HVRM Insight V3 User Guide【免费下载链接】hcommHCOMMHuawei Communication是HCCL的通信基础库提供通信域以及通信资源的管理能力。项目地址: https://gitcode.com/cann/hcomm1. OverviewHVRM Insight V3 currently mainly supportsDAGViewtask graph viewing. This document focuses on common operations around DAGView, including data preparation, page access, dataset selection, DAG graph browsing, and node search.Other pages and advanced linked capabilities are not yet implemented.2. PrerequisitesBefore using Insight V3, ensure Checker has been executed and Insight data has been generated.To configure Checker data output, ensure Insight dump is enabled in the Checker configuration file:# Configuration file located at /pathto/hccl_vm_install/plugin/checker/manifest.json { ... setting: { // Checker plugin configuration ... enable_insight_dump: true, // Whether to enable visualization data output (disabled by default) enable_memory_snapshot_dump: false // Whether to enable visualization memory snapshot data output (disabled by default, only supported by old Checker, requires enable_insight_dump to be enabled first) } }DAGView requires at least the following data files:dataset_name/ ├── manifest.json └── graph/ ├── graph.msgpack └── layout.msgpackmanifest.jsonis used for reading dataset metadata, whilegraph.msgpackandlayout.msgpackare used for rendering the DAG task graph.3. Compilation and Installation of Insight PluginThe Insight V3 frontend source code is located at:{hccl_vm directory}/src/plugin/insight/frontend_v3Before installing the Insight plugin, first complete the frontend compilation, then execute the HCCL VM build and installation process.Note: Compiling the Insight visualization frontend requiresNode.jsandnpm. It is recommended to useNode.js 20.19.0or later andnpm 10.xor later.Recommended steps:Enter the frontend directory:cd {hccl_vm directory}/src/plugin/insight/frontend_v3Install frontend dependencies:npm installCompile the frontend artifacts:npm run buildAfter frontend compilation completes, thesrc/plugin/insight/dist/directory is generated.Return to the HCCL_VM root directory and executebuild.shto compile and install hccl-vm:cd {HCCL_VM directory} bash build.sh --package-path ASCEND_CANN_PATH --hcomm-path HCOMM_CODE_PATHTo package the installation directory, append--pkg. If the current build scenario requires AICPU / AIV / FULL mode, append--aicpu,--aiv, or--fullas per the projects standard process.4. Insight Plugin Configuration, Start and StopThe Insight plugin configuration file is located at:hccl_vm_install/plugin/visualization/insight/manifest.jsonThe current default configuration is:{ name: insight, version: 1.0.0, entry: python3 server.py, dependency: { min_core_version: 1.0.0 }, setting: { dist_path: ./dist, data_path: ../../checker/data/insight, topo_config_path: ../../../../../asset/cluster_model/config/cluster, port: 8080 } }Common field descriptions:FieldDescriptionentryInsight plugin startup command; currently starts the service viapython3 server.pysetting.dist_pathFrontend static resource directory; reads./distby defaultsetting.data_pathInsight data directory; defaults to Checker outputdata/insightsetting.topo_config_pathCluster topology configuration directorysetting.portInsight service port; defaults to8080To change the port or data directory, directly edit thesettingfields in thismanifest.json.Warning: Insight binds to127.0.0.1by default viapython3 server.py. To listen on a specific IP instead of the defaultlocalhost (127.0.0.1), enter thehccl_vm_install/plugin/visualization/insightdirectory and manually execute:python3 server.py --host ip --port portFor example:cd hccl_vm_install/plugin/visualization/insight python3 server.py --host 0.0.0.0 --port 8080Plugin installation and uninstallation commands:# Install and start the Insight plugin hccl-vm plugin install insight # Uninstall the Insight plugin hccl-vm plugin uninstall insight5. Opening InsightIf the Insight service is already started, open the service address directly in your browser, for example:http://localhost:8080To start Insight via the plugin, execute:hccl-vm plugin install insightAfter the plugin starts, the terminal outputs the access address. Open the address in your browser to enter the Insight page.6. Selecting a DatasetAfter opening Insight, you enter theOverviewpage by default. Follow these steps to select a dataset for analysis:View the dataset list in the center dataset table.Click the target dataset.Confirm the ranks to view in the Rank tree on the left.If the DAG graph is large, reduce the rank selection range first.ClickEnter Correlation Viewon the right.It is recommended to start with a small number of ranks for initial analysis, then expand the rank range after confirming the analysis direction.7. Entering DAGViewAfter entering theCorrelationpage, focus on the lower halfsDAGView Task