[libre-riscv-dev] daily kan-ban update 08jun2020

Luke Kenneth Casson Leighton lkcl at lkcl.net
Mon Jun 8 23:13:07 BST 2020


On Mon, Jun 8, 2020 at 10:17 PM Cole Poirier <colepoirier at gmail.com> wrote:

> I'll try to put just a few very specific questions (such as the
> advantages/high level summary of Simple-V) for you and other technical
> project members to answer. I would try answer this myself given your
> present state, but unfortunately I don't think I can understand and
> explain something like Simple-V on my own in a day.

yyeah.  last time i explained it to someone (in detail) it was to
Alain.  it took six hours.

the summary: rather than have dozens of instructions ADD x1, x10 ADD
x2, x11 ADD x3, x12, instead you deploy SimpleV - with *one*
instruction - telling it (by setting VL) exactly how many registers,
in sequence, you would like that *ONE* instruction to push into the
instruction queue.

that's basically all there is to it.  it should be pretty obvious
therefore that it is effectively a very simple version of an
instruction "compression algorithm".

in effect, Simple-V is a hardware-level for-loop.  if you have heard
of Zero-Overhead Loops, it is a bit like that.  except, whilst ZOHLs
loop around a *group* of instructions, Simple-V gives the option to
apply the for-loop to *ANY* instruction, and does so *on* that
instruction *before* moving on to the next one.

it's like a Sub- Program Counter, in effect.


> Can I ask you to see if you're able to answer some of these questions
> today, if your illness permits?

it's fine, now: i took... i won't go into the medical details, there's
a procedure i follow which clears out toxins and it was effective.
also, a drink made from large quantities of ginger root, cardamons,
black pepper and chilis, caused me to cough and splutter and generally
hawk and spit profusely (as might be expected), and surpriise! i'm
feeling much better.

l.



More information about the libre-riscv-dev mailing list