This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: hopefully, last draft of altivec...
- To: Dale Johannesen <dalej at apple dot com>
- Subject: Re: hopefully, last draft of altivec...
- From: Aldy Hernandez <aldyh at redhat dot com>
- Date: 07 Nov 2001 16:06:04 -0500
- Cc: Geoff Keating <geoffk at cygnus dot com>, gcc patches <gcc-patches at gcc dot gnu dot org>
- References: <480C65B2-D3BC-11D5-9999-003065C86F94@apple.com>
ok guys. patches are all committed.
i suggest apple folks fix BIGGEST_ALIGNMENT in darwin.h if they need to.
> The problem is that reg_alloc_order[] in regclass.c has
> FIRST_PSEUDO_REGISTER
> entries, and will look through all of them. If you don't supply enough
> the compiler will stick in an extra 0 at the end, in addition to the 0
> that's
> there legitimately. This can cause reload to crash. (I chased down this
> bug in Apple's version, and it wasn't easy.)
i see. here's a patch to fix it.
i am retesting bootstrap and tests just because i'm paranoid. i will
commit when under the obviously correct rule unless someone speaks up.
(i also broke up that last long line).
--
Aldy Hernandez E-mail: aldyh@redhat.com
Professional Gypsy
Red Hat, Inc.
2001-11-07 Aldy Hernandez <aldyh@redhat.com>
* rs6000.h (REG_ALLOC_ORDER): Add vrsave.
Index: rs6000.h
===================================================================
RCS file: /cvs/uberbaum/gcc/config/rs6000/rs6000.h,v
retrieving revision 1.132
diff -c -p -r1.132 rs6000.h
*** rs6000.h 2001/11/07 20:49:41 1.132
--- rs6000.h 2001/11/07 21:00:45
*************** extern int rs6000_debug_arg; /* debug a
*** 772,778 ****
90, 89, 88, 87, 86, 85, 84, 83, 82, 81, 80, \
79, \
96, 95, 94, 93, 92, 91, \
! 108, 107, 106, 105, 104, 103, 102, 101, 100, 99, 98, 97 \
}
/* True if register is floating-point. */
--- 772,779 ----
90, 89, 88, 87, 86, 85, 84, 83, 82, 81, 80, \
79, \
96, 95, 94, 93, 92, 91, \
! 108, 107, 106, 105, 104, 103, 102, 101, 100, 99, 98, \
! 97, 109 \
}
/* True if register is floating-point. */