This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Semi-Latent Bug in tree vectorizer


On Fri, Apr 08, 2005 at 12:04:02PM -0600, Jeffrey A Law wrote:
> On Fri, 2005-04-08 at 13:55 -0400, Diego Novillo wrote:
> > On Fri, Apr 08, 2005 at 10:52:02AM -0600, Jeffrey A Law wrote:
> > 
> > > It would probably be wise to audit the other uses of 
> > > copy_virtual_operands.  We might also consider forcing statement
> > > rescans as part of our IL checking code to avoid these kinds of
> > > problems in the future.
> > > 
> > Yes, I've run into this problem and am about to commit fixes in
> > this area.
> Can you send me your pending changes it would be good to know if
> they're going to fix my problem or not.  Or are you just referring
> to changes which force statement rescans?
> 
Sure.  Sent in a separate message.  I should be committing this
patch later today.  I'm running final tests.

> > Both ivopts and the vectorizer were creating bad memory tags for
> > the new pointers.
> Which could be the root cause of my problem.
> 
It was causing trouble mainly because they would use real
variables as memory tags.

> >  Also, after ivopts, the whole CFG needs to be
> > re-scanned because the new alias relations it creates affect
> > statements that have not even been modified by the process.
> Wow.  Egad.
> 
Yeah, I'm not very happy with that either.  Now that we are
finally starting to improve the alias stuff, we should probably
evaluate how to be nicer wrt to forcing whole IL scans (I'm not
too sure we can avoid those, unfortunately).


Diego.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]