This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: How early are COND_EXEC patterns created?
- To: John Wehle <john at feith dot com>
- Subject: Re: How early are COND_EXEC patterns created?
- From: Michael Meissner <meissner at cygnus dot com>
- Date: Tue, 2 May 2000 00:25:14 -0400
- Cc: gcc at gcc dot gnu dot org
- References: <200005020410.AAA07635@jwlab.FEITH.COM>
On Tue, May 02, 2000 at 12:10:06AM -0400, John Wehle wrote:
> What's the earliest that COND_EXEC patterns are created?
> I'm currently using note_stores in the value range propagation
> code to determine when register values change, however COND_EXEC
> adds the wrinkle that the SET (or CLOBBER) which is passed to
> (*fun) by note_stores may not actually happen. If COND_EXEC
> doesn't appear until after gcse then there's no problem, otherwise
> I suppose I could hack note_stores so that COND_EXEC_TEST is
> also passed to (*fun). BTW, can COND_EXEC appear inside a PARALLEL?
Looking at toplev.c, there is a call to if_convert after the first jump pass
(in my original code, the cond_exec code was called after each of the two flow
passes).
--
Michael Meissner, Cygnus Solutions, a Red Hat company.
PMB 198, 174 Littleton Road #3, Westford, Massachusetts 01886, USA
Work: meissner@redhat.com phone: +1 978-486-9304
Non-work: meissner@spectacle-pond.org fax: +1 978-692-4482