Chris Radek wrote:

>On Sat, Dec 13, 2008 at 10:14:04PM -0600, Stuart Stevenson wrote:
>  
>
>>Gentlemen,
>>   which style of computation would run faster in the kinematics file?
>>    
>>
>
>If they are equivalent (and I sure as heck didn't check that), 
>
I didn't check for correctness either.  I did notice one thing that 
seems like a bad choice, but it may be intentional:

...
va2=atan(-bskew1)
....
va8=va7*tan(va2)
...

since tan(atan(x)) should equal x, it seems you should be able to just 
use "va8=-va7*bskew1".

>it
>will make little difference whether you split it up like #1 or have
>huge statements like #2.  The compiler will do approximately the same
>work
>
I was in the middle of writing a similar statement, with some general 
instructions on writing a test program, when I decided to just write the 
test program instead :)  I found that the first version runs 1,000,000 
loops in about 0.5 seconds (on a 2.2 GHz Core 2 Duo), whereas the second 
takes about 2 seconds.  By making my own "sqr" function, that was cut 
down to about 1.5 seconds (this surprised me because I would have 
expected common subexpression elimination to work there, but I suppose 
gcc can't know if any of the called math functions save some state and 
may therefore return different values from call to call).

>.  The only concern is which one you can understand better when
>you (or some other unlucky person?) come back to it later.
>  
>
That is a big concern, and the code runs pretty fast anyway.  Like I 
said, my 2.2GHz CPU can run the slower version 500,000 to 750,000 times 
a second, so even a 500 MHz CPU, that's not as well optimized for math, 
should be able to do it 100,000 times a second (or about 100 times as 
often as necessary)

- Steve



------------------------------------------------------------------------------
SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
The future of the web can't happen without you.  Join us at MIX09 to help
pave the way to the Next Web now. Learn more and register at
http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
_______________________________________________
Emc-users mailing list
Emc-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to