This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Questions on schedule-insns and IA32
- From: Jim Wilson <wilson at specifixinc dot com>
- To: Steve Snyder <swsnyder at insightbb dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: Wed, 17 Sep 2003 11:22:39 -0700
- Subject: Re: Questions on schedule-insns and IA32
- References: <200309170641.44520.swsnyder@insightbb.com>
Steve Snyder wrote:
1. In GCC v3.2.x is -fschedule-insns enabled by -O2 or not?
Depends on the target. For IA32, it is not enabled by default at -O2.
2. In GCC v3.2.x is -fschedule-insns known to cause problems
(mis-compilation or reduced performance) on IA32 or not?
It shouldn't cause mis-compilation. However, because it isn't used by
default, it gets little or no testing on IA32, and thus there is a small
risk of latent bugs.
Conventional wisdom is that it will reduce performance for targets with
small register sets, such as IA32, because of register pressure issues.
However, this is very dependent on the testcase, as most compiler
optimizations are. Some will see reduced performance, and some will see
increased performance.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com