On 01/03/2012 02:57 PM, Pádraig Brady wrote:
Just a comment on the accuracy of the numbers.
# ps_mem.py¹ | grep tracker
3.2 MiB + 730.5 KiB = 3.9 MiB tracker-miner-fs
3.9 MiB + 722.5 KiB = 4.6 MiB tracker-miner-flickr
5.5 MiB + 549.0 KiB = 6.0 MiB tracker-store
One of simple, yet least ambiguous methods of determining how many
pages is REALLY used only by a process is to nuke it in a mostly
quiescent system, and compare MemFree just before that and just after that:
# cat /proc/meminfo >/tmp/1; killall tracker-store; sleep 1; cat /proc/meminfo
>/tmp/2; cat /tmp/1 /tmp/2 | grep MemFree
MemFree: 1940372 kB
MemFree: 1963860 kB
As you see, killing it on my machine freed over 23 megs worth of pages.
Note that any shared pages (such as glibc) are not freed this way;
also, non-mapped pages (such as large, but unused malloced space, or large,
but unused file mappings) also do not contribute to MemFree increase.
IOW: this method does not give artificially inflated numbers.
IOW: I do not think I overestimated memory consumption in my original post.
If anything, I probably underestimated it.
I would appreciate if this discussion would not be derailed onto
off-topic tracks of "exactly how many bytes were in use?"
and "my memory measurement method is better than yours".
--
vda
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
