<mithro>
Can you specify timing constraints with PCF?
pie_ has quit [Read error: Connection reset by peer]
<mithro>
I also don't quite understand the LogicCell40 model - is lcout == lut clocked out? IE when using the ff and ltout when using the output directly?
<mithro>
They don't really map to anything in the bitdocs -- They have lutff_X/cout, lutff_X/lout and lutff_X/out ....
_whitelogger has joined ##openfpga
<rqou>
ok, i fuzzed all neighbor links and outputs from the row IOs
<rqou>
i think this is finally everything in the connectivity graph?
<rqou>
16348 routing pairs
<rqou>
unless something is still missing
<mithro>
rqou: Want to start importing into the arch-defs?
<rqou>
not yet
<mithro>
rqou: What more do you need?
<rqou>
flip-flops, for one
<mithro>
rqou: You probably don't need the actual bits for the bitstream to start getting things into vpr
<mithro>
rqou: Do you have a description of how the switch box / routing in the part works?
<rqou>
azonenberg: ping?
pie_ has joined ##openfpga
ZombieChicken has quit [Quit: Have a nice day]
<azonenberg>
ack
<rqou>
how'd your SAR mission go?
<rqou>
aaargh i _just_ discovered even more weird interconnect paths
<rqou>
the journey never ends
<rqou>
azonenberg: i think i need a better approach?
<rqou>
otherwise it seems like a perpetual game of "oh shit this is also connected to that"
<azonenberg>
Lol
<azonenberg>
fund my silicon reversing projects that ioa isn't
<azonenberg>
and develop tools to re the die and figure out 100% of legal routes? :p
<rqou>
lol i wish i had that kind of money
<rqou>
anyways, i just discovered that lut outputs can feed into the local tracks in a tile that _aren't_ the special designated ones for connecting luts to other luts
<rqou>
seems this chip has a lot of "oh, we have something extra, what should we put here?" features
<mithro>
ioa?
<rqou>
wtf
<rqou>
azonenberg: lut number 4 has a cascade connection to lut number 5
<rqou>
but these aren't physically adjacent
<rqou>
how does that work?
<rqou>
azonenberg: so the physical ordering of luts is 01234 98765
<rqou>
but somehow a wire runs from 4 to 5
<rqou>
any thoughts as to wtf they might be doing?
<mithro>
rqou: Could it be something like -> 01234\n98765 (IE 0 above 9, 1 above 8, etc?)
<mithro>
rqou: IE Two rows?
bitd has joined ##openfpga
<rqou>
low-res die photo i have says no
pie__ has joined ##openfpga
<azonenberg>
yeah a 2-col tile makes sense
<rqou>
but that doesn't match the photo i'm looking at
<daveshah>
Also note that at the corners, a couple the neighbour outputs that would correspond to non existing tiles in the corners are actually used for outputs from the PLL
<openfpga-github>
Glasgow/master 6cb4f99 awygle: Minor silkscreen cleanups.
<rqou>
wow it's surprisingly difficult to figure out how to activate the logic tile control signals
<rqou>
i also can't figure out how to activate global clocks
mumptai has joined ##openfpga
xdeller_ has quit [Remote host closed the connection]
xdeller_ has joined ##openfpga
<rqou>
PSA: You need `set_instance_assignment -name GLOBAL_SIGNAL "GLOBAL CLOCK" -to a` to force use of global wires
<gruetzkopf>
*nods*
<rqou>
and yes, you can force a "global" wire that only goes to one place
<rqou>
don't do this in production of course
<awygle>
Don't Do This In Production - The ##OpenFPGA Chronicles
<rqou>
aaaah shit
<awygle>
Don't Do This In Production would actually be a great name for a biography. I would be satisfied if my biography was called that.
