Perfetto UI enables you to view and analyze traces in the browser. It supports several different tracing formats, including the perfetto proto trace format and the legacy json trace format.
Sometimes you may want to insert some fake slices into the timeline to help with your understanding of the data. You can do so by inserting rows into a magic debug_slices
table.
debug_slices
table has five columns:
id
(integer) [optional] If present, Perfetto UI will use it as slice id to open the details panel when you click on the slices.name
(string) [optional] The displayed slice title.ts
(integer) [required] Start of the slice, in nanoseconds.dur
(integer) [required] Duration of the slice, in nanoseconds. Determines slice width.depth
(integer) [optional] The row at which the slice is drawn. Depth 0 is the first row.You can open the debug track by going to the “Sample queries” menu on the left, and clicking “Show Debug Track”. A debug slice track will become pinned to the top and will initially be empty. After you insert slices in the debug_slices
table, you can click the reload button on the track to refresh the information shown in that track.
Here is a simple example with random slices to illustrate the use:
CREATE VIEW rand_slices AS SELECT * FROM SLICE ORDER BY RANDOM() LIMIT 2000; INSERT INTO debug_slices(id, name, ts, dur, depth) SELECT id, name, ts, dur, depth FROM RAND_SLICES;
After you click the reload button, you should see the slices in the debug track.
Once you're done, you can click the X button to hide the track, and you can clear the debug_slices
table (DELETE FROM debug_slices
) to clear the track.
A more interesting example is seeing RAIL modes in chrome traces:
SELECT RUN_METRIC('chrome/rail_modes.sql'); -- Depth 0 is the unified RAIL Mode INSERT INTO debug_slices SELECT NULL, rail_mode, ts, dur, 0 FROM combined_overall_rail_slices; -- Depth 2+ are for each Renderer process with depth 1 left blank INSERT INTO debug_slices SELECT NULL, short_name, ts, dur, depth + 1 FROM rail_mode_slices, (SELECT track_id, row_number() OVER () AS depth FROM (SELECT DISTINCT track_id FROM rail_mode_slices)) depth_map, rail_modes WHERE depth_map.track_id = rail_mode_slices.track_id AND rail_mode=rail_modes.mode;
This produces a visualization like this:
Note: There is no equivalent debug counters feature yet, but the feature request is tracked on b/168886909).
To use pivot tables in the Perfetto UI, you will need to enable the “Pivot tables” feature flag in the “Flags” tab under “Support” in the Sidebar. You can pop up a pivot table over the entire trace when clicking “p” on your keyboard. The “Edit” button opens a pop up window to add/remove and reorder columns and change the default sorting of aggregations.
Clicking on “Query” generates a table with the selected columns. Table cells with the expand icon can be expanded to show the next column values. The “name (stack)” column displays top level slices that can be expanded to show their descendants down to the last child.
Area selection pops up a pre-filled pivot table restricted over the selected timestamps and track ids.
Some metrics execute at trace load time to annotate the trace with additional tracks and events. You can stop these metrics from running by disabling them in the ‘Flags’ page: