Query profiles with PerfettoSQL

You can query ProfilingManager profiles using PerfettoSQL and modules from the Perfetto Standard Library. Because ProfilingManager redacts system traces to include only your app's process data, system-wide standard library queries that depend on other processes aren't supported. However, most app-scoped queries and standard library modules work without modification. For additional ready-to-use queries, see Analyzing Android Traces.

ProfilingManager sample queries

Use the following sample queries directly or as templates for your own PerfettoSQL analysis.

Find the most duplicated slices

Repeated slices often point to redundant work. For details on function syntax, see Defining functions in PerfettoSQL. The following query defines a reusable PerfettoSQL function that finds slices matching a GLOB pattern and sorts them by frequency:

-- You only need to call this once in the session to create or update the function
CREATE OR REPLACE PERFETTO FUNCTION find_duplicates(pattern STRING) RETURNS
TABLE(name STRING, count_slice LONG) AS SELECT name, COUNT(dur) as count_slice FROM slice WHERE name GLOB $pattern GROUP BY name HAVING COUNT(name) >= 2 ORDER BY count_slice DESC;

-- Subsequent calls can just use the function to find dupes
SELECT * FROM find_duplicates('*Text*')

Jank queries

Use the following queries to identify slow or janky frames in a system trace.

Find slow frames

This query finds frames that take longer than 16.6 ms (60 Hz) to render. Because Perfetto stores slice durations in nanoseconds, dur is compared against 16660000:

INCLUDE PERFETTO module android.frames.timeline;
SELECT * FROM android_frames WHERE dur > 16660000;

Find jank-causing frames

This query uses the frame timeline module to find frames where the app missed its rendering deadline according to the device's refresh rate:

INCLUDE PERFETTO module android.frames.timeline;
SELECT * FROM actual_frame_timeline_slice WHERE jank_type GLOB '*App Deadline Missed*';

Find most duplicated objects

In a Java heap dump, this query returns the top 100 classes with two or more instances to help you spot unintended duplicates or caching opportunities:

INCLUDE PERFETTO MODULE android.memory.heap_graph.heap_graph_class_aggregation;

SELECT * FROM android_heap_graph_class_aggregation WHERE obj_count >= 2
ORDER BY obj_count DESC LIMIT 100

Cold startup latency

This query estimates cold startup time to first frame (TTFF) by measuring the duration between bindApplication and the first Choreographer#doFrame slice:

-- This function finds slices that match the given GLOB $pattern
CREATE OR REPLACE PERFETTO FUNCTION find_slices(pattern STRING) RETURNS
TABLE (name STRING, ts LONG, dur LONG) AS
SELECT name,ts,dur FROM slice WHERE name GLOB $pattern;

-- This function generates a slice that starts at $startSlicePattern and finishes at the slice matched by $endSlicePattern. If $inclusive is true, then the end slice dur will be added, otherwise, the end slice start time will be used.
CREATE OR REPLACE PERFETTO FUNCTION generate_start_to_end_slices(startSlicePattern STRING, endSlicePattern STRING, inclusive BOOL) RETURNS
TABLE(name STRING, ts LONG, dur LONG) AS
SELECT name, ts, MIN(startToEndDur) as dur
FROM
  (SELECT S.name as name, S.ts as ts, E.ts + IIF($inclusive, E.dur, 0) - S.ts as startToEndDur
  FROM find_slices($startSlicePattern) as S CROSS JOIN find_slices($endSlicePattern) as E
  WHERE startToEndDur > 0)
GROUP BY name, ts;

-- Using these functions we can estimate cold startup time by generating a slice between bindApplication and first frame.
SELECT * from generate_start_to_end_slices('bindApplication','*Choreographer#doFrame [0-9]*', true)