I have started running benchmarks on several of my NetBSD machines.

I found something very curious.

I booted the Kernel to stop before loading any daemons and performed the   
tests with drystone compiled from pkgsrc.

dry2reg performed only a bit faster than dry2 in some tests.  nice -20   
had little affect on performance.

Celeron-300
10,000,000 cycles
542,005 drystones - 1.8 sec. (dry2)
542,229 drystones - 1.8 sec. (nice -20 dry2)
541,125 drystones - 1.8 sec. (dry2reg)

Intel Pentium-75
1,000,000 cycles
100,300 drystones - 10 sec. (dry2)
100,100 drystones - 10 sec. (nice -20 dry2)
100,700 drystones - 9.9 sec. (dry2reg)

Alpha Multia UDB 166
1,000,000 cycles
56,000 drystones (dry2) (!?!?!)
(I forgot to write the rest down, but you get the idea)

Why is the Alpha so much slower that the P-75?
I realize that the Multia was a bit watered down, as Alphas go; but   
nearly 1/2 as fast as a P-75?

What gives?  Is it the compiler or NetBSD or the computer?
Can someone with more regular Alpha in the 166Mhz range run dry2 and   
reinstill my faith in the Alpha?

John A. Maier
Kemper Military School and College
Director of MIS
Voice: 660-882-5623
Fax: 660-882-3778
Email: johnam@kemper.org
Web: www.kemper.org