如果16个线程都执行，那么最后7个线程不是会产生越界错误？（我在CPU上执行是这样的，GPU上会得到准确答案）
如果这是一个异常的情况，数据不一定能够刚好等于线程数量，如何避免这种情况呢？谢谢！
__global__ void assignNumber(float* d_out, float dim)
{
int x=blockIdx.x*blockDim.x+threadIdx.x;
int y=blockIdx.y*blockDim.y+threadIdx.y;
int pos=y*dim+x;
//printf("%d\t%d\t%d\t%d\t%d\t%d\t%d\n",gridDim.x, gridDim.y, blockIdx.x, blockIdx.y, threadIdx.x, threadIdx.y, pos);
d_out[pos]=pos;
}
void main()
{
int dim=3;
float* d_out;
cudaMalloc((void**)&d_out, dim*dim*sizeof(float)));
dim3 thread(4,4);
dim3 grid( (dim-1)/thread.x+1, (dim-1)/thread.y+1);
assignNumber<<<grid, thread>>>(d_out,dim);
float* h_out=new float[dim*dim];
cudaMemcpy(h_out, d_out, dim*dim*sizeof(float), cudaMemcpyDeviceToHost);
//Check if pos equals the value
bool rst=true;
for(int i=0;i<dim;i++)
{
for(int j=0;j<dim;j++)
{
int pos=i*dim+j;
if(h_out[pos]!=pos)
{
rst=false;
break;
}
}
if(!rst) break;
}
if(rst)
{
cout<<"Success"<<endl;
}
else
{
cout<<"Failed"<<endl;
}
}
