cory-master/BabelStream
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add mul kernel

Browse Source
This commit is contained in:
Tom Deakin 2016-04-26 16:03:28 +01:00
parent c22b74ba47
commit dec0237353
1 changed files with 13 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
src/CUDAStream.cu
View File

@ -65,10 +65,21 @@ void CUDAStream<T>::copy()
check_error(); check_error();
} }
template <typename T>
__global__ void mul_kernel(T * b, const T * c)
{
const T scalar = 3.0;
const int i = blockDim.x * blockIdx.x + threadIdx.x;
b[i] = scalar * c[i];
}
template <class T> template <class T>
void CUDAStream<T>::mul() void CUDAStream<T>::mul()
{ {
return; mul_kernel<<<1024, 1024>>>(d_b, d_c);
check_error();
cudaDeviceSynchronize();
check_error();
} }
template <class T> template <class T>
@ -85,3 +96,4 @@ void CUDAStream<T>::triad()
template class CUDAStream<float>; template class CUDAStream<float>;
template class CUDAStream<double>; template class CUDAStream<double>;