.or. - expression

.or. - expression

Portrait de faidor
faidor
lun, 15/04/2002 - 15:34

Hello to all,

following question: Because of calculating speed, I would like to know the following. When using the .or.-expression in for example an if-else-statement. How is it internally parsed? Does Fortran first check the first expression and, if it is true, it skips checking for the second? Or does it check both expressions in any case?

Example
if ( (input .eq. 1.0) .or. (input .eq. 2) ) then
do something
end if

-> does Fortran check the frist expression (input=1?) and, if true skips checking the second (input=2?), or does it check both wether the first one is true or not?

I appreciate every answer, thanks a lot,

Andre

RSS Haut
7 posts / 0 nouveau(x)
Dernière contribution
Reportez-vous à notre Notice d'optimisation pour plus d'informations sur les choix et l'optimisation des performances dans les produits logiciels Intel.
Portrait de Steve Lionel (Intel)
Steve Lionel (Intel)
lun, 15/04/2002 - 15:52

The compiler may evaluate either or both subexpressions, in either order, as it sees fit. It might even not evaluate either if it can determine what the result is at compile-time. If you want to force the order of evaluation, or do 'short-circuit' evaluation, use nested IF-THEN.

Steve

Steve
Haut
Portrait de durisinm
durisinm
lun, 15/04/2002 - 15:57

According to the CVF Language Reference Manual, Section 4.1.4, "You should not write logical expressions whose results might depend on the evaluation order of subexpressions. The compiler is free to evaluate subexpressions in any order."

Mike

Haut
Portrait de faidor
faidor
lun, 15/04/2002 - 16:04

Thanks for the fast answer...
I was just too much programming Perl in the last time -> in Perl the order is defined by the programmed order ;-).
that was very helpful.

@Steve: I wanted to circumvent using nested if then, since comparisons like "if"-conditions and loops are the most computing power using elements (in general - afaik), so I wanted to reduce necessary comparisons. Anyhow, I do know something more - thanks a lot!

Andre

Haut
Portrait de Steve Lionel (Intel)
Steve Lionel (Intel)
lun, 15/04/2002 - 16:39

There really isn't any difference between nested IFs and compound expression IFs. If anything, the compiler has a better chance of optimizing well with nested IFs.

You might want to see if the SELECT CASE statement is appropriate to your application - if the case values are from a small set of integers, the compiler can generate special optimized code for the selection.

Steve

Steve
Haut
Portrait de faidor
faidor
lun, 15/04/2002 - 16:47

Thanks, Steve, I'll keep that in mind and check it out. Is this special optimized code also usable for non-equal cases with ocmparisons to real numbers? Like greater than, lower than? I would have to compare some values to be smaller than 1.00001 and bigger than -0.00001. Normally it would be ok to use 0 and 1, but I have to deal with rounding errors ;-).
Thanks a lot, greetz form germany,

Andre

Haut
Portrait de Steve Lionel (Intel)
Steve Lionel (Intel)
lun, 15/04/2002 - 17:18

If that's what you're doing, then SELECT CASE isn't for you. When comparing reals, you probably want something like this:

if (abs(x-1.0) < 0.00001) then ...



Steve

Steve
Haut

Connectez-vous pour laisser un commentaire.