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 HCCLV2_CCU_CONTEXT_REDUCE_SCATTER_MESH_2D_MEM2MEM_H_
      12              : #define HCCLV2_CCU_CONTEXT_REDUCE_SCATTER_MESH_2D_MEM2MEM_H_
      13              : 
      14              : #include <vector>
      15              : #include <ios>
      16              : #include "log.h"
      17              : #include "ccu_context_alg_base.h"
      18              : #include "ccu_datatype.h"
      19              : #include "ccu_assist.h"
      20              : 
      21              : namespace Hccl {
      22              : 
      23              : class CcuContextReduceScatterMeshMem2Mem2D : public CcuContextAlgBase {
      24              : public:
      25              :     CcuContextReduceScatterMeshMem2Mem2D(const CcuCtxArg &arg, 
      26              :                                          const std::vector<CcuTransport*> &transports,
      27              :                                          const CcuTransportGroup &group);
      28            0 :     ~CcuContextReduceScatterMeshMem2Mem2D() override {}
      29              : 
      30              :     void Algorithm() override;
      31              :     std::vector<uint64_t> GeneArgs(const CcuTaskArg &arg) override;
      32              : 
      33              : protected:
      34              :     void InitResources();
      35              : 
      36              : private:
      37              :     void CreateLocalCopyLoop();                                         // for loop-group local copy 
      38              :     void LocalCopyByLoopGroup(CcuRep::Memory dst, CcuRep::Memory src);  // for loop-group local copy
      39              :     std::vector<uint64_t> CalMeshChunkSlice(uint64_t dataSize, uint64_t sliceNum); // for mesh-chunk
      40              :     void PreSync();
      41              :     void PostSync(uint32_t signalIndex);
      42              :     void LoadArgs();
      43              :     void AxisSync(uint32_t signalIndex);
      44              :     void Step1Reduce();
      45              :     void Step2Reduce();
      46              : 
      47              :     // 在构造函数中初始化
      48              :     uint32_t axisId_{0};            // 值为 0 表示 x 轴, 值为 1 表示 y 轴
      49              :     std::vector<uint64_t> dimSize_; // [0] mesh2d 拓扑的列数
      50              :     std::vector<uint32_t> dimId_;   // [0] 当前 NPU 所在列编号, [1] 当前 NPU 所在行编号
      51              :     uint32_t localId_{0};           // 当前 rank 在指定 axisId_ 上的编号
      52              :     uint32_t localSize_{0};         // mesh2d 拓扑在指定 axisId_ 轴上的维度
      53              :     uint32_t oppsiteSize_{0};       // mesh2d 拓扑在与指定 axisId_ 轴相反的轴上的维度
      54              :     DataType dataType_;
      55              :     DataType outputDataType_;
      56              :     // load 进来的参数
      57              :     std::vector<CcuRep::Variable> input_;
      58              :     std::vector<CcuRep::Variable> output_;
      59              :     std::vector<CcuRep::Variable> token_;
      60              :     CcuRep::Variable xAxisSize_;
      61              :     CcuRep::Variable yAxisSize_;
      62              :     CcuRep::Variable step0BaseOffset_;
      63              :     CcuRep::Variable step0AddOffset_;
      64              :     CcuRep::Variable step1AddOffset_;
      65              :     CcuRep::Variable yAxisOffset_;
      66              :     GroupOpSize xAxisGroupOpSize_;
      67              :     GroupOpSize yAxisGroupOpSize_;
      68              :     GroupOpSize curGoSize_;                 // for loop-group local copy
      69              : 
      70              :     // For mesh-chunk
      71              :     std::vector<CcuRep::Variable> xlocalSlice_;
      72              :     std::vector<CcuRep::Variable> ylocalSlice_;
      73              : 
      74              :     CcuRep::Variable inputSize_;
      75              :     CcuRep::Variable offset_;
      76              :     CcuRep::Variable sliceOffset_;
      77              :     CcuRep::Variable strideSize_;
      78              : 
      79              :     // 跨轴同步信号
      80              :     std::string localAxisSignalName_;       // 由 axisId_ 初始化
      81              :     std::string anotherAxisSignalName_;     // 由 axisId_ 初始化
      82              :     CcuRep::MaskSignal localAxisSignal_;
      83              :     CcuRep::MaskSignal anotherAxisSignal_;
      84              : }; // class CcuContextReduceScatterMeshMem2Mem2D
      85              : }  // namespace Hccl
      86              : 
      87              : #endif // HCCLV2_CCU_CONTEXT_REDUCE_SCATTER_MESH_2D_MEM2MEM_H_