[libre-riscv-dev] spike-sv non-default element widths

[Jacob Lifshay]

On Fri, Oct 26, 2018 at 12:01 AM lkcl <lkcl at libre-riscv.org> wrote:

> On Fri, Oct 26, 2018 at 1:43 AM lkcl <lkcl at libre-riscv.org> wrote:
> >
> > okay!  at last, retrieving / storing to registers (integer only at the
> > moment) actually works, with a c.mv preliminary test, on bit-widths of
> > default (xlen), 8, 16 and 32.  i've yet to put in sign/zero-extension,
> > that's going to be interesting / tricky and i may need some help.
>
I must have missed something, as last I had heard the widths used were
default, default * 2, default / 2 and 8. I like the new widths better.

>
>  ok first couple of unit tests are in (add and addw), and it's looking
> good so far.  the unit tests picked up a bug where i had forgotten to
> mask off the high bits of the data being spliced into the result
> register.
>
>  the add test checks the zero-extension, the addw tests sign-extend.
>
Great!

Jacob