R C-R wrote:
You can't tell why memory is being paged out, nor if it is freeing up memory that is immediately used by some other process, at least not with tools like Activity Monitor or iStat. If you don't believe that, do your own research. There have been several references in this long thread to the developer documentation from Apple. If that is too complicated to follow, at least check out the Wikipedia entry for memory leak.
All this "supposed" & "possibly" stuff will get you nowhere. Tracking memory management is not easy even for programmers with access to source code. Just guessing about what is happening from tools never designed to show that is futile, at least if you are looking for answers that could solve your memory use issues.
Someone asked why I keep repeating this stuff or if I think it could help. To that I ask, if you don't know what is causing a problem, how do you expect to eliminate it?
That doesn't matter. I am not running an enterprise server. I have only a couple of applications open with well known "standard" resource requirements that in one case (Safari) more than doubles if compared to SL.
As far as the why goes, it is indeed what we are trying to understand. What we suggest is that it shouldn't happen for the simple reason it wasn't happening before and that it has been noticed immediately. As soon as I upgraded to Lion Safari and the paging went crazy and many other people Which to me is more than a reason for Apple to look into its own code if Apple cares about its wn reputation.
Yesterday I also used "leaks", "vmmap" and "vm_stat". These are the tools suggested by Aplle itself. I've got the link straight from the Activity Monitor help page: http://developer.apple.com/library/mac/#documentation/Performance/Conceptual/Man agingMemory/Articles/AboutMemory.html#//apple_ref/doc/uid/20001880-BCICIHAB
Safari and WebProcess still using little memory were reported by "leaks" to have some small leaks. But I should try with AdBlock on to see if it reports anything more serious. More interesting has been to run "vm_stat" in my last test. It showed how starting another application (Chrome again) the free memory went down, then up getting pages from the inactive memory and paging out some stuff (only a couple of hundreds MB this time). Let's say that this time it looked more like a reasonable behaviour. Thus I will need to repeat the test hoping to reproduce the worst case when Lion pages for nothing.
Obviously we don't have the tools to dig what really is going on. And we are not the Apple developers either. But some analysis is better than no analysis at all, isn't it? Especially if we desire to report the issue to Apple. Just denying it doesn't help at all. Simply because Lion in some machines is performing much worse than Windows 7 (not to mention Snow Leopard) doing the same activities. If it doesn't happen to someone it doesn't mean no bug is there either. There are people with more Macs saying that in one Mac Lion runs flawlessly, in another it is a nightmare.