[菜鸟每天来段CUDA_C]使用多个CUDA流提高程序执行效率

CUDA流表示一个GPU操作队列，并且该队列中的操作以添加到队列的先后顺序执行。使用CUDA流可以实现任务级的并行，比如当GPU在执行核函数的同时，还可以在主机和设备之间交换数据(前提是GPU支持重叠，property的deviceOverlay为true)。

cudaMemcpyAsync函数的功能是在GPU和主机之间复制数据。它是一个异步函数，即函数被调用后，只是放置一个请求，表示在流中执行一次内存复制操作。函数返回时，复制操作不一定启动或执行结束，只是该操作被放入执行队列，在下一个被放入流中的操作之前执行。

实验通过把一组数据分块复制到GPU执行，返回执行结果，来说明使用cuda流的使用能提高程序的执行效率。原理主要是使数据复制操作和核函数执行操作交叉执行，不用等到第一次核函数执行结束再开始第二轮的数据复制，以减少顺序执行带来的延迟(类似于编译中使用流水线在解决冲突的前提下提高效率)。

程序代码如下：

#include "cuda_runtime.h"
#include "cutil_inline.h"

#include <stdio.h>
#include <math.h>

static void HandleError( cudaError_t err,const char *file,int line ) 
{

	if (err != cudaSuccess)

	{

		printf( "%s in %s at line %d\n", cudaGetErrorString( err ),
			file, line );
		exit( EXIT_FAILURE );
	}
}
#define HANDLE_ERROR( err ) (HandleError( err, __FILE__, __LINE__ ))
#define N (1024*1024)
#define FULL_DATA_SIZE N*20

__global__ void kernel(int* a, int *b, int*c)
{
	int idx = blockIdx.x * blockDim.x + threadIdx.x;
	int offset = gridDim.x * blockDim.x;

	if (idx < N)
	{
		int idx1 = (idx + 1) % 256;
		int idx2 = (idx + 2) % 256;
		float as = (a[idx] + a[idx1] + a[idx2]) / 3;
		float bs = (b[idx] + b[idx1] + b[idx2]) / 3;
		c[idx] = (as + bs) / 2;
	}
}

int main()
{
	cudaDeviceProp prop;
	int devID;

	HANDLE_ERROR(cudaGetDevice(&devID));
	HANDLE_ERROR(cudaGetDeviceProperties(&prop, devID));

	if (!prop.deviceOverlap)
	{
		printf("No device will handle overlaps. so no speed up from stream.\n");
		return 0;
	}

	cudaEvent_t start, stop;
	float elapsedTime;
	HANDLE_ERROR(cudaEventCreate(&start));
	HANDLE_ERROR(cudaEventCreate(&stop));
	HANDLE_ERROR(cudaEventRecord(start, 0));

	cudaStream_t stream0;
	cudaStream_t stream1;
	HANDLE_ERROR(cudaStreamCreate(&stream0));
	HANDLE_ERROR(cudaStreamCreate(&stream1));

	int *host_a, *host_b, *host_c;
	int *dev_a0, *dev_b0, *dev_c0;
	int *dev_a1, *dev_b1, *dev_c1;

	HANDLE_ERROR(cudaMalloc((void**)&dev_a0, N*sizeof(int)));
	HANDLE_ERROR(cudaMalloc((void**)&dev_b0, N*sizeof(int)));
	HANDLE_ERROR(cudaMalloc((void**)&dev_c0, N*sizeof(int)));

	HANDLE_ERROR(cudaMalloc((void**)&dev_a1, N*sizeof(int)));
	HANDLE_ERROR(cudaMalloc((void**)&dev_b1, N*sizeof(int)));
	HANDLE_ERROR(cudaMalloc((void**)&dev_c1, N*sizeof(int)));

	HANDLE_ERROR(cudaHostAlloc((void**)&host_a, FULL_DATA_SIZE*sizeof(int), cudaHostAllocDefault));
	HANDLE_ERROR(cudaHostAlloc((void**)&host_b, FULL_DATA_SIZE*sizeof(int), cudaHostAllocDefault));
	HANDLE_ERROR(cudaHostAlloc((void**)&host_c, FULL_DATA_SIZE*sizeof(int), cudaHostAllocDefault));


	for (int i=0; i<FULL_DATA_SIZE; i++)
	{
		host_a[i] = rand();
		host_b[i] = rand();
	}


