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 COMM_H
      12              : #define COMM_H
      13              : 
      14              : #include "hccl_common.h"
      15              : #include "common.h"
      16              : #include "hccl_socket.h"
      17              : 
      18              : static constexpr uint32_t HCCL_ALG_MESH = 0b1U;
      19              : static constexpr uint32_t HCCL_ALG_SWITCH = (HCCL_ALG_MESH << 1U);
      20              : static constexpr uint32_t HCCL_ALG_RING = (HCCL_ALG_MESH << 2U);
      21              : static constexpr uint32_t HCCL_ALG_PAIRWISE = (HCCL_ALG_MESH << 3U);
      22              : enum class HcclTopoLevel {
      23              :     HCCL_TOPO_L0 = 0,
      24              :     HCCL_TOPO_L1,
      25              :     HCCL_TOPO_MAX,
      26              : };
      27              : 
      28              : namespace hccl {
      29              : enum class HcclCommState {
      30              :     IDLE = 0,   // 初始化完成，未使用
      31              :     BUILDING,   // 正在使用，且在建链
      32              :     INUSE,      // 正在使用，建链完成或未建链
      33              :     RESERVED
      34              : };
      35              : 
      36          235 : inline const char* HcclCommStateToString(HcclCommState state) {
      37          235 :     switch (state) {
      38          235 :         case HcclCommState::IDLE: return "IDLE";
      39            0 :         case HcclCommState::BUILDING: return "BUILDING";
      40            0 :         case HcclCommState::INUSE: return "INUSE";
      41            0 :         case HcclCommState::RESERVED: return "RESERVED";
      42            0 :         default: return "UNKNOWN";
      43              :     }
      44              : }
      45              : 
      46           15 : using HcclCommConnections = struct HcclCommConnectionsDef {
      47              :     bool isRoot{false};
      48              :     std::shared_ptr<HcclSocket> agentConnection{nullptr};
      49              :     std::map<u32, std::shared_ptr<HcclSocket>> serverConnections;
      50              : };
      51              : 
      52           15 : using HcclSocketPortConfig = struct HcclSocketPortConfigDef {
      53              :     // devPortSwitchOn 用于判断是否开启了用户配置的端口（通过环境变量配置的端口范围或者通过ranktable指定的端口）。
      54              :     // devPortSwitchOn开启时，将启用独立的vnic端口；即nic和vnic使用的端口可能不一致。
      55              :     bool devPortSwitchOn{ false };
      56              :     std::pair<std::shared_ptr<HcclSocket>, HcclNetDevCtx> devNicListen{ nullptr, nullptr }; // 抢占的device nic socket
      57              :     std::pair<std::shared_ptr<HcclSocket>, HcclNetDevCtx> devVnicListen{ nullptr, nullptr }; // 抢占的device vnic socket
      58              :     std::pair<std::shared_ptr<HcclSocket>, HcclNetDevCtx> backupDevNicListen{ nullptr, nullptr }; // 抢占的backup nic socket
      59              : };
      60              : 
      61              : using HcclCommParams = struct TagHCCLCollectiveParams {
      62              :     /**
      63              :     通信域的基本构建信息，通信域标识、节点数及本节点的编号
      64              :     通信域通过如下条件构建:
      65              :     1.用户在某个计算实体(rank)内调用hcclGetUniqueId作为本通信域的id
      66              :     2.将此id发往通信域的其它计算实体(rank)
      67              :     3.用户指定本comm对应的device, 本comm实例对应的device将会是用户set的device
      68              :     4.每个计算实体根据id, rank和total_ranks创建通信域
      69              :     */
      70              :     HcclRootInfo id; /* * 用于标识不同的通信域 */
      71              :     u32 rank;        /* * 用于标识通信域内不同节点 */
      72              :     u32 userRank;
      73              :     u32 totalRanks; /* * 用于指示通信域内的节点总数, rank范围[0, totalRanks-1] */
      74              :     s32 logicDevId;
      75              :     std::string serverId;
      76              :     DevType deviceType;  // 芯片类型信息
      77              :     HcomProfilingMode profilingMode;
      78              :     std::string profilingOption;
      79              :     bool profilingInitiated;
      80              :     HcclComm commHandle;
      81              :     bool isHeterogComm;
      82              :     bool hcomGroupNicInit;  // 在子group中对应world group NIC初始化标识
      83              :     CommAttr attr;
      84              :     WorkMode commWorkMode = WorkMode::HCCL_MODE_NORMAL;
      85              :     std::string identifier;
      86              :     std::string cclBuffName;
      87              :     u32 ranktableCrc;
      88              :     HcclCommConnections commConnections;
      89              :     HcclSocketPortConfig commPortConfig;
      90         4723 :     TagHCCLCollectiveParams()
      91         4723 :         : id{0}, rank(INVALID_VALUE_RANKID), userRank(INVALID_VALUE_RANKID), totalRanks(0xFFFFFFFF),
      92         9431 :           logicDevId(-1), deviceType(DevType::DEV_TYPE_COUNT), profilingMode(HcomProfilingMode::PROFILING_CLOSE),
      93         4723 :           profilingInitiated(false), commHandle(nullptr), isHeterogComm(false), hcomGroupNicInit(false),
      94        23555 :           identifier(""), cclBuffName(""), ranktableCrc(0)
      95              :     {
      96         4723 :     }
      97              : };
      98              : 
      99              : using WorldGroupInfo = struct worldGroupInfo {
     100              :     bool inlineReduceSwitchOn;
     101              :     DevType deviceType;
     102              :     s32 deviceLogicId;
     103              :     bool profilingInitiated;
     104              :     std::string serverId;
     105              :     std::unordered_map<std::string, std::map<u32, HcclIpAddress>> phyIdNicInfoMap;
     106              :     std::vector<RankInfo> worldRankInfoList;
     107              :     std::vector<u32> ranksPort;
     108              :     std::vector<u32> vnicRanksPort;
     109              :     bool devPortSwitchOn{ false };
     110              :     bool useSuperPodMode;
     111           38 :     worldGroupInfo()
     112           38 :         :inlineReduceSwitchOn(true), deviceType(DevType::DEV_TYPE_COUNT), deviceLogicId(-1), profilingInitiated(false),
     113           38 :         useSuperPodMode(false)
     114              :     {
     115           38 :     }
     116              : };
     117              : } // hccl
     118              : #endif // COMM_H