This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] resubmitted for review to fix modulo scheduling. COMMITTED
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: Kenneth Zadeck <zadeck at naturalbridge dot com>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Ian Lance Taylor <ian at airs dot com>, "Berlin, Daniel" <dberlin at dberlin dot org>, Janis Johnson <janis187 at us dot ibm dot com>, Bernd Schmidt <bernds_cb1 at t-online dot de>
- Date: Tue, 23 May 2006 09:43:01 -0400
- Subject: Re: [PATCH] resubmitted for review to fix modulo scheduling. COMMITTED
- References: <4472FCD8.3000107@naturalbridge.com>
Kenneth Zadeck writes:
> It apparently suffers from several problems:
> reload scews it up (what a shocker!), it relies on patterns that only
> occur on the ppc, and the instructions are expanded on other platforms
> too late to get a good schedule.
The modulo scheduling pass relies on the presence of doloop
patterns to recognize loops after many optimization passes have been
applied. Eventually this should rely on Tree-SSA loop discovery.
The doloop patterns are not limited to ppc. Please do not
inaccurately parrot comments from IRC chat.
Thanks, David