[libre-riscv-dev] status dec 2019

[Jacob Lifshay]

On Fri, Dec 6, 2019 at 11:53 PM Luke Kenneth Casson Leighton
<lkcl at lkcl.net> wrote:
>
> hi all, just a quick status update.
>
> am still here in canada, helping my associates start this business doing
> ebikes. http://www.teznic.com it's a long story, however yes down the line
> we want collision avoidance warnings (like the new damon electric
> motorcycle), self driving gyros, and a mesh network etc.

You mean networking is possible without centralized servers doing everything? :P
I've been thinking about implementing a P2P protocol for my game
(Voxels), once I get around to working on it again (probably be quite
a while as Libre-RISCV is taking up most of my programming time).

> funnily enough we'll need a decent processor for all that, so there is a
> way to bring in Canada R&D funding.  the general idea being, here: the
> company my associates are setting up, they "get it" and will be a major
> customer.

Yay!

> for NLnet, with help from everyone, i answered the questions for Michiel,
> on the applications, the main thing that happened was that the external
> reviewers were really surprised at the difference in the amounts from the
> first application.
>
> <snip>
>
> the strategy here is to get the video and 3D projects each working in
> software first (unoptimised) then add SimpleV then add accelerated opcodes
> (most likely one by one) to see if they improve efficiency and performance.
> this last part is what has had its budget cut, reducing the number of
> possible hardware-related iterations.
>
> we will hear soon enough and will have to start putting together some
> milestones, for each project.

Sounds good to me!

> the ones i took off the list are the opencl one and the gcc one. the gcc
> one realistically needs a quarter of a million dollars minimum, to attract
> people with the right level of expertise.  opencl can be done later.

GCC can also be done later, since GCC already supports running
programs on RISC-V and Power, just without using the new instructions.
Rustc/Clang/etc. work fine with LLVM, which will have support for all
the new instructions, though probably not in the front-ends right
away.

One other thing that needs to be worked on soon is to determine if
we're going to fosdem and if nlnet or someone else is willing to help
with expenses, there's only a few months left and that kind of thing
is usually better done well in advance.

Jacob