Skip to content

GitLab

Unverified Commit ec7c2e91 authored by Chao Liu's avatar Chao Liu Committed by GitHub
Browse files

Code refactor (#175) 

* format

* improving pipeline

* fix typo

* format

* adding thread group

* adding thread group

* adding thread group

* adding gemm pipeline

* tweak

* refactor

* refactor

* add missing type convert

* refactor

* refactor

* refactor

* clean

* fix build

* refactor

* format

* clean up

* use remove_cvref_t

* clean

* clean up

* clean up

* clean up
parent a3c910ac
Hide whitespace changes
Inline Side-by-side
Showing with 60 additions and 39 deletions
include/ck/utility/amd_xdlops.hpp
View file @ ec7c2e91
...@@ -266,8 +266,8 @@ struct intrin_mfma_i32_32x32x8i8<32, 32> ...@@ -266,8 +266,8 @@ struct intrin_mfma_i32_32x32x8i8<32, 32>
__device__ static void Run(const int8x4_t& reg_a, const int8x4_t& reg_b, FloatC& reg_c) __device__ static void Run(const int8x4_t& reg_a, const int8x4_t& reg_b, FloatC& reg_c)
{ {
reg_c.template AsType<int32x16_t>()(Number<0>{}) = reg_c.template AsType<int32x16_t>()(Number<0>{}) =
__builtin_amdgcn_mfma_i32_32x32x8i8(bit_cast<int>(reg_a), __builtin_amdgcn_mfma_i32_32x32x8i8(bit_cast<int32_t>(reg_a),
bit_cast<int>(reg_b), bit_cast<int32_t>(reg_b),
reg_c.template AsType<int32x16_t>()[Number<0>{}], reg_c.template AsType<int32x16_t>()[Number<0>{}],
0, 0,
0, 0,
...@@ -285,8 +285,8 @@ struct intrin_mfma_i32_16x16x16i8<16, 16> ...@@ -285,8 +285,8 @@ struct intrin_mfma_i32_16x16x16i8<16, 16>
__device__ static void Run(const int8x4_t& reg_a, const int8x4_t& reg_b, FloatC& reg_c) __device__ static void Run(const int8x4_t& reg_a, const int8x4_t& reg_b, FloatC& reg_c)
{ {
reg_c.template AsType<int32x4_t>()(Number<0>{}) = reg_c.template AsType<int32x4_t>()(Number<0>{}) =
__builtin_amdgcn_mfma_i32_16x16x16i8(bit_cast<int>(reg_a), __builtin_amdgcn_mfma_i32_16x16x16i8(bit_cast<int32_t>(reg_a),
bit_cast<int>(reg_b), bit_cast<int32_t>(reg_b),
reg_c.template AsType<int32x4_t>()[Number<0>{}], reg_c.template AsType<int32x4_t>()[Number<0>{}],
0, 0,
0, 0,
......
include/ck/utility/common_header.hpp
View file @ ec7c2e91
...@@ -28,6 +28,7 @@ ...@@ -28,6 +28,7 @@
#include "transpose_vectors.hpp" #include "transpose_vectors.hpp"
#include "inner_product.hpp" #include "inner_product.hpp"
#include "element_wise_operation.hpp" #include "element_wise_operation.hpp"
#include "thread_group.hpp"
#include "debug.hpp" #include "debug.hpp"
#include "amd_buffer_addressing.hpp" #include "amd_buffer_addressing.hpp"
......
include/ck/utility/get_id.hpp
View file @ ec7c2e91
...@@ -3,11 +3,15 @@ ...@@ -3,11 +3,15 @@
namespace ck { namespace ck {
__device__ constexpr index_t get_wave_size() { return CK_GPU_WAVE_SIZE; } __host__ __device__ constexpr index_t get_warp_size()
{
// warpSize is defined by HIP
return warpSize;
}
__device__ index_t get_thread_local_1d_id() { return threadIdx.x; } __device__ index_t get_thread_local_1d_id() { return threadIdx.x; }
__device__ index_t get_wave_local_1d_id() { return threadIdx.x / get_wave_size(); } __device__ index_t get_warp_local_1d_id() { return threadIdx.x / get_warp_size(); }
__device__ index_t get_block_1d_id() { return blockIdx.x; } __device__ index_t get_block_1d_id() { return blockIdx.x; }
......
include/ck/utility/thread_group.hpp 0 → 100644
View file @ ec7c2e91
#pragma once
#include "get_id.hpp"
namespace ck {
template <index_t ThreadPerBlock>
struct ThisThreadBlock
{
static constexpr index_t kNumThread_ = ThreadPerBlock;
__device__ static constexpr index_t GetNumOfThread() { return kNumThread_; }
__device__ static constexpr bool IsBelong() { return true; }
__device__ static index_t GetThreadId() { return get_thread_local_1d_id(); }
};
} // namespace ck
include/ck/utility/tuple.hpp
View file @ ec7c2e91
...@@ -21,9 +21,9 @@ struct TupleElement ...@@ -21,9 +21,9 @@ struct TupleElement
{ {
__host__ __device__ constexpr TupleElement() = default; __host__ __device__ constexpr TupleElement() = default;
template <typename T, template <
typename enable_if<!is_same<remove_reference_t<remove_cv_t<T>>, TupleElement>::value, typename T,
