tree ea570095cf843d8d434ef481e143e78e5a7bb771
parent b98184a9ff697b5706fb331675354a36c7c0fa41
author Lalit Maganti <lalitm@google.com> 1537273712 +0100
committer Lalit Maganti <lalitm@google.com> 1537273712 +0100

trace_processor: collapse sched slice table to a flat table

Instead of storing sched events per-cpu, store them all in a flat map.
This makes queries significantly faster on large traces as we do not
constantly need to search for the next index across CPUs.

Instead we can skip across the vector using std find on a vector<bool>
which is very fast.

While this happens, also remove ts_lower_bound, ts_clip and quantum.
quantum and ts_clip can either be expressed as a combination of a
span join with a window table or a quantum table.

ts_bound is used just for displaying tracks accurately in the UI
and can be replaced by clipping instead.

Change-Id: I1715658b17fcf83c13eebbd8ef01b46f85503b90