LWN: Comments on "Precursor: an open-source mobile hardware platform"

Precursor: an open-source mobile hardware platform

chipb — Sun, 27 Sep 2020 22:43:37 +0000

Assuming Symbiflow’s current status table is up to date, Project X-ray does not yet have (full?) support for block rams and no support for DSP blocks. These can be overcome by implementing memory and multipliers in LUTs, but the utilization profile would probably lead to a significantly different input design for that bitstream.

While I’m sure Bunnie has his eye on open source toolchains in the future, using the vendor tooling right now seems like a pragmatic choice.

Precursor: an open-source mobile hardware platform

Wol — Sun, 27 Sep 2020 13:21:55 +0000

I feel your pain ... my handwriting is crap too. Fortunately, my experience was being forced to use a fountain pen, which was easy to write with - to this day I hate biros precisely because I have to grip them too hard.

Luckily, I was able to specialise in Maths and Science, where I didn't have to write too much. One of my abiding memories is a crap RE teacher, who accused me of slacking when in reality I was writing flat out trying to keep up!

Computers weren't an option for me - I remember a lab technician bringing his brand new Sinclair ZX80 in to show us 6th-formers.

Cheers,
Wol

Precursor: an open-source mobile hardware platform

nix — Sun, 27 Sep 2020 11:07:00 +0000

Back in the days before handwriting became an obscure niche skill, we were probably attracted to computing in part because *they can write for you*. Suddenly you can write as much as you like and have it typeset in something that looks so good it shocks your teachers (hello, TeX) rather than turning out stuff even you can't read at perhaps four words per minute. And it doesn't hurt!

(yes, handwriting hurts if you're bad enough at it. Everything cramps in well under a minute because of how hard I had to grip the pen to get any control at all.)

Of course that meant I wrote far too much instead of far too little. What my teachers thought of *that* development is unrecorded.

It's a shame my school was so hidebound that this was all forbidden out of hand until I was in the sixth form. Five years of pointless pain and grades so permanently terrible that to this day I don't know how good I am at classical scholastic subjects, only that my handwriting was bad. Couldn't have the pupils forgetting how to write by hand! That's a more important skill than anything else they could possibly learn! (This period was 1987-1992, so word processors were not unknown unless you were stuck in a ten-year-deep time warp.)

Precursor: an open-source mobile hardware platform

Cyberax — Sat, 26 Sep 2020 20:39:17 +0000

On the contrary, doctors have _excellent_ handwriting. Imagine being able to handwrite many books' worth of notes during their medical school. It's just that it became so specialized that other people can't read it.

If anything, fine motor skills also need to be learned during childhood. That's why extinction of cursive writing is kinda worrying. It would be interesting to see how China where hand-writing and calligraphy are taught from the first grade would compare with other countries.

Precursor: an open-source mobile hardware platform

mpr22 — Sat, 26 Sep 2020 18:42:50 +0000

STEM in general is packed to the gunwales with people with absolutely terrible handwriting :)

Precursor: an open-source mobile hardware platform

ssmith32 — Sat, 26 Sep 2020 18:41:02 +0000

lol. So, based on their penmanship, I guess I have yet to meet a single doctor or pharmacist that has "developed their intellect". 🤔?

Precursor: an open-source mobile hardware platform

himi — Thu, 24 Sep 2020 03:22:11 +0000

Better than a guard dog, too - when was the last time you had a dog that laid eggs?

Precursor: an open-source mobile hardware platform

amacater — Wed, 23 Sep 2020 13:24:23 +0000

... and the proprietary software that drives them. In another part of the forest, I've friends who complain that their particular smart pens are no longer as readily supportable / capable of being updated because they rely on internet access and cloud technology.

Precursor: an open-source mobile hardware platform

amacater — Wed, 23 Sep 2020 13:08:36 +0000

As someone with diplegic cerebral palsy I cannot agree with nix's comment enough. It's coincidental that I drifted into computers and Linux after amateur radio - but the technical folk here seem familiar, understanding and understood and Linux has amounted to an enabling tool that bootstrapped a career. I would (now) identify myself more as neuro-diverse than physically disabled / mobility impaired (though all are true).

