<jn>
^ there's this, but it doesn't even have a license
<jn>
(or maybe it does now?)
<rqou>
ok, there's research papers i guess
<Wallbraker>
Ah cool thanks
<Wallbraker>
That looks quite dead, maybe something to start from.
cr1901_modern1 has quit [Read error: Connection reset by peer]
<azonenberg>
Wallbraker: So
<azonenberg>
I am not aware of any FPGAs that are f/oss, however
<azonenberg>
I do have a synthesizeable, fully functional, bitstream-compatible verilog model of a CPLD
<azonenberg>
the XC2C32A, to be specific
<azonenberg>
it's fairly inefficient and follows the actual silicon structure closely without any attempts to optimize for running on an FPGA, so it uses a lot of LUTs
<azonenberg>
But it does work and has been tested in hardware
<azonenberg>
fills about half of an xc7a100t, iirc, for 32 macrocells
<azonenberg>
and can run actual coolrunner bitstreams
<azonenberg>
I made it a while ago to verify that me and rqou actually understood the bitstream format
<Wallbraker>
azonenberg: Oh that's really cool!
<azonenberg>
Wallbraker: In general this channel is (so far) mostly about taking existing FPGAs with binary-blob compilers and making our own f/oss toolchains for them
<azonenberg>
however, open HDL designs to run on FPGAs (with proprietary or f/oss toolchains) come up in conversation fairly often
<azonenberg>
Creation of a fully open FPGA from scratch (i.e. new silicon) is very much on the long-term roadmap
<azonenberg>
however we're a long way from even thinking about specifics
<Wallbraker>
Yeah, that's what I figured. But I thought you guys might at least knew of some projects
<Wallbraker>
nods
<azonenberg>
Once we have a team with solid experience in EDA tool design and understanding of existing architectures, we can try and develop our own based on what we learned
<azonenberg>
I'm aware of one class project years ago that actually got fabbed but they didnt release full specifics that i know of ,and i think the toolchain was just manually bit twiddling
<azonenberg>
That plus the xilinx coolrunner-2 documentation should be enough to understand how it works
<azonenberg>
The HDL in the model is commented/structured with the assumption that the reader is already somewhat familiar with coolrunner microarchitecture so i didnt necessarily explain all of the subtleties of how various bits fit together
<Wallbraker>
I'm going to FOSDEM and I'm hoping to talk to the chips4makers person there. He is doing some retro CPU stuff for ASIC in order to setup a FOSS toolchain for bringing designs to ASICs. Dunno how much vaporware that is.
<azonenberg>
Sounds cool but i'm not holding my breath
<Wallbraker>
Ah thanks!
<azonenberg>
Right now if i want to get a design actually running any time soon, FPGA is the way to go
<Wallbraker>
Same
<azonenberg>
So i'm just trying to open that up as much as i can
<azonenberg>
I actually havent played much with icestorm and the other low-end open toolchains
<azonenberg>
Because they're too small for the stuff i have in mind
<azonenberg>
i work with large artixes and have plans for ultrascale designs down the road
<Wallbraker>
Tho he works for a research institute that does ASICs so he would be the guy to do it.
<Wallbraker>
Looks like the guys behind YoSys are working on Artix 7 support
<azonenberg>
It's not just the yosys folks
<azonenberg>
the group calls themselves "project x-ray"
<azonenberg>
most of them hang out in here
<Wallbraker>
Ah right
<azonenberg>
They're working on 7 series in general, artix and kintex are both in progress
<Wallbraker>
Cool
<azonenberg>
So far the team's main focus has been bitstream format RE but i hear that they have some VPR place-and-route in progress
<azonenberg>
We have a guy here (awygle) who is doing research into parallel PAR
<azonenberg>
not like 4-8 cores
<azonenberg>
we want thousands
<azonenberg>
no idea if/when anything will come from it but the dream is to fill a large ultrascale or something in minutes on a giant pile of ec2 systems etc
<Wallbraker>
Nice nice
<azonenberg>
we've met up and chatted bunch (he's local to me) about architecture and plans but i havent seen code yet
<azonenberg>
I'm in the middle of remodeling a house i bought and trying to get it livable before my current landlord kicks me out (he wants to move back into his house, coming home from wherever he's at)
<azonenberg>
so progress has been slow on my end too lol
<Wallbraker>
