[libre-riscv-dev] PowerISA, NLNet grants

Luke Kenneth Casson Leighton lkcl at lkcl.net
Sun Jan 19 01:26:32 GMT 2020


ok so let's be clear about the goals, here:

1.  the project shall be a hybrid CPU-GPU because if it is not, the
complexity involved in developing a split shared-memory CPU-GPU both at a
hardware and a software level will be so costly it will jeapordise the
project.

2. the project shall be commercial and mass-volume (100 million units and
above)

3. the project shall be entirely transparent so that end-users will be able
to trust it

4. the source code shall be available at all times for all components for
BUSINESS reasons, making development and use of SDKs dead simple and aiding
and assisting developers AND BUSINESSES in debugging and thus hugely saving
them money.

with these goals in mind - which are BUSINESS objectives - not "Libre
airy-fairy oh i'm a religious frothing moronic deluded zealot fighting for
freedom without actually thinking through how to achieve that" - when
presented with the intransigence of the RISC-V Foundation, we have a choice
as to whether to destroy / sacrifice / ignore any one of these objectives,
or whether to find an alternative option.

it should be fairly clear that the above goals are both reasonable and
inter-dependent.  any one is removed or not fulfilled, it makes the entire
set worthless and pointless to pursue.

thus it comes down to an analysis of the "blockers":

1. NDAs (which prevent and prohibit us from being fully transparent) are
directly incompatible with the above goals.

2. we have been told - repeatedly - to "get lost you stupid libre losers,
stop wasting everybody's time, and go write a Custom Extension".  Custom
Extensions are *specifically* for *private* proprietary usage (Western
Digital's private HDD / SSD customisation is a really good example, as is
NVIDIA's private customisation for use in their newer GPUs).

3. the RISC-V Mailing lists are entirely closed.  to meet the full
transparency business objectives, we would have to cc the libre lists.  by
doing so, when hitting "reply" to any such private discussion, we would not
only be violating the Terms and Conditions of Membership by sharing the
contents of those mailing lists with the outside world, we would be placing
ourselves at the risk of being sued for sharing other member's
"Confidential and Proprietary Information".

4. it would be irresponsible of us to pursue the "Custom Extension" route
because if not properly endorsed (the opcodes not properly reserved), it
would create a high-profile "Altivec" conflict situation in the ISA.  this
would not only destroy *our* reputation, it would destroy the reputation of
RISC-V entirely as a conflict-free ISA.

5. with the RISC-V Foundation's Directors being unable and unwilling to
follow the legal requirements of Trademark Law, we would have to engage in
a protracted legal battle just to get them out of their intransigence.  we
have better things to do.


by complete contrast, the discussions with the Director of the OpenPOWER
Foundation were, as i explained in the crowdsupply updates, extremely
positive and understanding.  everything that i raised as a concern, Hugh
had *already discussed* with the Open Power Members - IBM, Freescale/NXP,
and others.

this because Hugh is himself a long-standing member of the software libre
community, who keenly understands the *business* benefits of collaboration
and transparency.

l.


More information about the libre-riscv-dev mailing list