[libre-riscv-dev] Why The Dual ISA

[Adam Van Ymeren]

On 2020-01-19 10:57 AM, Luke Kenneth Casson Leighton wrote:
> On Sunday, January 19, 2020, Adam Van Ymeren <adam at vany.ca> wrote:
>
>> How firm are these commitments?
>
> one is USD 500 a month and as it is unconditional it is a big deal.  they
> very specifically gave us that donation on the basis of RISCV compatibility.
>
> their users are expecting to see a RISCV Libre processor.
>
> the conversation i had to have with them about the RISCV Foundation's
> intransigence, and the lack of action that had been taken by someone in
> SiFive who continues with defamation of the project despite formal
> complaints, was... difficult.

That does sound difficult, I don't envy you.  Just brainstorming here,
how much money roughly has been received so far?  Might be worth
considering what the financial downside is of backing out of that
agreement now, compared to the risk to the project it brings to stay
committed to that.  The community here might be able to raise some funds
if it seems worthwhile.

>
>
>
>>   The crowd funding campaign hasn't start
>> yet has it?
>
> there is more than the crowdfunding campaign.
>
>
>>  Although it sounds like you have a solid plan for dual-ISA,
>> I wouldn't want to jeopardize the success of the project by including
>> unnecessary complexity.
>
> we don't exactly have a choice.  committments have been made.
>
> now, if those turn out to be too difficult, after we have made every
> effort, i am fine with explaining why.
>
> but, "not trying" is definitely not ok.

That conversation may be easier to have now rather than after receiving
another year or two of funding.  It's not just the added work of
supporting dual-ISA.  Every decision made going forward on the design
will be influenced by this need to support dual-ISA whether consciously
or not.  A year spent trying to build dual-ISA _could_ be worth 2 years
spent building just one.

>
>  I'm reminded of the saga that was the mini-HDMI
>> connector on EOMA68 and how much pain and delays that were caused by
>> even a seemingly simple part.
>
> good attempt at an analogy, bad example :) Mid Mount Micro HDMI connectors
> are insanely difficult to get hold of.

:) maybe not the best analogy but I think its applicable.  The lack of
availability of those connectors put the whole project at risk.  It may
have saved a lot of time money and effort to just scrap that connector
entirely once the problems became known.  Innovative projects almost
always need to sacrifice scope to deliver what is the true vision.

Now I'm not a hardware designer, and it sounds like you have a plan that
should be fairly easy to implement to support both.  Its just that from
my experience in innovative software projects, I've seen how easily
scope creep can sneak up on you, and attempting compatibility with too
many existing systems can destroy any room for innovation and chance of
success.  That all being said, the brand name recognition of RISC-V
certainly generates some interest and evidently is generating some
funding, both of which are _also_ important to the success of the
project.  But if the customers for the 100 million units we're hoping to
ship really don't care about RISC-V vs POWER, the return from that is
far greater than the funding currently being received.  Obviously that
assumes we can get to that point without the current funding, which is
probably where this all breaks down ;)

>
>
>> This is such a critically important
> project, the world needs a commercially viable libre processor.
>> I'm assuming that the sponsor commitments are some of the NLnet proposals?
>
>  no, a separate backer (Purism)

Great!

>
>
>> This is also a good illustration of how seemingly innocuous early
>> decisions can have significant impact later in the project.  The goal
>> really isn't, or shouldn't have been to make a libre RISC-V SoC, but to
>> make just a libre SoC.
>
> sigh, hindsight.  if we had known what we know, we wouldn't have started
> from there.

Indeed.

>
>
>> RISC-V seemed like such an obvious choice early
>> on, and I'm sure the name helped raise the profile of this project to
>> begin with, but in the end was not the right choice.
>>
>> Unrelated question, are you up here in Canada?
>
> i was, i'm in cambridge uk now.
Hope you had a pleasant visit.