On Thu, 25 Mar 2021, Richard Sandiford wrote:
> Richard Biener writes:
> > Whether a vector DR is aligned or not depends on the vectorization
> > scheme but at the point of computing peeling for alignment
> > that has not been determined so we try to work with scalar DRs.
> >
> > For the uses by
Richard Biener writes:
> Whether a vector DR is aligned or not depends on the vectorization
> scheme but at the point of computing peeling for alignment
> that has not been determined so we try to work with scalar DRs.
>
> For the uses by vectorizable_load/store I pondered to pass
> the memory_acc
On Thu, 25 Mar 2021, Richard Biener wrote:
> This fixes bogus alignment computation for negative strided
> grouped accesses which we now can vectorize but only through SLP.
> The biasing for negative strided non-grouped accesses as supported
> via VMAT_CONTIGUOUS_REVERSE does not apply there and i
This fixes bogus alignment computation for negative strided
grouped accesses which we now can vectorize but only through SLP.
The biasing for negative strided non-grouped accesses as supported
via VMAT_CONTIGUOUS_REVERSE does not apply there and instead causes
alignment analysis to compute wrong th
