We encounter memory leak problems and the app profiler can't help. (we simply use windows resource monitor.)
A test code sample is given and the code structure is simple:
cl_kernel memset;
cl_command_queue Q1, Q2;
for(;;) {
cl_event Q1_evt[51];
cl_event Q2_evt[50];
clEnqueueNDRangeKernel(Q1, memset, ..., 0, NULL, &Q1_evt[0]);
for (int iter = 0; iter < 50; iter++) {
clEnqueueNDRangeKernel(Q2, memset, ..., 1, &Q1_evt[iter], &Q2_evt[iter]);
clEnqueueNDRangeKernel(Q1, memset, ..., 1, &Q2_evt[iter], &Q1_evt[iter+1]);
}
for (int i=0; i<50; i++) {
clReleaseEvent(Q1_evt);
clReleaseEvent(Q2_evt);
}
clReleaseEvent(Q1_evt[50]);
clFinish(Q1);
clFinish(Q2);
}
it is a twisted waiting structure that Q1 waits for Q2 and vise versa.
i guess clReleaseEvent() may not function right, so i create a thread here: http://devgurus.amd.com/message/1283971
my environment:
HD 7970, APP SDK 2.7, Catalyst 12.6 Beta (8.98-120522a-139735E-ATI), win7 64bit
