Welcome, Guest. Please login or register.

Author Topic: Bonnie disk benchmark, really slow "per char"  (Read 3101 times)

Description:

0 Members and 1 Guest are viewing this topic.

Offline FailureTopic starter

  • Lifetime Member
  • Sr. Member
  • ****
  • Join Date: Jun 2004
  • Posts: 332
    • http://awhitlock.net/
Bonnie disk benchmark, really slow "per char"
« on: July 12, 2009, 05:47:50 AM »
I've done an apparently mostly functioning port using ixemul of the old Bonnie disk benchmarking tool to AmigaOS as part of an effort to compare I/O throughput on a few different m68k OS on my A3000.  Not trying to prove anything, just learning.  The Amiga UNIX results are here.  I've done some preliminary trials with NetBSD and it's appreciably faster than Amiga UNIX.  I've wanted to get 3.9 too but I haven't had anything that worked until tonight.

My expectation was that AmigaOS would be fastest.  No real reason, just seemed like it should be.  The block output seems to bear that out.  But the "per char" output is slow with less than half the speed of Amix.  I'm getting about 75KB/sec writes, 26KB/sec rewrite (this is actually block), 103KB/sec read.  47.6% CPU utilization reported for read/write, and 3% only for rewrite.

I guess it's not as bad if I look at the amount of throughput for the CPU used, Amiga UNIX used basically 100% to do the per char tests.  But is this fair to the Amiga?  I didn't run the tests with any special consideration on Amiga UNIX or NetBSD, no renicing, etc.  I definitely haven't done anything with AmigaOS either, I'm such a newb I wouldn't really know where to start anyway.  Any thoughts on this?

Maybe a better port?  :-)  Mine crashes the Amiga when it finishes.  It even crashes WinUAE, heh.
You can\'t spell evil without "vi"
AMIX Wiki | AmixBP