[Libre-soc-dev] solution for what to do when svp64 gains 128-bit registers

[Luke Kenneth Casson Leighton]

On Wed, Jun 29, 2022 at 10:20 PM Jacob Lifshay via Libre-soc-dev
<libre-soc-dev at lists.libre-soc.org> wrote:

> Posting here so it doesn't get forgotten:
> https://libre-soc.org/irclog/%23libre-soc.2022-06-29.log.html#t2022-06-29T21:54:13

jacob: i have asked you multiple times now to drop discussion of
128-bit registers and greater than 128 registers.

it is an extremely complex topic that you do not have the Project
Management experience to assess how damaging to the project
at this critical time will be, should you get side-tracked even for
"just" one week on such a task for which there is *no payment*
available.

that one week, as i have told you multiple times, represents at the
time of writing TEN PERCENT the available time remaining until
the October deadline.

who is going to do the critical work that is allocated to you?

you are causing me severe distress that you keep not listening
when i ask you to drop certain technical topics from discussion
because from experience i know that it will damage the project
in *multiple* ways if they are investigated right now.

you did the exact same thing on the last critical deadline and
it made me seriously ill and very distressed that i had to be the
one to pick up the slack because you had committed your time to
unpaid work and ignored the obligations to NLnet which i had to
fulfil on your behalf as well as my own.

*please listen* when i say that this particular topic is too complex
to even investigate investigating.

i will put in the spec that there is space reserved for future encoding
but that it is firmly out of scope for this revision of the spec.  that
*has* to be the end of the discussion and i would appreciate it
very much if, should the topic come up again, you would also
inform people that it is out of scope, and leave it at that.

if they keep pressing then please explain to them that it is a
highly complex topic that requires considerable resources even
to discuss, and would they like to engage in a CONTRACT to
pay for your insights and expertise to assess it for them.

we need to protect ourselves from being "sponged off" just because
we are discussing and developing highly technical subjects that
would normally have multiple people being paid 5 figure salaries
for the sole benefit of the multi-billion-dollar company they work for.

even if this topic comes up as a question from the OPF ISA WG
or one of its stakeholders i am very specifically going to tell them
that 128-bit and >128 registers is firmly out of scope for discussion
until the first revision of the spec is ratified.

otherwise what will happen is that we end up doing work for IBM's
benefit, or other stakeholder's benefit, not just for zero money, but
worse than that, it *destroys* our own timescales and could stop us
from getting VC funding because it's far too unrealistic a workload
to complete *or even be able to assess* and such a huge distraction.

when approaching VCs it is critically important to have the
FULL knowledge of what can and cannot be delivered, and to
be able to truthfully say "yes, i can achieve this".

such a statement becomes a CONTRACT!

and as a Director of RED Semiconductor Ltd, i am then *legally*
committed to deliver on what i have said is achievable and
deliverable.

therefore i *have* to know every single technical detail about 128-bit
registers and >128 registers, and i have already done the assessment
that even to *assess* it is too great a risk to take on.

so please - *LISTEN* to what i am saying, that you are placing the
entire project at risk by continuing to raise a topic that i have told
you multiple times is too great a risk for multiple reasons.

please *drop* the topic

l.