perfetto-sql

To analyze Android Perfetto traces with SQL, you translate natural language intents into syntactically valid queries and execute them against a local trace file via the trace_processor. This allows you to extract timing, slice, and thread scheduling data directly.

The best way to extract total time spent in specific slices is by converting your intent into a Perfetto SQL query that aggregates slice durations. It enforces safe querying practices, including handling incomplete slices where duration equals -1.

Span_join in Perfetto works by joining tables based on overlapping time intervals to analyze thread scheduling. It requires strict use of schema-discovered tables, correct module includes, and robust join keys like utid or upid for accurate trace analysis.

Yes, you can use natural language to query counter and CPU frequency data from an Android trace. The process translates your intent into valid Perfetto SQL, executing counter lookups and event inspections around specific timestamps via the trace_processor.

You should use GLOB instead of LIKE for string matching in Perfetto SQL because the guided querying practices enforce GLOB-based syntax for pattern matching. This ensures syntactic correctness when filtering slices by name, such as matching RenderThread events.