	// tasks are put into stack for gpu execution
	for (int i=0; i<FULL_DATA_SIZE; i+=2*N)
	{
		HANDLE_ERROR(cudaMemcpyAsync(dev_a0, host_a+i, N*sizeof(int), cudaMemcpyHostToDevice, stream0));
		HANDLE_ERROR(cudaMemcpyAsync(dev_a1, host_a+i+N, N*sizeof(int), cudaMemcpyHostToDevice, stream1));

		HANDLE_ERROR(cudaMemcpyAsync(dev_b0, host_b+i, N*sizeof(int), cudaMemcpyHostToDevice, stream0));
		HANDLE_ERROR(cudaMemcpyAsync(dev_a1, host_a+i+N, N*sizeof(int), cudaMemcpyHostToDevice, stream1));

		kernel<<<N/256, 256, 0, stream0>>>(dev_a0, dev_b0, dev_c0);
		kernel<<<N/256, 256, 0, stream1>>>(dev_a1, dev_b1, dev_c1);

		HANDLE_ERROR(cudaMemcpyAsync(host_c+i, dev_c0, N*sizeof(int), cudaMemcpyDeviceToHost, stream0));
		HANDLE_ERROR(cudaMemcpyAsync(host_c+i+N, dev_c1, N*sizeof(int), cudaMemcpyDeviceToHost, stream1));

// 		HANDLE_ERROR(cudaMemcpyAsync(dev_a0, host_a+i, N*sizeof(int), cudaMemcpyHostToDevice, stream0));
// 		HANDLE_ERROR(cudaMemcpyAsync(dev_b0, host_b+i, N*sizeof(int), cudaMemcpyHostToDevice, stream0));
// 		kernel<<<N/256, 256, 0, stream0>>>(dev_a0, dev_b0, dev_c0);
// 		HANDLE_ERROR(cudaMemcpyAsync(host_c+i, dev_c0, N*sizeof(int), cudaMemcpyDeviceToHost, stream0));
// 
// 		HANDLE_ERROR(cudaMemcpyAsync(dev_a1, host_a+i+N, N*sizeof(int), cudaMemcpyHostToDevice, stream1));
// 		HANDLE_ERROR(cudaMemcpyAsync(dev_a1, host_a+i+N, N*sizeof(int), cudaMemcpyHostToDevice, stream1));
// 		kernel<<<N/256, 256, 0, stream1>>>(dev_a1, dev_b1, dev_c1);
// 		HANDLE_ERROR(cudaMemcpyAsync(host_c+i+N, dev_c1, N*sizeof(int), cudaMemcpyDeviceToHost, stream1));
	}

	// wait until gpu execution finish
	HANDLE_ERROR(cudaStreamSynchronize(stream0));
	HANDLE_ERROR(cudaStreamSynchronize(stream1));

	HANDLE_ERROR(cudaEventRecord(stop, 0));
	HANDLE_ERROR(cudaEventSynchronize(stop));
	HANDLE_ERROR(cudaEventElapsedTime(&elapsedTime, start, stop));

	printf("Time taken: %3.1f ms\n", elapsedTime);

	// free stream and mem
	HANDLE_ERROR(cudaFreeHost(host_a));
	HANDLE_ERROR(cudaFreeHost(host_b));
	HANDLE_ERROR(cudaFreeHost(host_c));
	HANDLE_ERROR(cudaFree(dev_a0));
	HANDLE_ERROR(cudaFree(dev_b0));
	HANDLE_ERROR(cudaFree(dev_c0));
	HANDLE_ERROR(cudaFree(dev_a1));
	HANDLE_ERROR(cudaFree(dev_b1));
	HANDLE_ERROR(cudaFree(dev_c1));
	HANDLE_ERROR(cudaStreamDestroy(stream0));
	HANDLE_ERROR(cudaStreamDestroy(stream1));
	return 0;
}

相对于顺序执行，使用两个cuda流使程序的执行时间少了20ms（由于数据量不大，所以使用流的优势不太明显）.