This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Simple Value Range Propagation
- From: Steven Bosscher <s dot bosscher at student dot tudelft dot nl>
- To: Josef Zlomek <zlomj9am at artax dot karlin dot mff dot cuni dot cz>
- Cc: gcc-patches at gcc dot gnu dot org, Diego Novillo <dnovillo at redhat dot com>
- Date: Sun, 08 Jun 2003 23:47:45 +0200
- Subject: Re: Simple Value Range Propagation
Josef Zlomek wrote:
>
> Is it really necessary to add new essentially high-level optimizers at
> the RTL level? This pass would be quite straightforward on tree-ssa.
I agree that it is better to do it on tree-ssa. But I think it is useful
on RTL level too. Some more optimizations may be written by using it and
maybe we could use it for branch predictions too.
There is an SSA VRP algorithm that Diego had plans with (see
tree-ssa project page). Especially if Honza's plans to keep
the CFG alive, there seems to be little reason to keep such
an expensive pass at the RTL level as well.
Also, if I read your patch correctly, you enable vrp at -O1.
How about moving it to -O2, or maybe even -O3, given the 2%
compile time performance degradation vs. only 0.5% runtime
performance improvement you reported?
Gr.
Steven