Linux Benchmark Software?

[William J. Brinkman]

<SNIP>

> lmbench only tries to measure operating system performance, though.  It 
> won't give you any idea how well applications will perform, which is 
> probably the kind of benchmark the original poster wanted.  (Unless the 
> applications of interest to the OP are the sort which spend almost all 
> their time in kernel mode doing the things lmbench does, in which case 
> lmbench is a fine benchmark.  :)

Good point, of course.
 
> P.S.  interesting because the scores are different than OS X in GUI 
> mode? or something else?  I've lmbenched OS X before, in GUI mode, and 
> the only notable thing was how poor the results were compared to Linux 
> on the same hardware.  But that was back before Apple had done much 
> performance optimization, I think it's gotten better since then.  
> They'll have a hard time matching the Linux lmbench scores since they 
> use Mach as part of their kernel; Mach seems to inherently have higher 
> latencies etc. than Linux.  But the scores in 10.0.x and 10.1.x were 
> really bad, so there was a lot of room for improvement.

You can see my results here: 
http://www.derandomized.org/index.php3?section=Home&subsection=System

spaceclam and localhost are both my G3 450 iBook1 SE. What I found 
interesting was how bad the file and VM latencies are under OS X 
10.1.x. When I first got these results, I assumed that lmbench had some 
kind of measurement problem under OS X, and that the numbers were wrong.

But in light of my experience... It seems to me that OS X is very snappy 
and responsive now that I have 310 megs of RAM, but the increase in RAM 
didn't seem to affect my Linux observed performance very much. I wonder if 
the file system really is enough slower in OS X that VM is mostly 
useless.

I really haven't dug into it enough to be certain, but the fact that the 
two memory latency measurements came out so close seems to indicate that 
lmbench knows how to time things in both OS's.

Bo