[英]CUDA 6.5 (Cap 1.1) cudaMemcpy array of structs results in “invalid argument”
码:
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;
}
它编译良好,但是当我运行它时,它返回“ 3:无效参数”。 很明显cudaMempy出了点问题。 我已经在这里查看过类似的帖子,但找不到我犯的错误。
谢谢你的帮助。
线
err = cudaMalloc( &gpu_nodearray, cnodes * sizeof(Octree_t) != cudaSuccess );
应该
err = cudaMalloc( &gpu_nodearray, cnodes * sizeof(Octree_t));
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.