Add clearing kernel

This commit is contained in:
2024-04-11 19:49:09 +02:00
parent b49dddf34a
commit e86e04f6d6
2 changed files with 16 additions and 0 deletions

View File

@@ -62,6 +62,11 @@ __global__ void vec_scalar_sub(
const unsigned int w const unsigned int w
); );
__global__ void clear(
float* __restrict__ d_vector,
const unsigned int w
);
} // namespace CUDANet::Kernels } // namespace CUDANet::Kernels
#endif // CUDANET_MATMUL_H #endif // CUDANET_MATMUL_H

View File

@@ -84,3 +84,14 @@ __global__ void Kernels::vec_scalar_sub(
} }
d_output[tid] = d_vector[tid] - d_scalar[0]; d_output[tid] = d_vector[tid] - d_scalar[0];
} }
__global__ void Kernels::clear(
float* __restrict__ d_vector,
const unsigned int w
) {
int tid = blockDim.x * blockIdx.x + threadIdx.x;
if (tid >= w) {
return;
}
d_vector[tid] = 0.0f;
}