There is a significant dispute as to whether cerebral palsy amounts to a developmental difference in utero (thus being differently wired from the outset) / a perinatal problem or even, potentially, something apparent only months after birth - it certainly renders my dyslexic/dyspraxic/high functioning autistic friends and colleagues more understandable to me.

Precursor: an open-source mobile hardware platform

Wol — Wed, 23 Sep 2020 10:09:54 +0000

:-)

Good advice for the tech apocalypse :-)

Cheers,
Wol

Precursor: an open-source mobile hardware platform

nix — Wed, 23 Sep 2020 09:00:56 +0000

Just keep a goose in the room. Don't think of it as a noisy, smelly bird that messes the place up and really wants to get out and have some nice grass: think of it as a quill supply!

Precursor: an open-source mobile hardware platform

nix — Wed, 23 Sep 2020 09:00:00 +0000

Yes, but... they're only "closely connected" in that people with general neurodevelopmental problems often exhibit damage to both cognitive and motor skills (get deprived of oxygen for long enough at birth and both are going to get damaged). You can also have only one damaged, or one damaged much more than the other. It is true that motor circuitry *is* used for other things, so people with seriously damaged motor skills will exhibit a curious pattern of damage elsewhere, but if your intellect is largely intact you can work around the non-motor aspects of this pretty well, well enough that people who don't know you well think you are unimpaired.

It is certainly *not* the case that motor damage equals intellectual impairment, thankyouverymuch: indeed in some conditions it is often the opposite, with motor damage and strong skills in various intellectual domains going together (heck, given the number of people with autistic-spectrum disorders in tech, it's probably getting on for the *common case* in some tech companies and quite possibly on LWN to, though that is not true in the general population).

Precursor: an open-source mobile hardware platform

Wol — Tue, 22 Sep 2020 23:11:27 +0000

Smart pens are only smart if they've got a charged battery.

Pencil, pen-knife, and you probably don't even need paper, and you can keep going for ages. Likewise a pen and full ink-bottle...

Cheers,
Wol

Precursor: an open-source mobile hardware platform

Cyberax — Tue, 22 Sep 2020 21:37:09 +0000

Penmanship (and other fine motor skills) appear to be closely connected to development of the intellect. And so far nothing I've tried can replicate the tactile feeling of a pen on paper.

Precursor: an open-source mobile hardware platform

nix — Tue, 22 Sep 2020 21:34:24 +0000

Or that any of our modern styli are superior to traditional tools like pen and paper.

Modern smart pens are able to do things like remember what you wrote and give you a jpeg; accept audio recordings (and, in future, no doubt, do TTS on them, and accept what you wrote, do handwriting analysis, and speak it back to you) and a whole bunch of other things that pen and paper just cannot do. (For that matter, if only it was more widely available with non-insanely-proprietary usage terms, and if only it wasn't quite so breakable, e-paper has paper beaten hollow for many uses.)

(Not that it matters to me because I can't handwrite in any meaningful fashion.)

Precursor: an open-source mobile hardware platform

geert — Tue, 22 Sep 2020 18:59:30 +0000

As the SoC is a softcore RISC-V, it can be extended with instructions for doing optimized EC curve encryption.

Precursor: an open-source mobile hardware platform

catilac — Tue, 22 Sep 2020 17:48:40 +0000

Wow. Two years before I was born haha. I'm glad we've come along way with it all.
I'm hopefully about the e-ink work being done. I think we can encode the tactile experience into the materials.

Anyways, I better just focus on learning how to write some drivers once my client work comes to an end next week :)

Precursor: an open-source mobile hardware platform

rgmoore — Tue, 22 Sep 2020 17:43:20 +0000

Touch control has been out there in some form for quite a while. The first computer I saw with touch control was an HP-150, which was released back in 1983. It wasn't very good touch control- it couldn't measure to a single character- but it was out there.

