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_ALLGATHER_EXECUTOR_H
      12              : #define COLL_ALLGATHER_EXECUTOR_H
      13              : #include "coll_comm_executor.h"
      14              : namespace hccl {
      15              : class CollAllGatherExecutor : public CollCommExecutor {
      16              : 
      17              : public:
      18              :     explicit CollAllGatherExecutor(const HcclDispatcher dispatcher, std::unique_ptr<TopoMatcher> &topoMatcher);
      19           15 :     ~CollAllGatherExecutor() override = default;
      20              : 
      21              :     HcclResult Orchestrate(OpParam& param, AlgResourceResponse& algRes) override;
      22              : protected:
      23              :     // AllGather Loop Executor公共接口
      24              :     virtual u64 CalcLoopMaxCount(const u64 cclBuffSize, const u32 unitSize);
      25              :     virtual bool IsHugeData(const u64 curSize);
      26              :     virtual bool IsSmallData(const u64 size);
      27              :     virtual u64 CalcTotalCount(const OpParam &param) const;
      28              :     HcclResult RunLoop(OpParam &param, AlgResourceResponse &algRes);    // non-virtual
      29              : 
      30              :     // agv
      31              :     virtual std::vector<u64> GetCounts(const OpParam &param) const;
      32              :     virtual std::vector<u64> GetDispls(const OpParam &param) const;
      33              :     virtual u64 GetCurrentCount(const OpParam &param, const std::vector<u64> &curCounts) const;
      34              :     virtual u64 CalcCurrentTotalCount(const OpParam &param, const std::vector<u64> &curCounts) const;
      35              :     virtual HcclOpMetaInfoDef GetOpMetaInfo(u32 algTypeLevel1, bool hugeData, bool smallData, bool dataSplit) const;
      36              :     virtual void UpdateOpParam(OpParam &param, std::vector<u64> &curCounts, std::vector<u64> &curDispls) const;
      37              :     bool CalcCountsDispls(const u64 maxTotalCount, std::vector<u64> &countsLeft,
      38              :         std::vector<u64> &displs, std::vector<u64> &curCounts, std::vector<u64> &curDispls);
      39              :     void PrintCountsDispls(bool finished, const std::vector<u64> &curCounts, const std::vector<u64> &curDispls);
      40              :     HcclResult RunLoopV(OpParam &param, AlgResourceResponse &algRes);    // non-virtual
      41              : 
      42              :     // 工具类
      43              :     HcclResult PrepareAllgatherSlice(u32 sliceNum, u64 inputMemSize, std::vector<Slice> &dataSegsSlice) const;
      44              : 
      45              :     HcclResult CalculateLevel1AllgatherSlice(u64 inputMemSize, u32 level0RankSize, u32 level1RankSize,
      46              :         std::vector<std::vector<Slice>> multRingsSliceZero, std::vector<std::vector<Slice>> &multRingsSlice) const;
      47              : 
      48              :     HcclResult CalculateLevel2AllgatherSlice(u64 inputMemSize, u32 level0RankSize,
      49              :     u32 level1RankSize, u32 level2RankSize, std::vector<std::vector<Slice>> multRingsSliceZero,
      50              :     std::vector<Slice> &level2DataSlice, u32 ringIndex) const;
      51              : 
      52              :     HcclResult AllGatherLevel2(const std::string &tag, DeviceMem &inputMem, DeviceMem &outputMem,
      53              :         u64 count, HcclDataType dataType, Stream &stream, HcomCollOpInfo *opInfo = nullptr);
      54              : 
      55              :     bool DMAReduceFlag_{false}; // 是否DMA消减的标志
      56              :     bool isAllGatherV_{false};
      57              : };
      58              : 
      59              : } // namespace hccl
      60              : 
      61              : #endif