ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-02 20:51:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] Image to Column kernel (#1532)

Browse Source 
* [CK_TILE] Image to Column kernel

* Fixes

* Vector loads and stores

* Fixes

* Fixes

* change test dir name
This commit is contained in:
Bartłomiej Kocot
2024-09-27 22:57:38 +02:00
committed by GitHub
parent 9d69a099a4
commit de3e3b6424
19 changed files with 1419 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
include/ck_tile/ops/image_to_column/pipeline/block_image_to_column_problem.hpp Normal file
View File

@@ -0,0 +1,27 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck_tile/core/utility/type_traits.hpp"
namespace ck_tile {
template <typename InDataType_,
typename OutDataType_,
typename BlockShape_,
index_t NDimSpatial_,
index_t AligmentIn_,
index_t AligmentOut_>
struct BlockImageToColumnProblem
{
using InDataType = remove_cvref_t<InDataType_>;
using OutDataType = remove_cvref_t<OutDataType_>;
using BlockShape = remove_cvref_t<BlockShape_>;
static constexpr index_t NDimSpatial = NDimSpatial_;
static constexpr index_t AligmentIn = AligmentIn_;
static constexpr index_t AligmentOut = AligmentOut_;
};
} // namespace ck_tile

32
include/ck_tile/ops/image_to_column/pipeline/tile_image_to_column_shape.hpp Normal file
View File

@@ -0,0 +1,32 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck_tile/core.hpp"
namespace ck_tile {
template <typename ThreadTile, // Sequence<...
typename WarpTile, // Sequence<...
typename BlockTile> // Sequence<...
struct TileImageToColumnShape
{
static constexpr index_t kMPerThread = ThreadTile::at(number<0>{});
static constexpr index_t kKPerThread = ThreadTile::at(number<1>{});
static constexpr index_t kMPerWarp = WarpTile::at(number<0>{});
static constexpr index_t kKPerWarp = WarpTile::at(number<1>{});
static constexpr index_t kMThreadPerWarp = kMPerWarp / kMPerThread;
static constexpr index_t kKThreadPerWarp = kKPerWarp / kKPerThread;
static constexpr index_t kMPerBlock = BlockTile::at(number<0>{});
static constexpr index_t kKPerBlock = BlockTile::at(number<1>{});
static constexpr index_t kMWarpPerBlock = kMPerBlock / kMPerWarp;
static constexpr index_t kKWarpPerBlock = kKPerBlock / kKPerWarp;
static constexpr index_t kBlockSize = warpSize * kMWarpPerBlock * kKWarpPerBlock;
};
} // namespace ck_tile