This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: Re: -ftree-vectorize failures on cygwin
- From: "Timothy C Prince" <tprince at myrealbox dot com>
- To: jblomqvi at cc dot hut dot fi
- Cc: tprince at computer dot org, fortran at gcc dot gnu dot org
- Date: Wed, 15 Feb 2006 01:10:40 +0000
- Subject: Re: Re: -ftree-vectorize failures on cygwin
-----Original Message-----
From: Janne Blomqvist <jblomqvi@cc.hut.fi>
To: tprince@computer.org
Date: Tue, 14 Feb 2006 20:40:23 +0200
Subject: Re: -ftree-vectorize failures on cygwin
On Tue, Feb 14, 2006 at 05:58:21AM -0800, Tim Prince wrote:
> I am getting several run-time failures when I use -ftree-vectorize on
> cygwin, where there is no problem with the same cases on 32-bit linux.
> Is there sufficient interest to justify problem reports?
I don't use cygwin, but I'd like to point out that matmul is compiled
with -ftree-vectorize, so it could be a problem. What kind of failures
are you seeing?
--
When I run outside the debugger, there is a Segmentation fault. Under the insight debugger, in the first problem subroutine, I can step along for a little while, then the debugger hangs (message "dbg busy"). It is an overlapping array assignment which requires either a temporary or a loop reversal to vectorize.
In matmul, only the loops which zero out arrays get vectorized, and those work fine, AFAIK.
Tim Prince