Computing Memory Heatmap

The visualize_data.bin file generated by msprof op can be visualized using MindStudio Insight. The page presents basic operator information, computing workload analysis data, and memory workload analysis data by resource, allowing developers to identify resource bottlenecks from a comprehensive perspective.

• To use MindStudio Insight, you need to install the MindStudio Insight software package separately. For details about the download link, see"Installation and Uninstallation".
• For details about how to import the visualize_data.bin file to MindStudio Insight, see Importing Profile Data.
• For details about how to use MindStudio Insight, see .

Figure 1 Details page 1

• Core Occupancy displays the time consumption, total data throughput, and cache hit ratio of each physical core in a data pane, allowing developers to improve the usage efficiency of physical cores.
  

  • This function is supported only by Atlas A3 training products / Atlas A3 inference products and Atlas A2 training products / Atlas A2 inference products .
  • The number of cores displayed depends on the hardware.
• Roofline: For details, see Roofline Bottleneck Analysis Chart.
• Compute Workload Analysis displays computing workload data in bar charts and data tables, helping developers analyze whether Cube/Vector computing resources are fully utilized.
• Memory Workload Analysis displays the active bandwidth of each MTE channel (the active bandwidth of MTE1 and MTE2 channels on the Cube is not displayed when dynamic instrumentation is disabled). The number of requests, transfer bandwidth, and usage of each channel are displayed in memory heatmaps and data panes, helping developers analyze channels that may have bottlenecks.
  

  • The content displayed in data panes varies depending on the operator type.
  • The active bandwidth display function is not supported by Atlas inference products .
  • The maximum bandwidth ratio display is not supported by Atlas A3 training products / Atlas A3 inference products .