View.The main areas of the page are:AreaFunctionLeft panelSelect ranks, search nodesCenter bottomDisplay the DAG task graphRight details panelView current node detailsIf the upper memory view is empty, it does not affect the viewing and use of the DAGView below.8. Viewing the DAG Task GraphThe DAG task graph consists of swimlanes, nodes, and arrows:ElementDescriptionSwimlaneRepresents an execution queue, typically displayed asRank / Stream / QueueNodeRepresents a task, such as data copy, Reduce, Record, Wait, etc.ArrowRepresents dependency relationships between tasksLoop dashed boxRepresents a Loop region, enclosed by a dashed box with aLooplabel outsideCommon node types:Node TypeMeaningTRANS_MEMData copy taskREDUCEReduce taskRECORDNotify record taskWAITNotify wait taskCCU_GRAPHCCU graph taskAIV_GRAPHAIV graph taskRecommended viewing approach:Follow the arrow direction to view task dependency order.Click on a task node of interest.View the nodes Rank, Stream, Queue, and detailed information in the right details panel.Continue viewing the nodes parent and child nodes to trace upstream and downstream dependencies.For Loop display, pay additional attention to the following:If a Loop exists in the DAG, the page automatically overlays a Loop dashed box on the task graph.The dashed box covers the Loops start, end, and internal nodes, making it easy to identify loop boundaries.Clicking the Loop capsule directly locates the corresponding Loop Start node.After selecting a Loop internal node, the right details panel shows the Loop information and nested chain for that node.9. Canvas OperationsThe DAGView canvas supports the following operations:OperationDescriptionMove canvasDrag the blank areaZoom in / outUse the mouse wheel, or click/-in the bottom rightReset zoomClick the percentage button in the bottom rightSelect nodeClick the target nodeWhen the graph is large, first reduce the rank selection range, then zoom into the local area to view dependencies.10. Viewing Node DetailsAfter clicking a DAG node, the right details panel displays node information. Common information includes:SectionDescriptionNode OverviewNode ID, task type, Rank, Stream, QueueLoop InfoThe Loop the current node belongs to, Loop count, instruction range, Loop boundaries, etc.Parent NodesUpstream nodes the current node depends onChild NodesDownstream nodes that depend on the current nodeNode SemanticsNotify, Task metadata, Memory Slices, etc.Raw JSONComplete raw information for the current nodeWhen investigating dependency relationships, prioritize checkingParent NodesandChild Nodes. When investigating loop structures, focus onLoop Info. When investigating data copy issues, focus onMemory Slices.TheMemory Slicessection has the following display rules:Normal memory slices displayrank / type / offset / size.If the slice belongs toMS_CCU, Insight V3 preferentially displays theMSIDinstead of the underlying abstract offset.For batch tasks with multipleMS_CCUslices, they are automatically merged into a summary card.The summary card shows a description like8 MSIDs used in total.ExpandingMSID Detailsshows eachMSIDs correspondingidandsize.11. Searching for NodesThe leftSearchpanel can be used to quickly locate DAG nodes.Supported search fields:FieldUse CasetaskIdUse when the node ID is knowntaskTypeSearch by task type, e.g.,TRANS_MEMnotifyIdSearch by notify IDSteps:Select a search field in the left search panel.Enter the complete keyword.Click the search result.DAGView automatically locates and selects the corresponding node.The current search is exact match. If no results are found, confirm the keyword is complete and that the target rank is selected.12. Recommended Analysis FlowWe recommend the following DAGView analysis flow:Select a dataset on theOverviewpage.Select a small number of ranks in the left Rank tree.ClickEnter Correlation View.View the overall nodes and dependency arrows in DAGView.Click nodes of interest and view details on the right.Trace upstream and downstream dependencies through parent and child nodes.Use the search function to quickly locate known nodes.13. Frequently Asked Questions13.1 No Data on the PageConfirm the Insight service address is correct and that Checker has generated Insight data.13.2 DAG is Empty After Entering Correlation PageConfirm the following files exist in the dataset:graph/graph.msgpack graph/layout.msgpack13.3 DAG Graph is Too DenseFirst reduce the number of selected ranks on the left, view only some ranks, then zoom into local areas for analysis.13.4 Cannot Find a Node via SearchConfirm the search keyword is complete and that the target nodes rank is selected.【免费下载链接】hcommHCOMMHuawei Communication是HCCL的通信基础库提供通信域以及通信资源的管理能力。项目地址: https://gitcode.com/cann/hcomm创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考
