Kurti 64Bit Bench

[Omnicron]

Hi, ich wollte mal die Geschwindigkeitsunterschiede zwischen 32 und 64 Bit Operationen testen. Jetzt habe ich noch ein paar Sachen dazugepackt und es Bench genannt. ;D

Der Bench berechnet folgende Sachen:

Bei Integer64 Addition, Subtraktion, Multiplikation und Dividerung mit 64 Bit Werten. Bei der 32 Bit Version müssen die natürlich durch mehrere Befehle gemacht werden, dementsprechend langsamer.

Bei Integer32 einfach mit 32 Bit Werten, bei beiden Versionen gleich.

Triangle berechnet die Länge einer Dreiecksseite, 32 Bit Werte, aber 64 Bit CPUs haben hier den Vorteil ihrer zusätzlichen Register um Werte zwischenzuspeichern.

Mem copy kopiert 50 MB im Speicher hin und her, wenn vorhanden mit SSE2, andernfalls mit SSE.

Mem chksum berechnet eine 64 Bit checksumme aus den 50 MB. 64 Bit CPUs haben hier wieder einen Vorteil.

FPU Planets berechnet das Volumen der Planeten in unserem Sonnensystem.
FPU Box-Muller berechnet Zufallsvariablen nach der Box-Muller Formel.
In beiden Versionen gleich.

In dem Zip Archiv ist die 32 und 64 Bit Version. Die 64er läuft natürlich nur unter Win64 mit 64 Bit CPUs.

http://rapidshare.de/files/13071636/Kurti64.rar.html

Die Ergebnisse werden auch in eine result32/64.txt geschrieben.

Hier sind auf meinem Opteron@2600Mhz einmal die 64 und 32 Ergebnisse:


Kurti64 V1.0 Results

Integer64:
===========
>>> Addition: 0.967237 secs
>>> Subtraction: 0.932058 secs
>>> Multiplication: 1.756382 secs
>>> Division: 9.253612 secs

> Total: 12.909289 secs
Integer32:
===========
>>> Addition: 0.823172 secs
>>> Subtraction: 0.813242 secs
>>> Multiplication: 1.317948 secs
>>> Division: 5.047155 secs

> Total: 8.001516 secs

>>> Triangle: 1.629597 secs

Memory:
========
>>> Mem copy: 7.629754 secs | 1311 mb/s
>>> Mem chksum: 2.982232 secs

> Total: 10.611986 secs

FPU:
=====
>>> Planets: 5.291297 secs
>>> Box-Muller: 8.792708 secs

> Total: 14.084005 secs


>>>>>>> TOTAL: 47.236393 secs


Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 1.398841 secs
>>> Subtraction: 1.399580 secs
>>> Multiplication: 5.284890 secs
>>> Division: 19.650450 secs

> Total: 26.334181 secs

Integer32:
===========
>>> Addition: 0.815032 secs
>>> Subtraction: 0.852972 secs
>>> Multiplication: 1.278787 secs
>>> Division: 5.049225 secs

> Total: 7.996015 secs

>>> Triangle: 1.868646 secs

Memory:
========
>>> Mem copy: 7.450896 secs | 1342 mb/s
>>> Mem chksum: 4.163993 secs

> Total: 11.614889 secs

FPU:
=====
>>> Planets: 5.330923 secs
>>> Box-Muller: 8.761853 secs

> Total: 14.092777 secs


>>>>>>> TOTAL: 61.906508 secs

[misterh]

hier unter WinXP-32

A64 3000+ @ 2502Mhz
DDR400 @ DDR500 2-2-2-5


Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 1.474402 secs
>>> Subtraction: 1.475242 secs
>>> Multiplication: 6.161316 secs
>>> Division: 20.510647 secs

> Total: 28.146366 secs

Integer32:
===========
>>> Addition: 0.859996 secs
>>> Subtraction: 0.858667 secs
>>> Multiplication: 1.350138 secs
>>> Division: 5.269942 secs

> Total: 8.338743 secs

>>> Triangle: 1.971733 secs

Memory:
========
>>> Mem copy: 3.985388 secs | 2509 mb/s
>>> Mem chksum: 3.315388 secs

> Total: 7.300775 secs

FPU:
=====
>>> Planets: 5.474223 secs
>>> Box-Muller: 9.246392 secs

> Total: 14.720614 secs


>>>>>>> TOTAL: 60.478231 secs

[Jasch]

