Some positive feedback on the responsiveness of bzr(explorer)
Alexander Belchenko
bialix at ukr.net
Wed May 11 07:44:34 UTC 2011
Martitza Mendez пишет:
>
>
> On Mon, May 9, 2011 at 10:20 PM, John Arbash Meinel
> <john at arbash-meinel.com <mailto:john at arbash-meinel.com>> wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> ...
> >> Maximum memory consumption during commit was 315MB. Total page
> faults
> >> about 800,000 (on a machine with 2GB RAM).
> >
> > What is "Total page faults"?
>
> A page fault is when a program wants to access a bit of virtual memory
> but it isn't actually present in memory.
> http://en.wikipedia.org/wiki/Page_fault
>
> This happens in a variety of contexts. One is during "swapping", when
> the OS runs out of memory and writes the virtual pages to disk. Then
> when the program accesses them again, it has to read back from disk.
>
> I *think* it might also happen based on how you allocate. I know in
> Linux you can do tons of mallocs(), but until you touch the memory it
> doesn't actually assign real memory to your process. I believe you don't
> hit a page fault during malloc() just during the later access.
>
>
> This set of observations was on WinXP (I should have mentioned that).
> And I'm actually not sure whether or how 'lazy' memory allocation is on
> WIn32. Good question. I suppose I should know. I'll have to test that
> when I get a few minutes.
Which tools are you using to get such statistics? Process Explorer?
--
All the dude wanted was his rug back
More information about the bazaar
mailing list