[lug] Really off topic

[Sebastian Sobolewski ( Zeb )]

         Hi, I've been writing an Acoustics Analysis program at work and 
here are a few things that I've found:

         My own tests with standard optimized FFT routines vs. ones using 
MMX to speed up processing showed me that the MMX routines are 2-3 times 
faster.

         More then likely SETI at Home was compiled and is using the Intel® 
Signal Processing Library.  This library has highly optimized functions for 
every Intel CPU.  On PII, PII and Celeron CPU'S it uses MMX based functions 
for Fourier and most other floating point intensive calculations. 
(Wavelets, filters ..etc)

         The MMX instruction set is also much faster at this then the 
standard Floating Point unit in AMD K-series chips (Which is actually 
slower then a PII @ same clock rate); with the exception of the Athelon 
CPU. The Athelon has a lightning fast FPU and additional SIMD/MMX 
instructions.

         So more then likely it's because your K6-2 CPU is simply slower at 
Floating Point OPS.

Sebastian


At 05:58 PM 2/24/00 , you wrote:
>I know this doesn't relate to Linux, but I don't think I could find a
>more knowledgeable group to put this question to.  I am running an AMD
>K6-2 400 with 64 meg at home, at work Im using a pentium II 266 64 meg
>running win NT 4.  I started processing data from seti at home at work, and
>it took ~ 15 hours per work unit.  I thought it would be neat to see
>what it would do on my home machine (on both Linux and NT). I was
>disappointed to say the least... the AMD running NT took 25 hours, and
>running Linux it took ~ 14 hours.  As I understand they are doing a
>fourier analysis on the data.  Now with all that out of the way... does
>anyone know why the AMD might perform so poorly?
>It seems to everything else pretty well.
>
>Hope I didn't upset to many with this question.
>
>Whit
>
>
>_______________________________________________
>Web Page:  http://lug.boulder.co.us
>Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug