This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Flow analysis and optimization with computed gotos
- To: matzmich at cs dot tu-berlin dot de (Michael Matz)
- Subject: Re: Flow analysis and optimization with computed gotos
- From: Brad Lucier <lucier at math dot purdue dot edu>
- Date: Tue, 5 Dec 2000 02:30:15 -0500 (EST)
- Cc: lucier at math dot purdue dot edu (Brad Lucier), gcc at gcc dot gnu dot org, feeley at iro dot umontreal dot ca, hosking at cs dot purdue dot edu
Michael Matz wrote:
> It still isn't correct though. E.g. if the expression didn't come from the
> destination block. The real solution is, that abnormal critical edges
> (s->d) kill all expressions from which it's not provable that they don't
> trap in block d.
For many of these large routines, gcse is disabled because
the number of basic blocks is > 1000 and the number of edges/basic
block is >= 20. In these cases I'd be happier if these abnormal
critical edges were just ignored, if that would be possible, and gcse
performed on the subgraphs that do not involve these edges; if
we agreed that abnormal critical edges killed *all* expressions across
them, could they be ignored for purposes of gcse?
Brad