I'm always bit skeptical to super many core systems. Looking at the amazing number of hoops GPUs jump through save memory bandwidth
<azonenberg>
We wanted to do a distributed memory architecture
<azonenberg>
like, racks of servers each doing different parts of the problem
<azonenberg>
talking back and forth over ethernet etc
<azonenberg>
using something like MPI
<azonenberg>
this gives you massively more ram bandwidth (among other things) than a single system
<azonenberg>
But with higher latency as you bounce from node to node
<awygle>
I am still working on this but I've been disrupted by my recent health issues
<azonenberg>
The hard part is structuring your program such that the increased CPU and bandwidth buys you more performance than you lose from latency
<awygle>
Hi Wallbraker, welcome
<azonenberg>
awygle: No worries, take your time recovering
<Wallbraker>
Right
<azonenberg>
i'm getting over dental surgery myself
<azonenberg>
Had three wisdom teeth pulled a while back and still cant open my mouth all the way without pain
<azonenberg>
awygle: oh, so my new house is starting to look like a proper... something, now
<Wallbraker>
GPUs just run multiple "programs" at the same time and have lots of read and writes going at the same time. So they switch quickly between them
<awygle>
azonenberg: yeah I saw, congrats
<Wallbraker>
Thanks o/
<azonenberg>
Wallbraker: normally its one program and you just run different thread contexts in and out
<azonenberg>
And they switch to latency-hide as soon as you get a LLC miss
<azonenberg>
awygle: That U-shaped piece you see there was a PAIN to build
<awygle>
GPUs are weird but fun
<azonenberg>
so many cuts and bends
<azonenberg>
then i had to get in with the angle grinder to polish all of the cut marks because i dont have the right kind of bolt cutters to get a smooth flat finish
<Wallbraker>
azonenberg: Yeah same program ;)
<awygle>
Anyway I'm going back to bed (and turning off phone notifications lol)
<azonenberg>
so i have to grind out all of the sharp edges that would damage the cables
<azonenberg>
awygle: go sleep lol
<Wallbraker>
Spent a large part of my masters optimising code for AMD GPUs.
<azonenberg>
Wallbraker: oh cool
<azonenberg>
I built a parallel password bruteforcer back in... wow was it really ten years ago?
<azonenberg>
just when CUDA was getting hot
<Wallbraker>
Cool cool
<azonenberg>
at the time i had the best performance of any known tool for BSD-MD5 hashes
<Wallbraker>
Nice!
<azonenberg>
did some fun dirty tricks like storing "strings" packed 4 bytes at a time in registers
<azonenberg>
i implemented strcat() with bitshifts and stuff on registers
<azonenberg>
never touched ram
<Wallbraker>
Oh nice
<azonenberg>
oh and using duff's device so i only had one conditional branch during the entire function
<azonenberg>
not one per loop iteration
<azonenberg>
Wonder if i still have the code somewhere
<azonenberg>
that was my freshman year in a parallel computing class
<Wallbraker>
I spent a large time battling the compiler to emit the code I wanted it to emit. :P
<azonenberg>
I segfaulted the compiler more than a few times
<azonenberg>
and got it into infinite loops
<azonenberg>
And caused kernel panics and corrupted framebuffers more times than i could count
<azonenberg>
unsure if it was the drivers or the GPUs lacking memory protection but overwriting framebuffers and/or kernel ram from a bad pointer on the GPU was stupid-easy in those days
<azonenberg>
One bad poitner and your display fills with colored snow
<azonenberg>
if you're lucky X crashes and your system reboots
<azonenberg>
your x server8
<Wallbraker>
XP?
<Wallbraker>
Haha
<azonenberg>
if less lucky you have to hard reset the system
<azonenberg>
no, ubuntu... 8.04 or thereabouts
<azonenberg>
i'm a debian shop these days
<Wallbraker>
Ah okay
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
<Wallbraker>
I'm still on Ubuntu, I don't want to tinker with the OS I'm using anymore. :)
<Wallbraker>
Just have it work
<azonenberg>
That was why i switched to debian
<azonenberg>
ubuntu kept trying to pretend to be osx
<azonenberg>
while actually being linux
<azonenberg>
and it ended up failing hard at being either
<azonenberg>
that and the frequent updates meant things broke fairly often
<azonenberg>
With debian i can go a couple of years before doing any updates other than security patches
<Wallbraker>
I do like having up to date stuff. And i grew to like Unity.
