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              : 
      12              : #ifndef __COLL_COMM_AICPU_H__
      13              : #define __COLL_COMM_AICPU_H__
      14              : 
      15              : #include "common.h"
      16              : #include "aicpu_init_param.h"
      17              : #include "topo_matcher.h"
      18              : #include "hcomm_primitives.h"
      19              : #include "transport_pub.h"
      20              : #include "thread.h"
      21              : #include "local_notify.h"
      22              : #include "ub_transport_lite_impl.h"
      23              : #include "p2p_transport_lite_impl.h"
      24              : #include "task_exception.h"
      25              : #include "aicpu_launch_manager.h"
      26              : #include "channel_param.h"
      27              : #include "hdc_pub.h"
      28              : #include "ns_recovery/aicpu/ns_recovery_lite.h"
      29              : #include <atomic>
      30              : #include "hcclCommDfxLite.h"
      31              : #include "error_message_v2.h"
      32              : #include "kfc.h"
      33              : #include "aicpu_hdc.h"
      34              : #include "roce_transport_lite_impl.h"
      35              : #include "hccl/hccl_types.h"
      36              : 
      37              : using namespace hccl;
      38              : 
      39              : class CollCommAicpu {
      40              : public:
      41              :     ~CollCommAicpu();
      42              :     HcclResult InitAicpuIndOp(CommAicpuParam *commAicpuParam);
      43              :     HcclResult InitThreads(ThreadMgrAicpuParam *param);
      44              :     HcclResult AllocChannelResource(HcclChannelUrmaRes *commParam);
      45              :     HcclResult NotifyFree(NotifyMgrAicpuParam *param);
      46              :     HcclResult NotifyAlloc(NotifyMgrAicpuParam *param);
      47            2 :     const std::vector<std::shared_ptr<Thread>>& GetAllThread() { return threads_; };
      48            2 :     const HcclTopoInfo& GetTopoInfo() { return topoInfo_; }
      49           13 :     const std::string& GetIdentifier() { return identifier_; }
      50              : 
      51              :     // taskException
      52            0 :     bool IsErrorReported() { return isErrorReported_; }
      53            1 :     void SetErrorReported(bool isErrorReported) { isErrorReported_ = isErrorReported; }
      54              :     HcclResult SendErrorMessageReportToHost(Hccl::ErrorMessageReport& errMsgInfo);
      55              :     HcclResult RegisterProfCallBack();
      56           14 :     HcclCommDfxLite* GetHcclCommDfxLite() { return &dfx_; };
      57            2 :     std::shared_mutex& GetThreadMutex() { return threadMutex_; }
      58            4 :     u32 GetDevId() { return devId_; }
      59              : 
      60              :     // h2d - d2h通道信息交互
      61              :     HcclResult BackGroundGetCmd(Hccl::KfcCommand &cmd);
      62              :     HcclResult BackGroundSetStatus(Hccl::KfcStatus state);
      63              :     u32 UpdateIndex();
      64              : 
      65            6 :     HcclCommStatus GetCommmStatus() { return commStatus_; }
      66              :     void SetCommmStatus(HcclCommStatus status);
      67              : 
      68              :     // N秒快恢
      69              :     hccl::NsRecoveryLitePtr GetNsRecoveryLitePtr();
      70              :     HcclResult Clean();
      71              :     HcclResult Resume(HcclChannelUrmaRes *commParam);
      72              : 
      73              :     HcclResult CheckIndOpExecStatus(bool timeout);
      74              : 
      75              : private:
      76              :     // 初始化
      77              :     void InitIndopEnv(CommAicpuParam *commAicpuParam);
      78              :     HcclResult InitHDCommunicate(CommAicpuParam *commAicpuParam);
      79              : 
      80              :     HcclResult InitUrmaChannel(HcclChannelUrmaRes *commParam);
      81              :     HcclResult ParsePackData(std::vector<char> &data, ChannelHandle &handle);
      82              :     HcclResult RegisterChannelAddDfxTaskInfo(ChannelHandle channel);
      83              :     HcclResult RegisterThreadAddDfxTaskInfo(ThreadHandle thread);
      84              :     void InitBackGroundThread();
      85              :     HcclResult ResumePackData(std::vector<char> &data, ChannelHandle &handle);
      86              :     HcclResult ProcessUrmaRes(HcclChannelUrmaRes *commParam, bool isInit);
      87              : 
      88              :     u32 devId_{0};
      89              :     //通用的通道
      90              :     std::shared_ptr<hccl::HDCommunicate> kfcControlTransferH2D_{nullptr};
      91              :     std::shared_ptr<hccl::HDCommunicate> kfcStatusTransferD2H_{nullptr};
      92              : 
      93              :     std::string identifier_;
      94              :     HcclCommStatus commStatus_{HcclCommStatus::HCCL_COMM_STATUS_INVALID};
      95              :     HcclTopoInfo topoInfo_;
      96              :     std::shared_mutex threadMutex_;
      97              :     std::vector<std::shared_ptr<Thread>> threads_;
      98              :     std::vector<std::unique_ptr<LocalNotify>> notifys_;
      99              :     // A5 独立算子
     100              :     std::unordered_map<ChannelHandle, std::unique_ptr<Hccl::UbTransportLiteImpl>> ubTransportMap_;
     101              :     std::unordered_map<ChannelHandle, std::unique_ptr<Hccl::P2PTransportLiteImpl>> p2pTransportMap_;
     102              :     std::unordered_map<ChannelHandle, std::unique_ptr<Hccl::RoceTransportLiteImpl>> roceTransportMap_;
     103              : 
     104              :     // N秒快恢相关
     105              :     hccl::NsRecoveryLitePtr nsRecoveryLitePtr_{nullptr};
     106              : 
     107              :     // dfx
     108              :     bool isErrorReported_{false}; // 是否上报了taskException信息
     109              :     HcclCommDfxLite dfx_;
     110              :     u32 index_{0};
     111              : };
     112              : 
     113              : #endif // __COLL_COMM_AICPU_H__