On 09.07.19 11:59, Vinzenz Vietzke wrote:
you guys work
exclusively with x86-processors, am I right?
Currently, yes. But we also have an eye on ARM and are trying to extend
our laptop portfolio towards it.
A nice idea would be a cluster of several ARM SoCs, for example using
pcie-endpoint or MII (w/ separate switch chip).
I'm thinking about patching up qemu and docker to run workfloads on
remote CPU nodes, while using faster platform specific interconnects
than just plain network. Haven't done actual planning yet, but my
crystal ball suggests something in the scale of 1 man-year for such
OTOH, we could also consider an heterogenous NUMA-like architecture,
that can even schedule tasks across several nodes (okay, that's bit
more tricks, but certainly possible)
Can you give
us some words about what you are trying to do with a
potential developer in the future with your Firmware? Getting
(further) rid of IME?
That's nothing really fixed yet and partly depends on the dev's
knowledge and interest. The overall vision is to reduce proprietary,
privacy intriguing and insecure features as much as possible. That of
course includes IME but is not limited to it. Anything putting control
back into the hands of the hardware owner is something we'd love to see.
With x86/intel, we're pretty much lost at that front - unless somebody
manages to crack the IMEs bootloader signature privkey, so we could
run our own firmware there.
That also would give us a great opportunity for a much cleaner way of hw
init / early boot: let IME bring the board into a state where a kernel
can directly run, and also let it do low-level hardware management like
temperature control, etc. ... but for now, that's all just phantasy :(
I think the best idea would be just moving away from x86.
Enrico Weigelt, metux IT consult
Free software and Linux embedded engineering
info(a)metux.net -- +49-151-27565287