Hi;
May be this is simple but I need PPC mathematical performans in many cases. Would you please run and post me results with your configuration info (both OS & hardware).

Here are compiled versions:
MOS-any:
http://yyoru.com/mos/TestCPUmath.elf

MOS-PPC:
http://yyoru.com/mos/TestCPUmathPPC

CygWin:
http://yyoru.com/mos/TestCPUmathCYG.exe

DOS:
http://yyoru.com/mos/TestCPUmath.exe



Also you may compile source code;
http://yyoru.com/mos/TestCPUmath.c


PS: SCALE should be 1000

Thanx..
email:yilmaz.yoru@pegasos-tr.com

Hello yyoru,


I've got a PegII, 512MB, 80Gb (8MB cache) HD and Radeon9000Pro 64MB GFX.

Just tried your code.
Meanwhile I was still browsing, so I don't know if that affects your measurements. I got an error-message at the end.
The results from Holland


Timetest CPU Test Program 0.1 by YILMAZ YORU
SCALE (Number of MOPS, Default=1000) :
1000

Wait please ...
Test-1:Long Int Add ... 14.280 sec. 70.028 MOPS
Test-2:Long Int Sub ... 14.260 sec. 70.126 MOPS
Test-3:Long Int Mul ... 14.060 sec. 71.124 MOPS
Test-4:Long Int Div ... 35.340 sec. 28.297 MOPS

Test-5:Double Float Add ... 24.880 sec. 40.193 MOPS
Test-6:Double Float Sub ... 24.740 sec. 40.420 MOPS
Test-7:Double Float Mul ... 24.760 sec. 40.388 MOPS
Test-8:Double Float Div ... 52.680 sec. 18.983 MOPS
Ram Disk:timetest.elf failed returncode 30

Did that test again, this time not touching anything:

Timetest CPU Test Program 0.1 by YILMAZ YORU
SCALE (Number of MOPS, Default=1000) :1000

Wait please ...
Test-1:Long Int Add ... 14.060 sec. 71.124 MOPS
Test-2:Long Int Sub ... 14.080 sec. 71.023 MOPS
Test-3:Long Int Mul ... 14.060 sec. 71.124 MOPS
Test-4:Long Int Div ... 34.180 sec. 29.257 MOPS

Test-5:Double Float Add ... 25.080 sec. 39.872 MOPS
Test-6:Double Float Sub ... 25.100 sec. 39.841 MOPS
Test-7:Double Float Mul ... 25.100 sec. 39.841 MOPS
Test-8:Double Float Div ... 52.240 sec. 19.142 MOPS
Ram Disk:timetest.elf failed returncode 30

Still the error.

Bye,

Spidey

the "error" is because main() doesn't return any value.

yes, u are right, but i dont want to change code, cuz this one is the most friendly one.

Test done on a 2.4 Ghz P4.
It is amazing to see how faster the Cygwin version is for the 2 first integer tests

D:\temp>testcpumath
TestCPUmath CPU Test Program 0.2 by YILMAZ YORU
SCALE (Number of MOPS, Default=1000) :1000

Wait please ...
Test-1:Long Int Add ... 3.797 sec. 263.366 MOPS
Test-2:Long Int Sub ... 3.781 sec. 264.480 MOPS
Test-3:Long Int Mul ... 2.625 sec. 380.952 MOPS
Test-4:Long Int Div ... 25.469 sec. 39.263 MOPS

Test-5:Double Float Add ... 10.078 sec. 99.226 MOPS
Test-6:Double Float Sub ... 10.047 sec. 99.532 MOPS
Test-7:Double Float Mul ... 10.484 sec. 95.383 MOPS
Test-8:Double Float Div ... 18.594 sec. 53.781 MOPS

D:\temp>TestCPUmathCYG.exe
TestCPUmath CPU Test Program 0.1 by YILMAZ YORU
SCALE (Number of MOPS, Default=1000) :1000

Wait please ...
Test-1:Long Int Add ... 2.594 sec. 385.505 MOPS
Test-2:Long Int Sub ... 2.578 sec. 387.898 MOPS
Test-3:Long Int Mul ... 3.453 sec. 289.603 MOPS
Test-4:Long Int Div ... 24.938 sec. 40.099 MOPS

Test-5:Double Float Add ... 10.453 sec. 95.666 MOPS
Test-6:Double Float Sub ... 10.453 sec. 95.666 MOPS
Test-7:Double Float Mul ... 10.407 sec. 96.089 MOPS
Test-8:Double Float Div ... 18.531 sec. 53.964 MOPS
Test Done!

