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_ALLREDUCE_COMM_EXECUTOR_H
      12              : #define COLL_ALLREDUCE_COMM_EXECUTOR_H
      13              : 
      14              : #include "coll_comm_executor.h"
      15              : 
      16              : namespace hccl {
      17              : class CollAllReduceExecutor : public CollCommExecutor {
      18              : public:
      19              :     CollAllReduceExecutor(const HcclDispatcher dispatcher, std::unique_ptr<TopoMatcher> &topoMatcher);
      20           38 :     ~CollAllReduceExecutor() override = default;
      21              : 
      22              :     HcclResult Orchestrate(OpParam& param, AlgResourceResponse& algRes) override;
      23              : protected:
      24              :     /* *************** 算法编排 *************** */
      25              :     // AllReduce Loop Executor公共接口
      26              :     virtual u64 CalcLoopMaxCount(const u64 cclBuffSize, const u32 unitSize);
      27              :     virtual bool IsHugeData(const u64 curSize);
      28              :     virtual bool IsSmallData(const u64 totalSize, const u64 curSize);
      29              :     HcclResult RunLoop(OpParam &param, AlgResourceResponse &algRes);
      30              :     HcclResult AvoidSubgraphLoop(OpParam &param, AlgResourceResponse &algRes);
      31              : 
      32              :     // 工具类
      33              :     HcclResult GetSliceNum(const u64 totalSize, const bool isSmallData, u64& sliceNum, u32 unitSize=0);
      34              :     bool IsAllReduceSmallData(u64 size);
      35              :     HcclResult PrepareSliceDataWithAlignSize(u64 totalSize, u32 sliceNum,
      36              :         u64 piplineOffset, std::vector<Slice>& dataSlice, u64 alignSize);
      37              :     HcclResult PrepareAivBuffers(u32 rankSize, u32 rankId, u32 rankOffset,
      38              :         DeviceMem &inputMem, DeviceMem &outputMem, std::vector<LINK> &links, void **dataBuffers, void **flagBuffers,
      39              :         UserMemType dataMemType, UserMemType flagMemType, u32 dataMemOffset, u32 flagMemOffset);
      40              : 
      41              :     bool CCLMemSlice_{true};    // 每次Loop是否需要对CCLMem进行切片
      42              :     bool DMAReduceFlag_{false}; // 是否DMA消减
      43              : private:
      44              :     virtual HcclResult RunLoopInner(OpParam &param, const ReduceType &reduceType, ExecMem &execMem);
      45              : };
      46              : 
      47              : } // namespace hccl
      48              : 
      49              : #endif