Function TimedSort

test/OpenCL/RadixSortBenchmark/main.cpp:103–183

 */
template <typename K>
void TimedSort(
	unsigned int num_elements,
	K *h_keys,
	unsigned int iterations)
{
	printf("Keys only, %d iterations, %d elements\n", iterations, num_elements);

	int max_elements = num_elements;
	b3AlignedObjectArray<unsigned int> hostData;
	hostData.resize(num_elements);
	for (int i = 0; i < num_elements; i++)
	{
		hostData[i] = h_keys[i];
	}

	b3RadixSort32CL sorter(g_cxMainContext, g_device, g_cqCommandQueue);

	b3OpenCLArray<unsigned int> gpuData(g_cxMainContext, g_cqCommandQueue);
	gpuData.copyFromHost(hostData);
	//sorter.executeHost(gpuData);
	sorter.execute(gpuData);

	b3AlignedObjectArray<unsigned int> hostDataSorted;
	gpuData.copyToHost(hostDataSorted);

	clFinish(g_cqCommandQueue);

	{
		//printf("Key-values, %d iterations, %d elements", iterations, num_elements);

		// Create sorting enactor

		// Perform the timed number of sorting iterations
		double elapsed = 0;
		float duration = 0;
		b3Clock watch;

		//warm-start
		gpuData.copyFromHost(hostData);
		clFinish(g_cqCommandQueue);
		sorter.execute(gpuData);

		watch.reset();

		for (int i = 0; i < iterations; i++)
		{
			// Move a fresh copy of the problem into device storage
			gpuData.copyFromHost(hostData);
			clFinish(g_cqCommandQueue);

			// Start GPU timing record
			double startMs = watch.getTimeMicroseconds() / 1e3;

			// Call the sorting API routine
			sorter.execute(gpuData);

			clFinish(g_cqCommandQueue);