Pivot Tables are a way to summarize and aggregate information about selected slices in a configurable way. It is available in canary and autopush channels by default, and can be enabled in stable by toggling a corresponding flag (Support > Flags > Pivot tables V2).
To create a pivot table, you need to select a timeline area that contains slices. The table will then show an aggregation of the selected slices.
Pivot tables have two types of columns: pivots and aggregations. Conceptually, the UI requests all the data from columns of both types and then computes aggregate values for aggregation columns for every distinct tuple of pivot column values.
Pivots are hierarchical, and the aggregate values are also computed for all the prefixes of pivot column values. For example, if you select process name, category, and event name (in that order) as pivots and have duration sum as the only aggregation columns, the following aggregate values will be computed:
The table rows are appropriately nested in the UI, from more general to more specific. Portions of the table can be collapsed and expanded.
Pivot tables can be configured using dropdown menus in the table header cells. These can be used to: