[PATCH] D14664: [ARM] Match VABDL from log2 shuffles.

[Renato Golin via llvm-commits]

On 16 November 2015 at 20:10, Tim Northover <t.p.northover at gmail.com> wrote:
> I don't follow this. The incoming IR takes some bit-patterns and
> produces (hopefully well-defined) output bit-patterns. Whether the
> types involved were originally signed or unsigned in the source
> language is intentionally irrelevant. If you can produce the same
> black-box function starting from signed types, good on you and we'll
> optimise it anyway.

That was just a wondering question, not a hard-evidence one. I was
thinking in terms of gene splicing, but I can't think how that could
apply here. Sorry, nothing to see here, move along... :)

As usual, whatever bad happens, it's a bug and we fix it.

LGTM.

cheers,
--renato