<cr1901_modern>
(Count12 test fails when I try to run tests. No time to really debug in detail)
rohitksingh_work has joined ##openfpga
Bike has quit [Quit: Lost terminal]
StCypher has joined ##openfpga
digshadow has joined ##openfpga
<feuerrot>
gruetzkopf: you also put the storage into the rack?
StCypher has quit [Read error: Connection reset by peer]
StCypher has joined ##openfpga
sielicki has quit [*.net *.split]
anuejn has quit [*.net *.split]
sielicki has joined ##openfpga
m_t has joined ##openfpga
<azonenberg>
cr1901_modern: ping me in a day or two
<azonenberg>
Just got home from a multi-day sar op, have a few minuts to restock food and batteries and then grab a bit of sleep before re-deploying in the morning
<azonenberg>
cr1901_modern: the not implemented message is intentional and only applies to the sim model
<azonenberg>
it should be fully supported in hardware but i have no time to debug
<openfpga-github>
[Glasgow] whitequark commented on issue #33: So, I've had an issue several times that the Mouser part number didn't match the value of a resistor (less a problem with capacitors because we only have like one capacitor whose value actually matters, the rest is decoupling). It's a thing that's *really* easy to miss with manual review, so I want to automate it.... https://github.com/whitequark/Glasgow/iss
wpwrak has quit [Read error: Connection reset by peer]
wpwrak has joined ##openfpga
rohitksingh has quit [Quit: Leaving.]
rohitksingh has joined ##openfpga
<rqou>
finally my immune system decided it was going to kick in
<rqou>
my throat finally hurts less and my nose is now full of nice thick green mucus
<awygle>
And now I feel sick
<awygle>
Glad you're feeling better
<rqou>
whoops :P
<rqou>
disease transmission over IRC :P :P
<jn__>
don't catch that internet virus!
<rqou>
ILOVEYOU.vbs
<whitequark>
awygle: looks like I have fought setup/hold timings for one tiny instance and won
<awygle>
oh?
<whitequark>
I have one synchronous FIFO piping correct data into the host
<awygle>
hooray!
<daveshah>
whitequark: awesome
<whitequark>
azonenberg: there's been some pulseview change and it is REALLY fast now
<daveshah>
yeah I've noticed that too
<whitequark>
like "protocol decoder triggering on every edge on 1M samples of a 8 bit data bus also changing on every tick"
<whitequark>
and it still renders smoothly
<rqou>
azonenberg: how do you feel about packing some XeF2 into a chip to make it self-destructing?
<whitequark>
what
<whitequark>
you're aware that XeF2 is explosive, right
<whitequark>
or do you mean that and not etching
<pie_>
> self destructing
<whitequark>
at least use something less sensitive, like transition metal azides
<whitequark>
(note: do not actually do this)
<awygle>
lol
<pie_>
"things i will not work with". unrelated, rqou, i managed to lose the link, who was the electron microscopbe bird site dude you mentioned yesterday
<whitequark>
do they realize it'll going to blow if you so much as hit it with your finger
<whitequark>
XeF2 is absurdly sensitive
<pie_>
whats a mote
<awygle>
i'm gonna guess SmartDust stuff
rohitksingh has quit [Quit: Leaving.]
<pie_>
Meh, I dont see a contact for nanographs, I want to ask him if the "awesome microscope manuals" are scanned anywhere and don't want to make a tweeter...
digshadow has quit [Ping timeout: 246 seconds]
<whitequark>
awygle: ha, the 4-FIFO version of arbiter doesn't pass timing
<whitequark>
26.9/30 MHz
<whitequark>
actually no
<whitequark>
that's 1-FIFO version
<whitequark>
the 4-FIFO version is... faster?!
<awygle>
o_o
<awygle>
better seed i guess??
<whitequark>
man arachne is bullshit
<rqou>
man, i really want to play with mems now
<awygle>
yup
<rqou>
azonenberg get homecmos working pl0x
<awygle>
whitequark: something to think about - at what point does it become not worth your time to keep trying to make this work in a 5k?
