[Libre-soc-dev] setvl lost its reserved fields (was: setvl gaining CTR mode)

Jacob Lifshay programmerjake at gmail.com
Sun Jul 3 10:52:16 BST 2022


On Sun, Jul 3, 2022, 02:29 lkcl <luke.leighton at gmail.com> wrote:
> when i said "it is too late" it means, "it is too late". you should have
raised this (and offered to help) 6 to 24 months ago.

I did, as i mentioned in the last message, which you claim you ignored. i
will copy that part of it again in hopes that you will read it this time:
I wrote (added emphasis):
> Also, this is effectively changing back to the design *we already agreed
on* iirc (see link below), that had been lost presumably when setvl was
crammed into the bitmanip major op.

https://bugs.libre-soc.org/show_bug.cgi?id=535

please read the bottom paragraph of this email for an alternative idea that
requires no implementation changes whatsoever, only spec. text (assuming
our implementation was correct in the first place...).


third time of repeating, you have not listened the first time and you have
> not listened the second time.
>
> i have not read anything that you have written in the past two messages.


that is a *major* problem, if you don't listen, why should I bother trying
to convey anything? it doesn't help either of us. please do read messages.
I do at least read your messages even if i disagree with their contents...

i asked for an opinion on the minor incremental change to add CTR mode.
>

what i requested in the last message is a even more minor change than your
adding ctr mode....all that is happening is a constant-0 bit is renamed. no
changes to assembly syntax, what each bit of the instruction encoding
currently does, unit tests, etc.

alternatively:

If you don't like renaming fields at all, then we can do what is 100%
equivalent (except in clarity) -- state that all values of SVi with the msb
set must trap and are reserved for future extensions (with a
*non-normative* note explaining what could be done -- non-normative meaning
it explains what the idea is, but it isn't a part of the official standard
that imposes any requirements).

Jacob


More information about the Libre-soc-dev mailing list