规约算法 GPU优化方法

规约算法是用于将大量数据聚合为一个值,例如计算数组中所有元素的总和、最大值、最小值等，可以利用到GPU的并行特性。

数组求和算法优化过程：

比如一个求数组和的方法，其方法为：

//reduce 测试
__global__ void reduce1(int* g_idata, int* g_odata) 
{
   
	extern __shared__ int sdata[1024];

	// each thread loads one element from global to shared mem
	unsigned int tid = threadIdx.x;
	unsigned int i = blockIdx.x * blockDim.x + threadIdx.x;
	sdata[tid] = g_idata[i];
	
	__syncthreads();

	// do reduction in shared mem
	for (unsigned int s = 1; s < blockDim.x; s *= 2) {
   
		if (tid % (2 * s) == 0) {
   
			sdata[tid] += sdata[tid + s];
		}
		__syncthreads();
	}
	
	// write result for this block to global mem
	if (tid == 0)
	{
   
		g_odata[blockIdx.x] = sdata[0];
	}
}

但该方法中的取余操作比较慢，因此改为：

__global__ void reduce2(int* g_idata, int* g_odata)
{
   
	extern __shared__ int sdata[1024];

	// each thread loads one element from global to shared mem
	unsigned int tid = threadIdx.x;
	unsigned int i = blockIdx.x * blockDim.x + threadIdx.x;
	sdata[tid] = g_idata[i];

	__syncthreads();

	// do reduction in shared mem
	for (unsigned int s = 1; s < blockDim.x; s *= 2) {
   
		int index = 2 * s * tid;
		if (index < blockDim.x) 
		{
   
			sdata[index] += sdata[index + s];
		}
		__syncthreads();
	}


	// write result for this block to global mem
	if (tid == 0)
	{
   
		g_odata[blockIdx.x] = sdata[0];
	}
}

在RTX 4080 GPU上，性能提升十分恐怖，效率直接翻倍。

但这个实现仍然有其问题，即bank冲突，这里需要将for循环逆着来，改为：

__global__ void reduce3(int* g_idata, int* g_odata)
{
   
	extern __shared__ int sdata[1024];

	// each thread loads one element from global to shared mem
	unsigned int tid = threadIdx.x;
	unsigned int i = blockIdx.x * blockDim.x + threadIdx.x;
	sdata[tid] = g_idata[i];

	__syncthreads();

	// do reduction in shared mem