bool>::type = false> typename enable_if<!is_same<remove_cvref_t<T>, TupleElement>::value, bool>::type = false>
__host__ __device__ constexpr TupleElement(T&& v) : mData(std::forward<T>(v)) __host__ __device__ constexpr TupleElement(T&& v) : mData(std::forward<T>(v))
{ {
} }
...@@ -60,7 +60,7 @@ struct TupleImpl<Sequence<Is...>, Xs...> : TupleElement<TupleElementKey<Is>, Xs> ...@@ -60,7 +60,7 @@ struct TupleImpl<Sequence<Is...>, Xs...> : TupleElement<TupleElementKey<Is>, Xs>
template <typename Y, template <typename Y,
typename enable_if<sizeof...(Is) == 1 && sizeof...(Xs) == 1 && typename enable_if<sizeof...(Is) == 1 && sizeof...(Xs) == 1 &&
!is_same<remove_reference_t<remove_cv_t<Y>>, TupleImpl>::value, !is_same<remove_cvref_t<Y>, TupleImpl>::value,
bool>::type = false> bool>::type = false>
__host__ __device__ constexpr TupleImpl(Y&& y) __host__ __device__ constexpr TupleImpl(Y&& y)
: TupleElement<TupleElementKey<Is>, Xs>(std::forward<Y>(y))... : TupleElement<TupleElementKey<Is>, Xs>(std::forward<Y>(y))...
...@@ -101,8 +101,7 @@ struct Tuple : detail::TupleImpl<typename arithmetic_sequence_gen<0, sizeof...(X ...@@ -101,8 +101,7 @@ struct Tuple : detail::TupleImpl<typename arithmetic_sequence_gen<0, sizeof...(X
__host__ __device__ constexpr Tuple() = default; __host__ __device__ constexpr Tuple() = default;
template <typename Y, template <typename Y,
typename enable_if<sizeof...(Xs) == 1 && typename enable_if<sizeof...(Xs) == 1 && !is_same<remove_cvref_t<Y>, Tuple>::value,
!is_same<remove_reference_t<remove_cv_t<Y>>, Tuple>::value,
bool>::type = false> bool>::type = false>
__host__ __device__ constexpr Tuple(Y&& y) : base(std::forward<Y>(y)) __host__ __device__ constexpr Tuple(Y&& y) : base(std::forward<Y>(y))
{ {
......
library/include/ck/library/reference_tensor_operation/cpu/reference_gemm.hpp
View file @ ec7c2e91
#ifndef REFERENCE_GEMM_HPP #pragma once
#define REFERENCE_GEMM_HPP
#include <iostream> #include <iostream>
#include <sstream> #include <sstream>
#include "device_base.hpp" #include "device_base.hpp"
...@@ -129,4 +127,3 @@ struct ReferenceGemm : public device::BaseOperator ...@@ -129,4 +127,3 @@ struct ReferenceGemm : public device::BaseOperator
} // namespace host } // namespace host
} // namespace tensor_operation } // namespace tensor_operation
} // namespace ck } // namespace ck
#endif
library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
View file @ ec7c2e91
#include <stdlib.h> #include <stdlib.h>
#include "config.hpp" #include "config.hpp"
#include "device_gemm_xdl_c_shuffle.hpp" #include "device_gemm_xdl_cshuffle.hpp"
#include "element_wise_operation.hpp" #include "element_wise_operation.hpp"
#include "device_operation_instance.hpp" #include "device_operation_instance.hpp"
...@@ -20,26 +20,28 @@ using S = ck::Sequence<Is...>; ...@@ -20,26 +20,28 @@ using S = ck::Sequence<Is...>;
using PassThrough = ck::tensor_operation::element_wise::PassThrough; using PassThrough = ck::tensor_operation::element_wise::PassThrough;
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
// Compilation parameters for a[m, k] * b[n, k] = c[m, n] // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances = std::tuple< using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances = std::tuple<
// clang-format off // clang-format off
//#####################| AData| BData| CData| AccData| CShuffle| ALayout| BLayout| CLayout| A| B| C| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| Num| //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer|
//#####################| Type| Type| Type| Type| DataType| | | | Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector| Prefetch| //#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector|
//#####################| | | | | | | | | Operation| Operation| Operation| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl| _NWaveNPerXdl| | //#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl|
