Re: [RFC] Optimization for use-once pages

[Patrick Dreker <patrick@dreker.de>]

Am Dienstag, 24. Juli 2001 22:32 schrieb Rik van Riel:
> On Tue, 24 Jul 2001, Patrick Dreker wrote:
>
> 	[snip program with mmap()]
>
> > I have tested this on my Athlon 600 with 128 Megs of RAM, and it
> > does not make any difference whether I use plain 2.4.7 or
> > 2.4.5-use-once.
>
> As expected. Only programs using generic_file_{read,write}()
> will be impacted at the moment.
D'oh... got some thing wrong there it seems :-(
Anyways, I still have some old routines in the program doing the same via 
read(). I have tried that (although the program is pretty silly, as it reads 
the 240megs in 4 byte chunks.... the call overhead probably is the bigger 
problem there...) and it improved the runtime by aproximately 20% using the 
use_once patch.

linux-2.4.7:
22.300u 135.310s 2:41.38 97.6%  0+0k 0+0io 110pf+0w

linux-2.4.5-use_once:
14.980u 108.870s 2:09.79 95.4%  0+0k 0+0io 200pf+0w

Both measurements taken after reboot and while running KDE2.2
As stated: I am still willing to do further experiments... (read()ing larger 
chunks at once?)

-- 
Patrick Dreker