On Thu, 20 Sep 2012 02:29:18 +0200 lee <l...@yun.yagibdah.de> wrote: > Celejar <cele...@gmail.com> writes: > > > Your numbers are much too high. > > Maybe it's because I've been more looking at the virtual memory that top > shows. What's actually resident can be much less. Still: > > > ,---- [ top ] > | PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > > | 14529 lee 20 0 4755m 4.1g 12m S 1.3 52.9 1:27.93 gimp > > | 3956 lee 20 0 1794m 886m 56m S 4.3 11.1 10:29.61 seamonkey-bin > > | 5101 lee 20 0 1010m 866m 49m R 93.1 10.8 64:43.58 x3 > > `---- > > > That's gimp with a scan of a double-page of a paperback (a lot smaller > than an A4 page), converted to rgb colorspace after loading, my normal > seamonkey which I have pretty much always open and X3 running for a > short while. I have "vm.swappiness=80" in /etc/sysctl.conf because I > rather have unused stuff swapped out than keeping it in memory.
Not familiar with gimp or x3; sorry. > X3 will go up to over 3GB virtual and, IIRC, about 1.2--1.5GB resident > if you play long enough. (Don't buy it, it's got too many bugs.) Thanks for the tip ;) ... > If you want to play, compile the attachment with 'gcc -O2 mem.c -o mem' > and run something like './mem 20 20'. That allocates 20MB and does > nothing with it and another 20MB it fills with char 58. That will show > that you can basically allocate as much memory as you want as long as > you don't use it without anything happening and that the system will > start swapping when you allocate enough RAM *and* use it. > > You can try to bring your system down with that, just allocate enough > memory and wait for a critical process to be killed. I've made such > experiments 16 years ago and it worked :) I once filed a bug against the perl script Mail::Box for trying to allocate arbitrarily large amounts of memory when dealing with arbitrarily large mail folders, quickly bringing the system to a crawl. [My 512MB of RAM would be quickly exhausted when browsing a folder with 30,000 messages.] I never waited for the OOM killer to kick in, though ... http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=443259 Celejar -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20120919214221.9168af9c.cele...@gmail.com
