CUDA 6.5（Cap 1.1）結構的cudaMemcpy數組導致“無效參數”

Question

碼：

Octree_t* tarr = (Octree*)malloc(nodes * sizeof(Octree_t));
//nodearray[2]->test = 42;
for (int i = 0; i < cnodes; i++)
{
    tarr[i] = *nodearray[i];
}
//printf(" test: %d\n",tarr[2].test); //returns 42

cudaError_t err;
Octree_t*     gpu_nodearray;

//allocate storage on gpu
err = cudaMalloc( &gpu_nodearray, cnodes * sizeof(Octree_t) != cudaSuccess  );
if ( err != cudaSuccess)
{
    printf("1: %s\n", cudaGetErrorString(err));
    return;
}
err = cudaMemcpy(gpu_nodearray, tarr,  cnodes * sizeof(Octree_t), cudaMemcpyHostToDevice);
if ( err != cudaSuccess)
{
    printf("3: %s\n", cudaGetErrorString(err));
    return;
}

它編譯良好，但是當我運行它時，它返回“ 3：無效參數”。 很明顯cudaMempy出了點問題。 我已經在這里查看過類似的帖子，但找不到我犯的錯誤。

謝謝你的幫助。

Answer 1

線

err = cudaMalloc( &gpu_nodearray, cnodes * sizeof(Octree_t) != cudaSuccess  );

應該

err = cudaMalloc( &gpu_nodearray, cnodes * sizeof(Octree_t));