1 |
Paul de Vrieze posted <200602090955.31941.pauldv@g.o>, excerpted |
2 |
below, on Thu, 09 Feb 2006 09:55:25 +0100: |
3 |
|
4 |
> Besides the fact that memory use is negligeable, you should keep into account |
5 |
> that scripts (even oneliners) use one memory page per script. Aliasses |
6 |
> however are stored by bash in a way that multiple aliases fit into one block |
7 |
> of memory. And when the memory is needed, bash will be bumped out of memory |
8 |
> too. But the idea is that those small aliasses will not actually need more |
9 |
> memory. |
10 |
|
11 |
Ahh... now /that/ makes sense, and is something I hadn't read before. |
12 |
|
13 |
Just to extent your reasoning, are you saying that cached hard drive |
14 |
memory is organized one page (minimum) per file, and that the same general |
15 |
reasoning would therefore apply to caching of /any/ small file (assuming |
16 |
the alternative non-cache memory is tigher packed than a page per file)? |
17 |
|
18 |
I prefer cause/effect relationships of that sort, if possible, to arbitary |
19 |
rules, as cause/effect seems to fit my brain organization and stick with |
20 |
me, where arbitrary rules like bash aliases are more efficient in their |
21 |
memory usage than cached scripts, just don't stick with me, as they have |
22 |
no anchor-point to the cause-effect logical framework I build up to store |
23 |
factoids in. |
24 |
|
25 |
Is there a definitive kernel reference somewhere that states such a thing |
26 |
-- one page per file cached? Also, is that standard 4K pages (and is the |
27 |
AMD64 page size also 4K or is it 8K or something)? |
28 |
|
29 |
-- |
30 |
Duncan - List replies preferred. No HTML msgs. |
31 |
"Every nonfree program has a lord, a master -- |
32 |
and if you use the program, he is your master." Richard Stallman in |
33 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
34 |
|
35 |
|
36 |
-- |
37 |
gentoo-amd64@g.o mailing list |