[libre-riscv-dev] extremely busy crowdsupply update started

[Luke Kenneth Casson Leighton]

On Fri, Mar 27, 2020 at 4:35 PM Tobias Platen
<libre-soc at platen-software.de> wrote:
>
> On Fri, 27 Mar 2020 15:49:45 +0000
> Luke Kenneth Casson Leighton <lkcl at lkcl.net> wrote:
>
> I've cloned the repository and now I could write something to fill one or more TODOs in the file 023_2020mar26_decoder_emulator_started.mdwn probably either Coriolis2 experimentation started or POWER ISA decoder and Simulator.

sure - i've added you to the repo so you can git push to it.

> I think I'll work on the POWER ISA decoder and Simulator next.

i had an idea about that, based on this:

http://lists.mailinglist.openpowerfoundation.org/pipermail/openpower-hdl-cores/2020-March/000007.html

my feeling is, we're wasting our time trying to "manually" create both
a processor and also a simulator, from "scratch", when there are
perfectly good, machine-readable expressions (at least for the integer
instructions), like this:

lbz         RT,D(RA)
    34      RT      RA     D
0        6       11     16   31
if RA = 0 then b <= 0
else           b <= (RA)
EA <= + EXTS(D)
RT <= 56^0 || MEM(EA, 1)

i'm advocating, therefore, that we *literally* parse the PDF spec,
create an "interpreter" for those fragments, then auto-generate code
and HDL from them.

mulhwu      RT,RA,RB                      (Rc=0)
mulhwu.     RT,RA,RB                      (Rc=1)
    31      RT      RA      RB    /    11    Rc
0        6       11      16      21 22       31

prod0:63 <- (RA)32:63 x (RB)32:63
RT32:63 <- prod0:31
RT0:31 <- undefined

those really don't look very hard, to turn into code, do they?

FP is a little more challenging because they use "words" to describe
most aspects of the hardware.  oh wait!  Appendix A p775 v3.0B
contains FP to rounding algorithms, and a few more besides.

i would not be at all surprised if, should we do document-parsing, we
had an entire simulator for the integer V3.0 system in a week, and the
hardware in around 2.

l.