Intel Fortran Compiler

[Joe Griffin]

Hao,

If the code is simple, perhaps 20% is all you
can get.  Regarding accuracy, have you
tried the precision options:

-pc32 -pc64 -pc80
Enables floating-point significand precision control as
follows:
 -pc32 to 24-bit significand
 -pc64 to 53-bit significand, and
 -pc80 to 64-bit significand

Also, if you are on a Xeon, have you tried -axW

Regards,
Joe


 Hao He wrote:
> Dear friends,
> 
> I am trying the Intel Frotran compiler 6.0 for linux on my Pentium 4 Xeon 
> based system now. To my simple testing fortran program, which just 
> contains some multiplication of float point data, the generated code is 
> very fast. However, when I use double precision (real*8) or -r8 option,
> I found that the results showed error after the first 8 significant digits. 
> When I enforced the -O0 option to disable the default optimizaitions,
> the results seemed correct. Unfortunately, the performance droped
> a lot, just a little bit faster than g77 (20% or so).
> 
> Any one has experience on this prooduct?
> My question is: where does the error come from in optimization and
> how can I get better performance while not losing precision?
> Appreciate for your help.
> 
> Best regards,
> Hao He
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 
>