Precursor: an open-source mobile hardware platform

excors — Tue, 22 Sep 2020 17:24:11 +0000

They say step 2 is using their "custom SoC", not a standard one. That's the step that will cost "a couple million dollars", and should achieve their goal of protecting against attacks at the RTL level as well as at the software level (and some protection at the physical level, and not much protection at the manufacturing level).

For a few dollars you can get e.g. a SiLabs EFR32MG21 which has secure key storage, crypto acceleration hardware, physical tamper resistance, DPA countermeasures, etc. That's not going to provide the security that Betrusted wants, since it requires trusting SiLabs and any governments who can force or trick SiLabs into creating backdoors. But it sounds like good enough security for a usable prototype device, and I suspect it's better physical security than they could implement on an FPGA, and it's a lot cheaper and easier.

(https://betrusted.io/betrusted-architecture/ says for "rendering a UI rich enough to handle the linguistic diversity of humans, several megabytes of RAM are required", and that chip only has 96KB RAM. Admittedly I haven't done any calculations but I'm surprised it would need that much RAM for the UI on a 336x536 monochrome display - surely you just need a bitmap font in flash (not RAM), and some very simple code to render it? You can use off-chip flash (with signature verification to prevent tampering) so you have effectively unlimited storage for the font, so you can still provide full CJK support with extremely low RAM usage.)

Precursor: an open-source mobile hardware platform

ledow — Tue, 22 Sep 2020 15:31:46 +0000

Step 2 ironically is using a standard SoC.

Step 3 is apparently a device which they cannot verify matches their submitted design.

This all appears very backwards to achieving something. Surely you'd prototype with off-the-shelf hardware, then customise down when you know it works and what you need, not the other way around?

And the end admissions over not actually having any control over the design of the final "secure enclave" they're designing - something which presumably would be of great interest for certain places to compromise - sounds a really bad way to end the manufacturing for such a product.

ASLR isn't going to fix compromised hardware.

Precursor: an open-source mobile hardware platform

nix — Tue, 22 Sep 2020 14:44:01 +0000

Yeah. Touch control really needs something you can hold like a book, which means you need compactness. It seems to me that the dread 'gorilla arm' syndrome will always apply to larger, vertically-held screens (but then my coordination is so bad that I get the gorilla arm syndrome at all times, so I could be wrong :) ).

Precursor: an open-source mobile hardware platform

catilac — Tue, 22 Sep 2020 14:34:13 +0000

I know that future generations will expect touch screens. Fine. What I still struggle with is the idea that a non-tactile interface is supposedly a good idea. Or that any of our modern styli are superior to traditional tools like pen and paper. I'm not saying we should go back, but I think we should go further and find a way to build something that doesn't completely throw out a crucial dimension of the experience.

Maybe I shouldn't worry so much when everyone is obsessing over their mechanical keyboards :)

Precursor: an open-source mobile hardware platform

excors — Tue, 22 Sep 2020 14:32:02 +0000

I get the impression that it's not really aimed at any market, it just exists because it's step 1 of https://betrusted.io/dev-plan/ - i.e. Betrusted is the thing they really want to make, but that will take years and cost millions, so they're starting with an FPGA-based prototype that can demonstrate and test the user experience. (That doesn't sound an unreasonable plan. Though why not use a cheap SoC instead of an FPGA in the prototype, which would be faster and cheaper and easier to develop for, so they can get wider feedback on the user experience in a device with an imperfect but decent level of security (which should be more than good enough for a prototype), in parallel with doing the custom silicon development with regular non-portable FPGA dev kits?)

Now they're calling the prototype Precursor and crowdfunding it. That doesn't seem unreasonable either - maybe it could be useful or interesting to some as-yet-unknown group of other people, and if it is then they can share the development costs, and if not then the attempt hasn't really cost anything, so might as well try it. But "expensive FPGA with a battery and tiny keyboard" does sound like an extremely narrow niche, so it's hard to imagine exactly who would find it useful outside of Betrusted prototyping.

