/**
* prints some timing values of the kernel
*
* @param kernelEvent
*/
public void profile(cl_event kernelEvent) {
long submitTime[] = new long[1];
long queuedTime[] = new long[1];
long startTime[] = new long[1];
long endTime[] = new long[1];
CL.clGetEventProfilingInfo(
kernelEvent, CL.CL_PROFILING_COMMAND_QUEUED, Sizeof.cl_ulong, Pointer.to(queuedTime), null);
CL.clGetEventProfilingInfo(
kernelEvent, CL.CL_PROFILING_COMMAND_SUBMIT, Sizeof.cl_ulong, Pointer.to(submitTime), null);
CL.clGetEventProfilingInfo(
kernelEvent, CL.CL_PROFILING_COMMAND_START, Sizeof.cl_ulong, Pointer.to(startTime), null);
CL.clGetEventProfilingInfo(
kernelEvent, CL.CL_PROFILING_COMMAND_END, Sizeof.cl_ulong, Pointer.to(endTime), null);
submitTime[0] -= queuedTime[0];
startTime[0] -= queuedTime[0];
endTime[0] -= queuedTime[0];
// System.out.println("Queued : " + String.format("%8.3f", 0.0) + " ms");
// System.out.println("Submit : "
// + String.format("%8.3f", submitTime[0] / 1e6) + " ms");
// System.out.println("Start : "
// + String.format("%8.3f", startTime[0] / 1e6) + " ms");
// System.out.println("End : "
// + String.format("%8.3f", endTime[0] / 1e6) + " ms");
// long duration = endTime[0] - startTime[0];
// System.out.println("Time : " + String.format("%8.3f", duration / 1e6)
// + " ms");
System.out.printf(" | took : %8.3f milliseconds\n", endTime[0] / 1e6);
}
public void reinit() throws OpenClException {
// Obtain the platform IDs and initialize the context properties
cl_platform_id platforms[] = new cl_platform_id[1];
clGetPlatformIDs(platforms.length, platforms, null);
cl_context_properties contextProperties = new cl_context_properties();
contextProperties.addProperty(CL_CONTEXT_PLATFORM, platforms[0]);
// Create an OpenCL context on a GPU device
context =
CL.clCreateContextFromType(contextProperties, CL.CL_DEVICE_TYPE_GPU, null, null, null);
if (context == null) {
// If no context for a GPU device could be created,
// try to create one for a CPU device.
context =
CL.clCreateContextFromType(contextProperties, CL.CL_DEVICE_TYPE_CPU, null, null, null);
if (context == null) {
throw new OpenClException("Unable to create a context");
}
}
// Enable exceptions and subsequently omit error checks in this sample
CL.setExceptionsEnabled(true);
// Get the list of GPU devices associated with the context
long numBytes[] = new long[1];
CL.clGetContextInfo(context, CL.CL_CONTEXT_DEVICES, 0, null, numBytes);
// Obtain the cl_device_id for the first device
int numDevices = (int) numBytes[0] / Sizeof.cl_device_id;
cl_device_id[] devices = new cl_device_id[numDevices];
CL.clGetContextInfo(context, CL.CL_CONTEXT_DEVICES, numBytes[0], Pointer.to(devices), null);
device = devices[0];
// Create a command-queue
commandQueue = clCreateCommandQueue(context, devices[0], CL.CL_QUEUE_PROFILING_ENABLE, null);
}