Welcome, Guest. Please login or register.

Author Topic: Bogomips in AmigaOne Linux  (Read 3624 times)

Description:

0 Members and 1 Guest are viewing this topic.

  • Guest
Re: Bogomips in AmigaOne Linux
« Reply #29 from previous page: January 09, 2003, 07:29:40 PM »
Quote

Ryu wrote:
LOL the funny thing is as far as I am aware naTmeg wrote AmigaMARK, heck I betatested it for him under 68k here.


Well in that case ok ;) But then it must not be the same algorithm than the one use in Linux 2.4.18 or maybe Linux use some G3/G4 optimisations (compiling or in the algorithm) to get the best result but maybe these optimizations are not yet implemented in the AmigaMARK MOS version of this test...

At least I know that on PC the Bogomips test is performed using different algorithms that are each optimised for MMX, SSE or other CPU specific instructions and keep the best result.
 

Offline cdfr

  • Full Member
  • ***
  • Join Date: Jul 2002
  • Posts: 131
    • Show only replies by cdfr
Re: Bogomips in AmigaOne Linux
« Reply #30 on: January 09, 2003, 07:39:02 PM »
Right, the Morphos version exists and is available for a limited number of betatesters.  ;-)
Benchmarking is very difficult because there are a lot of different things to test and different ways to do it. I remember that Bytemarks were always better for PPC and SPARC, that SPEC (sponsored by Intel) was better for Intel (how strange!) ...
It really depends what compiler your use, what optimization are done etc.
That's why fist you need to compare apples with apples: execute the same code in the same conditions.
For example I don't think that comparing a system speed and an emulated system speed specialy with JIT) make any sense to have an idea of general speed difference.
Even in general, you can not say system X is 4 X faster than system Y.
For what ? CPU intense stuff ? My CPU is idling 99% of the time anyway !
I remember that I had to do a search replace in a huge text file at work and that the only tool I had was MS word. It completely froze my computer for hours. I took the same file and did the search / replace on my amiga with CED and it was finished in a few minutes.
Who in the PC world could think that some people still use 50 Mhz machine to browse internet, do mail etc ?
What counts is the user experience on the machine. Does it do the job ? Is it responsive. Does it get me mad or not ?
BTW Natmeg is doing an excelent job. Amigamark is going to be THE Amiga benchmark.   :-D  
Benchmark are fun but never make any general statement after looking at a benchmark. For example look at the bogomips test and the G3/604 and then look at dhrystones. What does it tells you ?
 

Offline naTmeg

  • Newbie
  • *
  • Join Date: Jan 2003
  • Posts: 3
    • Show only replies by naTmeg
    • http://member.ycn.com/~hausrup
Re: Bogomips in AmigaOne Linux
« Reply #31 on: January 09, 2003, 08:24:31 PM »
to be exact here the routines used in the two version.
its a assembler code build-in in the c-source.
here the functions:


m68k:

static void delay(ULONG loops)
{
   __asm__ __volatile__ ("\n\tmovel %0,%/d0\n1:\tsubql #1,%/d0\n\t"
   "bpls 1b\n"
               : /*no outputs*/
               : "g" (loops)
               : "d0");

}


ppc:

static void delay(ULONG loops)
{
   __asm volatile ("cmpwi 0,%0,0\n\t"
                   "mtctr %0\n\t"
                   "beqlr\n\t"
                   ".loop: bdnz .loop"
                   :
                   : "r" (loops)
                   : "cc");
}


hope this is clear now :)

naTmeg,










 
 

Offline cdfr

  • Full Member
  • ***
  • Join Date: Jul 2002
  • Posts: 131
    • Show only replies by cdfr
Re: Bogomips in AmigaOne Linux
« Reply #32 on: January 09, 2003, 09:50:18 PM »
hum, it looks like the 604 results are comparable to what we get on Linux/Apus.
The G3 results are more than suspicious.  :-?
 

Offline Seehund

  • Hero Member
  • *****
  • Join Date: Feb 2002
  • Posts: 1230
    • Show only replies by Seehund
    • http://AmigaPOP.8bit.co.uk/
Re: Bogomips in AmigaOne Linux
« Reply #33 on: January 09, 2003, 11:06:12 PM »
Quote

lorddef wrote:
My Dual P100 system gets 492 bogomips, My AMD Athlon 1.4Ghz gets 5870 bogomips.


You got some super-CPUs there? :)
One P100 gets about 40 BogoMIPS, and my dual AthlonMP@1752MHz get 3500 BogoMIPS per CPU.

Check out the BogoMIPS ratings listed in the BogoMIPS mini-HOWTO.

I can't believe we're discussing busy-loop timings under Linux! :)
[color=0000FF]Maybe it\\\'s still possible to [/color]save AmigaOS [color=0000FF][/size][/color]  :rtfm:......
 

Offline Hammer

  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 1996
  • Country: 00
    • Show only replies by Hammer
Re: Bogomips in AmigaOne Linux
« Reply #34 on: February 13, 2003, 10:36:27 PM »
Has any one compared JIT-68K speeds between  PPC-JIT-68K VS X86-JIT-68K?
Amiga 1200 PiStorm32-Emu68-RPI 4B 4GB.
Ryzen 9 7900X, DDR5-6000 64 GB, RTX 4080 16 GB PC.
 

Offline Hammer

  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 1996
  • Country: 00
    • Show only replies by Hammer
Re: Bogomips in AmigaOne Linux
« Reply #35 on: February 14, 2003, 04:46:45 AM »
Just a side note; my old AthlonXP1800+/WinXP-SP1**/JIT-WinUAE(0.8.22.R4, JIT-020/881) delivers ~725.2 Bogo  MIPS(via 68K Amiga Mark 2003 beta).

**with XP Pro's bloatware services turned on.
Amiga 1200 PiStorm32-Emu68-RPI 4B 4GB.
Ryzen 9 7900X, DDR5-6000 64 GB, RTX 4080 16 GB PC.