Re: [Fink-users] slow comparative execution time with mpfr on MBP vs Linux

[Martin C. <cos...@wa...>]

S. Newhouse wrote:
> Hello,
> I am running an MBP  with OSX 10.4.11. This has an Intel Core 2 duo at 
> 2.13GHz  I also have a Linux box with Fedora Core 6 and an Intel Core 2 
> duo at 2.13 GHz.   Both have 3 GB RAM.
> I compiled the attached code 'sample_bat.c'  a sample program using gmp 
> and mpfr on both machines with the following execution times.  The code 
> on the  MBP was compiled with cc -O3 and on the Linux bax with cc -O.
> Here are the execution times:
> 
> MBP:  [sen@basie MPFR]$ time ./sample_bat
> Sum is 2.7182818284590452353602874713526624977572470936999
> 
> real    0m2.384s
> user    0m2.242s
> sys     0m0.025s
> 
> 
> Linux:
> [sen@gumbie MPFR]$ time sample_bat
> Sum is 2.7182818284590452353602874713526624977572470936999
> 
> real    0m1.057s
> user    0m1.056s
> sys     0m0.000s
> 
> 
> Anyone know why the MBP is less than half the speed of the Linux box?   

On my new MBP with 2.4GHz I get

real	0m0.896s
user	0m0.892s
sys	0m0.003s

So it's not a general Apple slowness. My guess is that the bus speed on 
your MBP might have something to do with it. You might also consider 
upgrading to Leopard.

I did this also on a slightly older iMac with a core2 duo 2.16 GHz and 
got the same speed as on your linux box. The time was halved when I used 
the libmpfr1-64bit variants:

real	0m0.525s
user	0m0.521s
sys	0m0.002s

-- 
Martin