* 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 kernel_operator_mm_load2d_impl.h
* \brief
*/
#if !defined(__ASCENDC_INCLUDE_INTERNAL_HEADERS__)
#pragma message("impl/basic_api/kernel_operator_mm_load2d_impl.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 \"basic_api/kernel_operator_mm_intf.h\"\" and use public functions or variables defined in interface headers files.")
#define __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#define __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_KERNEL_OPERATOR_MM_LOAD2D_IMPL_H__
#endif
#ifndef ASCENDC_MODULE_OPERATOR_MM_LOAD2D_H
#define ASCENDC_MODULE_OPERATOR_MM_LOAD2D_H
#include "kernel_struct_mm.h"
namespace AscendC {
* LoadData 2d *
* ************************************************************************************************* */
* @ingroup DataLoad
* @brief Cube data loading
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor
* @param [in] loadDataParams.startIndex Fractal matrix ID
* @param [in] loadDataParams.repeatTimes repeat times
* @param [in] loadDataParams.srcStride src block stride
* @param [in] loadDataParams.sid SMMU SID
* @param [in] loadDataParams.dstGap interval between the previous tail and the next fractal head
* @param [in] loadDataParams.ifTranspose enable parameters of transpose function
*/
template <typename T>
__aicore__ inline void LoadDataImpl(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LoadData2DParams& loadDataParams)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFuncLoadData2d(dst, src, loadDataParams, "LoadData with LoadData2DParams")) {
ASCENDC_REPORT_CHECK_ERROR("LoadData with LoadData2DParams", KernelFuncType::NONE_MODE);
}
#endif
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
if (dstScope == Hardware::L0A) {
LoadData2DL12L0ACal((__ca__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if (dstScope == Hardware::L0B) {
LoadData2DL12L0BCal((__cb__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
}
}
template <typename T>
__aicore__ inline __inout_pipe__(MTE2) void LoadDataImpl(const LocalTensor<T>& dst, const GlobalTensor<T>& src,
const LoadData2DParams& loadDataParams)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFuncLoadData2d(dst, src, loadDataParams, "LoadData with LoadData2DParams")) {
ASCENDC_REPORT_CHECK_ERROR("LoadData with LoadData2DParams", KernelFuncType::NONE_MODE);
}
#endif
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
if (dstScope == Hardware::L0A) {
LoadData2DGM2L0ACal((__ca__ PrimT<T>*)dst.GetPhyAddr(), (__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if (dstScope == Hardware::L0B) {
LoadData2DGM2L0BCal((__cb__ PrimT<T>*)dst.GetPhyAddr(), (__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if (dstScope == Hardware::L1) {
LoadData2DGM2L1Cal((__cbuf__ PrimT<T>*)dst.GetPhyAddr(), (__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
}
}
#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
template <TPosition Dst, TPosition Src, typename T>
__aicore__ inline void LoadDataImpl(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const Load2DBitModeParam& loadDataParams)
{
CheckTensorAlign<T>(src, ONE_BLK_SIZE, "src", "LoadData with LoadData2DParams");
CheckTensorAlign<T>(dst, VALUE_512, "dst", "LoadData with LoadData2DParams");
if constexpr (Src != TPosition::A1 && Src != TPosition::A2) {
ASCENDC_CHECK_TPOSITION(false, "src", "A1 / B1",
"LoadData with LoadDataBitModeParams",
ConstDefiner::Instance().logicNameMap.at(static_cast<uint8_t>(src.GetPosition())));
};
if constexpr (Dst == TPosition::A2) {
LoadData2DL12L0ACal((__ca__ PrimT<T>*)dst.GetPhyAddr(),
(__cbuf__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if constexpr (Dst == TPosition::B2) {
