This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: patch [RFC] fix for an -msse2 failure
- From: Roger Sayle <roger at eyesopen dot com>
- To: Fariborz Jahanian <fjahanian at apple dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 12 Apr 2005 21:04:53 -0600 (MDT)
- Subject: Re: patch [RFC] fix for an -msse2 failure
On Tue, 12 Apr 2005, Fariborz Jahanian wrote:
> Is this the patch that I can pursue?
This patch seems like a reasonable workaround in the short-term,
i.e. suitable for 4.0.1. But I'd be interested in understanding
why the "const_vector:V4SI [(const_int 0)*]" still ICE'd, as part
of a longer term solution.
I'm also curious about the call-stack for the trunc_int_for_mode ICE;
one possibility is to allow implicit int->vector conversions at the
RTL-level (less than ideal), but I'd like to understand where we try
to truncate const0_rtx to a vector mode.
I would suggest that you investigate "CONST0_RTX (mode)" instead
of const0_rtx in this transformation, but the other latent problems
you describe make this unlikely to work.
Roger
--