CANN/hcomm HVRM Insight V3用户指南
HVRM Insight V3 User Guide【免费下载链接】hcommHCOMMHuawei Communication是HCCL的通信基础库提供通信域以及通信资源的管理能力。项目地址: https://gitcode.com/cann/hcomm1. OverviewHVRM Insight V3 currently mainly supportsDAGViewtask graph viewing. This document focuses on common operations around DAGView, including data preparation, page access, dataset selection, DAG graph browsing, and node search.Other pages and advanced linked capabilities are not yet implemented.2. PrerequisitesBefore using Insight V3, ensure Checker has been executed and Insight data has been generated.To configure Checker data output, ensure Insight dump is enabled in the Checker configuration file:# Configuration file located at /pathto/hccl_vm_install/plugin/checker/manifest.json { ... setting: { // Checker plugin configuration ... enable_insight_dump: true, // Whether to enable visualization data output (disabled by default) enable_memory_snapshot_dump: false // Whether to enable visualization memory snapshot data output (disabled by default, only supported by old Checker, requires enable_insight_dump to be enabled first) } }DAGView requires at least the following data files:dataset_name/ ├── manifest.json └── graph/ ├── graph.msgpack └── layout.msgpackmanifest.jsonis used for reading dataset metadata, whilegraph.msgpackandlayout.msgpackare used for rendering the DAG task graph.3. Compilation and Installation of Insight PluginThe Insight V3 frontend source code is located at:{hccl_vm directory}/src/plugin/insight/frontend_v3Before installing the Insight plugin, first complete the frontend compilation, then execute the HCCL VM build and installation process.Note: Compiling the Insight visualization frontend requiresNode.jsandnpm. It is recommended to useNode.js 20.19.0or later andnpm 10.xor later.Recommended steps:Enter the frontend directory:cd {hccl_vm directory}/src/plugin/insight/frontend_v3Install frontend dependencies:npm installCompile the frontend artifacts:npm run buildAfter frontend compilation completes, thesrc/plugin/insight/dist/directory is generated.Return to the HCCL_VM root directory and executebuild.shto compile and install hccl-vm:cd {HCCL_VM directory} bash build.sh --package-path ASCEND_CANN_PATH --hcomm-path HCOMM_CODE_PATHTo package the installation directory, append--pkg. If the current build scenario requires AICPU / AIV / FULL mode, append--aicpu,--aiv, or--fullas per the projects standard process.4. Insight Plugin Configuration, Start and StopThe Insight plugin configuration file is located at:hccl_vm_install/plugin/visualization/insight/manifest.jsonThe current default configuration is:{ name: insight, version: 1.0.0, entry: python3 server.py, dependency: { min_core_version: 1.0.0 }, setting: { dist_path: ./dist, data_path: ../../checker/data/insight, topo_config_path: ../../../../../asset/cluster_model/config/cluster, port: 8080 } }Common field descriptions:FieldDescriptionentryInsight plugin startup command; currently starts the service viapython3 server.pysetting.dist_pathFrontend static resource directory; reads./distby defaultsetting.data_pathInsight data directory; defaults to Checker outputdata/insightsetting.topo_config_pathCluster topology configuration directorysetting.portInsight service port; defaults to8080To change the port or data directory, directly edit thesettingfields in thismanifest.json.Warning: Insight binds to127.0.0.1by default viapython3 server.py. To listen on a specific IP instead of the defaultlocalhost (127.0.0.1), enter thehccl_vm_install/plugin/visualization/insightdirectory and manually execute:python3 server.py --host ip --port portFor example:cd hccl_vm_install/plugin/visualization/insight python3 server.py --host 0.0.0.0 --port 8080Plugin installation and uninstallation commands:# Install and start the Insight plugin hccl-vm plugin install insight # Uninstall the Insight plugin hccl-vm plugin uninstall insight5. Opening InsightIf the Insight service is already started, open the service address directly in your browser, for example:http://localhost:8080To start Insight via the plugin, execute:hccl-vm plugin install insightAfter the plugin starts, the terminal outputs the access address. Open the address in your browser to enter the Insight page.6. Selecting a DatasetAfter opening Insight, you enter theOverviewpage by default. Follow these steps to select a dataset for analysis:View the dataset list in the center dataset table.Click the target dataset.Confirm the ranks to view in the Rank tree on the left.If the DAG graph is large, reduce the rank selection range first.ClickEnter Correlation Viewon the right.It is recommended to start with a small number of ranks for initial analysis, then expand the rank range after confirming the analysis direction.7. Entering DAGViewAfter entering theCorrelationpage, focus on the lower