I tried it on a P3 500 and the integer results were still faster than the G4 1000

What compiler did you use ? did you enable optimization ?

Hi, tested it on my Pegasos II G4 1GHz here:

Eagle:Download> TestCPUmathPPC
TestCPUmath CPU Test Program 0.2 by YILMAZ YORU
SCALE (Number of MOPS, Default=1000) :1000

Wait please ...
Test-1:Long Int Add ... 14.320 sec. 69.832 MOPS
Test-2:Long Int Sub ... 14.440 sec. 69.252 MOPS
Test-3:Long Int Mul ... 14.220 sec. 70.323 MOPS
Test-4:Long Int Div ... 35.960 sec. 27.809 MOPS

Test-5:Double Float Add ... 24.500 sec. 40.816 MOPS
Test-6:Double Float Sub ... 24.780 sec. 40.355 MOPS
Test-7:Double Float Mul ... 24.320 sec. 41.118 MOPS
Test-8:Double Float Div ... 52.500 sec. 19.048 MOPS
Test Done!

---

Then I compiled it myself...

Eagle:Download> gcc TestCPUmath.c -o ram:Test
TestCPUmath.c: In function `main':
TestCPUmath.c:35: warning: return type of `main' is not `int'

And tried it again... seems it's the same numbers but 10 times faster so I guess there is some error with one decimal or extra zero in the compiled version on the page? Obviously a G4 at 1GHz has more computing power than a Athlon at 1,466GHz!

So here are the new tests:

RamDisk:> Test
TestCPUmath CPU Test Program 0.2 by YILMAZ YORU
SCALE (Number of MOPS, Default=1000) :1000

Wait please ...
Test-1:Long Int Add ... 1.408 sec. 710.227 MOPS
Test-2:Long Int Sub ... 1.418 sec. 705.219 MOPS
Test-3:Long Int Mul ... 1.410 sec. 709.220 MOPS
Test-4:Long Int Div ... 3.498 sec. 285.878 MOPS

Test-5:Double Float Add ... 2.480 sec. 403.226 MOPS
Test-6:Double Float Sub ... 2.472 sec. 404.531 MOPS
Test-7:Double Float Mul ... 2.500 sec. 400.000 MOPS
Test-8:Double Float Div ... 5.236 sec. 190.985 MOPS
Test Done!

---

And on my PC, a 1.466GHz Athlon (XP1700+) and the DOS version I got:

Wait please ...
Test-1:Long Int Add ... 3.922 sec. 254.972 MOPS
Test-2:Long Int Sub ... 2.813 sec. 355.492 MOPS
Test-3:Long Int Mul ... 3.547 sec. 281.928 MOPS
Test-4:Long Int Div ... 34.906 sec. 28.648 MOPS

Test-5:Double Float Add ... 22.969 sec. 43.537 MOPS
Test-6:Double Float Sub ... 24.812 sec. 40.303 MOPS
Test-7:Double Float Mul ... 22.375 sec. 44.693 MOPS
Test-8:Double Float Div ... 31.391 sec. 31.856 MOPS
Test Done!

I think you made a mistake while compiling. timer_interval should be
100 for AmigaOS, 1000 for WIN/DOS and 100000 for linux.

Black Eagle Do you put 100 for the timer_interval?
G4 still slow. If you dont belive what is interval then take a
stopwatch and measure what any step takes time.
The first parameter (seconds) should be aproximately same as your
stopwatch.

Unfortunately we specially I have a trouble, i know it
should be faster. But what should I have to do. Specially what kinda
of options should I set ?

Would someone able to explain what is wrong, or what kinda code should
I wrote for G4 for the fastest calculation?

Currently still I have 70 MOPS about and 40 MOPS about for the
floating operations with my PegII/G4 1ghz. :(

Thanx.