[lno] Two vectorizer ICEs

[Dorit Naishlos]

> I can still reproduce it with 3.5-tree-ssa-lno 20040224.

Looks like it's a target specific problem then.
Is the vectorizer able to vectorize the loop at all? it shouldn't be able
to handle structs, so I wouldn't expect it to change the code at all.
Therefore I assume that the ICE happens during vectorization; Do you want
to try to provide some more details, since I haven't been able to reproduce
it?

thanks,
dorit



                                                                                                                                            
                      Falk Hueffner                                                                                                         
                      <falk.hueffner@student.uni-tu        To:       Dorit Naishlos/Haifa/IBM@IBMIL                                         
                      ebingen.de>                          cc:       gcc-bugs@gcc.gnu.org                                                   
                                                           Subject:  Re: [lno] Two vectorizer ICEs                                          
                      24/02/2004 14:52                                                                                                      
                                                                                                                                            




Dorit Naishlos <DORIT@il.ibm.com> writes:

> I cannot reproduce the problem in lno-bug1.c with the current lno
> snapshot, perhaps it was already fixed (by the way, the vectorizer
> doesn't vectorize this loop because it doesn't handle structs).

I can still reproduce it with 3.5-tree-ssa-lno 20040224.

> I guess Devang's patch (
> http://gcc.gnu.org/ml/gcc-patches/2004-02/msg01588.html) solved the
problem
> in lno-bug2.c? In any case - it seem to pass compilation (and
> vectorization) successfully now.

Yes, that one is fixed for me.

--
             Falk