Re: interface definitions of BLAS
- From: "James Van Buskirk" <not_valid@xxxxxxxxxxx>
- Date: Wed, 8 Feb 2006 18:23:37 -0700
"Joost" <jv244@xxxxxxxxx> wrote in message
news:1139427332.928994.91380@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
I've looked at it now, but as you say, it won't really do what I need
(i.e. explicit interface blocks). It is interesting though, any plans
to do something similar for LAPACK :-) ?
Just thinking about it, I think you could get it to do what
you want in about 10 minutes: just go to each *.f file that
defines the BLAS procedures (they are significantly reduced
in number due to the genericity) and wipe out all executable
statements, perhaps renaming each file to *.inc in the process.
Then go back to the module files and change the CONTAINS
statement to an INTERFACE statement and put an END INTERFACE
statement just before the END MODULE statement. Well, maybe
20 minutes, but somewhat easier than starting from scratch,
I think.
Oh, about LAPACK: f95's rules about TKR matching and LAPACK's
usage of sequence association across subroutine calls means
that this approach is unworkable. Maybe you could run a little
test to see if
!DEC$ ATTRIBUTES REFERENCE
allows the 'R' in TKR to be ignored. I think
!DEC$ ATTRIBUTES NO_ARG_CHECK
(or however it's spelled) is too heavy-handed to work in this
context.
--
write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, &
6.0134700243160014d-154/),(/'x'/)); end
.
- Follow-Ups:
- Re: interface definitions of BLAS
- From: James Van Buskirk
- Re: interface definitions of BLAS
- References:
- interface definitions of BLAS
- From: Joost
- Re: interface definitions of BLAS
- From: James Van Buskirk
- Re: interface definitions of BLAS
- From: Joost
- interface definitions of BLAS
- Prev by Date: Re: Fortran
- Next by Date: Re: Pointers to procedures in Intel Fortran
- Previous by thread: Re: interface definitions of BLAS
- Next by thread: Re: interface definitions of BLAS
- Index(es):
Relevant Pages
|