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_INSTRUCTION_ALL_TO_ALL_MESH_2D_H
      12              : #define HCCLV2_CCU_INSTRUCTION_ALL_TO_ALL_MESH_2D_H
      13              : 
      14              : #include "template_utils.h"
      15              : #include "instruction.h"
      16              : #include "ins_queue.h"
      17              : #include "ccu_context_utils.h"
      18              : #include "ccu_ctx_signature.h"
      19              : #include "ccu_ins.h"
      20              : #include "ccu_rank_group.h"
      21              : 
      22              : namespace Hccl {
      23              : 
      24              : // 为AllGatherMesh1D实现的CCUIns、CCUCtxArg与CCUTaskArg
      25              : class CcuCtxArgAlltoAllMesh2D : public CcuCtxArg {
      26              : public:
      27            3 :     CcuCtxArgAlltoAllMesh2D(const std::vector<uint32_t> &dSize, uint32_t rId, uint32_t aId,
      28            3 :         const CollAlgOperator &op, const std::vector<std::vector<RankId>> &tempVTopo) :
      29            3 :             CcuCtxArg(), dimSize(dSize), rankId(rId), axisId(aId), op(op), tempVTopo(tempVTopo) {}
      30              : 
      31            3 :     ~CcuCtxArgAlltoAllMesh2D() override {}
      32              : 
      33            7 :     CcuCtxSignature GetCtxSignature() const override
      34              :     {
      35            7 :         CcuCtxSignature signature;
      36            7 :         GenerateCcuCtxSignature(signature, CcuInstType::CCU_ALLTOALL_MESH_2D_DIRECT, op, tempVTopo);
      37           21 :         HCCL_INFO("[CcuCtxArgAlltoAllMesh2D][GetCtxSignature] signature[%s]", signature.GetData().c_str());
      38            7 :         return signature;
      39            0 :     }
      40              : 
      41              :     // 需要存储，传递给算法
      42              :     std::vector<uint32_t> dimSize;
      43              :     uint32_t rankId;
      44              :     uint32_t axisId;
      45              : 
      46              :     const CollAlgOperator &op;
      47              :     const std::vector<std::vector<RankId>> &tempVTopo;
      48              : };
      49              : 
      50              : class CcuTaskArgAlltoAllMesh2D : public CcuTaskArg {
      51              : public:
      52           24 :     explicit CcuTaskArgAlltoAllMesh2D(uint64_t inputAddr, uint64_t outputAddr, uint64_t scratchAddr,
      53              :         uint64_t sendStride, uint64_t recvStride, uint64_t sendLength, uint64_t aSize, uint64_t bSize,
      54           24 :         uint64_t baseOffset, uint64_t token) :
      55           24 :             CcuTaskArg(), inputAddr(inputAddr), outputAddr(outputAddr), scratchAddr(scratchAddr),
      56           24 :             sendStride(sendStride), recvStride(recvStride), sendLength(sendLength), aSize(aSize), bSize(bSize),
      57           24 :             baseOffset(baseOffset), token(token) {}
      58              : 
      59              :     uint64_t inputAddr;
      60              :     uint64_t outputAddr;
      61              :     uint64_t scratchAddr;
      62              :     uint64_t sendStride;
      63              :     uint64_t recvStride;
      64              :     uint64_t sendLength;
      65              :     uint64_t aSize;  // X方向第一轮传输的数据量
      66              :     uint64_t bSize;
      67              :     uint64_t baseOffset;  // 多轮执行时的基础偏移，等于step*(aSize+bSize)
      68              :     uint64_t token;
      69              : };
      70              : 
      71              : class CcuInstructionAlltoAllMesh2D : public CcuInstruction {
      72              : public:
      73            4 :     CcuInstructionAlltoAllMesh2D(const CollAlgOperator &op, const std::vector<uint32_t> &dimSize,
      74            4 :         const std::vector<std::vector<RankId>> &tempVTopo) :
      75            4 :         CcuInstruction(), op_(op), dimSize_(dimSize), tempVTopo_(tempVTopo) {}
      76              : 
      77            4 :     void Init(uint32_t rankId, uint64_t inputAddr, uint64_t outputAddr, uint64_t scratchAddr, uint64_t axisId,
      78              :         uint64_t sendStride, uint64_t recvStride, uint64_t sendLength, uint64_t aSize, uint64_t bSize,
