* Copyright (c) 2025 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
* \file matmul_param.h
* \brief
*/
#if !defined(__ASCENDC_INCLUDE_INTERNAL_HEADERS__)
#pragma message( \
"impl/adv_api/detail/matmul/utils/matmul_param.h is an internal header file and must not be used directly. Functions or variables defined in this file may be removed in the future. Please use \"#include \"adv_api/matmul/matmul.h\"\" and use public functions or variables defined in interface headers files.")
#define __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#define __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_DETAIL_MATMUL_UTILS_MATMUL_PARAM_H__
#endif
#ifndef IMPL_MATMUL_UTILS_MATMUL_PARAM_H
#define IMPL_MATMUL_UTILS_MATMUL_PARAM_H
#include "kernel_macros.h"
#include "kernel_basic_intf.h"
#include "include/adv_api/matmul/tiling.h"
#include "../dfx/dfx_proxy.h"
#include "../resource/cube_in_buffer/global_cache.h"
#include "../../../tiling/matmul/matmul_constant_tiling_impl.h"
#include "matmul_type_def.h"
#include "matmul_utils.h"
#include "matmul_config_utils.h"
namespace AscendC {
namespace Impl {
namespace Detail {
* MatmulParamsBase *
* ************************************************************************************************* */
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsBase {
__aicore__ inline MatmulParamsBase(){};
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG, typename = void>
struct MatmulParamsNorm : public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsNorm(){};
using SrcT = typename A_TYPE::T;
using SrcBT = typename B_TYPE::T;
using DstT = typename C_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
TQue<TPosition::C1, QUEUE_DEPTH> qidBias_;
#if defined(__NPU_ARCH__) && __NPU_ARCH__ == 1001
TQue<TPosition::A2, QUEUE_DEPTH> qidA2_;
TQue<TPosition::B2, QUEUE_DEPTH> qidB2_;
#endif
LocalTensor<BiasT> cacheHeadBias_;
SrcT aScalar_;
SrcBT bScalar_;
LocalTensor<TensorTrait<SrcT>> leftMatrix_;
LocalTensor<TensorTrait<SrcBT>> rightMatrix_;
LocalTensor<TensorTrait<BiasT>> inputBias_;
__gm__ SrcT* aGlobal_;
__gm__ SrcBT* bGlobal_;
__gm__ BiasT* biasGlobal_;
TPipe* tpipe_;
int baseUseM_;
int baseUseK_;
int baseUseN_;
int blockUseM_;
int blockUseK_;
int blockUseN_;
bool isFirstIter_;
bool enableBias_;
int tailM_, tailK_, tailN_;
int curM_, curN_;
int curStepM_, curStepN_;
int stepMIdx_, stepNIdx_;
bool enHF32Mode_;
int32_t hf32TransMode_;
int baseMN_;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsNorm<
A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG,
enable_if_t<isNormEnableScheduler<A_TYPE, MM_CFG> || IsBasicBlockEnable<MM_CFG> || IsIntrablock<MM_CFG>>>
: public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsNorm(){};
using SrcT = typename A_TYPE::T;
using SrcBT = typename B_TYPE::T;
using DstT = typename C_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
TPipe* tpipe_;
int32_t hf32TransMode_;
int baseMN_;
bool enHF32Mode_;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsMDL : public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsMDL(){};
using SrcT = typename A_TYPE::T;
using SrcBT = typename B_TYPE::T;
using DstT = typename C_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
TPipe* tpipe_;
bool enHF32Mode_;
int32_t hf32TransMode_;
int baseMN_;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsMDLSparse : public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsMDLSparse(){};
using SrcT = typename A_TYPE::T;
using SrcBT = typename B_TYPE::T;
using DstT = typename C_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
TPipe* tpipe_;
bool enHF32Mode_;
int32_t hf32TransMode_;
int baseMN_;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsMxNorm : public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsMxNorm(){};
using SrcBT = typename B_TYPE::T;
using SrcT = typename A_TYPE::T;
using DstT = typename C_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
TPipe* tpipe_;
bool enHF32Mode_;
int32_t hf32TransMode_;
int baseMN_;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsMxMDL : public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsMxMDL(){};
using SrcT = typename A_TYPE::T;
using SrcBT = typename B_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
using DstT = typename C_TYPE::T;
TPipe* tpipe_;
bool enHF32Mode_;
int32_t hf32TransMode_;
int baseMN_;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsBasicBlock : public MatmulParamsNorm<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
__aicore__ inline MatmulParamsBasicBlock(){};
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParamsIBShareNorm : public MatmulParamsBase<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG> {
using L0cT = typename GetMmDstType<typename A_TYPE::T>::Type;
__aicore__ inline MatmulParamsIBShareNorm(){};
using SrcT = typename A_TYPE::T;
using SrcBT = typename B_TYPE::T;
using DstT = typename C_TYPE::T;
using BiasT = typename BIAS_TYPE::T;
#if defined(__NPU_ARCH__) && __NPU_ARCH__ == 1001
TQue<TPosition::A2, QUEUE_DEPTH> qidA2_;
TQue<TPosition::B2, QUEUE_DEPTH> qidB2_;
#endif
TPipe* tpipe_;
bool enHF32Mode_;
int32_t hf32TransMode_;
int baseMN_;
};
* MatmulParams *
* ************************************************************************************************* */
template <
class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG, MatmulMode MM_VER,
class ENABLE_SPARSE = void>
struct MatmulParams {
__aicore__ inline MatmulParams(){};
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<
A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(CFG_NORM),
enable_if_t<!HasScalePosition<A_TYPE>::value && !HasScalePosition<B_TYPE>::value>> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsNorm<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<
A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(CFG_MDL),
enable_if_t<!HasSparseIndex<B_TYPE>() && (!HasScalePosition<A_TYPE>::value && !HasScalePosition<B_TYPE>::value)>> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsMDL<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<
A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(CFG_MDL),
enable_if_t<
HasSparseIndex<B_TYPE>() && DoMatmulMDL(MM_CFG) &&
(!HasScalePosition<A_TYPE>::value && !HasScalePosition<B_TYPE>::value)>> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsMDLSparse<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<
A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(CFG_NORM),
enable_if_t<DoMatmulNorm(MM_CFG) && HasScalePosition<A_TYPE>::value && HasScalePosition<B_TYPE>::value>> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsMxNorm<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<
A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(CFG_MDL),
enable_if_t<HasScalePosition<A_TYPE>::value && HasScalePosition<B_TYPE>::value>> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsMxMDL<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(MM_CFG_BB)> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsBasicBlock<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const auto& MM_CFG>
struct MatmulParams<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(CFG_IBSHARE_NORM)> {
__aicore__ inline MatmulParams(){};
using PARAMS = MatmulParamsIBShareNorm<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG>;
};
}
}
}
#endif
#if defined(__UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_DETAIL_MATMUL_UTILS_MATMUL_PARAM_H__)
#undef __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#undef __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_DETAIL_MATMUL_UTILS_MATMUL_PARAM_H__
#endif