ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-07 16:26:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

Move the generic kernel instance to be the first of the instance list for elementwise op of normalization

Browse Source
This commit is contained in:
Qianfeng Zhang
2023-05-30 22:20:46 +00:00
parent 240900b1c0
commit de0bb3c26d
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
library/src/tensor_operation_instance/gpu/elementwise/device_normalize_instance.cpp
View File

@@ -28,10 +28,10 @@ using Normalize = ck::tensor_operation::element_wise::Normalize;
using device_normalize_from_mean_squaremean_f16_f32_f32_f16_f16_instances = std::tuple<
// clang-format off
//###################|<in, mean, square_mean, gamma, beta>| <out>| functor| NDim| MPerThread| <in, mean, square_mean, gamma, beta ScalarPerVector>| <out ScalarPerVector>|
DeviceElementwiseImpl<Tuple<F16, F32, F32, F16, F16>, Tuple<F16>, Normalize, 2, 1, Sequence<1, 1, 1, 1, 1>, Sequence<1> >,
DeviceElementwiseImpl<Tuple<F16, F32, F32, F16, F16>, Tuple<F16>, Normalize, 2, 8, Sequence<8, 1, 1, 8, 8>, Sequence<8> >,
DeviceElementwiseImpl<Tuple<F16, F32, F32, F16, F16>, Tuple<F16>, Normalize, 2, 4, Sequence<4, 1, 1, 4, 4>, Sequence<4> >,
DeviceElementwiseImpl<Tuple<F16, F32, F32, F16, F16>, Tuple<F16>, Normalize, 2, 2, Sequence<2, 1, 1, 2, 2>, Sequence<2> >,
DeviceElementwiseImpl<Tuple<F16, F32, F32, F16, F16>, Tuple<F16>, Normalize, 2, 1, Sequence<1, 1, 1, 1, 1>, Sequence<1> >
DeviceElementwiseImpl<Tuple<F16, F32, F32, F16, F16>, Tuple<F16>, Normalize, 2, 2, Sequence<2, 1, 1, 2, 2>, Sequence<2> >
// clang-format on
>;