LoadData2DL12L0BCal((__cb__ PrimT<T>*)dst.GetPhyAddr(),
(__cbuf__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else {
ASCENDC_CHECK_TPOSITION(false, "dst", "A2 / B2",
"LoadData with LoadData2DParams",
ConstDefiner::Instance().logicNameMap.at(static_cast<uint8_t>(dst.GetPosition())));
}
}
#endif
* LoadData 2d with transpose *
* ************************************************************************************************* */
* @ingroup DataLoad
* @brief Cube data loading
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor
* @param [in] loadDataParams.startIndex Fractal matrix ID in unit of fractal nums depending on dtype
* @param [in] loadDataParams.repeatTimes repeat times
* @param [in] loadDataParams.srcStride src block stride in unit of fractal nums depending on dtype
* @param [in] loadDataParams.dstGap interval between the previous tail and the next fractal head in unit of one 512byte
* fractal
* @param [in] loadDataParams.dstFracGap dst fractal gap in unit of one 512byte fractal
*/
template <typename T>
__aicore__ inline void LoadDataWithTransposeImpl(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LoadData2dTransposeParams& loadDataParams)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFuncLoadDataTranspose(dst, src, loadDataParams, "LoadDataWithTranspose")) {
ASCENDC_REPORT_CHECK_ERROR("LoadDataWithTranspose with LoadData2dTransposeParams", KernelFuncType::NONE_MODE);
}
#endif
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
if (dstScope == Hardware::L0A) {
LoadData2DL12L0ATransposeCal((__ca__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(),
loadDataParams);
} else if (dstScope == Hardware::L0B) {
LoadData2DL12L0BTransposeCal((__cb__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(),
loadDataParams);
}
}
* @ingroup DataLoad
* @brief Cube data loading
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor
* @param [in] loadDataParams.startIndex Fractal matrix ID in unit of one 512byte fractal
* @param [in] loadDataParams.repeatTimes repeat times
* @param [in] loadDataParams.srcStride src block stride in unit of one 512byte fractal
* @param [in] loadDataParams.dstGap interval between the previous tail and the next fractal head in unit of one 512byte
* fractal
* @param [in] loadDataParams.dstFracGap dst fractal gap in unit of one 512byte fractal
* @param [in] loadDataParams.srcFracGap src fractal gap in unit of one 512byte fractal
*/
template <typename T>
__aicore__ inline void LoadDataWithTransposeImpl(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LoadData2dTransposeParamsV2& loadDataParams)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFuncLoadDataTranspose(dst, src, loadDataParams, "LoadDataWithTranspose")) {
ASCENDC_REPORT_CHECK_ERROR("LoadDataWithTranspose with LoadData2dTransposeParamsV2", KernelFuncType::NONE_MODE);
}
#endif
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
if (dstScope == Hardware::L0B) {
LoadData2DL12L0BTransposeCal((__cb__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(),
loadDataParams);
}
}
* LoadData 2dV2 *
* ************************************************************************************************* */
* @ingroup DataLoad
* @brief Cube data loading
* @param [out] dst output LocalTensor
* @param [in] src input LocalTensor/GlobalTensor
* @param [in] loadDataParams.mStartPosition m start position
* @param [in] loadDataParams.kStartPosition k start position
* @param [in] loadDataParams.srcStride src block stride
* @param [in] loadDataParams.dstStride dst block stride
* @param [in] loadDataParams.mStep m step
* @param [in] loadDataParams.kStep k step
