On Mon, 9 Jan 2017, Jakub Jelinek wrote:
> On Mon, Jan 09, 2017 at 11:08:08AM +0100, Richard Biener wrote:
> > > > > There is one thing my patch doesn't do but should for efficiency, if
> > > > > loop1
> > > > > (outer loop) is not successfully outer-loop vectorized, then we
> > > > > should mar
On Mon, Jan 09, 2017 at 11:08:08AM +0100, Richard Biener wrote:
> > > > There is one thing my patch doesn't do but should for efficiency, if
> > > > loop1
> > > > (outer loop) is not successfully outer-loop vectorized, then we should
> > > > mark
> > > > loop2 (its inner loop) as dont_vectorize i
On Mon, 9 Jan 2017, Jakub Jelinek wrote:
> On Mon, Jan 09, 2017 at 10:08:24AM +0100, Richard Biener wrote:
> > > if if-conversion thinks outer loop vectorization might be successful.
> > > In this case, loop2 is if-converted. This works well if the outer loop
> > > versioning is subsequently succ
On Mon, Jan 09, 2017 at 10:08:24AM +0100, Richard Biener wrote:
> > if if-conversion thinks outer loop vectorization might be successful.
> > In this case, loop2 is if-converted. This works well if the outer loop
> > versioning is subsequently successful, doesn't work at all if it is
> > unsuccess
On Mon, 9 Jan 2017, Jakub Jelinek wrote:
> On Sat, Jan 07, 2017 at 12:46:26PM +0100, Richard Biener wrote:
> > >The following patch tweaks tree-if-conv.c so that when it wants to
> > >version
> > >an outer loop, it actually transforms:
> > > loop1
> > > loop2
> > >into:
> > >
On Sat, Jan 07, 2017 at 12:46:26PM +0100, Richard Biener wrote:
> >The following patch tweaks tree-if-conv.c so that when it wants to
> >version
> >an outer loop, it actually transforms:
> > loop1
> > loop2
> >into:
> > if (LOOP_VECTORIZED (1, 3))
> > {
>
On January 6, 2017 7:56:10 PM GMT+01:00, Jakub Jelinek wrote:
>Hi!
>
>If-conversion can't easily predict whether outer loop vectorization
>will be
>successful or not. In GCC 6, we'd version and guard with
>LOOP_VECTORIZED
>only the inner loop, which is beneficial if the outer loop
>vectorization
