Re: 2.1.32 - performance problems

Peter Leif Rasmussen (peter@udgaard.isgtec.com)
Sun, 6 Apr 1997 17:25:05 -0400

Pavel Machek wrote:
>
>It looks like 2.1.32 has big performance problems on my machine - kernel
>compilation undeer 2.0.29 takes 25 min, on 2.1.32 it takes over 40
>minutes. I have 486DX2/80 with 20MB of ram.
>
I'll second that, and will extend that to the latest 2.1.x kernels. At least
since 2.1.26 have the performance been slower than on 2.0.29. For example a
2.0.29 kernel compile with a 2.0.29 kernel:

Start: Sun Apr 6 13:06:15 EDT 1997
Finish: Sun Apr 6 13:44:07 EDT 1997 Delta Time = 37m 52s

Doing totally the same (date>beg;make dep;make clean;make bzImage;date>end)
when running a 2.1.30 kernel gives:

Start: Sun Apr 6 15:23:38 EDT 1997
Finish: Sun Apr 6 16:33:50 EDT 1997 Delta Time = 1h 10m 12s

This is on a Gateway 2000 486DX2/66 machine with PCI, i.e. a CMD640 chip
controlling a WDC2700H and an AHA1542 controlling a ST1239NS disk. There is
no swapping because of 40MB RAM (70nS DRAM). The basic Linux installation is
Slackware-96.

I am quite surprised by this as I thought Linus mentioned in the beginning of
the 2.1.x series that some interrupt handling had been improved? I would have
expected that to increase performance.

If someone can explain why performance is so and that it was anticipated, I
will be quiet :-)

Thanks,

Peter

--
main(){char*s="O_>>^PQAHBbPQAHBbPOOH^^PAAHBJPAAHBbPA_H>BB";int i,j,k,l,m,n;
for(j=0;j<7;j++)for(l=0;m=l-6+j,i=m/6,n=j*6+i,k=1<<m%6,l<41-j;l++)
putchar(l<6-j?' ':l==40-j?'\n':k&&s[n]&k?'*':' ');}