public inbox for gsl-discuss@sourceware.org
 help / color / mirror / Atom feed
* Re: [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch
       [not found] <87hfde1idc.fsf@portacipria.lanl.gov>
@ 2000-04-07 11:24 ` Gerard Jungman
  2000-04-09  3:14   ` Peter Hopfgartner
  2000-04-10 20:54   ` Christopher R. Gabriel
  0 siblings, 2 replies; 4+ messages in thread
From: Gerard Jungman @ 2000-04-07 11:24 UTC (permalink / raw)
  To: GSL discussion list

I have replied to the full list, to try and clear this up again.

No, it should not be in the general library. There are two
separate blas implementations, one native GSL and one implemented
as a wrapper over a conformant cblas implementation.

People choose which one they want at _application_ link time.
So all applications that use linear algebra must
specify either -lgslblasnative or -lgslblascblas,
as well as -lgsl. When they link against -lgslblascblas,
they will also need to link -lcblasfoo, where cblasfoo
is their local cblas conformant implementation.


Mark Galassi wrote:
> 
> [dude, should blas be in the general library?]
> 
> From: cgabriel@linux.it (Christopher R. Gabriel)
> Subject: gsl configure.in patch
> To: rosalia@lanl.gov
> Date: 05 Apr 2000 04:41:03 +0200
> 
> I think you need this, if the blas lib is not something just for
> testing cvs.
> 
> *** configure.in        Wed Apr  5 05:36:04 2000
> --- configure.in.old    Wed Apr  5 05:37:15 2000
> *************** esac
> *** 212,218 ****
>   # Now restore our (possibly augmented) CFLAGS.
>   CFLAGS="$save_cflags"
>   GSL_CFLAGS="$CFLAGS -I$prefix/include/gsl"
> ! GSL_LIBS="-L$prefix/lib/gsl -lgsl -lgslblascblas -lgslblasnative"
> 
>   AC_SUBST(GSL_CFLAGS)
>   AC_SUBST(GSL_LIBS)
> --- 212,218 ----
>   # Now restore our (possibly augmented) CFLAGS.
>   CFLAGS="$save_cflags"
>   GSL_CFLAGS="$CFLAGS -I$prefix/include/gsl"
> ! GSL_LIBS="-L$prefix/lib/gsl -lgsl"
> 
>   AC_SUBST(GSL_CFLAGS)
>   AC_SUBST(GSL_LIBS)
> 
> --
> Christopher R. Gabriel              |    bez0(x) = (1 - x)**3
> http://www.linux.it/~cgabriel/      |    bez1(x) = 3 * (1 - x)**2 * x
> Short messages: cgabriel@sms.tin.it |    bez2(x) = 3 * (1 - x) * x**2
>                                          bez3(x) = x**3
>                                          set title "Ecco che ride!"

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch
  2000-04-07 11:24 ` [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch Gerard Jungman
@ 2000-04-09  3:14   ` Peter Hopfgartner
  2000-04-10 20:54   ` Christopher R. Gabriel
  1 sibling, 0 replies; 4+ messages in thread
From: Peter Hopfgartner @ 2000-04-09  3:14 UTC (permalink / raw)
  To: Gerard Jungman; +Cc: GSL discussion list

It's nice to see that more and more people is on the mailing list.
Obviously, some questions might come up every once in a while. So it might
be the time to start a FAQ.

Peter

Gerard Jungman wrote:

