Line data    Source code

       1              : /**
       2              :  * Copyright (c) 2025 Huawei Technologies Co., Ltd.
       3              :  * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
       4              :  * CANN Open Software License Agreement Version 2.0 (the "License").
       5              :  * Please refer to the License for details. You may not use this file except in compliance with the License.
       6              :  * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
       7              :  * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
       8              :  * See LICENSE in the root of the software repository for the full text of the License.
       9              :  */
      10              : 
      11              : #ifndef COLL_ALL_REDUCE_RING_FOR_910_93_EXECUTOR_H
      12              : #define COLL_ALL_REDUCE_RING_FOR_910_93_EXECUTOR_H
      13              : 
      14              : #include "coll_all_reduce_executor.h"
      15              : 
      16              : namespace hccl {
      17              : class CollAllReduceRingFor91093Executor : public CollAllReduceExecutor {
      18              : 
      19              : public:
      20              :     CollAllReduceRingFor91093Executor(const HcclDispatcher dispatcher, std::unique_ptr<TopoMatcher> &topoMatcher);
      21            6 :     ~CollAllReduceRingFor91093Executor() override = default;
      22              : 
      23              : private:
      24              :     /* *************** 资源计算 *************** */
      25              :     HcclResult CalcStreamNum(u32& streamNum) override;
      26              :     HcclResult CalcCommInfo(std::vector<LevelNSubCommTransport>& opTransport) override;
      27              :     HcclResult CalcLevel0CommInfo(TransportMemType inputType,
      28              :         TransportMemType outputType,
      29              :         std::vector<LevelNSubCommTransport>& opTransport) override;
      30              :     HcclResult CalcLevel2CommInfo(TransportMemType inputType,
      31              :         TransportMemType outputType,
      32              :         std::vector<LevelNSubCommTransport>& opTransport) override;
      33              :     HcclResult CalcTransportMemType(TransportMemType &inputType, TransportMemType &outputType);
      34              : 
      35              :     /* *************** 算法编排 *************** */
      36              :     bool IsHugeData(const u64 curSize) override;
      37              :     bool IsSmallData(const u64 totalSize, const u64 curSize) override;
      38              :     virtual HcclResult RunIntraSeverReduceScatter(const std::string &tag, DeviceMem &inputMem, DeviceMem &outputMem,
      39              :         const u64 count, const HcclDataType &dataType, const HcclReduceOp &reductionOp,
      40              :         const std::vector<std::vector<Slice>> &multRingsSliceZero, const Stream &stream, s32 profStage,
      41              :         const u64 baseOffset = 0, const HcomCollOpInfo *opInfo = nullptr,
      42              :         const std::vector<std::vector<Slice>> &multRingsUserMemSlice = std::vector<std::vector<Slice>>(0),
      43              :         const bool disableDMAReduce = false);
      44              :     virtual HcclResult RunIntraSeverAllGather(const std::string &tag, DeviceMem &inputMem, DeviceMem &outputMem,
      45              :         const u64 count, const HcclDataType &dataType,
      46              :         const std::vector<std::vector<Slice>> &multRingsSliceZero, const Stream &stream,
      47              :         s32 profStage, const u64 baseOffset = 0, const HcomCollOpInfo *opInfo = nullptr,
      48              :         const std::vector<std::vector<Slice>> &multRingsUserMemSlice = std::vector<std::vector<Slice>> (0));
      49              :     HcclResult KernelRun(const OpParam &param, ExecMem &execMem) override;
      50              :     HcclResult Getlevel1CommRank(SubCommInfo& level1CommInfo) override;
      51              :     HcclResult SelectTempAlg(std::unique_ptr<AlgTemplateBase> &level1TempAlg, u32 level1RankSize) override;
      52              :     virtual HcclResult GetLevelCommInfo();
      53              :     HcclResult GetNicList(std::vector<u32> &mockNicList);
      54              :     HcclResult PrepareARSLevel1CommInfo(u32 &segmentIdx, u32 &commIndex, u64 &hdSize,
      55              :                                                   const SubCommInfo &commInfo,
      56              :                                                   const std::vector<std::vector<Slice>> &multRingsSliceZero,
      57              :                                                   const std::string &tag, const std::vector<u32>& nicList);
      58              : protected:
      59              :     SubCommInfo logicalLevel0CommInfo_ = {0, 0, {}, {}};
      60              :     SubCommInfo logicalLevel1CommInfo_ = {0, 0, {}, {}};
      61              :     CommPlane logicalLevel0plane_ = COMM_LEVEL_RESERVED;
      62              :     CommPlane logicalLevel1plane_ = COMM_LEVEL_RESERVED;
      63              : };
      64              : 
      65              : } // namespace hccl
      66              : 
      67              : #endif