Precursor: an open-source mobile hardware platform

ledow — Tue, 22 Sep 2020 13:28:04 +0000

Once had a laptop brought to me with a smashed screen. Quite clearly a large finger-sized pressure point, from which all the broken cracked glass originated.

Turns out it was given to a four year old who just assume everything was like their tablet, so they kept pressing the screen harder until it responded. Or, in this case, broke.

The next gen are never going to understand that for my generation touch-control started off in Star Trek and only came to life many, many years later.

Precursor: an open-source mobile hardware platform

ledow — Tue, 22 Sep 2020 13:24:34 +0000

I don't get the market for this?

The unit price of the main FGPA chip alone from an electronics supplier in the UK is £55.

That's without the SRAM, the secondary chip, the FLASH (128Mb), the interconnects, the keyboard, the screen, the PCB, the wifi chip, the USB-C port, the battery, accelerometer, etc.

It's going to cost a lot, surely - maybe £200 / $200? Especially if the runs aren't large. And it's going to be a poor match for many use cases. It's going to cost more than even an RPi Zero and an FGPA addon-board, for example.

I wouldn't be on here if it wasn't for wanting open-systems, etc. but I can't see the use-case for something that's expensive and very limited in speed/capability. This is an incredibly niche product.

I could see it as maybe an chip-emulation device, plug in to replace an existing out-of-production chip which has an FGPA module available for it, but I can't see the utility of a handheld version of that with tiny keyboard and screen running a ~100MHz chip of an odd architecture in 2020.

It seems to be focused on secure messaging and voice chat, and I wonder how a 100MHz chip is going to handle some of the latest EC curve encryptions, etc. - and that it isn't a phone because it lacks crucial things like a mic. It appears to be a wifi-only Blackberry-like thing, with a 100MHz chip. Then it launches into talking about emulating old sound chips, and 2FA.

I can't see the target market here? Hackers? Criminals? Programmers? Electronics tinkerers? Retro-gamers? It seems to jump around while not really being suited to any of them, or having questionable utility for them.

Could someone enlighten me? What would you use it for, and how would it be better than what already exists to do that?

Precursor: an open-source mobile hardware platform

leromarinvit — Mon, 21 Sep 2020 19:25:34 +0000

> Excuse me for cooling down your and even more mine excitement, but it seems that blob tools equivalent does not exist for Xilinx yet

It does: https://symbiflow.github.io/

I haven't tried it yet, so I can't say if it would work for this project in its current state, but the open FPGA toolcahin situation certainly seems to be getting better.

Precursor: an open-source mobile hardware platform

nix — Mon, 21 Sep 2020 19:19:59 +0000

You're not the only one. I remember when my niece, aged perhaps two and a half, picked a book off my bookshelves and, on the basis of the size and the obvious display, tried to turn the pages by swiping the cover :)

(and now, six years later, she's devouring books faster than me.)

Precursor: an open-source mobile hardware platform

martin.langhoff — Mon, 21 Sep 2020 18:41:10 +0000

One of Bunnie's prior projects -- Precursor's precursor :-) -- is Novena, which is a reasonably small-but-powerful laptop board, which includes an FPGA.

Precursor: an open-source mobile hardware platform

catilac — Mon, 21 Sep 2020 15:56:38 +0000

Oh wow, I totally misunderstood what this is. It's "not looking to replace your smartphone." :)
Funny how quickly I assume things are just phones because they're small and have a keyboard.

Precursor: an open-source mobile hardware platform

pizza — Mon, 21 Sep 2020 15:46:23 +0000

FWIW, this is explicitly not a "phone", and it also deliberately lacks a built-in microphone.

Now there is no inherent reason why it couldn't be used as a VOIP endpoint, but only over WiFi using protocols that are suitably securable.

