[Libre-soc-dev] results of phone call about simple-v prefix

[Jacob Lifshay]

On Fri, Nov 27, 2020, 15:30 Cole Poirier <colepoirier at gmail.com> wrote:

> On Friday, November 27, 2020, Jacob Lifshay <programmerjake at gmail.com>
> wrote:
> >
> >
> > Turns out that primary opcode 0 is used for implementation-specific
> > opcodes. primary opcode 1 and 5 are defined to cause illegal instruction
> > traps for the encodings we want to use. We could potentially move to
> using
> > primary opcode 6 for 32/48-bit instructions -- 1/8 is currently used (new
> > in v3.1), the rest causes illegal instruction traps. This would reduce
> the
> > space for svp32 somewhat, svp48 would have just as much space as before.
> >
>
> Isn’t this behaviour for PO 0, 1, 5 exactly what we were hoping for?
>

Yes for 1 and 5, no for 0. PO 0 instructions can do implementation-specific
things on older processors, basically it's where they put things that
aren't part of the spec, such as hardware debugging helper instructions, or
processor self-test instructions.

Jacob