      79              :         uint64_t baseOffset, uint64_t token)
      80              :     {
      81            4 :         rankId_ = rankId;
      82            4 :         inputAddr_ = inputAddr;
      83            4 :         outputAddr_ = outputAddr;
      84            4 :         scratchAddr_ = scratchAddr;
      85            4 :         axisId_ = axisId;
      86            4 :         sendStride_ = sendStride;
      87            4 :         recvStride_ = recvStride;
      88            4 :         sendLength_ = sendLength;
      89            4 :         aSize_ = aSize;
      90            4 :         bSize_ = bSize;
      91            4 :         baseOffset_ = baseOffset;
      92            4 :         token_ = token;
      93           12 :         HCCL_INFO("[CcuInstructionAlltoAllMesh2D][Init] rankId[%u] inputAddr[%llu] outputAddr[%llu] scratchAddr[%llu],\
      94              : axisId[%u], sendStride[%llu], recvStride[%llu], sendLength[%llu], aSize[%llu], bSize[%llu], baseOffset[%llu], \
      95              : dimSize.size[%u], tempVTopo.size[%u]", rankId_, inputAddr_, outputAddr_, scratchAddr_, axisId_,
      96              :             sendStride_, recvStride_, sendLength_, aSize_, bSize_, baseOffset_, dimSize_.size(),
      97              :             tempVTopo_.size());
      98            4 :         return;
      99              :     }
     100              : 
     101            0 :     CcuInstType GetInstType() const override
     102              :     {
     103            0 :         HCCL_INFO("CcuInstructionAllGatherMesh1D instype is CCU_ALLTOALL_MESH_2D_DIRECT.");
     104            0 :         return instType_;
     105              :     }
     106              : 
     107            0 :     std::string Describe() const override
     108              :     {
     109            0 :         return StringFormat("[CcuInstructionAllGatherMesh1D]RankId[%u] Ins[%s]", rankId_, instType_.Describe().c_str());
     110              :     }
     111              : 
     112            0 :     std::unique_ptr<CcuCtxArg> GetCtxArg() const override
     113              :     {
     114            0 :         HCCL_INFO("[CcuInstructionAlltoAllMesh2D][GetCtxArg] dimSize.size[%u], rankId[%u], axisId[%u], tempVTopo.size[%u]",
     115              :             dimSize_.size(), rankId_, axisId_, tempVTopo_.size());
     116            0 :         return std::make_unique<CcuCtxArgAlltoAllMesh2D>(dimSize_, rankId_, axisId_, op_, tempVTopo_);
     117              :     }
     118              : 
     119              :     void SetInstType(CcuInstType instType) 
     120              :     { 
     121              :         instType_ = instType; 
     122              :     }
     123              : 
     124            0 :     std::unique_ptr<CcuTaskArg> GetTaskArg() const override
     125              :     {
     126            0 :         return std::make_unique<CcuTaskArgAlltoAllMesh2D>(inputAddr_, outputAddr_, scratchAddr_, sendStride_,
     127            0 :                 recvStride_, sendLength_, aSize_, bSize_, baseOffset_, token_);
     128              :     }
     129              : 
     130              : private:
     131              :     CollAlgOperator op_;
     132              :     std::vector<uint32_t> dimSize_;
     133              :     std::vector<std::vector<RankId>> tempVTopo_;
     134              : 
     135              :     CcuInstType instType_ = CcuInstType::CCU_ALLTOALL_MESH_2D_DIRECT;
     136              : 
     137              :     uint32_t rankId_{0};
     138              :     uint64_t inputAddr_{0};
     139              :     uint64_t outputAddr_{0};
     140              :     uint64_t scratchAddr_{0};
     141              :     uint64_t sendStride_{0};
     142              :     uint64_t recvStride_{0};
     143              :     uint64_t axisId_{0};
     144              :     uint64_t sendLength_{0};  // 多轮时的单个数据块总大小
     145              :     uint64_t aSize_{0};
     146              :     uint64_t bSize_{0};
     147              :     uint64_t baseOffset_{0};
     148              :     uint64_t token_{0};
     149              : };
     150              : 
     151              : }
     152              : #endif // HCCLV2_CCU_INSTRUCTION_ALL_TO_ALL_MESH_2D_H