<awygle>
(also, synthing it for an HX8k and seeing what timing you can hit would be an interesting data point)
<whitequark>
awygle: well
<whitequark>
there's a lot of avenues for improvement
<whitequark>
timing-driven placement for arachne is one
<whitequark>
I could also just use a PLL and run it much slower
<rqou>
heh, also somebody else who has realized that silver epoxy can be used for wirebonds
<whitequark>
wtf why did yosys infer two BRAMs here
<daveshah>
whitequark: yosys' heuristic is to an infer a block ram if it will be more than 2% used
<whitequark>
ah, because I asked it to do four and two were redundant
<whitequark>
aha, if I use a buffered FIFO then it works better
<rqou>
TIL MEMS doesn't have good simulation tools?!
rohitksingh has joined ##openfpga
<pie_>
mems sounds like youd have to simulate ALL the physics
<esden>
I will have to get in touch with them directly, as I want the color order to be more specific than the one they have randomly there ... :D also have to ask about the crimps and heatshrink on the other side... as I said before on my streams, samtec main business are odd custom connectors and wire harnesses. Most of the stuff they sell you can not even select on their website... :D
<esden>
I just wanted to get close to what I want with the samples ... now the fun will begin :D
cr1901_modern1 has joined ##openfpga
cr1901_modern has quit [Ping timeout: 240 seconds]
<tnt>
esden: oh nice, I missed that on digikey, should have gotten the colored one.
<balrog>
rqou: I thought Tanner EDA was primarily focused on MEMS
<balrog>
(We used to have it here for academic use before Mentor bought them out)
digshadow has quit [Ping timeout: 276 seconds]
<tnt>
esden: btw, no streams recently ? :)
<shapr>
I'd love to see whitequark streaming glasgow dev
<whitequark>
shapr: I resent how my voice sounds so no chance of that
<shapr>
aw, too bad
<tnt>
whitequark: do it like nurdrage and use a voice changer :p
<whitequark>
tnt: won't help
<whitequark>
it's not so much just pitch as the entire set of parameters
<qu1j0t3>
get a friend to dub over it
<qu1j0t3>
they see only google's AI closed captions
<qu1j0t3>
voice firewalled
<whitequark>
that's not really livestreaming
<qu1j0t3>
could be!
<qu1j0t3>
you've seen live interpreters at conferences
<whitequark>
meh, I don't really see much point anyway
<whitequark>
I get distracted a *lot*
<shapr>
I understand that, I have many chunks of 45 minutes of productivity.
<qu1j0t3>
i have few chunks of four minutes of productivity.
<awygle>
i keep planning to stream stuff
<pointfree>
I can never get myself to listen to my own talks. I guess I just sound different outside of my inner ear. A friend of mine (Sam Falvo) listens to his own talks and his are much smoother than mine. Maybe I should try listening to my own talks on youtube as well, there could be something to it.
<awygle>
then convincing myself it's uninteresting
<awygle>
i did some boring javascript streams back when i was unemployed and trying to improve my resume
<awygle>
maybe i'll stream glasgow rev C pcb stuff
<awygle>
"watch me drag traces around in kicad"
<shapr>
awygle: yes plz
<shapr>
I learn fastest when pairing with someone on something, whatever that thing might be
<rqou>
everyone knows the actual reason to watch streams is because cats :P
<shapr>
awygle's cats are cute
<rqou>
yeah
<whitequark>
awygle: ha, do it
<sorear>
pointfree: you know kc5tja?
<jn__>
rqou: c3voc needs more cats
<rqou>
don't they already have one on every podium? :P
<awygle>
i will 100% have a cat on my lap, they love it when i do pcb stuff
indy has joined ##openfpga
<jn__>
rqou: but Maneki-neko (apparently that's the right translation of Winkekatze) aren't real cats :D
<awygle>
yeah ok i'll give it a shot, why not
<rqou>
are your cats fat and lazy, or are they extremely energetic like scanlime's?
<qu1j0t3>
haha
<pointfree>
sorear: Yes, I see him almost every month at svfig. I've learned a lot from conversations with him. sorear, do you know him through risc-v stuff?
<q3k>
some day I'll get a second monitor so I can watch streams ,_,
<q3k>
i feel like I'm missing out
<awygle>
rqou: one is fat and one is skinny, one is lazy and one is energetic, but the pairing is not what you'd expect
<zkms>
good morning
<rqou>
lol
<rqou>
afternoon zkms
<sorear>
pointfree: yes, and I also touched kestrel many years ago
<sorear>
back when it was a 65816 project
<awygle>
apparently i should be at this "teardown" conference in portland?
<esden>
tnt: Digikey has a very limited selection of samtec stuff... you have to look on the samtec site directly and order samples/products from there. Digikey is also puting quite a heavy markup on samtec stuff...
<rqou>
yeah people seem to be flooding my TL about it but i don't really "get" it
<awygle>
well, people i think are cool will be there, and it's 3h away by car. so i feel like maybe i should head down there this weekend.
<rqou>
timing is a little bit awkward anyways
<esden>
you all did not hear about teardown until today? o_O
<awygle>
i didn't, no
<awygle>
is it a Big Deal?
