0
我開始閱讀「Cuda By Example」書籍,我一直在使用「共享內存」的點示例問題。我從書中複製粘貼示例,然後設置:N = x * 1024; threadsPerBlock = 32; blocksPerGrid = 8.其中我用2,3,4,5測試「x」值。 如果我設置x = 3,結果不好,但是當我使用x = 2,4,5時,一切正常。我不明白問題在哪裏。該代碼是:Cuda的點積產品通過示例不適用於我
#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <stdio.h>
#define imin(a, b) (a<b?a:b)
#define sum_squares(x) (x*(x+1)*(2*x+1)/6)
const int x = 3;
const int N = 3 * 1024;
const int threadsPerBlock = 32;
const int blocksPerGrid = 8;
__global__ void dot(float *a, float *b, float *c)
{
__shared__ float cache[threadsPerBlock];
int tid = threadIdx.x + blockIdx.x * blockDim.x;
int cacheIndex = threadIdx.x;
float temp = 0;
while (tid < N)
{
temp += a[tid] * b[tid];
tid += blockDim.x * gridDim.x;
}
cache[cacheIndex] = temp;
__syncthreads();
int i = blockDim.x/2;
while (i != 0)
{
if (cacheIndex < i)
cache[cacheIndex] += cache[cacheIndex + i];
__syncthreads();
i /= 2;
}
if (cacheIndex == 0)
c[blockIdx.x] = cache[0];
}
int main()
{
float *a, *b, *partial_c, result;
float *d_a, *d_b, *d_partial_c;
a = (float *)malloc(N * sizeof(float));
b = (float *)malloc(N * sizeof(float));
partial_c = (float *)malloc(blocksPerGrid * sizeof(float));
cudaMalloc((void **)&d_a, N * sizeof(float));
cudaMalloc((void **)&d_b, N * sizeof(float));
cudaMalloc((void **)&d_partial_c, blocksPerGrid * sizeof(float));
for (int i = 0; i < N; i++)
{
a[i] = i;
b[i] = 2 * i;
}
cudaMemcpy(d_a, a, N * sizeof(float), cudaMemcpyHostToDevice);
cudaMemcpy(d_b, b, N * sizeof(float), cudaMemcpyHostToDevice);
dot << <blocksPerGrid, threadsPerBlock >> >(d_a, d_b, d_partial_c);
cudaMemcpy(partial_c, d_partial_c, blocksPerGrid * sizeof(float), cudaMemcpyDeviceToHost);
result = 0;
for (int i = 0; i < blocksPerGrid; i++)
result += partial_c[i];
if (2 * sum_squares((float)(N - 1)) == result)
printf(":)\n");
else
printf(":(\n");
cudaFree(d_a);
cudaFree(d_b);
cudaFree(d_partial_c);
free(a);
free(b);
free(partial_c);
getchar();
return 0;
}
好的,我改變了浮動倍數,工作正常。但我不明白爲什麼在我的情況下,x = 4,5(結果很大)沒有問題,但x = 3(較長的結果)失敗。 –
@PavelAngelMendozaVillafane你可以打印你正在比較的兩個浮點數,以及精確的結果來查看原因。 '2 * sum_squares((float)(N - 1))'不一定會給'float'帶來正確的結果。 – kangshiyin