>> Hmm, right, you mean punning with HImode again? Yeah, that would work and
>> we
>> could get rid of the strided fallback for that case. I'll do that in v2.
>
> It can be just put within BFmode should be fine, I mean
> load/store/move/insert/extract could just use normal vector
> instruction with SEW=16,
> Only tha casting from/to other data types need real BF16 instruction.
> ('real' might not be the right term to describe, but I guess you
> should be able to get my point here :P)
>
> no casting to/from HImode, so that we can use the same pattern/logic
> between w/ and w/o vector BF16 extension.
>
> And I think we won't ever have vfmv.bf16.v.f / vfmv.bf16.f.v since
> they just play the exact same operation as vfmv.v.f / vfmv.f.v with
> SEW=16.
Just getting back to this now. I'm not sure I fully got it.
Which instruction sequence do you have in mind for a BF broadcast?
just e.g.
vsetivli zero,8,e16,m1
vfmv.v.f v2,f0?
I somehow expected this to vill but if it works and is legal/safe, sure :)
--
Regards
Robin
