Gentoo Archives: gentoo-amd64

From: Duncan <1i5t5.duncan@×××.net>
To: gentoo-amd64@l.g.o
Subject: [gentoo-amd64] Re: Re: Upgrade from single to dual core CPU
Date: Thu, 26 Jan 2006 09:06:24
Message-Id: pan.2006.01.26.09.04.16.969814@cox.net
In Reply to: Re: [gentoo-amd64] Re: Upgrade from single to dual core CPU by "Hemmann
1 Hemmann, Volker Armin posted
2 <200601252311.22037.volker.armin.hemmann@××××××××××××.de>, excerpted
3 below, on Wed, 25 Jan 2006 23:11:22 +0100:
4
5 > On Wednesday 25 January 2006 20:37, Duncan wrote:
6 >>
7 >> With a gig of physical memory here, and swap entirely turned off for
8 >> some time, my issues weren't with kdebase or kdelibs, but with kdepim,
9 >> before the spit ebuilds, or kmail, since them. With
10 >> USE=kdeenablefinal, there's one spot in the kmail build that takes over
11 >> 700MB for a single build-thread! I watched as the single thread
12 >> gobbled that much, in top.
13 >
14 > swap does not help.
15 >
16 > I had several ooms with tons of swap free...
17 >
18 > (I have posted one of these to lkml and asked for an explanation, but
19 > got none).
20 >
21 > So I use -j1 now.
22 > Has the advantage, that my single core, single cpu box is much more
23 > responsive while compiling away in the background ;)
24
25 Were you the one that posted to that effect some time ago? At the time, I
26 had swap turned off, so I couldn't verify one way or another, but like I
27 said, I've run up to 5 merges at a time, at -j3, including the big kmail
28 with its 700 meg requirement, actually sort of trying to duplicate that
29 OOM someone mentioned before (that it wouldn't go to swap), and somewhat
30 to my surprise, everything went thru.
31
32 I do know that memory isn't as simple as you have it or you don't. There
33 are different zones of memory, and locked memory that can't swap as
34 compared to regular memory that can. At the time of the earlier report, I
35 assumed something was running that couldn't swap. <shrug>
36
37 There's one other factor with KDE, as well. Regular (GNU?) make doesn't
38 make as efficient a use of parallel jobs as unsermake does. My tests
39 above were with regular make, because I haven't been able to get unsermake
40 running correctly such that the KDE builds would use it, for some time. I
41 believe it broke with KDE 3.4, upline, anyway, and they basically gave up
42 on it. From what I've read, 4.0 will use a different make system again,
43 not GNU-make, not unsermake, something else. They gave up on unsermake.
44 Regular GNU-make will AFAIK be usable, as it is with KDE 3, but it's just
45 not terribly efficient on C++ code and when you are compiling that much...
46
47 --
48 Duncan - List replies preferred. No HTML msgs.
49 "Every nonfree program has a lord, a master --
50 and if you use the program, he is your master." Richard Stallman in
51 http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html
52
53
54 --
55 gentoo-amd64@g.o mailing list