|[6 earlier articles]|
|Re: Loop jamming!? email@example.com (Stanley Chow) (1997-07-21)|
|Re: Loop jamming!? firstname.lastname@example.org (Christopher Glaeser) (1997-07-22)|
|Re: Loop jamming!? email@example.com (1997-07-27)|
|Re: Loop jamming!? firstname.lastname@example.org (1997-07-28)|
|Re: Loop jamming!? email@example.com (Steve Simmons) (1997-07-29)|
|Re: Loop jamming!? firstname.lastname@example.org (Stanley Chow) (1997-07-31)|
|Re: Loop jamming!? email@example.com (Jan Vorbrueggen) (1997-07-31)|
|Re: Loop jamming!? cliff.click@Eng.Sun.COM (cliffc) (1997-08-07)|
|Re: Loop jamming!? firstname.lastname@example.org (Mike Kent) (1997-08-07)|
|Re: Loop jamming!? email@example.com (Stanley Chow) (1997-08-09)|
|From:||Jan Vorbrueggen <firstname.lastname@example.org>|
|Date:||31 Jul 1997 19:35:16 -0400|
|Organization:||Institut fuer Neuroinformatik, Ruhr-Universitaet Bochum, Germany|
|References:||97-07-089 97-07-102 97-07-108 97-07-118|
|Keywords:||optimize, parallel, Fortran|
email@example.com (Gene Wagenbreth) writes:
> If you are after high performance, f90 array syntax can get in your way. The
> compiler must translate the f90 to the equivalent of an f77 DO loop while
> generating code. It must perform loop jamming, and sometimes array demotion.
> Performance depends on how the compiler does this. The user has no control,
> and usually can not find out what the compiler has done. Rewriting the code
> by hand in f77 Do loops is often the best thing to do.
Well, the the whole point of using f90's array operations is lost,
isn't it? But I'm sure the APL/J crowd will point out to you that
they did this type of optimization oh, three decades ago. If your f90
compiler doesn't do its work properly, return it to the manufacturer
while complaining loudly.
Return to the
Search the comp.compilers archives again.