pthread and opencl ?

Accelerated Computing CUDA CUDA Programming and Performance
1

hello,

im trying to develop a simple test program. theres loop in which I run one opencl kernel (kernel_1) and another kernel - kernel_2 (or group of kernels) is running in background using pthreads ( saying in background i mean to execute kernel_2 clEnqueueNDRangeKernel less frequent than kernel_1).

unfortunatelly there are errors cl_out_of resources thrown on both kernels when calling clEnqueueNDRangeKernel and clEnqueueReadBuffer.

is there a solution to use pthread and opencl (or tbb+opencl) together without problems ?

full source at:

code - http://pastebin.com/sGz0tkEt

kernel - http://pastebin.com/jQQ4FXE2

most important part below:

[codebox]

pthread_mutex_t region_mutex = PTHREAD_MUTEX_INITIALIZER;

int main(int argc, char *argv) {

void *moj(void*);

pthread_t moj_thread; 

prepare_CL();



pthread_create(&moj_thread,NULL,moj,NULL);



for(int i=0; i < 100; i++) {

	ciErr1 = clEnqueueNDRangeKernel(cqCommandQue, kernel_fold2, 1, NULL, &worksize1D_global, &worksize1D_local, 0, NULL, &ee);

	if (ciErr1 != CL_SUCCESS) {printf("Error in clEnqueueNDRangeKernel !!!\n\n");}

	clFinish(cqCommandQue);

}



read_CL();



for(int j=0;j<size;j++) {

	cout << "thread: " << data_fold_out[j] << " main: " <<  data_fold_out2[j] << endl;

}

}

void moj(void) {

pthread_mutex_lock(&region_mutex);

ciErr1 = clEnqueueNDRangeKernel(cqCommandQue, kernel_fold, 1, NULL, &worksize1D_global, &worksize1D_local, 0, NULL, NULL);

  if (ciErr1 != CL_SUCCESS) {printf("Error in clEnqueueNDRangeKernel thread!!!\n\n"); pthread_exit(NULL);}

clFinish(cqCommandQue);

pthread_mutex_unlock(&region_mutex);

pthread_exit(NULL);

}

[/codebox]

2

**
oh… my mistake… post can be deleted