Reduce pss collection amount, improve logging.
Tuned rates that we collect PSS, to reduce how much we do that heavy operation. Added a new way to determine whether a process has changed to a state for the "first" time -- now this is when it has gone to that state for the first time since it was in a lower state. This will reduce the amount of time we consider a process to be first to only when it has previously gone into a higher state than it had before. Keep track of more fine-grained information about why we collect a PSS sample (not just internal, but for a single process, all processes because of a mem state change, all processes because of a poll). Started collecting RSS in various places, so we can start looking at that w.r.t. PSS and see about transitioning to it is a new primary metric. Added logging for many of the places where the system writes its configuration files, so we can more easily see any bad behavior going on in those areas. Added some currently disabled code to read smaps directly instead of using fgets(). Probably won't help, but want tot test. Bug: 70859548 Test: atest CtsAppTestCases Change-Id: I400dba0f3ae9c024df51c946cfa592561028b598
Loading
Please register or sign in to comment