ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-02 04:31:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

CK Tile GEMM CICD fixed & register block method refactor (#1776)

Browse Source 
* refactor the block_gemm_areg_breg_creg_v1 and add the v2 policy with 2x2 warp gemm

* Finished the 2x2 warp gemm policy and the block selection mechanism

* Clang format

* address poyen's comment

* Address feedbacks

* Fixed the compilation issue

* Change the function name
This commit is contained in:
Thomas Ning
2025-01-12 21:10:44 -08:00
committed by GitHub
parent 0b8f117f1a
commit 5d671a5fc4
8 changed files with 109 additions and 107 deletions
Download Patch File Download Diff File Expand all files Collapse all files

120
include/ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_v1.hpp
View File

@@ -21,7 +21,62 @@ struct BlockGemmARegBRegCRegV1
using CDataType = remove_cvref_t<typename Problem::CDataType>;
using BlockGemmShape = remove_cvref_t<typename Problem::BlockGemmShape>;
static constexpr index_t kBlockSize = Problem::kBlockSize;
static constexpr index_t kBlockSize = Problem::kBlockSize;
static constexpr index_t MPerBlock = BlockGemmShape::kM;
static constexpr index_t NPerBlock = BlockGemmShape::kN;
static constexpr index_t KPerBlock = BlockGemmShape::kK;
static constexpr auto config = Policy::template GetWarpGemmMWarpNWarp<Problem>();
using WG = remove_cvref_t<decltype(config.template at<0>())>;
static constexpr index_t MWarp = config.template at<1>();
static constexpr index_t NWarp = config.template at<2>();
static constexpr index_t MIterPerWarp = MPerBlock / (MWarp * WG::kM);
static constexpr index_t NIterPerWarp = NPerBlock / (NWarp * WG::kN);
static constexpr index_t KIterPerWarp = KPerBlock / WG::kK;
CK_TILE_DEVICE static constexpr auto MakeABlockDistributionEncode()
{
constexpr auto a_block_outer_dstr_encoding =
tile_distribution_encoding<sequence<NWarp>,
tuple<sequence<MIterPerWarp, MWarp>, sequence<KIterPerWarp>>,
tuple<sequence<1, 0>>,
tuple<sequence<1, 0>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto a_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
a_block_outer_dstr_encoding, typename WG::AWarpDstrEncoding{});
return a_block_dstr_encode;
}
CK_TILE_DEVICE static constexpr auto MakeBBlockDistributionEncode()
{
constexpr auto b_block_outer_dstr_encoding =
tile_distribution_encoding<sequence<MWarp>,
tuple<sequence<NIterPerWarp, NWarp>, sequence<KIterPerWarp>>,
tuple<sequence<0, 1>>,
tuple<sequence<0, 1>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto b_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
b_block_outer_dstr_encoding, typename WG::BWarpDstrEncoding{});
return b_block_dstr_encode;
}
CK_TILE_DEVICE static constexpr auto MakeCBlockDistributionEncode()
{
constexpr auto c_block_outer_dstr_encoding = tile_distribution_encoding<
sequence<>,
tuple<sequence<MIterPerWarp, MWarp>, sequence<NIterPerWarp, NWarp>>,
tuple<sequence<1, 2>>,
tuple<sequence<1, 1>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto c_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
c_block_outer_dstr_encoding, typename WG::CWarpDstrEncoding{});
return c_block_dstr_encode;
}
// C += A * B
template <typename CBlockTensor, typename ABlockTensor, typename BBlockTensor>
@@ -34,54 +89,11 @@ struct BlockGemmARegBRegCRegV1
std::is_same_v<CDataType, remove_cv_t<typename CBlockTensor::DataType>>,
"wrong!");
constexpr index_t MPerBlock = BlockGemmShape::kM;
constexpr index_t NPerBlock = BlockGemmShape::kN;
constexpr index_t KPerBlock = BlockGemmShape::kK;
constexpr auto a_block_dstr_encode = MakeABlockDistributionEncode();
constexpr auto config = Policy::template GetWarpGemmMWarpNWarp<Problem>();
constexpr auto b_block_dstr_encode = MakeBBlockDistributionEncode();
using WG = remove_cvref_t<decltype(config.template at<0>())>;
constexpr index_t MWarp = config.template at<1>();
constexpr index_t NWarp = config.template at<2>();
constexpr index_t MIterPerWarp = MPerBlock / (MWarp * WG::kM);
constexpr index_t NIterPerWarp = NPerBlock / (NWarp * WG::kN);
constexpr index_t KIterPerWarp = KPerBlock / WG::kK;
// M->N Warp
constexpr auto a_block_outer_dstr_encoding =
tile_distribution_encoding<sequence<NWarp>,
tuple<sequence<MIterPerWarp, MWarp>, sequence<KIterPerWarp>>,
tuple<sequence<1, 0>>,
tuple<sequence<1, 0>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto b_block_outer_dstr_encoding =
tile_distribution_encoding<sequence<MWarp>,
tuple<sequence<NIterPerWarp, NWarp>, sequence<KIterPerWarp>>,
tuple<sequence<0, 1>>,
tuple<sequence<0, 1>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto c_block_outer_dstr_encoding = tile_distribution_encoding<
sequence<>,
tuple<sequence<MIterPerWarp, MWarp>, sequence<NIterPerWarp, NWarp>>,
tuple<sequence<1, 2>>,
tuple<sequence<1, 1>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto a_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
a_block_outer_dstr_encoding, typename WG::AWarpDstrEncoding{});
constexpr auto b_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
b_block_outer_dstr_encoding, typename WG::BWarpDstrEncoding{});
constexpr auto c_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
c_block_outer_dstr_encoding, typename WG::CWarpDstrEncoding{});
constexpr auto c_block_dstr_encode = MakeCBlockDistributionEncode();
// check ABC-block-distribution
static_assert(
@@ -159,20 +171,6 @@ struct BlockGemmARegBRegCRegV1
CK_TILE_DEVICE static constexpr auto MakeCBlockTile()
{
constexpr index_t MPerBlock = BlockGemmShape::kM;
constexpr index_t NPerBlock = BlockGemmShape::kN;
constexpr auto config = Policy::template GetWarpGemmMWarpNWarp<Problem>();
using WG = remove_cvref_t<decltype(config.template at<0>())>;
constexpr index_t MWarp = config.template at<1>();
constexpr index_t NWarp = config.template at<2>();
constexpr index_t MIterPerWarp = MPerBlock / (MWarp * WG::kM);
constexpr index_t NIterPerWarp = NPerBlock / (NWarp * WG::kN);
// constexpr index_t KIterPerWarp = KPerBlock / WG::kK;
constexpr auto c_block_outer_dstr_encoding = tile_distribution_encoding<
sequence<>,
tuple<sequence<MIterPerWarp, MWarp>, sequence<NIterPerWarp, NWarp>>,

