This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: RMS's aliasing suggestion appears unimplementable
- To: amylaar at cygnus dot co dot uk
- Subject: Re: RMS's aliasing suggestion appears unimplementable
- From: Mike Stump <mrs at wrs dot com>
- Date: Fri, 17 Sep 1999 19:00:48 -0700 (PDT)
- Cc: egcs at egcs dot cygnus dot com, law at cygnus dot com, zack at bitmover dot com
> From: Joern Rennecke <amylaar@cygnus.co.uk>
> To: mrs@wrs.com (Mike Stump)
> Date: Fri, 17 Sep 1999 23:32:14 +0100 (BST)
> Cc: law@cygnus.com, zack@bitmover.com, egcs@egcs.cygnus.com
> > mem(r5) does not alias mem(r4+4)
> Did you mean mem:SI(r5) does not alias mem:SI(r5+4),
> for GET_MODE_SIZE (SImode) == 4 ?
No.
> Otherwise, the above doesn't make sense to me. Without more context
> information, mem(r5) might or might not alias mem(r4+4).
Good, you precisely understood my point, my example and what I was
trying to say. It is ok that we think that they don't alias (for the
purposes of the warning), even if they do. If we can later, somehow
determine (value prop) they are the same (or near the same) value,
then we can `fix' the compiler and make it warn for this additional
case. Initially we should ignore this case (don't let it cause us to
not solve the sub problem). Or, anyway, that would be rms's take.