> I have replied to the full list, to try and clear this up again.
>
> No, it should not be in the general library. There are two
> separate blas implementations, one native GSL and one implemented
> as a wrapper over a conformant cblas implementation.
>
> People choose which one they want at _application_ link time.
> So all applications that use linear algebra must
> specify either -lgslblasnative or -lgslblascblas,
> as well as -lgsl. When they link against -lgslblascblas,
> they will also need to link -lcblasfoo, where cblasfoo
> is their local cblas conformant implementation.
>
> Mark Galassi wrote:
> >
> > [dude, should blas be in the general library?]
> >
> > From: cgabriel@linux.it (Christopher R. Gabriel)
> > Subject: gsl configure.in patch
> > To: rosalia@lanl.gov
> > Date: 05 Apr 2000 04:41:03 +0200
> >
> > I think you need this, if the blas lib is not something just for
> > testing cvs.
> >
> > *** configure.in        Wed Apr  5 05:36:04 2000
> > --- configure.in.old    Wed Apr  5 05:37:15 2000
> > *************** esac
> > *** 212,218 ****
> >   # Now restore our (possibly augmented) CFLAGS.
> >   CFLAGS="$save_cflags"
> >   GSL_CFLAGS="$CFLAGS -I$prefix/include/gsl"
> > ! GSL_LIBS="-L$prefix/lib/gsl -lgsl -lgslblascblas -lgslblasnative"
> >
> >   AC_SUBST(GSL_CFLAGS)
> >   AC_SUBST(GSL_LIBS)
> > --- 212,218 ----
> >   # Now restore our (possibly augmented) CFLAGS.
> >   CFLAGS="$save_cflags"
> >   GSL_CFLAGS="$CFLAGS -I$prefix/include/gsl"
> > ! GSL_LIBS="-L$prefix/lib/gsl -lgsl"
> >
> >   AC_SUBST(GSL_CFLAGS)
> >   AC_SUBST(GSL_LIBS)
> >
> > --
> > Christopher R. Gabriel              |    bez0(x) = (1 - x)**3
> > http://www.linux.it/~cgabriel/      |    bez1(x) = 3 * (1 - x)**2 * x
> > Short messages: cgabriel@sms.tin.it |    bez2(x) = 3 * (1 - x) * x**2
> >                                          bez3(x) = x**3
> >                                          set title "Ecco che ride!"

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch
  2000-04-07 11:24 ` [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch Gerard Jungman
  2000-04-09  3:14   ` Peter Hopfgartner
@ 2000-04-10 20:54   ` Christopher R. Gabriel
  2000-04-11  8:56     ` Brian Gough
  1 sibling, 1 reply; 4+ messages in thread
From: Christopher R. Gabriel @ 2000-04-10 20:54 UTC (permalink / raw)
  To: gsl-discuss

>>>>> "Gerard" == Gerard Jungman <jungman@lanl.gov> writes:

    Gerard> I have replied to the full list, to try and clear this up
    Gerard> again.

    Gerard> No, it should not be in the general library. There are two
    Gerard> separate blas implementations, one native GSL and one
    Gerard> implemented as a wrapper over a conformant cblas
    Gerard> implementation.
    Gerard> People choose which one they want at _application_ link
    Gerard> time.  So all applications that use linear algebra must
    Gerard> specify either -lgslblasnative or -lgslblascblas, as well
    Gerard> as -lgsl. When they link against -lgslblascblas, they will
    Gerard> also need to link -lcblasfoo, where cblasfoo is their
    Gerard> local cblas conformant implementation.

Thanks for the explanation. Could be a possibile solution to avoid
problems with library users adding an option to the gsl-config script?
something like 'gsl-config --libs-without-native-blas' to let users
add their preferred blas implementation?

-- 
Christopher R. Gabriel              |    bez0(x) = (1 - x)**3
http://www.linux.it/~cgabriel/      |    bez1(x) = 3 * (1 - x)**2 * x	
Short messages: cgabriel@sms.tin.it |    bez2(x) = 3 * (1 - x) * x**2
                                         bez3(x) = x**3
                                         set title "Ecco che ride!"

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch
  2000-04-10 20:54   ` Christopher R. Gabriel
@ 2000-04-11  8:56     ` Brian Gough
  0 siblings, 0 replies; 4+ messages in thread
From: Brian Gough @ 2000-04-11  8:56 UTC (permalink / raw)
  To: Christopher R. Gabriel; +Cc: Mark Galassi

That sounds reasonable.

Christopher R. Gabriel writes:
 > 
 > Thanks for the explanation. Could be a possibile solution to avoid
 > problems with library users adding an option to the gsl-config script?
 > something like 'gsl-config --libs-without-native-blas' to let users
 > add their preferred blas implementation?

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2000-04-11  8:56 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <87hfde1idc.fsf@portacipria.lanl.gov>
2000-04-07 11:24 ` [cgabriel@linux.it (Christopher R. Gabriel)] gsl configure.in patch Gerard Jungman
2000-04-09  3:14   ` Peter Hopfgartner
2000-04-10 20:54   ` Christopher R. Gabriel
2000-04-11  8:56     ` Brian Gough

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).