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 "coll_all_reduce_single_rank_executor.h"
      12              : 
      13              : namespace hccl {
      14              : 
      15            0 : CollAllReduceSingleRankExecutor::CollAllReduceSingleRankExecutor(const HcclDispatcher dispatcher,
      16            0 :                                                                  std::unique_ptr<TopoMatcher> &topoMatcher)
      17            0 :     : CollAllReduceExecutor(dispatcher, topoMatcher)
      18              : {
      19            0 : }
      20              : 
      21            0 : HcclResult CollAllReduceSingleRankExecutor::KernelRun(const OpParam &param, ExecMem &execMem)
      22              : {
      23            0 :     HCCL_CONFIG_INFO(HCCL_ALG, "[CollAllReduceSingleRankExecutor][KernelRun] AllReduce single rank");
      24            0 :     u64 totalSize = execMem.count * SIZE_TABLE[param.DataDes.dataType];
      25            0 :     ReduceType reduceType =
      26            0 :         ((param.reduceType != HCCL_REDUCE_PROD) && (param.DataDes.dataType != HCCL_DATA_TYPE_INT64)) ?
      27              :         ReduceType::INLINE_REDUCE : ReduceType::TBE_REDUCE;
      28            0 :     auto originalAlgTypeLevel1 = static_cast<u32>(algType_.algoLevel1);
      29            0 :     bool hugeData = totalSize > SDMA_SEND_MAX_SIZE;
      30            0 :     if (execMem.inputPtr == execMem.outputPtr) {
      31            0 :         auto opMeta = HcclOpMetaInfo::GetOneForAllReduce(originalAlgTypeLevel1, param.DataDes.dataType, reduceType,
      32              :             totalSize <= HCCL_SMALL_COUNT_128_KB, 1, hugeData, CopyPattern::ZCOPY); // 通过CopyPattern字段区分不同子图
      33            0 :         CHK_RET(InitTask(dispatcher_, const_cast<Stream&>(param.stream), opMeta.isEnableCache, opMeta.GetCacheKey()));
      34              :     } else {
      35            0 :         auto opMeta = HcclOpMetaInfo::GetOneForAllReduce(originalAlgTypeLevel1, param.DataDes.dataType, reduceType,
      36              :             totalSize <= HCCL_SMALL_COUNT_128_KB, 1, hugeData, CopyPattern::BCOPY); // 通过CopyPattern字段区分不同子图
      37            0 :         CHK_RET(InitTask(dispatcher_, const_cast<Stream&>(param.stream), opMeta.isEnableCache, opMeta.GetCacheKey()));
      38              :         // ranksize = 1; input、output地址不同，input->output
      39            0 :         DeviceMem srcMem(execMem.inputPtr, totalSize);
      40            0 :         DeviceMem dstMem(execMem.outputPtr, totalSize);
      41            0 :         CHK_RET(HcclD2DMemcpyAsync(dispatcher_, dstMem, srcMem, const_cast<Stream&>(param.stream)));
      42            0 :     }
      43            0 :     CHK_RET(LaunchTask(dispatcher_, const_cast<Stream&>(param.stream)));
      44            0 :     return HCCL_SUCCESS;
      45              : }
      46              : 
      47              : REGISTER_EXEC("AllReduceSingleExecutor", AllReduceSingleRank, CollAllReduceSingleRankExecutor);
      48              : 
      49              : } // namespace hccl