* @param [in] loadDataParams.sid SMMU SID
* @param [in] loadDataParams.ifTranspose enable parameters of transpose function
*/
template <typename T>
__aicore__ inline void LoadDataImpl(const LocalTensor<T>& dst, const LocalTensor<T>& src,
const LoadData2DParamsV2& loadDataParams)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFuncLoadData2dv2(dst, src, loadDataParams, "LoadData with LoadData2DParamsV2")) {
ASCENDC_REPORT_CHECK_ERROR("LoadData with LoadData2DParamsV2", KernelFuncType::NONE_MODE);
}
#endif
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
if (dstScope == Hardware::L0A) {
LoadData2DL12L0ACal((__ca__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if (dstScope == Hardware::L0B) {
LoadData2DL12L0BCal((__cb__ PrimT<T>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
}
}
#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
template <typename T, typename U>
__aicore__ inline void LoadDataImpl(const LocalTensor<U>& dst, const LocalTensor<T>& src,
const LocalTensor<fp8_e8m0_t>& srcMx, const LoadData2DParamsV2& loadDataParams,
const LoadData2DMxParams& loadMxDataParams)
{
CheckTensorPos<T>(src, Hardware::L1, "src", "A1 / B1",
"LoadData with LoadData2DParamsV2");
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
if (dstScope == Hardware::L0A) {
LoadData2DL12L0ACal((__ca__ PrimT<U>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(),
(__cbuf__ fp8_e8m0_t*)srcMx.GetPhyAddr(), loadDataParams, loadMxDataParams);
} else if (dstScope == Hardware::L0B) {
LoadData2DL12L0BCal((__cb__ PrimT<U>*)dst.GetPhyAddr(), (__cbuf__ PrimT<T>*)src.GetPhyAddr(),
(__cbuf__ fp8_e8m0_t*)srcMx.GetPhyAddr(), loadDataParams, loadMxDataParams);
} else {
ASSERT(false);
}
}
#endif
template <typename T>
__aicore__ inline __inout_pipe__(MTE2) void LoadDataImpl(const LocalTensor<T>& dst, const GlobalTensor<T>& src,
const LoadData2DParamsV2& loadDataParams)
{
#if ASCENDC_CPU_DEBUG
if (!CheckFuncLoadData2dv2(dst, src, loadDataParams, "LoadData with LoadData2DParamsV2")) {
ASCENDC_REPORT_CHECK_ERROR("LoadData with LoadData2DParamsV2", KernelFuncType::NONE_MODE);
}
#endif
const Hardware dstScope = GetPhyType((TPosition)dst.GetPosition());
#if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
const uint8_t cacheMode = ExtractCacheMode(src);
if (dstScope == Hardware::L0A) {
LoadData2DGM2L0ACal((__ca__ PrimT<T>*)dst.GetPhyAddr(),
(__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams, cacheMode);
} else if (dstScope == Hardware::L0B) {
LoadData2DGM2L0BCal((__cb__ PrimT<T>*)dst.GetPhyAddr(),
(__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams, cacheMode);
} else if (dstScope == Hardware::L1) {
LoadData2DGM2L1Cal((__cbuf__ PrimT<T>*)dst.GetPhyAddr(),
(__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams, cacheMode);
} else {
ASCENDC_CHECK_TPOSITION(false, "dst", "A1 / B1 / A2 / B2",
"LoadData with LoadData2DParamsV2",
ConstDefiner::Instance().logicNameMap.at(static_cast<uint8_t>(dst.GetPosition())));
}
#else
if (dstScope == Hardware::L0A) {
LoadData2DGM2L0ACal((__ca__ PrimT<T>*)dst.GetPhyAddr(), (__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if (dstScope == Hardware::L0B) {
LoadData2DGM2L0BCal((__cb__ PrimT<T>*)dst.GetPhyAddr(), (__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
} else if (dstScope == Hardware::L1) {
LoadData2DGM2L1Cal((__cbuf__ PrimT<T>*)dst.GetPhyAddr(), (__gm__ PrimT<T>*)src.GetPhyAddr(), loadDataParams);
}
#endif
}
}
#endif
#if defined(__UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_KERNEL_OPERATOR_MM_LOAD2D_IMPL_H__)
#undef __ASCENDC_INCLUDE_INTERNAL_HEADERS__
#undef __UNDEF_ASCENDC_INCLUDE_INTERNAL_HEADERS_KERNEL_OPERATOR_MM_LOAD2D_IMPL_H__
#endif
