ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-13 01:36:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

change file extension to hip.hpp and hip.cpp

Browse Source
This commit is contained in:
Chao Liu
2019-02-15 02:13:21 -06:00
parent a414e3fdf8
commit b2888adfbe
35 changed files with 212 additions and 234 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
src/include/ConstantMatrixDescriptor.hip.hpp Normal file
View File

@@ -0,0 +1,59 @@
#pragma once
#include "common.hip.hpp"
template <unsigned NRow_, unsigned NCol_, unsigned RowStride_>
struct ConstantMatrixDescriptor
{
__host__ __device__ constexpr ConstantMatrixDescriptor()
{
static_assert(NCol_ <= RowStride_, "wrong! NCol > RowStride!");
}
__host__ __device__ constexpr unsigned NRow() const { return NRow_; }
__host__ __device__ constexpr unsigned NCol() const { return NCol_; }
__host__ __device__ constexpr unsigned RowStride() const { return RowStride_; }
__host__ __device__ constexpr auto GetLengths() const { return Sequence<NRow_, NCol_>{}; }
__host__ __device__ constexpr unsigned GetElementSize() const { return NRow_ * NCol_; }
__host__ __device__ constexpr unsigned GetElementSpace() const { return NRow_ * RowStride_; }
__host__ __device__ unsigned Get1dIndex(unsigned irow, unsigned icol) const
{
return irow * RowStride_ + icol;
}
template <unsigned SubNRow, unsigned SubNCol>
__host__ __device__ constexpr auto MakeSubMatrixDescriptor(Number<SubNRow>,
Number<SubNCol>) const
{
return ConstantMatrixDescriptor<SubNRow, SubNCol, RowStride_>{};
}
};
template <unsigned NRow, unsigned NCol>
__host__ __device__ constexpr auto make_ConstantMatrixDescriptor(Number<NRow>, Number<NCol>)
{
return ConstantMatrixDescriptor<NRow, NCol, NCol>{};
}
template <unsigned NRow, unsigned NCol, unsigned RowStride>
__host__ __device__ constexpr auto
make_ConstantMatrixDescriptor(Number<NRow>, Number<NCol>, Number<RowStride>)
{
return ConstantMatrixDescriptor<NRow, NCol, RowStride>{};
}
template <class TDesc>
__host__ __device__ void print_ConstantMatrixDescriptor(TDesc, const char* s)
{
const auto desc = TDesc{};
constexpr auto I0 = Number<0>{};
constexpr auto I1 = Number<1>{};
printf("%s NRow %u NCol %u RowStride %u\n", s, desc.NRow(), desc.NCol(), desc.RowStride());
}