On 10 Jun 2005 at 14:00, A-NO-NE Music wrote: > I have a suspicion at Tiger itself quite a while. There are at least > 2 other apps I run displays problem similar to what you describe. The > app slows down when Tiger is doing some other task, and the app never > regain speed until it is restarted (not reboot). I have pinged this > to many places but has not received any response yet. Workaround is > easy as restarting the offending app.
Now, I'm not a Mac user, so take this for what it's worth (exactly ZERO), but I know a little bit about how the theory of virtual memory management is supposed to work. What you describe sounds to me like a problem with an application that isn't recovering well from not having a certain portion of the total CPU usage. MIDI, for instance, has real time requirements, and is thus supposed to get pretty high priority (though not as high as something like disk writes will get). What you describe sounds like the OS swapping some memory pages (i.e., triggering a disk write), which is going to be higher priority than the application, and the application is never recovering back to the level of efficiency it had before the OS grabbed CPU cycles for swapping. One thing that has also been something to try on Windows machines with massive amounts of RAM is to play around with the size of the VMM's swap file. Sometimes with very large amounts of real memory, the overhead required to manage it can cause *more* swapping than there would be with less physical RAM. One strategy is to try making the swap file very small, so that things basically stay in RAM. Another setting that many VMMs have is to pre-allocate the file space, rather than making the swap file sizing dynamic. This is usually done by setting a minimum size so that the sectors on the disk are permanently reserved for pages swapped out of memory. Of course, different VMMs are, well, different, and tuning of particular parameters may have different effects (positive on one VMM, negative on another). But I would also suspect the apps that exhibit the problem of not being engineered robustly enough if they can't recover from the OS taking over for memory paging (I'm assuming it's not those apps whose memory pages are being swapped, which would be problematic and shouldn't happen if the apps are properly written). -- David W. Fenton http://www.bway.net/~dfenton David Fenton Associates http://www.bway.net/~dfassoc _______________________________________________ Finale mailing list Finale@shsu.edu http://lists.shsu.edu/mailman/listinfo/finale