<rqou>
not every mux is the same like i thought
<rqou>
wait wut
<rqou>
something weird is going on
<rqou>
at this point "something weird is going on" should probably be the motto of Project Chibi
<qu1j0t3>
awygle: +1
<qu1j0t3>
mine can be called Testing in Production
<rqou>
hmm how does this work? the normally two-hot muxes become magically a one-hot mux when a global wire is used
<daveshah>
Maybe it reduces delay or skew marginally
<daveshah>
By only needing one switch stage
<rqou>
but how is the silicon built to recognize this?
<awygle>
Maybe it skips a layer, yeah. Wasn't there one bit pattern that wasn't used?
<daveshah>
So I imagine the two hot multiplexers are two one hot multiplexers cascaded
<daveshah>
Except that the final multiplexer has som direct global inputs
<rqou>
except that global nets _reuse_ bits that also appear in other patterns
<daveshah>
Without going through a first stage
<daveshah>
Oh, that's much weirder
<daveshah>
So you mean like one of the normal inputs, but with only one bit set?
<rqou>
yeah
<daveshah>
or cleared, as I think max v is?
Marex has quit [Ping timeout: 240 seconds]
Marex has joined ##openfpga
<gruetzkopf>
more like "You did what now with WHAT?" for me
<sorear>
Maybe what you think is the first stage is actually the second
<gruetzkopf>
including a not-quite-yet-working clone of the pokemon gen2 engine to centrifuge controllers.
<pie_>
<rqou> and yes, you can force a "global" wire that only goes to one place
<pie_>
yay antenna \o/ ?
<gruetzkopf>
(well, civil tech that uses the same graphics subsystem)
<pie_>
ITAR-ed pokemon games
<rqou>
pie_: not a very good antenna since there's buffers in the middle that turn off unused spurs
<rqou>
(although i think i know which bits control them)
<pie_>
aww
<pie_>
gruetzkopf, actually i just realized, how did you even get the software for the video card over laptop pcie thing working? im guessing the pcie stuff is more flexible than i suspect
<gruetzkopf>
the "cursed gaming" adapter?
<gruetzkopf>
just works
<pie_>
ah.
<pie_>
i need to move to cccac
<pie_>
rqou keeps telling me to move
<pie_>
halp
<gruetzkopf>
(with the laptop in the video having a firmware bug, which turns off the intel gpu because when there's a nvidia GPU present it's clearly the integrated one)
<rqou>
pie_: i mean, HU isn't really the best tech/hacker hub ever :P
<pie_>
stahp :(
<pie_>
gruetzkopf, almost surprised you didnt fix that too :P
<gruetzkopf>
works fine for my normal choice of thinkpads
<gruetzkopf>
even more so today than half a year ago
<rqou>
pie_: i don't even have fonts for that
<pie_>
apparently i dont either
<pie_>
shame.
<awygle>
that's a cool alphabet
<rqou>
oh wtf: "In LaTeX, the double acute accent is typeset with the \H{} (mnemonic for Hungarian) command."
<rqou>
why can't latex support unicode like a normal program?
<egg|zzz|egg>
meow
<egg|zzz|egg>
rqou: use XeLaTeX
<egg|zzz|egg>
rqou: and LaTeX doesn't because it's from 1983 and based on TeX which is from 78 :-p
<rqou>
it can't get upgraded?
<pie_>
rqou, lualatex and xelated do
<pie_>
there are ?hacks? for getting latex to work with fancy characters
<pie_>
i never really looked into how that shit works
<pie_>
seems unnecessarily arcane
<rqou>
that describes all of latex though
<pie_>
on one hand *latex is pretty cool, on the other hand, good god WHY
<pie_>
qu1j0t3 may be able to say otherwie
<pie_>
otherwise
<awygle>
i like the implication that "normal programs" support unicode
<awygle>
it seems to me more an aspiration than a truth though
<rqou>
"normal*" *void in US/JP
<pie_>
my friends game saves kept getting put in oddly named folders because his windows account name had an accent in it
<pie_>
i dont know if it was a windows or a game issue
<rqou>
fine then, "normal*" *void on win32, especially in US/JP
<pie_>
i should have known better but i spent half a week trying to get a game to run on different machine, for probably the same reasons
<pie_>
then it hit me.