<zkms>
what is teardown
<q3k>
i haven't either
<rqou>
i don't want to have to rush up and back down in time for commencement
<tnt>
esden: yeah. I didn't look but since I'm in europe I expect shipping might be killing me if I order from samtec just a few connectors :p
<esden>
it is the first time they are doing it... but it is promising to be on par with supercon ... so yes very big deal :D
<rqou>
one week later would have been perfect
<esden>
and in my backyard so perfect :P
<rqou>
i don't know about supercon either
<awygle>
hm maybe i'll get down there for sat/sun
<esden>
rqou: well now you do :D
<rqou>
also, idk who "they" are or why they matter
<awygle>
rqou: hackaday
<awygle>
hackaday supercon
<rqou>
hackaday is hosting teardown?
<esden>
no crowd supply is teardown
<esden>
hackaday is supercon
<rqou>
yeah somehow i don't really know anything about these
<rqou>
other than people (mostly @pdp7) flooding my TL about it
<esden>
so far supercon is the best US based hacker related conference I have been to
<q3k>
i went to the hackaday unconference in dublin, it was pretty decent
<q3k>
although very fast paced and draining for me
<rqou>
not BH/DC?
<q3k>
FUCK defcon
<q3k>
it's garbage
<esden>
wtf no both suck
<pointfree>
sorear: Yeah kc5tja has given an enourmous number of talks at svfig meetups. Always something new to apply. I'm tempted to write an irc client and gopher client for one of the kestrel forths.
<jn__>
yay, gopher \o/
<esden>
rqou: do not waste your time on BH/DC ... unless you have a very good commercial reason to go there do not waste your time or money
<q3k>
even the ctf at defcon doesn't make up for the shitty conference
<q3k>
and it's a good ctf.
<whitequark>
sign, so many off by ones
<whitequark>
sigh*
<daveshah>
whitequark: I feel in some contexts that could even be a Freudian slip
<whitequark>
hahaha
<pointfree>
jn__: irc and email don't take much code to implement. That just leaves the web. If I were to set up a gopher proxy to the web, based on the readability bookmarklet maybe I could use the kestrel or a microcontroller such as a PSoC 5LP for most of my daily computing needs.
<whitequark>
"email doesn't take much code to implement" uh
<whitequark>
really
<pointfree>
whitequark: I used to send emails with telnet.
<daveshah>
If you're going that route, go the full open hog and go for risc-v on ice40
<daveshah>
With external hyperram it should easily be able to handle all that stuff
<whitequark>
pointfree: I can send emails with telnet
<whitequark>
but it's one thing to do that and the other is to have a compliant implementation
<Ultrasauce>
i sure as shit wouldn't want to use my own tcp/ip implementation on a daily driver
<rqou>
I've sent email using telnet, but going through a local MTA first to fix all of the noncompliance :P
<whitequark>
Ultrasauce: i totally would
<whitequark>
my tcp/ip implementation is very solid :P
<esden>
q3k: yeah I do not attend the ctfs ... I heard those are good. But vegas in general is not a place for humans to be. The drinking party culture at DC is also not really fun. I prefer the Mate fueled culture of CCC, over booze fueled DC...
<Ultrasauce>
solid but very much a subset is it not?
<q3k>
yep.
<q3k>
i mean, ccc isn't sober either
<esden>
no definitely not... but that is nothing compared to DC
<q3k>
but you don't go there to get waster and P A R T Y A T T H E P O O L organized by CLOWNSTRIKE
<q3k>
*wasted
<whitequark>
Ultrasauce: define "subset"
<rqou>
no ipv6 :P :P :P
<whitequark>
it has ipv6
<rqou>
oh?
<whitequark>
we merged that recently
<q3k>
whitequark: by the by, what congestion control algorithm did you employ in smoltcp?
<rqou>
nice
<whitequark>
q3k: lololol it doesn't have any
<q3k>
oh
<whitequark>
that's probably one notable omission out of three
<rqou>
<troll> no tcp/ip over classic bluetooth PAN </troll>
<whitequark>
the other two are window scaling and probing zero windows
<Ultrasauce>
well i mean your readme is full of big bold "it doesnt do this" lists
<rqou>
but seriously, no non-Ethernet transports
<whitequark>
Ultrasauce: sure, because i like being very explicit
<whitequark>
rqou: no one wanted it badly enough to implement?
<rqou>
i can totally imagine somebody wanting smoltcp to be usable for TCP over SLIP/PPP?
<whitequark>
we used to have PPP in ARTIQ
<whitequark>
but it wasn't important enough to fund during smoltcp migration
<daveshah>
I ran PPP over LoRa once
<daveshah>
It was actually usable, at about 1kByte/s
<rqou>
also no .1q
<daveshah>
Might be interesting for certain low bandwidth embedded applications
<pointfree>
whitequark: Yeah I'd probably just write an imap4 forth wordset. I wouldn't write an email server right now.
