This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: Problem with altivec.h? [was: Elena -- where is altivec.h?]
- From: Elena Zannoni <ezannoni at redhat dot com>
- To: Michael Elizabeth Chastain <mec at shout dot net>
- Cc: ezannoni at redhat dot com, gdb-patches at sources dot redhat dot com, msnyder at redhat dot com
- Date: Fri, 17 May 2002 21:00:38 -0400
- Subject: Re: Problem with altivec.h? [was: Elena -- where is altivec.h?]
- References: <200205172258.g4HMwfJ19933@duracef.shout.net>
Michael Elizabeth Chastain writes:
> > In case this is not obvious, the altivec C testsfiles (or any altivec
> > programs) need to include this <altivec.h> gcc file because it has
> > all the necessary #defines for vector types.
>
> include/gdb is for building gcc and gdb, not for building target
> programs. Consider a cross-compiler + cross-debugger + simulator:
> host=i686-pc-linux-gnu, target=rs6000-unknown-whatever. So that
> directory is right out.
>
Yes, true. Long day, sorry.
> gcc should be installing altivec.h with the other target include files.
> AldyH would know about that.
>
Yes, that's what happens.
> Perhaps you are testing out of a build directory, rather than an install
> directory, and the dejagnu machinery is not producing enough -I flags?
> (I am studying this area right now because I'm working on Petr's
> f77/f95 patch).
>
No, I have found no problems with the tests. I was trying to address
MichaelS concerns.
> Suggestion:
>
> I did a find in the gcc source:
>
> % find * -name '*altivec*' -print
> gcc/config/rs6000/altivec.h
> gcc/config/rs6000/altivec-defs.h
> gcc/config/rs6000/eabialtivec.h
> gcc/config/rs6000/linuxaltivec.h
> gcc/testsuite/gcc.dg/altivec-1.c
> gcc/testsuite/gcc.dg/altivec-2.c
> gcc/testsuite/gcc.dg/altivec-3.c
> gcc/testsuite/gcc.dg/altivec-4.c
> gcc/testsuite/gcc.dg/altivec-5.c
> gcc/testsuite/gcc.dg/altivec-7.c
> gcc/testsuite/gcc.dg/altivec-8.c
>
> Most of the altivec-*.c test files do not use altivec.h. They simply
> use builtin keywords directly. Have a look at altivec-4.c.
> You could write the tests like that.
>
Yes, actually I asked Aldy and that is what he suggested doing, just put the
necessary #define (just one) in the .c files. I'll do that.
Elena
> Michael C