X2@2,7 Winxp x64


Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 1.336105 secs
>>> Subtraction: 1.336151 secs
>>> Multiplication: 5.330131 secs
>>> Division: 20.282108 secs

> Total: 26.948344 secs

Integer32:
===========
>>> Addition: 0.779252 secs
>>> Subtraction: 0.779190 secs
>>> Multiplication: 1.225291 secs
>>> Division: 4.787720 secs

> Total: 7.571453 secs

>>> Triangle: 1.770258 secs

Memory:
========
>>> Mem copy: 5.469393 secs | 1828 mb/s
>>> Mem chksum: 3.621283 secs

> Total: 9.090677 secs

FPU:
=====
>>> Planets: 4.999724 secs
>>> Box-Muller: 8.318294 secs

> Total: 13.318018 secs


>>>>>>> TOTAL: 58.698749 secs

Kurti64 V1.0 Results

Integer64:
===========
>>> Addition: 0.890825 secs
>>> Subtraction: 0.890751 secs
>>> Multiplication: 1.670948 secs
>>> Division: 9.784962 secs

> Total: 13.237487 secs

Integer32:
===========
>>> Addition: 0.779128 secs
>>> Subtraction: 0.779945 secs
>>> Multiplication: 1.224791 secs
>>> Division: 4.787390 secs

> Total: 7.571254 secs

>>> Triangle: 1.558694 secs

Memory:
========
>>> Mem copy: 5.629267 secs | 1776 mb/s
>>> Mem chksum: 2.937280 secs

> Total: 8.566547 secs

FPU:
=====
>>> Planets: 4.977398 secs
>>> Box-Muller: 8.326869 secs

> Total: 13.304267 secs


>>>>>>> TOTAL: 44.238249 secs


[/B][/QUOTE]

[VoodooJack]

Kurti64 V1.0 Results

Integer64:
===========
>>> Addition: 0.792241 secs
>>> Subtraction: 0.794056 secs
>>> Multiplication: 1.490935 secs
>>> Division: 7.932170 secs

> Total: 11.009402 secs

Integer32:
===========
>>> Addition: 0.692680 secs
>>> Subtraction: 0.691182 secs
>>> Multiplication: 1.087370 secs
>>> Division: 4.249368 secs

> Total: 6.720600 secs

>>> Triangle: 1.382760 secs

Memory:
========
>>> Mem copy: 4.866767 secs | 2055 mb/s
>>> Mem chksum: 2.445887 secs

> Total: 7.312654 secs

FPU:
=====
>>> Planets: 4.349952 secs
>>> Box-Muller: 7.377723 secs

> Total: 11.727675 secs

>>>>>>> TOTAL: 38.153091 secs

FX-57 @3.05 - WinXP x64

[RoNsOn Xs]

Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 1.817105 secs
>>> Subtraction: 2.246523 secs
>>> Multiplication: 7.786283 secs
>>> Division: 22.976505 secs

> Total: 32.579893 secs

Integer32:
===========
>>> Addition: 1.112694 secs
>>> Subtraction: 1.108738 secs
>>> Multiplication: 1.940439 secs
>>> Division: 5.974625 secs

> Total: 10.136496 secs

>>> Triangle: 3.282489 secs

Memory:
========
>>> Mem copy: 15.501284 secs | 645 mb/s
>>> Mem chksum: 10.995821 secs

> Total: 26.497105 secs

FPU:
=====
>>> Planets: 6.103201 secs
>>> Box-Muller: 10.103114 secs

> Total: 16.206315 secs


>>>>>>> TOTAL: 88.702298 secs

[AnarchX]

P4 3300Mhz
220Mhz FSB

Win XP 32bit

Integer64:
===========
>>> Addition: 1.631048 secs
>>> Subtraction: 1.652855 secs
>>> Multiplication: 13.874236 secs
>>> Division: 29.889871 secs

> Total: 45.395156 secs

Integer32:
===========
>>> Addition: 0.792061 secs
>>> Subtraction: 0.794950 secs
>>> Multiplication: 4.950114 secs
>>> Division: 3.025967 secs

> Total: 9.563092 secs

>>> Triangle: 4.396018 secs

Memory:
========
>>> Mem copy: 4.947455 secs | 2021 mb/s
>>> Mem chksum: 3.966496 secs

