Function Performance¶
When you open a JSON file whose name contains name_map, the extension reads callable_id_to_name and combines it with swimlane data in the same directory to generate a function-level performance table.
Prerequisite¶
The directory containing name_map*.json must also contain the following swimlane file:
chip_swimlane_records.json.
Example:
If no swimlane data is found, the view displays an error. This feature is not a standalone Func ID-to-name mapping browser.
Table columns¶
| Column | Meaning |
|---|---|
| Func ID | Callable function identifier |
| Function Name | Function name from callable_id_to_name |
| Count | Number of task executions for the function |
| Max Duration | Maximum execution duration |
| Min Duration | Minimum execution duration |
| Avg Duration | Average execution duration |
Durations are measured in microseconds. Statistics subtract local setup time and include only actual kernel execution. You can sort columns and adjust their widths.
Recommended workflow¶
- Sort by Avg Duration to find functions with consistently long executions;
- compare Max Duration with Avg Duration to identify duration variability;
- use Count to distinguish expensive individual calls from high invocation counts;
- search for the function name in Chip Swimlane to inspect individual task instances and their scheduling context.