变更摘要

该 PR 将旧的 TopoMatch1D/TopoMatchMultilevel 拓扑匹配切换为新的 V2 分层匹配体系：新增 TopoMatchBaseV2 基类及 TopoMatchOneLevel、TopoMatchTwoLevel、TopoMatchThreeLevel、TopoMatchConcurrentV2 实现，MatchTopo 接口增加 AlgorithmProfile 参数；新增基于 RankGraph 的 NetInstance/TopoInstance 合成的物理层级链 physicalLevels（含构建、归一化、一致性校验与序列化），并让各 executor 的注册与 CalcAlgHierarchyInfo 调用随之适配；同时补充 POD 机型形态检测与老 CANN/老驱动的兼容降级处理。

主要改动

• 新增 V2 拓扑匹配框架: 新增 TopoMatchBaseV2 基类与 TopoMatchOneLevel/TopoMatchTwoLevel/TopoMatchThreeLevel/TopoMatchConcurrentV2（含 topo_match_base_v2.cc/h 等新文件），MatchTopo 统一增加 const AlgorithmProfile& profile 参数；各 executor 注册由 TopoMatch1D/TopoMatchMultilevel 改为 TopoMatchOneLevel/TopoMatchTwoLevel/TopoMatchThreeLevel，并通过 if constexpr (std::is_base_of_v<TopoMatchBaseV2, AlgTopoMatch>) 分支调用新签名或旧签名。

• 新增物理层级链数据结构与构建流程: alg_param.h 新增 PhysicalLevelInfo、PhysicalLevelView、PhysicalSourceRef、PhysicalLevelIndex、AlgoType 等结构，TopoInfoWithNetLayerDetails 新增 physicalLevels、physicalLevelNum、isPod 字段并接入 Serialize/DeSerialize（带 PHYSICAL_LEVEL_NUM_LIMIT 阈值校验与 EndpointDesc 可平凡复制断言）；新增 physical_level_build.cc/physical_level_normalize.cc 实现 BuildPhysicalLevels、NormalizePhysicalLevels、ValidatePhysicalLevels，任何失败一律降级为空视图，不影响 CalcTopoShape 返回与旧路径。

• 算法画像与调用链打通: 新增 BuildAlgorithmProfile 按算法名推断 algoTypes/engine，executor_v2_base 增加 SetAlgName/algName_，op_common.cc 在 CalcAlgHierarchyInfo 前调用 executor->SetAlgName(param.algName)；AlgHierarchyInfoForAllLevel 新增 physicalIdxForAlgoLevels 并纳入 AlgResourceCtxSerializable 序列化。

• POD 机型形态检测与兼容降级: topo_host.cc 新增 CalcDeviceFormFactor 查询 ACL_DEV_ATTR_DEVICE_FORM_FACTOR 写入 isPod 并接入 CalcTopoShape；hcalrtGetDeviceInfo 增加 quiet 参数（失败改按 WARNING 降噪），adapter_acl.h 以 ACL_DEVICE_FORM_FACTOR_POD 宏探测 HCCL_SUPPORT_DEV_FORM_FACTOR 以兼容老 CANN。

• 构建接入: scatter_aicpu_kernel.cmake 将新增的 topo_match_base_v2.cc、topo_match_one_level.cc、topo_match_two_level.cc、topo_match_three_level.cc、topo_match_concurrent_v2.cc 加入 AICPU 内核编译源文件列表。

代码审查

✅ 未发现问题

Thanks for your pull-request.
The full list of commands accepted by me can be found at here.
You can get sig-info at here.
You can self-configure the PR merge rules for this repository. For more details, please refer to Here.
For more, you also can visit HICANN.

PR Approval Progress

⚠️ This PR does not yet meet the following requirements:lgtm (requires ≥ 2 person(s) per module)、approve (requires ≥ 1 person(s) per module)

Module Approval Details

💡 Tip:

• Committer can comment /approve or /lgtm
• Commenting /approve implies both code review (lgtm) and intent to merge (approve)

CLA Signature Pass

acoolntly, thanks for your pull request. All authors of the commits have signed the CLA. 👍