> Total: 8.913952 secs

FPU:
=====
>>> Planets: 22.345493 secs
>>> Box-Muller: 16.311645 secs

> Total: 38.657139 secs


>>>>>>> TOTAL: 106.925357 secs


:uponder:

[Raff]

Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 1.686710 secs
>>> Subtraction: 2.078084 secs
>>> Multiplication: 7.263171 secs
>>> Division: 21.404168 secs

> Total: 30.354049 secs

Integer32:
===========
>>> Addition: 1.037661 secs
>>> Subtraction: 1.037243 secs
>>> Multiplication: 1.815360 secs
>>> Division: 5.576419 secs

> Total: 9.466684 secs

>>> Triangle: 3.047467 secs

Memory:
========
>>> Mem copy: 12.537179 secs | 798 mb/s
>>> Mem chksum: 9.510589 secs

> Total: 22.047768 secs

FPU:
=====
>>> Planets: 5.712118 secs
>>> Box-Muller: 9.421824 secs

> Total: 15.133942 secs


>>>>>>> TOTAL: 80.049911 secs

AXP Tbred @ ~210x11 = 2315 MHz, WinXP Home.

Kommt mir zu schnell vor.

MfG,
Raff

[Crazy_Chris]

Es wäre schön wenn du auch den Quellcode veröffentlichen würdest. :biggrin:


Ist ja krass wie lahm der P4 ist. :eek:

[Raff]

Ok, jetzt noch mit dem Maximum:

Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 1.612579 secs
>>> Subtraction: 1.988713 secs
>>> Multiplication: 6.947544 secs
>>> Division: 20.472297 secs

> Total: 29.032420 secs

Integer32:
===========
>>> Addition: 0.992170 secs
>>> Subtraction: 0.992242 secs
>>> Multiplication: 1.736563 secs
>>> Division: 5.333804 secs

> Total: 9.054779 secs

>>> Triangle: 2.914717 secs

Memory:
========
>>> Mem copy: 12.695743 secs | 788 mb/s
>>> Mem chksum: 9.388523 secs

> Total: 22.084267 secs

FPU:
=====
>>> Planets: 5.438847 secs
>>> Box-Muller: 9.013288 secs

> Total: 14.452135 secs


>>>>>>> TOTAL: 77.538318 secs

AXP Tbred @ ~210x11 = 2420 MHz, WinXP Home.

MfG,
Raff

[Black-Scorpion]

Venice 3000+ (1800MHz) und WinXP Home
Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 3.606316 secs
>>> Subtraction: 3.629894 secs
>>> Multiplication: 6.383709 secs
>>> Division: 28.306962 secs

> Total: 38.296986 secs

Integer32:
===========
>>> Addition: 1.171310 secs
>>> Subtraction: 1.187834 secs
>>> Multiplication: 1.834911 secs
>>> Division: 7.218560 secs

> Total: 11.412615 secs

>>> Triangle: 2.685272 secs

Memory:
========
>>> Mem copy: 5.615220 secs | 1781 mb/s
>>> Mem chksum: 4.439630 secs

> Total: 10.054849 secs

FPU:
=====
>>> Planets: 7.509736 secs
>>> Box-Muller: 12.746653 secs

> Total: 20.256388 secs


>>>>>>> TOTAL: 82.706111 secs

[AnarchX]

PIII 0.5 Ghz

XP 32bit



Kurti32 V1.0 Results

Integer64:
===========
>>> Addition: 12.754888 secs
>>> Subtraction: 13.348752 secs
>>> Multiplication: 34.672624 secs
>>> Division: 116.800114 secs

> Total: 164.227627 secs

Integer32:
===========
>>> Addition: 7.018648 secs
>>> Subtraction: 7.037024 secs
>>> Multiplication: 10.322372 secs
>>> Division: 24.025037 secs

> Total: 48.403082 secs

>>> Triangle: 15.649182 secs

Memory:
========
>>> Mem copy: 66.292592 secs | 151 mb/s
>>> Mem chksum: 60.706153 secs

> Total: 126.998745 secs

FPU:
=====
>>> Planets: 15.912256 secs
>>> Box-Muller: 33.936216 secs

> Total: 49.848472 secs


>>>>>>> TOTAL: 405.127106 secs

>>> Planets: 22.345493 secs


Bei FPU\Planets scheint definitv etwas nicht zu stimmen ;)