//#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | //#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 16, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 4>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 16, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 64, 64, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 16, 1, 1, 4>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 64, 64, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 128, 128, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 32, 1, 1, 4>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 128, 128, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 128, 32, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 16, 1, 1, 8>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 128, 32, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 64, 64, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 16, 1, 1, 4>, 8, 2>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 64, 64, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8>,
DeviceGemmXdl_C_Shuffle< F16, F16, F16, F32, F16, Row, Col, Row, PassThrough, PassThrough, PassThrough, 64, 32, 64, 32, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, 1, 1, S<1, 1, 16, 1, 1, 4>, 8, 2> DeviceGemm_Xdl_CShuffle< Row, Col, Row, F16, F16, F16, F32, F16, PassThrough, PassThrough, PassThrough, GemmDefault, 2, 64, 32, 64, 32, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8>
// clang-format on // clang-format on
>; >;
......
profiler/include/profile_gemm_impl.hpp
View file @ ec7c2e91
...@@ -421,7 +421,7 @@ void profile_gemm_impl(int do_verification, ...@@ -421,7 +421,7 @@ void profile_gemm_impl(int do_verification,
std::size_t flop = std::size_t(2) * M * N * K; std::size_t flop = std::size_t(2) * M * N * K;
std::size_t num_btype = std::size_t num_btype =
sizeof(ADataType) * M * K + sizeof(BDataType) * K * M + sizeof(CDataType) * M * N; sizeof(ADataType) * M * K + sizeof(BDataType) * K * N + sizeof(CDataType) * M * N;
float tflops = static_cast<float>(flop) / 1.E9 / ave_time; float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
......
test/gemm/gemm_bf16.cpp
View file @ ec7c2e91
...@@ -15,7 +15,7 @@ ...@@ -15,7 +15,7 @@
#include "host_gemm.hpp" #include "host_gemm.hpp"
#include "device_tensor.hpp" #include "device_tensor.hpp"
#include "device_gemm_xdl.hpp" #include "device_gemm_xdl.hpp"
#include "device_gemm_xdl_c_shuffle.hpp" #include "device_gemm_xdl_cshuffle.hpp"
#include "element_wise_operation.hpp" #include "element_wise_operation.hpp"
#include "reference_gemm.hpp" #include "reference_gemm.hpp"
#include "gemm_specialization.hpp" #include "gemm_specialization.hpp"
......
test/gemm/gemm_fp16.cpp
View file @ ec7c2e91
...@@ -13,7 +13,7 @@ ...@@ -13,7 +13,7 @@
#include "host_gemm.hpp" #include "host_gemm.hpp"
#include "device_tensor.hpp" #include "device_tensor.hpp"
#include "device_gemm_xdl.hpp" #include "device_gemm_xdl.hpp"
#include "device_gemm_xdl_c_shuffle.hpp" #include "device_gemm_xdl_cshuffle.hpp"
#include "element_wise_operation.hpp" #include "element_wise_operation.hpp"
#include "gemm_specialization.hpp" #include "gemm_specialization.hpp"
......
test/gemm/gemm_fp32.cpp
View file @ ec7c2e91
...@@ -15,7 +15,7 @@ ...@@ -15,7 +15,7 @@
#include "host_gemm.hpp" #include "host_gemm.hpp"
#include "device_tensor.hpp" #include "device_tensor.hpp"
#include "device_gemm_xdl.hpp" #include "device_gemm_xdl.hpp"
#include "device_gemm_xdl_c_shuffle.hpp" #include "device_gemm_xdl_cshuffle.hpp"
#include "element_wise_operation.hpp" #include "element_wise_operation.hpp"
#include "reference_gemm.hpp" #include "reference_gemm.hpp"
#include "gemm_specialization.hpp" #include "gemm_specialization.hpp"
......
test/gemm/gemm_int8.cpp
View file @ ec7c2e91
...@@ -15,7 +15,7 @@ ...@@ -15,7 +15,7 @@
#include "host_gemm.hpp" #include "host_gemm.hpp"
#include "device_tensor.hpp" #include "device_tensor.hpp"
#include "device_gemm_xdl.hpp" #include "device_gemm_xdl.hpp"
#include "device_gemm_xdl_c_shuffle.hpp" #include "device_gemm_xdl_cshuffle.hpp"
#include "element_wise_operation.hpp" #include "element_wise_operation.hpp"
#include "reference_gemm.hpp" #include "reference_gemm.hpp"
#include "gemm_specialization.hpp" #include "gemm_specialization.hpp"
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment