Re: IO speed slipping again?

[Benjamin Kosnik <bkoz at redhat dot com>]

>Hi, all.  For a little while, the mainline library has been giving
>output speeds close to gcc 2.95 (within 10-20%).  Some time in the
>last week or two, the library slowed way down again.  Do any of you
>know what might have happened?  Fallout from Paolo's work?  If not,
>I'll try to do a search and locate the patch.

I saw the slowdown too, but it seems to have come up again. (I see this
across two machines). I suggest trying to isolate libc/kernel/compiler
bits. In particular, somebody need to try the current library versus a
mid-march compiler.

P4/1.8/512MB RHL 9
%rm foo.txt ; time gcc-2.96-sh.exe
6.760u 0.390s 0:07.16 99.8%     0+0k 0+0io 140pf+0w

%rm foo.txt ; time 20030301.exe
6.160u 0.300s 0:07.31 88.3%     0+0k 0+0io 90pf+0w

%rm foo.txt ; time 20030513.exe
6.290u 0.460s 0:11.05 61.0%     0+0k 0+0io 89pf+0w

%rm foo.txt ; time 20030613-sh.exe
5.860u 0.250s 0:06.50 94.0%     0+0k 0+0io 214pf+0w


Athlon/.8/768MB RHL 7.3
%time gcc-2.96.exe 
6.380u 0.760s 0:10.41 68.5%	0+0k 0+0io 29pf+0w

%rm foo.txt ; time 20030316.exe 
7.020u 1.110s 0:12.77 63.6%	0+0k 0+0io 86pf+0w

%rm foo.txt ; time 20030406.exe 
7.040u 1.130s 0:11.99 68.1%	0+0k 0+0io 81pf+0w

%rm foo.txt ; time 20030501.exe 
7.450u 1.230s 0:13.12 66.1%	0+0k 0+0io 83pf+0w

%rm foo.txt ; time 20030608.exe 
7.440u 0.850s 0:10.80 76.7%	0+0k 0+0io 84pf+0w

%rm foo.txt ; time 20030613.out 
7.330u 0.790s 0:08.21 98.9%	0+0k 0+0io 81pf+0w


I'll look at making a make-check rule that records some of the
performance testing automagically so that it's easier for us to compare
notes.

best,
benjamin