This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Proposed patch for loop.c
- To: Bernd Schmidt <crux at pool dot informatik dot rwth-aachen dot de>, egcs at cygnus dot com
- Subject: Re: Proposed patch for loop.c
- From: Richard Henderson <rth at cygnus dot com>
- Date: Tue, 2 Jun 1998 16:10:59 -0700
- References: <Pine.SOL.3.90.980528110457.28555H-100000@matlock.informatik.rwth-aachen.de>
- Reply-To: Richard Henderson <rth at cygnus dot com>
On Thu, May 28, 1998 at 11:26:21AM +0200, Bernd Schmidt wrote:
> I found two problems. First, general_induction_var returns the benefit of
> a giv, but it always returns 1 even if it calculated a benefit of zero.
While my previous loop patches would already handle this particular
example, I think your point about the benefit calculation is sound,
and have incorporated it into my jumbo loop patch.
Of course, benefit calculation based on the cost of the address
calculation in any form is not going to be as useful as benefit
calculation based on register usage across the entire loop. IMHO.
r~