This fixes PR50494 by avoiding to increase alignment of decls
that are in the constant pool by the vectorizer.
Bootstrap & regtest pending on powerpc64-linux-gnu, with
the older fix reverted.
Richard.
2013-03-05 Richard Biener
PR middle-end/50494
* tree-vect-data-refs.c (vec
On Thu, Feb 14, 2013 at 12:53:47PM +0100, Richard Biener wrote:
> > (completely untested). Because, if we force the constant to be aligned
> > more than it would be by default, the RTL optimizers should be told about
> > that too.
>
> At the moment the vectorizer refuses to bump alignment for any
On Thu, 14 Feb 2013, Jakub Jelinek wrote:
> On Wed, Feb 13, 2013 at 01:04:04PM +0100, Richard Biener wrote:
> > 2013-02-13 Richard Biener
> >
> > PR lto/50494
> > * varasm.c (output_constant_def_1): Get the decl representing
> > the constant as argument.
> > (output_constant_de
On Wed, Feb 13, 2013 at 01:04:04PM +0100, Richard Biener wrote:
> 2013-02-13 Richard Biener
>
> PR lto/50494
> * varasm.c (output_constant_def_1): Get the decl representing
> the constant as argument.
> (output_constant_def): Wrap output_constant_def_1.
> (make_dec
This should fix PR50494 - when we gimplify a local initializer
via a block copy and emit a constant to the constant pool we
stream the representative VAR_DECL with LTO, including its
eventually changed alignment. When we then lookup RTL for this
constant at
/* If this variable belongs to the g
