[Libre-soc-dev] clamping/saturation semantics

[Jacob Lifshay]

On Fri, Dec 11, 2020, 13:29 Luke Kenneth Casson Leighton <lkcl at lkcl.net>
wrote:

> On 12/11/20, Jacob Lifshay <programmerjake at gmail.com> wrote:
>
> > mv.swizzle would be a 1 src 1 dest op.
>
> this is probably a good idea to have regardless of whether svp64 has it.
>
>
> > I think you might misunderstand:
>
> probably :)
>
> > from what I recall from playing around with audio code,
> > the operation required would be the following translated to a sequence
> of 1
> > or more instructions:
> > read inputs at src size
> > scale to account for src/dest size mismatch // if src size==dest size,
> this
> > step isn't needed
> > perform arith at higher precision
> > clamp result to dest size // not src size, not any other size
> > write result to dest
>
> ok so in that context, how critical would it be to have separate
> mv.saturate ops
>

I think we'd want these either way.

>
> vs
>
> arith.saturate being able to specify a saturation depth less than that
> of the arithmetic operation elwidth
>

I've never heard of an ISA that saturates to a different size than the dest
size, so I'd expect it to be mostly unused.

that said, having an op with different src and dest sizes could be quite
handy.

Jacob