------- Comment #4 from rakdver at gcc dot gnu dot org 2006-03-29 01:03 -------
(In reply to comment #1)
> Note that we in principle know the number of iterations - just we cannot prove
> the loop runs at least once in number of iterations analysis. Of course we
> know this because we did loop header copying on the loop and no other pass
> interfered with this. Maybe we should remember the BB of the loop header copy
> so we can verify
> later that it still dominates the loop header.
I do not see much difference wrto the current system (where we simply check all
blocks that dominate the loop header); checking just the one known to be the
copy of the original header would be a bit more efficient, but it might also
cause us to miss some opportunities to learn something about the bounds of the
loop from the other conditions.
I do not think remembering what the copied loop header is affects the need to
deal with symbolic range simplification in any way. I think I can however
improve the # of iterations analysis to be less demanding to fold, though.
--
rakdver at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
AssignedTo|unassigned at gcc dot gnu |rakdver at gcc dot gnu dot
|dot org |org
Status|UNCONFIRMED |ASSIGNED
Ever Confirmed|0 |1
Last reconfirmed|0000-00-00 00:00:00 |2006-03-29 01:03:55
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26900