halfsDAGView Task View.The main areas of the page are:AreaFunctionLeft panelSelect ranks, search nodesCenter bottomDisplay the DAG task graphRight details panelView current node detailsIf the upper memory view is empty, it does not affect the viewing and use of the DAGView below.8. Viewing the DAG Task GraphThe DAG task graph consists of swimlanes, nodes, and arrows:ElementDescriptionSwimlaneRepresents an execution queue, typically displayed asRank / Stream / QueueNodeRepresents a task, such as data copy, Reduce, Record, Wait, etc.ArrowRepresents dependency relationships between tasksLoop dashed boxRepresents a Loop region, enclosed by a dashed box with aLooplabel outsideCommon node types:Node TypeMeaningTRANS_MEMData copy taskREDUCEReduce taskRECORDNotify record taskWAITNotify wait taskCCU_GRAPHCCU graph taskAIV_GRAPHAIV graph taskRecommended viewing approach:Follow the arrow direction to view task dependency order.Click on a task node of interest.View the nodes Rank, Stream, Queue, and detailed information in the right details panel.Continue viewing the nodes parent and child nodes to trace upstream and downstream dependencies.For Loop display, pay additional attention to the following:If a Loop exists in the DAG, the page automatically overlays a Loop dashed box on the task graph.The dashed box covers the Loops start, end, and internal nodes, making it easy to identify loop boundaries.Clicking the Loop capsule directly locates the corresponding Loop Start node.After selecting a Loop internal node, the right details panel shows the Loop information and nested chain for that node.9. Canvas OperationsThe DAGView canvas supports the following operations:OperationDescriptionMove canvasDrag the blank areaZoom in / outUse the mouse wheel, or click/-in the bottom rightReset zoomClick the percentage button in the bottom rightSelect nodeClick the target nodeWhen the graph is large, first reduce the rank selection range, then zoom into the local area to view dependencies.10. Viewing Node DetailsAfter clicking a DAG node, the right details panel displays node information. Common information includes:SectionDescriptionNode OverviewNode ID, task type, Rank, Stream, QueueLoop InfoThe Loop the current node belongs to, Loop count, instruction range, Loop boundaries, etc.Parent NodesUpstream nodes the current node depends onChild NodesDownstream nodes that depend on the current nodeNode SemanticsNotify, Task metadata, Memory Slices, etc.Raw JSONComplete raw information for the current nodeWhen investigating dependency relationships, prioritize checkingParent NodesandChild Nodes. When investigating loop structures, focus onLoop Info. When investigating data copy issues, focus onMemory Slices.TheMemory Slicessection has the following display rules:Normal memory slices displayrank / type / offset / size.If the slice belongs toMS_CCU, Insight V3 preferentially displays theMSIDinstead of the underlying abstract offset.For batch tasks with multipleMS_CCUslices, they are automatically merged into a summary card.The summary card shows a description like8 MSIDs used in total.ExpandingMSID Detailsshows eachMSIDs correspondingidandsize.11. Searching for NodesThe leftSearchpanel can be used to quickly locate DAG nodes.Supported search fields:FieldUse CasetaskIdUse when the node ID is knowntaskTypeSearch by task type, e.g.,TRANS_MEMnotifyIdSearch by notify IDSteps:Select a search field in the left search panel.Enter the complete keyword.Click the search result.DAGView automatically locates and selects the corresponding node.The current search is exact match. If no results are found, confirm the keyword is complete and that the target rank is selected.12. Recommended Analysis FlowWe recommend the following DAGView analysis flow:Select a dataset on theOverviewpage.Select a small number of ranks in the left Rank tree.ClickEnter Correlation View.View the overall nodes and dependency arrows in DAGView.Click nodes of interest and view details on the right.Trace upstream and downstream dependencies through parent and child nodes.Use the search function to quickly locate known nodes.13. Frequently Asked Questions13.1 No Data on the PageConfirm the Insight service address is correct and that Checker has generated Insight data.13.2 DAG is Empty After Entering Correlation PageConfirm the following files exist in the dataset:graph/graph.msgpack graph/layout.msgpack13.3 DAG Graph is Too DenseFirst reduce the number of selected ranks on the left, view only some ranks, then zoom into local areas for analysis.13.4 Cannot Find a Node via SearchConfirm the search keyword is complete and that the target nodes rank is selected.【免费下载链接】hcommHCOMMHuawei Communication是HCCL的通信基础库提供通信域以及通信资源的管理能力。项目地址: https://gitcode.com/cann/hcomm创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考
