| # Trace Processor Library | |
| This directory contains a Python library which can deserialize and extract metrics from binary FXT traces and legacy JSON trace formats. | |
| Library modules include: | |
| 1. `trace_importing` - Code to deserialize legacy Fuchsia trace JSON into an in-memory data model. | |
| 1. `trace_importing_fxt` - Direct ingestion path supporting binary FXT traces utilizing Perfetto Trace Processor SQL queries. | |
| 1. `trace_model` - The definition of the in-memory trace data model. | |
| 1. `trace_time` - Nanosecond-resolution time types for use by the trace data model. | |
| 1. `trace_utils` - Utilities to extract and filter trace events from a data model. | |
| The library is based on the original version of this code located at | |
| `sdk/testing/sl4f/client/lib/src/trace_processing/` and `sdk/testing/sl4f/client/test/` | |
| Binary module includes: | |
| 1. `run_cpu_breakdown` - Standalone processing on a Fuchsia trace JSON that outputs information about CPU metrics. | |
| How to use `run_cpu_breakdown`: | |
| 1. `fx set` with the flag: `--with-host //src/performance/lib/trace_processing:run_cpu_breakdown` | |
| 1. `fx build` | |
| 1. `fx run_cpu_breakdown by_cpu <path to trace JSON> <path to output>` | |
| How to use `run_cpu_breakdown` in aggregate mode: | |
| 1. `fx set` with the flag: `--with-host //src/performance/lib/trace_processing:run_cpu_breakdown` | |
| 1. `fx build` | |
| 1. `fx run_cpu_breakdown by_freq <path to trace JSON> <path to output> --hardware_profile="vim3"` | |
| How to run tests for `cpu_breakdown`: | |
| 1. `fx set` with the flag: `--with-host //src/performance/lib:tests,//src/performance/lib/trace_processing:run_cpu_breakdown` | |
| 1. `fx build` | |
| 1. `fx test --host` |