This is the mail archive of the
mailing list for the GCC project.
Re: [new-regalloc-branch] bugfix for get_free_reg
- From: Denis Chertykov <denisc at overta dot ru>
- To: Michael Matz <matz at kde dot org>
- Cc: Denis Chertykov <denisc at overta dot ru>, <gcc-patches at gcc dot gnu dot org>
- Date: 21 Dec 2001 23:35:56 +0300
- Subject: Re: [new-regalloc-branch] bugfix for get_free_reg
- References: <Pine.GSO.4.33.0112211049040.23792-100000@platon>
Michael Matz <firstname.lastname@example.org> writes:
> On Fri, 21 Dec 2001, Denis Chertykov wrote:
> > Thu Dec 20 23:57:05 2001 Denis Chertykov <email@example.com>
> > * ra.c (get_free_reg): Check hardregs for dont_begin_colors.
> No. dont_begin_colors is named correctly.
> It contains the hardreg
> numbers on which the pseudo in question should not begin. I.e. testing
> the following regs in a multi-reg mode is wrong. There only free_colors
> matters. If there is some bug when selecting a color it is in the
> calculation of dont_begin in colorize_one_web().
> (The difference becomes important with partial conflicts between
> multi-word webs)
I will try to fix a bug in colorize_one_web.
PS: Few weeks ago you promised to commit a big good changes. How about it ?