7
include/ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_mem.hpp
View File

@@ -104,9 +104,10 @@ struct GemmPipelineAgBgCrMem : public BaseGemmPipelineAgBgCrMem<Problem>
using CLayout = remove_cvref_t<typename Problem::CLayout>;
using BlockGemm = remove_cvref_t<decltype(Policy::template GetBlockGemm<Problem>())>;
using I0 = number<0>;
using I1 = number<1>;
using I2 = number<2>;
using I0 = number<0>;
using I1 = number<1>;
using I2 = number<2>;
static constexpr index_t MPerBlock = BlockGemmShape::kM;
static constexpr index_t NPerBlock = BlockGemmShape::kN;

4
include/ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_v1.hpp
View File

@@ -23,6 +23,8 @@ struct GemmPipelineAGmemBGmemCRegV1
using BLayout = remove_cvref_t<typename Problem::BLayout>;
using CLayout = remove_cvref_t<typename Problem::CLayout>;
using BlockGemm = remove_cvref_t<decltype(Policy::template GetBlockGemm<Problem>())>;
static constexpr index_t BlockSize = Problem::kBlockSize;
static constexpr index_t kMPerBlock = BlockGemmShape::kM;
@@ -126,7 +128,7 @@ struct GemmPipelineAGmemBGmemCRegV1
b_lds_block, make_tuple(number<kNPerBlock>{}, number<kKPerBlock>{}), {0, 0});
// Block GEMM
auto block_gemm = Policy::template GetBlockGemm<Problem>();
auto block_gemm = BlockGemm();
// Acc register tile
auto c_block_tile = decltype(block_gemm(a_lds_gemm_window, b_lds_gemm_window)){};

9
include/ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_v1_default_policy.hpp
View File

@@ -12,8 +12,11 @@ namespace ck_tile {
// Default policy class should not be templated, put template on member functions instead
struct GemmPipelineAGmemBGmemCRegV1DefaultPolicy
{
static constexpr auto I0 = number<0>{};
static constexpr auto I1 = number<1>{};
static constexpr auto I2 = number<2>{};
static constexpr bool TransposeC = false;
static constexpr bool TransposeC = true;
#if 0
// 2d
@@ -491,10 +494,6 @@ struct GemmPipelineAGmemBGmemCRegV1DefaultPolicy
template <typename Problem>
CK_TILE_HOST_DEVICE static constexpr auto GetBlockGemm()
{
constexpr auto I0 = number<0>{};
constexpr auto I1 = number<1>{};
constexpr auto I2 = number<2>{};
using AccDataType = float;
using BlockWarps = typename Problem::BlockGemmShape::BlockWarps;
using WarpTile = typename Problem::BlockGemmShape::WarpTile;

1
include/ck_tile/ops/gemm/pipeline/gemm_universal_pipeline_ag_bg_cr_policy.hpp
View File

@@ -11,7 +11,6 @@ namespace ck_tile {
// UniversalGemm Policy
struct UniversalGemmPipelineAgBgCrPolicy
{
static constexpr auto I0 = number<0>{};
static constexpr auto I1 = number<1>{};
static constexpr auto I2 = number<2>{};