ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-13 17:55:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

tweak

Browse Source
This commit is contained in:
Chao Liu
2019-08-03 15:05:25 -05:00
parent 4a1e97cf86
commit adc1008836
6 changed files with 27 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
composable_kernel/include/tensor_operation/blockwise_generic_tensor_slice_copy.hpp
View File

@@ -447,17 +447,19 @@ struct BlockwiseGenericTensorSliceCopy_v2
}
template <class T, bool PositiveDirection>
__device__ void MoveSrcSlicingWindow(T step_sizes, integral_constant<bool, PositiveDirection>)
__device__ void
MoveSrcSlicingWindow(T step_sizes,
integral_constant<bool, PositiveDirection> positive_direction)
{
mThreadwiseLoad.MoveSrcSlicingWindow(step_sizes,
integral_constant<bool, PositiveDirection>{});
mThreadwiseLoad.MoveSrcSlicingWindow(step_sizes, positive_direction);
}
template <class T, bool PositiveDirection>
__device__ void MoveDstSlicingWindow(T step_sizes, integral_constant<bool, PositiveDirection>)
__device__ void
MoveDstSlicingWindow(T step_sizes,
integral_constant<bool, PositiveDirection> positive_direction)
{
mThreadwiseLoad.MoveDstSlicingWindow(step_sizes,
integral_constant<bool, PositiveDirection>{});
mThreadwiseLoad.MoveDstSlicingWindow(step_sizes, positive_direction);
}
private:

6
composable_kernel/include/tensor_operation/threadwise_generic_tensor_slice_copy.hpp
View File

@@ -227,9 +227,9 @@ struct ThreadwiseGenericTensorSliceCopy_v2
template <class T, bool PositiveDirection>
__device__ void MoveDstSlicingWindow(T step_sizes, integral_constant<bool, PositiveDirection>)
{
static_if<PositiveDirection>([&](auto) { mDstSliceOrigin += step_sizes; }).Else([&](auto) {
mDstSliceOrigin -= step_sizes;
});
static_if<PositiveDirection>{}([&](auto) {
mDstSliceOrigin += step_sizes;
}).Else([&](auto) { mDstSliceOrigin -= step_sizes; });
}
// private: