convolveImageHoriz(__global float *ppp, __global float *kernel, float sum, int width)
^
:1:33: note: to match this ‘(’
__kernel void convolveImageHoriz(__global float *ppp, __global float *kernel, float sum, int width)
^
:1:71: error: parameter name omitted
__kernel void convolveImageHoriz(__global float *ppp, __global float *kernel, float sum, int width)
^
:5:19: error: use of undeclared identifier ‘width’
if (k >= 0 && k < width)
^
:7:1: error: use of undeclared identifier ‘sum’
sum = ppp[width-k-1] * kernel[k];
^
:7:11: error: use of undeclared identifier ‘width’
sum = ppp[width-k-1] * kernel[k];
^
:7:24: error: expected expression
sum = ppp[width-k-1] * kernel[k];

You cannot use keyword “kernel” in your OpenCL code.Try to use the fixed code:

__kernel void convolveImageHoriz(__global float *ppp, __global float *Kernel, float sum, int width)
{
int k = get_global_id(0);

if (k >= 0 && k < width)
{
sum = ppp[width-k-1] * Kernel[k];
}
}
3

Hi.
Thanks for your reply.

I replaced “kernel” with “Kernel” and got the code to work properly.
Thanks for your help.
:slight_smile: :smiley: