See issue #337, especially the comments #337 (comment) and #337 (comment).
Currently the activity graphs in the timeline are computed from the inverted tree: as a result, when we click on "invert tree", we have to compute the inverted trees for all threads. This doesn't seem necessary.
However we need the call node for the activity graphs, and the call node is computed out of the inverted stack table. One comment in #337 mentions that we could invert the call node instead of invert the stack chart, and this sounds like a splendid idea.
As a bonus we could also base the flamegraph from this uninverted call node, and have it always displayed, instead of having to switch the inverted bit back to uninverted.
Possibly the problem is elsewhere too: is it useful to have an inverted stack chart? Do we see inverted stacks in marker tooltips, or in the sample tooltips?
┆Issue is synchronized with this Jira Task
See issue #337, especially the comments #337 (comment) and #337 (comment).
Currently the activity graphs in the timeline are computed from the inverted tree: as a result, when we click on "invert tree", we have to compute the inverted trees for all threads. This doesn't seem necessary.
However we need the call node for the activity graphs, and the call node is computed out of the inverted stack table. One comment in #337 mentions that we could invert the call node instead of invert the stack chart, and this sounds like a splendid idea.
As a bonus we could also base the flamegraph from this uninverted call node, and have it always displayed, instead of having to switch the inverted bit back to uninverted.
Possibly the problem is elsewhere too: is it useful to have an inverted stack chart? Do we see inverted stacks in marker tooltips, or in the sample tooltips?
┆Issue is synchronized with this Jira Task