Speed Test Results

[catventure]

Hello Eros,

Here are speed test results for Win98 PentiumII, 333mhz 64mb RAM
Shows differences betweeen latest preview I just d/loaded and the
preview from 14/10/06

TestSpeed
----------
14/10/06
100000 FOR = 1616msec
100000 WHILE = 2187msec
100000 DO WHILE 2317msec
100000 DO UNTIL 2410msec
total = 8410msec
final msg = 4msec

17/10/06
100000 FOR = 1814msec
100000 WHILE = 2308msec
100000 DO WHILE 2488msec
100000 DO UNTIL 2525msec
total = 9135
final msg = 2msec

TestSpeedCountOneMillion
------------------------
14/10/06
7.640000secs

17/10/06
7.850000secs

TestSpeedCountOneMillionNoFunction
----------------------------------
14/10/06
8.180000secs

17/10/06
8.190000secs

TestSpeed2
-----------
14/10/06
2133msecs

17/10/06
2090msecs

TestSpeed2Function
-------------------
14/10/06
8.520000secs

17/10/06
8.240000secs

TestSpeed2Typed
---------------
14/10/06
3.190000secs

17/10/06
3.080000secs

Cheers,
catventure.

[Petr Schreiber]

Hi,

On my brother box, PII 400 MHz, 256MB RAM, I can see speed improvements very similar to the catventures PC.
Very good !

... but my computer needs exorcist !
On my AMD Duron 850 MHz, 320 MB RAM I still get the same or few milliseconds worse results, see here ( old version, new version):


TestSpeed
----------
100000 FOR = 675,742 msec
100000 WHILE = 795,785 msec
100000 DO WHILE 810,785 msec
100000 DO UNTIL 810,790 msec
total = 3090,3102 msec
final msg = 2,0 msec

TestSpeedCountOneMillion
-------------------------
3.35, 3.35 secs

TestSpeedCountOneMillionNoFunction
-----------------------------------
3.18, 3.20 secs

TestSpeed2
-----------
885,957 msecs

TestSpeed2Function
-------------------
2.52,2.75 msecs

TestSpeed2Typed
----------------
1.27,1.31 msecs

Any ideas ? I know Durons were punished by smaller cache than Athlons and other minor breaks, is it possible I'm reaching limits on my box ? I'm absolutely ok with current thinBASIC execution speed, just wonder why it does not move :'(


Bye,
Petr

[ErosOlmi]

Well it seems I got a negative effect >:(

I've worked mainly on changing from memory movements/copying to pointers handling. Most of the time it takes longer to copy some data from here to there rather than declaring a new pointer and using it to access data.

I think the improvements I've got on my PC (about 15% improve) are due to Intel Core Duo technology and better cache handling: it is very effecient in Core Duo processors.

So I have to reconsider some changes.

[Petr Schreiber]

Eros,

you did it good - Intel processors have really noticable improvement !

The "worse" results are caused more by GetTickCount unprecision than anything else ! I think on my Duron it does not have negative effect. Also consider I have old RAM running at 100 MHz only !

Thanks,
Petr