This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: test patch for computed gotos
- From: Brad Lucier <lucier at math dot purdue dot edu>
- To: rth at redhat dot com (Richard Henderson)
- Cc: lucier at math dot purdue dot edu (Brad Lucier), gcc-patches at gcc dot gnu dot org, feeley at iro dot umontreal dot ca
- Date: Sat, 15 Feb 2003 18:09:43 -0500 (EST)
- Subject: Re: test patch for computed gotos
> On Fri, Feb 14, 2003 at 05:49:03PM -0800, Richard Henderson wrote:
> > I'll investigate further. Probably we want to drop the cutoff
> > from 100 edges to like 10 edges, and make it a --param.
>
> No, the problem is a bug in how we checked for cutoff. We'd not cut
> off at all if a fallthru edge was found early in the list of edges.
The part of the patch with invoke.texi didn't bootstrap on my machine;
sorry, I don't have the error message.
The performance results with the new patches are at
http://www.math.purdue.edu/~lucier/gcse-test3.html
Overall, it seems to be a win at -O1 -fno-gcse; it would probably be good
if other people running programs like threaded interpreters would test this.
Thanks.
Brad