<whitequark>
rqou: no one wanted it badly enough etc
<rqou>
wat
<rqou>
i actually want .1q
<whitequark>
hm?
<whitequark>
sure, write a PR :P
<rqou>
lol at some point
<q3k>
802.1q sounds trivial to implement
<q3k>
unless I'm missing something
<whitequark>
sure
<q3k>
if you leave that unimplemented I might do it if I ever touch smoltcp
<rqou>
it is if you don't care about the priority part
<rqou>
(which i personally don't)
<q3k>
i don't think anyone does
<rqou>
BRCM does
<rqou>
my father did
<whitequark>
hit BRCM with an ICBM imo
<rqou>
lol
<rqou>
i actually have used DiffServ though
<q3k>
nuke broadcom from orbit indeed
<rqou>
i was doing a bulk upload to crashplan and a housemate complained that this made overwatch very laggy :P
<whitequark>
i just used linux tc on the router
<rqou>
poke the diffserv classes a bit and linux automagically improves it
<rqou>
explicit messing with tc isn't required
<rqou>
also seriously we have way too freaking many ways to mark "class of service" type of information
<zkms>
i just used fq_codel or whatever the heck that's called
<rqou>
yeah well crashplan's gui just has a little dropdown that pokes the diffserv label
<rqou>
so that was even easier
<whitequark>
oh that makes sense
<q3k>
small brain: use 802.11p
<q3k>
normal brain: manually assign TC classes to devices
<zkms>
whats 11p
<q3k>
cosmic brain: overprovision the fuck out of your uplink
<q3k>
*.1p
<q3k>
derp
<awygle>
11p is vehicles
<rqou>
oh yeah wifi has yet another way to encode class of service information
<rqou>
that's distinct from .1p or diffserv
<rqou>
honestly a home network only really needs three traffic classes: bulk, normal, and real-time
<zkms>
tbh you can get good results with fq/sfq-ish things that isolate flows and a queueing mechanism that does head-drop and measures the latency it imposes (and keeps it under a certain threshold)
<zkms>
without any sort of manual classification
<awygle>
Manual classification always feels like you've already lost
<awygle>
Train a neutral network to measure perceived latency using some kind of human factors model, te
<awygle>
*tweak queueing params based on output
<rqou>
lol
<sorear>
Are there good implementations of real-time? I mostly hear complaints about USB isochronous
<rqou>
yeah i don't get what the deal is with USB isoc
<rqou>
afaict on the bus level it's not that much different
<rqou>
but apparently in the driver layer everything is fucked?
<whitequark>
it's identical on the bus level because it only affects how the host schedules packets
<whitequark>
well, almost identical
<whitequark>
there's no ACKs and some stupid token toggling
<rqou>
yeah, scanlime keeps talking about some complicated AF logic about how USB allocates bandwidth but i haven't bothered yet to look into wtf she's talking about
<rqou>
e.g. telling people they're doing it wrong by calling read() in a loop or something?
zkms has quit [Quit: WeeChat 2.1]
<pie_>
is there such a thing as a reference implementation
<pie_>
(which kind of implies it shouldnt suck >_>)
zkms has joined ##openfpga
<rqou>
idk
<rqou>
i kinda agree with azonenberg that Ethernet is the proper future
<kc8apf>
he seemed to quickly realize that his sales pitch wasn't going to work on me
<awygle>
That's a lot of people, burn rate must suck
<kc8apf>
has GV backing so they've at least convinced someone that they have a viable business plan
Bike has joined ##openfpga
<zkms>
speaking of RF links i found out that apparently there's HFT people using transatlantic HF/shortwave links to get faster latency than fibre; i knew they did this with microwave but i didn't know they'd actually gone to HF >_>
<rqou>
just attended a dissertation talk about doing deep learning on complex-valued data
<rqou>
essentially it's "yeah, you can do that, but the devil is in the details"
<Bike>
i d on't understand how that's different from just having two dimensions?
<rqou>
doing it that way doesn't quite work as well
<rqou>
so e.g. if you ReLU both channels independently, that's not really physically meaningful for your system
<whitequark>
woohooo i think i have conquered half of the FIFOs
<Bike>
that sounds like it involves a system more specific than "complex-valued"
<rqou>
not really?
<rqou>
use case was MRI postprocessing and similar "signals" tasks
<whitequark>
rqou: quite amazingly, my transfer function seems to work
<rqou>
"transfer function"?
<whitequark>
between FIFO addresses
<whitequark>
what we discussed yesterday
<rqou>
oooh that kind of transfer function
<rqou>
too many overloaded terms
<balrog>
rqou: we had graduation ceremonies today
<balrog>
when's yours?
<rqou>
next week
indy has quit [Ping timeout: 240 seconds]
<Bike>
swear the whole area around campus in the city was traffic hell for it