[libre-riscv-dev] uniform instruction format

Luke Kenneth Casson Leighton lkcl at lkcl.net
Sun Jun 16 08:45:12 BST 2019

On Sun, Jun 16, 2019 at 7:39 AM Jacob Lifshay <programmerjake at gmail.com> wrote:

> One part I particularly want to keep in the current proposal is that
> register fields are 8-bits wide, allowing future extension to treat
> the integer and FP registers as a combined register file.

yes... great... however why are we finding that out only just now,
after two, almost three weeks of time have passed when we have limited
time and budget (and this particular task is allocated *zero* budget),
and the project's charter says that we run along unanimous
decision-making lines?

if you'd consulted me i would have suggested, again, leaving the
32-bit format embedded in the 64+ bit one, and having 3 "extension"
bits (instead of just the one) per rs1/2/3/rd to take them up to 8

this would have left us with plenty of time to work *together* to
solve the 48-bit to 64+-bit issue of the (non-fitting,
non-32-bit-line) 48-bit opcodes, several days ago.

do you notice how i always describe what i am doing?  it's not just
there "because it's fun to do so", i describe what i am doing because
i expect you - and everyone else - to tell me if it's a direction that
we want, as a *team*, making *unanimous decisions* - to take.

whilst i appreciate that in many cases it takes actually "doing" to
ascertain what "needs to be done", can you *please* describe the
*idea*, with the expectation that we - as a *team* - will *evaluate*
the idea, before spending critical resources of time and money on
going ahead and doing something that has *not been discussed*?

it may be brilliant, it may be exactly what we want to do, however
it's not how we agreed that the project is to be run, is it?


More information about the libre-riscv-dev mailing list