[opus] [OPUS] celt_inner_prod() and dual_inner_prod() NEON intrinsics

[Bjorn Reese]

On 06/06/2017 08:40 PM, Jean-Marc Valin wrote:
> OK, so it looks like the issue is with denormalized numbers. I did a bit
> of searching, but it's not clear whether Neon handles then properly (vs
> setting then to zero). If that's indeed the issue, then I'm OK with your
> original patch (using VERY_SMALL, not 1.0+xy), but I'd like to make sure
> that it's indeed the issue.

Check with "gcc -dM -E -x c /dev/null | grep DENORM"\