CUDA 6.5（第1.1）的結構的結果cudaMemcpy陣列中「無效參數」

Question

代碼：

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; 
} 

它編譯細但是當我運行它，則返回「3：無效參數」。很顯然，cudaMempy有些問題。我已經在這裏看過類似的帖子，我找不到我犯的錯誤。

感謝您的任何幫助。

Answer 1

線

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

應該

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