Precursor: an open-source mobile hardware platform

catilac — Mon, 21 Sep 2020 15:23:01 +0000

This is really exciting. I would love to work on a project like this. Phones that aren't surveillance tools against the user shouldn't be a specialty device.

Precursor: an open-source mobile hardware platform

dskoll — Mon, 21 Sep 2020 15:08:23 +0000

I meant a modern CPU like the ones you typically find in mobile devices. I'm not saying it's useless by any means, but you have to have reasonable expectations.

Precursor: an open-source mobile hardware platform

tlamp — Mon, 21 Sep 2020 14:46:39 +0000

Agree here.
I made some games and applications with 8 to 16 MHz 8-bit AVR processors, with 2 to 8 KiB RAM and 32 KiB program flash, a few years ago. Compared to those MCUs this is a performance monster.

One can really do a lot with a few tens of MHz, especially with custom tailored software.

At 100 MHz + MMU even Linux is a real feasible option, a cut down kernel made for that exact HW can run quite efficiently.

The VexRiscV actually has documented Linux support https://github.com/litex-hub/linux-on-litex-vexriscv

Precursor: an open-source mobile hardware platform

ppisa — Mon, 21 Sep 2020 14:45:33 +0000

Excuse me for cooling down your and even more mine excitement, but it seems that blob tools equivalent does not exist for Xilinx yet

From https://github.com/betrusted-io/betrusted-soc

Building

Check out this repo with git clone --recurse-submodules <repo>.
Ensure you have Python 3.5 or newer installed.
Ensure you have make installed.
Download the Risc-V toolchain from https://www.sifive.com/products/tools/ and put it in your PATH.
Go to https://www.xilinx.com/support/download.html and download All OS installer Single-File Download
Do a minimal Xilinx install to /opt/Xilinx/, and untick everything except Design Tools / Vivado Design Suite / Vivado and Devices / Production Devices / 7 Series
Go to https://www.xilinx.com/member/forms/license-form.html, get a license, and place it in ~/.Xilinx/Xilinx.lic
Run ./betrusted-soc.py (or python3 ./betrusted-soc.py)

But even with the Viado use, it is great project and opens many followup designs.

Precursor: an open-source mobile hardware platform

pizza — Mon, 21 Sep 2020 13:43:15 +0000

"A cool development board, but performance won't be anywhere near a real CPU core."

That's a a bit of a non-sequiter.

Sure, at 100MHz it won't be a particularly _fast_ CPU core, but it's hard to claim that it's not somehow a _real_ core.

The VexRISC-V core they chose is quite configurable -- scaling from 0.52 DMIPS/MHz to about 1.44 DMIPS/MHz, and it's not clear what options were chosen for the Precursor design. But even with unlimited FPGA (or ASIC!) resources thrown at that core, it doesn't seem to be capable of exceeding about 200MHz.

(As a point of comparison, the wildly successful ARM Cortex-M0+ core is 0.95-1.36DMIPS/MHz, and has a hard upper limit of 64MHz, due to the utter lack of pipelining in the multiplier. Does that relatively low max clock make it any less of a "real" CPU?)

Precursor: an open-source mobile hardware platform

dskoll — Mon, 21 Sep 2020 12:08:10 +0000

It's a Xilinx Spartan 7 FPGA with about 52K logic elements. I'm impressed they can generate a bitstream without having to use Xilinx's proprietary tools (at least, that's how I read it.)

A cool development board, but performance won't be anywhere near a real CPU core.

Precursor: an open-source mobile hardware platform

atai — Sun, 20 Sep 2020 23:56:52 +0000

the core notable:

_Precursor is powered by the software-defined hardware of a Field Programmable Gate Array (FPGA). FPGAs are a sea of basic logic units that users can wire up using a “bitstream”. Precursor comes pre-loaded with a bitstream that makes the FPGA behave like a RISC-V CPU, but you’re free to load up (or code up) any CPU you like, be it a 6502, an lm32, an AVR, an ARM, or something else_