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              : #include "scatter_operator.h"
      12              : #include "coll_alg_utils.h"
      13              : 
      14              : namespace hccl {
      15              : 
      16            1 : ScatterOperator::ScatterOperator(AlgConfigurator* algConfigurator, CCLBufferManager &cclBufferManager,
      17            1 :     HcclDispatcher dispatcher, std::unique_ptr<TopoMatcher> &topoMatcher)
      18            1 :     : CollAlgOperator(algConfigurator, cclBufferManager, dispatcher, topoMatcher, HcclCMDType::HCCL_CMD_SCATTER)
      19              : {
      20              :     // 由于scatter只支持server间ring、nb和nhr，其他算法需要重定向到ring
      21            2 :     if (!(algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_NHR) &&
      22            2 :         !(algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_NB) &&
      23            2 :         !(algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_RING)) {
      24            2 :         algType_.algoLevel1 = AlgTypeLevel1::ALG_LEVEL1_RING;
      25            2 :         HCCL_INFO("[ScatterOperator][ScatterOperator] algType[%s] is not supported, reset algType=ring",
      26              :             AlgTypeToStr(algType_).c_str());
      27              :     }
      28            4 : }
      29              : 
      30            8 : ScatterOperator::~ScatterOperator()
      31              : {
      32            8 : }
      33              : 
      34            4 : HcclResult ScatterOperator::SelectAlg(const std::string& tag, const OpParam& param, std::string& algName,
      35              :     std::string& newTag)
      36              : {
      37            4 :     if (isDiffDeviceType_) {
      38            0 :         HCCL_ERROR("[ScatterOperator][SelectAlg] Scatter not support diffDeviceType");
      39            0 :         return HCCL_E_NOT_SUPPORT;
      40              :     }
      41            4 :     if (userRankSize_ == 1) {
      42            0 :         algName = "ScatterSingleExecutor";
      43            0 :         HCCL_INFO("[SelectAlg] Scatter SelectAlg is algName [%s]", algName.c_str());
      44            0 :         return HCCL_SUCCESS;
      45              :     }
      46            4 :     newTag = param.tag;
      47            8 :     if (GetWorkflowMode() == HcclWorkflowMode::HCCL_WORKFLOW_MODE_OP_BASE && (algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_HD
      48            4 :         || algType_.algoLevel2 == AlgTypeLevel2::ALG_LEVEL2_HD)) {
      49            0 :         std::string appendTag = "";
      50            0 :         u32 serverNumPerSuperPod = superPodNum_ == 0 ? moduleNum_ : moduleNum_ / superPodNum_;
      51            0 :         if (algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_HD) {
      52            0 :             u32 part1Size = FACTOR_TWO * (serverNumPerSuperPod - (1 << static_cast<u32>(log2(serverNumPerSuperPod))));
      53            0 :             u32 rootId = param.root / deviceNumPerAggregation_ % serverNumPerSuperPod;
      54            0 :             appendTag += "L1_" + std::to_string((rootId >= part1Size) || ((rootId % FACTOR_TWO) == 0));
      55              :         }
      56            0 :         HCCL_DEBUG("[ScatterOperator]SelectAlg for algoLevel1");
      57            0 :         if (algType_.algoLevel2 == AlgTypeLevel2::ALG_LEVEL2_HD) {
      58            0 :             u32 part1Size = FACTOR_TWO * (superPodNum_ - (1 << static_cast<u32>(log2(superPodNum_))));
      59            0 :             u32 rootId = param.root / deviceNumPerAggregation_ / serverNumPerSuperPod;
      60            0 :             appendTag += (appendTag.empty() ? "L2_" : "_L2_") + std::to_string((rootId >= part1Size) || ((rootId % FACTOR_TWO) == 0));
      61              :         }
      62            0 :         HCCL_DEBUG("[ScatterOperator][SelectAlg]tag is [%s]", tag);
      63            0 :         newTag = newTag + '_' + appendTag;
      64            0 :         if (GetExternalInputHcclEnableEntryLog() && param.opBaseAtraceInfo != nullptr) {
      65            0 :             CHK_RET(param.opBaseAtraceInfo->SavealgtypeTraceInfo(appendTag, param.tag));
      66              :         }
      67            0 :     }
      68              : 
      69              :     // 由于scatter只支持server间ring,nb和NHR，如果不是需要重定向到ring；910_93仅支持server间ring
      70            4 :     if (!(algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_NHR) &&
      71            4 :         !(algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_NB) &&
      72            4 :         !(algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_RING)) {
      73            0 :         HCCL_INFO("[ScatterOperator][Scatter] algType[%s] is not supported, reset algType=ring",
      74              :             AlgTypeToStr(algType_).c_str());
      75            0 :         algType_.algoLevel1 = AlgTypeLevel1::ALG_LEVEL1_RING;
      76              :     }
      77              : 
      78            0 :     bool isMeshTopo = topoType_ == TopoType::TOPO_TYPE_NP_MESH || topoType_ == TopoType::TOPO_TYPE_4P_MESH ||
      79            4 :         topoType_ == TopoType::TOPO_TYPE_2P_MESH || topoType_ == TopoType::TOPO_TYPE_1P_MESH;
      80            8 :     bool isRingTopo = topoType_ == TopoType::TOPO_TYPE_NP_SINGLE_RING || topoType_ == TopoType::TOPO_TYPE_8P_RING ||
      81            4 :         topoType_ == TopoType::TOPO_TYPE_NP_DOUBLE_RING;
      82              : 
      83            4 :     if (multiModuleDiffDeviceNumMode_ || multiSuperPodDiffServerNumMode_) {
      84            0 :         algName = "ScatterCommExecutor";
      85            4 :     } else if (isMeshTopo) {
      86            4 :         algName = "ScatterMeshExecutor";
      87            0 :     } else if (isRingTopo) {
      88            0 :         if (deviceType_ == DevType::DEV_TYPE_910_93) {
      89            0 :             algName = "ScatterRingFor91093Executor";
      90              :         } else {
      91            0 :             algName = "ScatterRingExecutor";
      92              :         }
      93              :     } else {
      94            0 :         algName = "ScatterCommExecutor";
      95              :     }
      96            4 :     if (GetWorkflowMode() == HcclWorkflowMode::HCCL_WORKFLOW_MODE_OP_BASE) {
      97            4 :         newTag = newTag + algName;
      98            3 :         HCCL_INFO("[SelectAlg] Scatter newTag is [%s] algName is [%s]", newTag.c_str(), algName.c_str());
      99              :     }
     100            4 :     newTag += (param.aicpuUnfoldMode ? "_device" : "_host");
     101            4 :     return HCCL_SUCCESS;
     102              : }
     103              : 
     104              : REGISTER_OP(HcclCMDType::HCCL_CMD_SCATTER, Scatter, ScatterOperator);
     105              : }