This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Reassociation
- To: law at cygnus dot com, egcs at cygnus dot com
- Subject: Re: Reassociation
- From: Stephen L Moshier <moshier at mediaone dot net>
- Date: Sat, 19 Dec 1998 08:30:27 -0500 (EST)
- Reply-To: moshier at mediaone dot net
Some languages (ie fortran) allow reassociation as long as we honor
paren boundaries. We need some way to find those boundaries. How do
you suggest we find them?
By looking at the rtl. For example, in the rtl for a*b*c*d all the
set destinations are the same, a giveaway that the parentheses were
omitted. I suggest asking Wilson about this; he would probably know
whether this is a feasible plan in general.
See fold-const.c It does most of the same transformations on trees
that cse.c is doing on RTL.
That's fine, but the optimizations in point are now being squelched
specifically in cse.c.