On Sun, 2016-09-11 at 14:56 +0300, Tanu Kaskinen wrote: > On Sun, 2016-09-11 at 13:30 +0200, Pali Rohár wrote: > > > > On Sunday 11 September 2016 13:10:53 Ralph Benzinger wrote: > > > > > > Actually looking at "what is wrong with system mode?", all security > > > issues are moot for me, as I own all the users on the system. But > > > the point > > > > > >     When in system mode, shared memory data transport is disabled for > > >     security reasons, which means: much higher memory usage and CPU > > >     load in system mode > > > > > > does sound bad. Could someone quantify this statement so that one > > > might see if this is an actual issue (rather than a mere fact) on a > > > modern desktop system? > > I doubt that the statement in the wiki is based on any real benchmarks. > I believe the performance benefit of shm is negligible in most cases. > With very low latencies it starts to matter, though. > > For example, I worked on the Nokia N900, a smartphone with 600 MHz > single-core ARM processor, and we used PulseAudio in the system mode > without shared memory, and the lack of shm caused no performance > problems. The same thing with the later N9 phone. (Sorry for sending so many mails...) Now I realized that benchmarks wouldn't matter anyway - we always disable SHM when the server and client are under different users. -- Tanu
