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_reduce_scatter_v_ring_for_910_93_executor.h"
      12              : #include <numeric>
      13              : 
      14              : namespace hccl {
      15              : 
      16            0 : CollReduceScatterVRingFor91093Executor::CollReduceScatterVRingFor91093Executor(const HcclDispatcher dispatcher,
      17            0 :     std::unique_ptr<TopoMatcher> &topoMatcher)
      18            0 :     : CollReduceScatterRingFor91093Executor(dispatcher, topoMatcher)
      19              : {
      20            0 :     isReduceScatterV_ = true;
      21            0 :     desc_.level1SupportedAlgos = {
      22              :         AlgTypeLevel1::ALG_LEVEL1_NHR,
      23              :         AlgTypeLevel1::ALG_LEVEL1_NB,
      24              :         AlgTypeLevel1::ALG_LEVEL1_RING
      25            0 :     };
      26            0 : }
      27              : 
      28            0 : u64 CollReduceScatterVRingFor91093Executor::CalcLoopMaxCount(const u32 unitSize)
      29              : {
      30              :     // 中转内存单次最多能够接受的output count，这里不除以RankSize，因为每次循环可能会减少需要参与通信的Rank
      31            0 :     return inCCLbufferSize_ / HCCL_MIN_SLICE_ALIGN * HCCL_MIN_SLICE_ALIGN / unitSize;
      32              : }
      33              : 
      34            0 : bool CollReduceScatterVRingFor91093Executor::IsHugeData(const u64 curSize, OpParam *param)
      35              : {
      36              :     u32 level2RankSize;
      37            0 :     if (algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_AHC ||
      38            0 :         algType_.algoLevel1 == AlgTypeLevel1::ALG_LEVEL1_AHC_BROKE) {
      39              :         // AHC非对称场景下没有L2
      40            0 :         level2RankSize = 1;
      41              :     } else {
      42              :         // 多QP哈希散列开启且RDMA通信下，强制刷新子图
      43              :         // 这里如果CheckCommSize返回ERROR，相当于HugeData true，防止GetSubCommInfo越界
      44            0 :         CHK_RET(CheckCommSize(COMM_LEVEL2, COMM_INDEX_0 + 1));
      45            0 :         SubCommInfo level2CommInfo = GetSubCommInfo(COMM_LEVEL2, COMM_INDEX_0);
      46            0 :         level2RankSize = level2CommInfo.localRankSize;
      47            0 :     }
      48              : 
      49            0 :     const HcclDataType dataType = param->GetDataType();
      50            0 :     const u64 TBE_REDUCE_MAX_COUNT = INT32_MAX;
      51            0 :     u64 curCount = curSize / SIZE_TABLE[dataType];
      52            0 :     bool issupportRDMAInlineReduce = IsSupportRDMAReduce(dataType, param->reduceType);
      53            0 :     bool hugeData = (curSize * level2RankSize > RDMA_SEND_MAX_SIZE) || (curSize > SDMA_SEND_MAX_SIZE) ||
      54            0 :         ((!isSupportSDMAReduce_) && (curCount > TBE_REDUCE_MAX_COUNT)) ||
      55            0 :         ((!issupportRDMAInlineReduce) && (curCount * level2RankSize > TBE_REDUCE_MAX_COUNT));
      56            0 :     return hugeData;
      57              : }
      58              : 
      59              : REGISTER_EXEC("ReduceScatterVRingFor91093Executor", ReduceScatterVRingFor91093, CollReduceScatterVRingFor91093Executor);
      60              : }