[Libre-soc-dev] complex numbers as matrices

[Hendrik Boom]

On Sat, Jun 26, 2021 at 10:22:04AM +0100, Luke Kenneth Casson Leighton wrote:
> On Sat, Jun 26, 2021 at 3:18 AM Jacob Lifshay <programmerjake at gmail.com>
> wrote:
> 
> >
> > hmm, I would say it might be better to instead add a complex mul scalar op
> > and maybe even complex div and mul-add/sub scalar ops. that means you won't
> > have to constantly reprogram remap if you just need 1 or 2 muls/divs.
> > complex add/sub are covered by 2-element sv subvectors, so don't need
> > scalar ops.
> >

This is what I thought too.

> 
> which makes for quite a lot of new operations, operating on top of say vec2.
> that's definitely a "phase 2" workload.

Are we still running short of opcode space?

-- hendrik