Re: Semi-Latent Bug in tree vectorizer

[Jeffrey A Law <law at redhat dot com>]

On Fri, 2005-04-08 at 13:39 -0400, Daniel Berlin wrote:
> On Fri, 2005-04-08 at 11:08 -0600, Jeffrey A Law wrote:
> > On Fri, 2005-04-08 at 13:04 -0400, Daniel Berlin wrote:
> > > > When we rescan the operands, we get a different set of V_MAY_DEFS,
> > > > specifically we lose the V_MAY_DEF for SFT.3_20.
> > > 
> > > Why?
> > > 
> > > It should be copying subvars to the new vectorizer variable too.
> > > At least, i believe i added that.
> > Where would I look for that code?  I'm totally unfamiliar with the
> > vectorizer code.
> > jeff
> 
>  /** (2) Add aliasing information to the new vector-pointer:
>           (The points-to info (SSA_NAME_PTR_INFO) may be defined later.)
> **/
> 
>   tag = STMT_VINFO_MEMTAG (stmt_info);
>   gcc_assert (tag);
>   get_var_ann (vect_ptr)->type_mem_tag = tag;
>   get_var_ann (vect_ptr)->subvars = STMT_VINFO_SUBVARS (stmt_info);
> 
> from tree-vect-transform.c:vect_create_data_ref_ptr (the
> STMT_VINFO_SUBVARS  is set in a few routines in tree-vect-analyze.c).
> 
> It's quite possible i missed a place in analyze or something that sets
> the subvars, or some place that copies the subvars.
> 
> The code was not easy to follow, as i imagine you have discovered as
> well :)
Well, in this particular case STMT_VINFO_SUBVARS is null.  I have no
idea if that's correct or not.

jeff