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 "dev_aicpu_ts_hccs_channel.h"
12 : #include "dispatcher_ctx.h"
13 : #include "adapter_hal_pub.h"
14 :
15 : namespace hccl {
16 0 : DevAicpuTsHccsChannel::~DevAicpuTsHccsChannel()
17 : {
18 0 : for (auto &pair : slots_) {
19 0 : if (pair.second.transport != nullptr) {
20 0 : (void)pair.second.transport->DeInit();
21 0 : pair.second.transport.reset();
22 : }
23 : }
24 0 : slots_.clear();
25 0 : }
26 :
27 0 : HcclResult DevAicpuTsHccsChannel::SetTransportMachinePara(hccl::MachinePara &machinePara,
28 : const HcclChannelHccsRes &channelHccsRes)
29 : {
30 0 : machinePara.linkAttribute = 0x03; /* 0x03同时支持目的端和源端发起 */
31 :
32 0 : machinePara.remoteDeviceId = channelHccsRes.remoteDevicePhyId;
33 0 : machinePara.localDeviceId = channelHccsRes.localDevicePhyId;
34 0 : machinePara.deviceLogicId = channelHccsRes.localDeviceLogicId;
35 :
36 0 : machinePara.deviceType = static_cast<DevType>(channelHccsRes.deviceType);
37 0 : machinePara.tag = channelHccsRes.channelTag;
38 0 : machinePara.machineType = channelHccsRes.machineType;
39 :
40 : // 非910_93 2die sio与hccs并发场景，specifyLink设置为RESERVED_LINK_TYPE，平台层将按实际链路类型建链
41 0 : machinePara.specifyLink = LinkTypeInServer::RESERVED_LINK_TYPE;
42 0 : machinePara.isNewOneSide = true;
43 :
44 0 : machinePara.localBufSize = channelHccsRes.localBufSize;
45 0 : machinePara.remoteBufSize = channelHccsRes.remoteBufSize;
46 0 : machinePara.localBufMem = channelHccsRes.localBufMem;
47 0 : machinePara.remoteBufMem = channelHccsRes.remoteBufMem;
48 :
49 0 : HCCL_INFO("%s success, linkAttribute[%x], localUserRank[%u], remoteWorldRank[%u], "
50 : "remoteUserrank[%u], deviceLogicId[%d], localDeviceId[%d], deviceType[%d], newTag[%s], "
51 : "specifyLink[%d], machineType[%u], localBufSize[%u],remoteBufSize[%u], localBufMem[%p], remoteBufMem[%p]",
52 : __func__, machinePara.linkAttribute, machinePara.localUserrank,
53 : machinePara.remoteWorldRank, machinePara.remoteUserrank, machinePara.deviceLogicId, machinePara.localDeviceId,
54 : machinePara.deviceType, machinePara.tag.c_str(), machinePara.specifyLink,
55 : static_cast<u32>(machinePara.machineType), machinePara.localBufSize,
56 : machinePara.remoteBufSize, machinePara.localBufMem, machinePara.remoteBufMem);
57 0 : return HCCL_SUCCESS;
58 : }
59 :
60 0 : HcclResult DevAicpuTsHccsChannel::Create(const void *blob, u64 blobBytes,
61 : const HcommDeviceInfo &deviceInfo, ChannelHandle &outHandle)
62 : {
63 0 : CHK_PTR_NULL(blob);
64 0 : if (blobBytes < sizeof(HcclChannelHccsRes)) {
65 0 : HCCL_ERROR("[DevAicpuTsHccsChannel][Create] blob too small[%llu]",
66 : static_cast<unsigned long long>(blobBytes));
67 0 : return HCCL_E_PARA;
68 : }
69 :
70 0 : const HcclChannelHccsRes &channelHccsRes = *static_cast<const HcclChannelHccsRes *>(blob);
71 0 : const HcclChannelP2p &channelP2p = channelHccsRes.channelP2p;
72 :
73 : // 创建Transport对象
74 0 : MachinePara machinePara;
75 0 : CHK_RET(SetTransportMachinePara(machinePara, channelHccsRes));
76 0 : machinePara.notifyNum = channelHccsRes.p2pNotifyNum;
77 :
78 0 : TransportDeviceP2pData transDevP2pData;
79 0 : transDevP2pData.inputBufferPtr = nullptr;
80 0 : transDevP2pData.outputBufferPtr = nullptr;
81 :
82 : // 获取transportAttr信息
83 0 : transDevP2pData.transportAttr = channelP2p.transportAttr;
84 :
85 : // 创建Transport对象
86 0 : TransportPara para{};
87 0 : const std::unique_ptr<hccl::NotifyPool> notifyPool;
88 :
89 0 : u32 devId = 0;
90 0 : CHK_RET(hrtDrvGetLocalDevIDByHostDevID(channelHccsRes.localDevicePhyId, &devId));
91 : // for data dispatcher read/write with DEFAULT_DISPATCH_NAME
92 0 : DispatcherCtxPtr dispatcherCtx{nullptr};
93 0 : if (!FindDispatcherByCommId(&dispatcherCtx, DEFAULT_DISPATCH_NAME)) {
94 0 : CHK_RET(CreateDispatcherCtx(&dispatcherCtx, devId, DEFAULT_DISPATCH_NAME));
95 : }
96 0 : CHK_PTR_NULL(dispatcherCtx);
97 :
98 0 : DispatcherCtx *ctx = static_cast<DispatcherCtx *>(dispatcherCtx);
99 0 : CHK_PRT(ctx->SetDispatcherHcclQos(channelHccsRes.channelP2p.qos)); // 调度器添加hcclQos
100 0 : CHK_PTR_NULL(ctx);
101 :
102 0 : std::shared_ptr<Transport> transport;
103 0 : transport.reset(new (std::nothrow) Transport(
104 : TransportType::TRANS_TYPE_DEVICE_P2P, para, ctx->GetDispatcher(),
105 0 : notifyPool, machinePara, transDevP2pData));
106 0 : CHK_SMART_PTR_NULL(transport);
107 :
108 0 : CHK_RET(transport->Init()); // 初始化需要增加远端用户注册内存
109 :
110 0 : outHandle = reinterpret_cast<ChannelHandle>(transport.get());
111 0 : HccsSlot slot;
112 0 : slot.dispatcherCtx = dispatcherCtx;
113 0 : slot.transport = std::move(transport);
114 0 : slot.tag = channelHccsRes.channelTag;
115 : {
116 0 : std::lock_guard<std::mutex> lock(mutex_);
117 0 : slots_.emplace(outHandle, std::move(slot));
118 0 : }
119 0 : HCCL_INFO("[DevAicpuTsHccsChannel][%s] transport[%p] create done", __func__, outHandle);
120 0 : return HCCL_SUCCESS;
121 0 : }
122 :
123 0 : bool DevAicpuTsHccsChannel::Destroy(ChannelHandle handle)
124 : {
125 0 : HccsSlot slot;
126 : {
127 0 : std::lock_guard<std::mutex> lock(mutex_);
128 0 : auto it = slots_.find(handle);
129 0 : if (it == slots_.end()) {
130 0 : return false;
131 : }
132 0 : slot = std::move(it->second);
133 0 : slots_.erase(it);
134 0 : }
135 0 : if (slot.transport != nullptr) {
136 0 : (void)slot.transport->DeInit();
137 0 : slot.transport.reset();
138 : }
139 0 : HCCL_DEBUG("[DevAicpuTsHccsChannel][Destroy] destroyed handle[0x%llx]", handle);
140 0 : return true;
141 0 : }
142 : }
|
