[Libre-soc-dev] v3.1B prefix

[Luke Kenneth Casson Leighton]

long write-up, will reply more in detail tomorrow, alexander.

On Wed, Dec 16, 2020 at 11:23 PM Alexandre Oliva <oliva at gnu.org> wrote:

> A much denser representation would be achieved by using one bit per
> upcoming insn, to tell whether or not it's compressed, as I suggested

what we feared about that was that the dropping back to 32-bit mode
would only be able to return to 16-bit by way of a 10-bit encoding.
if a *small* modification to the analyser can show that that's
unfounded, *great*.

l.