<Wallbraker>
Still prefer it to Gnome-Shell, which I feel is a big step in design
<Wallbraker>
Step back
<azonenberg>
yeah thats what made me leave
<azonenberg>
I run XFCE now
<Wallbraker>
Heh okay
<azonenberg>
with clearlooks it has a nice clean simple look somewhat like win7 but less brightly colored and without the fancy glass
<pointfree>
cyrozap had concerns about silver particle-based ink being used for high frequency antennas. There is a company, Liquid X Printed Metal that makes a particle-free inkjet ink http://www.liquid-x.com/
<qu1j0t3>
pointfree: Oh, brilliant, thanks for looking that up
<pointfree>
some papers I read about inkjet printer antennas say particle-free ink is better for high frequency
m_t has quit [Quit: Leaving]
<pointfree>
qu1j0t3: I've moved on to drawing circuits on paper with graphite and graphine. I just ordered a box of these pencils https://www.amazon.com/Drawing-Pencils-Set-Piece-Sketching/dp/B073B5G1JW ...14B is the softest graphite pencil I've ever seen (I thought they only made them as soft as 9B)
<pointfree>
softer pencil graphite is more conductive, harder pencil graphite is more resistive
<pointfree>
One can draw working capacitors on paper as two parallel lines with the paper between serving as the dialectric
<qu1j0t3>
nice!
<pointfree>
qu1j0t3: a diode can be drawn as a funnel and box
<qu1j0t3>
that is fascinating. would make a lovely video
<pointfree>
I have have also seen a working pencil and paper JFET transistor
Dolu has quit [Ping timeout: 240 seconds]
<pointfree>
btw graphene is a better conductor than copper, silver, and silicon ...graphene is a single layer of graphite. You can get it with piece of scotch tape.
jhol has quit [Quit: Coyote finally caught me]
jhol has joined ##openfpga
<pointfree>
qu1j0t3: So there you have it: resistors, diodes, capacitors, transistors with pencil, paper, paper clips for terminals, and perhaps some scotch tape for getting a graphene layer. I have it in my mind to do a pencil & paper fpga YouTube video series.
genii has joined ##openfpga
* genii
slides qu1j0t3 a fresh mug of the good stuff
<pointfree>
This morning I was drawing a Hilbert space filling curve to serve a compact, high capacity capacitor possibly for use as a battery for my pencil and paper computer.
<qu1j0t3>
genii: thanks
<qu1j0t3>
pointfree: hah
<qu1j0t3>
pointfree: soon you'll want a plotter with the pencil fitted
<pointfree>
So I could draw a capacitive keyboard for input and use capacitors for memory and hopefully put a Forth interpreter on it. Not sure how to do a display though...
<pointfree>
qu1j0t3: I thought I saw pencil based printer a while ago... it was kind of a joke project at the time, but now I want one.
<genii>
Do you guys think conductive ink on Kapton is feasible for a flexible flat cable? Would be a tiny BGA mounted with a fanout and a few pulldown resistors
<pointfree>
I wish I had thought of this when I was still in school.
<genii>
qu1j0t3: But before it went kablooie it got warm enough beam to make holding your hand in front of it very uncomfortable
<qu1j0t3>
be careful with your eyes and UV
<qu1j0t3>
i would be using goggles
<genii>
I used my welding goggles
<qu1j0t3>
hmmm
<qu1j0t3>
they may not be designed to block UV
<genii>
( high enough rated for watching solar eclipses )
<qu1j0t3>
yeah, but that's visible light
<qu1j0t3>
you can't be sure it's blocking UV
<genii>
Hm, interesting point
<pointfree>
So clay is mixed into graphite increase hardness/resistivity. I'd like to fill an inkjet cartridge with the clay and have the printer mix that with the other graphene ink (like color mixing) to control resistivity.
<genii>
Clay is comprised of fairly large particles though
<pointfree>
yeah..
<genii>
Why not carbon, they used to silkscreen carbon resistors directly onto boards before
<pointfree>
genii: Good idea. They happen to mix carbon into graphite pencils as well.
<pointfree>
genii: It seems pure cabon inkjet ink is already easy to get.
<pointfree>
*carbon
<genii>
I think the suspension medium might be the key there
<genii>
If you had a medium which entirely evaporates you'd know it was only carbon remaining
<genii>
Or if you're printing on something like plexi acetone could etch the carbon into the surface
digshadow has quit [Ping timeout: 276 seconds]
m_t has joined ##openfpga
cr1901_modern has joined ##openfpga
theMagnumOrange has quit [Quit: Leaving]
soylentyellow has quit [Ping timeout: 240 seconds]
carl0s has joined ##openfpga
<carl0s>
Hi, has anybody tried to compile icestorm/arachne on Ubuntu 17.10? seems like glibc got updated to 2.26 and it dropped support for xlocale header file.