This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: patch: x86 generic simd fixlet
On Tue, Jul 08, 2003 at 05:31:46PM -0700, Richard Henderson wrote:
> On Tue, Jul 08, 2003 at 08:29:08PM -0400, Aldy Hernandez wrote:
> > Yeah sure, with SSE the arguments may go in SIMD registers. Without
> > it, they go in GPRs. Completely different ABIs ;-).
>
> Um, no. The bit you're changing is checking for alignment
> on the stack. The x86 abi uses *no* SSE registers for passing
> parameters. The amd64 abi does; don't confuse the two.
Aight. As discussed on irc, here is a patch removing the abort.
Tested on x86 linux.
Ok?
2003-07-08 Aldy Hernandez <aldyh@redhat.com>
PR/11144
* config/i386/i386.c (ix86_function_arg_boundary): Remove abort.
Index: config/i386/i386.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/i386/i386.c,v
retrieving revision 1.587
diff -c -p -r1.587 i386.c
*** config/i386/i386.c 8 Jul 2003 18:01:46 -0000 1.587
--- config/i386/i386.c 9 Jul 2003 00:52:08 -0000
*************** contains_128bit_aligned_vector_p (tree t
*** 2610,2617 ****
return false;
}
! /* Gives the alignment boundary, in bits, of an argument with the specified mode
! and type. */
int
ix86_function_arg_boundary (enum machine_mode mode, tree type)
--- 2610,2617 ----
return false;
}
! /* Gives the alignment boundary, in bits, of an argument with the
! specified mode and type. */
int
ix86_function_arg_boundary (enum machine_mode mode, tree type)
*************** ix86_function_arg_boundary (enum machine
*** 2642,2649 ****
if (!contains_128bit_aligned_vector_p (type))
align = PARM_BOUNDARY;
}
- if (align != PARM_BOUNDARY && !TARGET_SSE)
- abort();
}
if (align > 128)
align = 128;
--- 2642,2647 ----