<awygle>
it annoys me that it's not win64 now
<egg|zzz|egg>
rqou: I mean, yes, normal programs written sensibly in the past couple of decades by *are* Unicode-aware
<pie_>
"<rqou> nothing is good enough for you people"
<egg|zzz|egg>
but those built up from the seventies, less so
<rqou>
egg|zzz|egg: except on windows in US/JP? :P
<egg|zzz|egg>
I don't know, I'm not in US/JP
<egg|zzz|egg>
I do use Windows though, and I'm quite happy with its Unicode support :-p
<pie_>
im just glad i was born in a time unicode exists
<pie_>
cant imagine the encoding fuckery otherwise ( and yet there is still plenty of encoding fuckery
<rqou>
egg|zzz|egg: i consistently learn about new wtfery in windows caused by microsoft digging in their heels with utf-16
<awygle>
fuckin' utf-16. worst encoding possible.
<gruetzkopf>
stay far away with your surrogate pairs
<egg|zzz|egg>
rqou: it's hardly Microsoft-specific, see also Java
<egg|zzz|egg>
it made sense at the time when Unicode was going to be the BMP
<egg|zzz|egg>
then it eggspanded :-p
<awygle>
it's used in a lot of places unfortunately
<awygle>
often under the unqualified name "unicode"
<awygle>
just to make it worse
<egg|zzz|egg>
*that* is annoying
<awygle>
yarp
<egg|zzz|egg>
meow
bitd has quit [Remote host closed the connection]
<awygle>
apropos of nothing - i came across some code this week that sends a 32-bit value over the network in little endian order, then calls le2hl on the other side. this irritates me, because it implies knowing about network order, and deciding not to use it.
<qu1j0t3>
egg|zzz|egg | rqou: and LaTeX doesn't because it's from 1983 and based on TeX which is from 78 :-p // It was rewritten for "TeX 82"
<egg|zzz|egg>
yes
<egg|zzz|egg>
but what if you want to use a non-ASCII ISO 646 encoding :-p
<qu1j0t3>
well, TeX includes a mapping table so you could probably hack it, but that's an '80s hobby
<egg|zzz|egg>
:-p
<egg|zzz|egg>
I mean we *are* discussing this over a protocol which has ISO 646 legacy
<qu1j0t3>
i'm sure there are like 150 TUGboat articles from European users on such hacks :)
<egg|zzz|egg>
try /nick egg\zzz\egg
<rqou>
another altera wtf: afaict GCLK0/1 are "better" than GCLK2/3
<pie_>
rqou, i like how i atched a couple dragon maid videos and now my entire youtube suggestions is that, but when i listen to some good music my suggestions are full of random crap that just never ever goes away
<rqou>
altera's arch is full of tons of _weird_ limitations
<rqou>
yeah, youtube's algorithms are _awful_
<rqou>
i've complained to my housemate about this many tiles
<rqou>
*times
<awygle>
i find that youtube's algorithms are the least awful of all the recommendation engines
<rqou>
no way
<awygle>
but that they get really stuck on certain themes for no good reason
<pie_>
NIGHTCORE 1 HOUR MIX
<rqou>
i.e. alt-right conspiracy theories?
<awygle>
and then there's the whole radicalization thing of course
<rqou>
at $WORK i don't have adblock installed, and i use youtube to listen to music
<rqou>
most of which is video game soundtracks/OSTs
<awygle>
but the front page "here's shit you'd care about" page has always been pretty decent for me
<rqou>
and i get the _worst_ ads
<pie_>
also using youtube for music is so wasteful, bandwidth pls
<awygle>
oh yeah the ads are horrible
<rqou>
e.g. blatantly sexist "gamer" crap
<awygle>
but all ads are horrible
<pie_>
its kind of weird having been around to see "e-sports" take off
<pie_>
then again its not like i ever saw a lan party before that so i wouldnt really know
<pie_>
maybe its just more visible now
<awygle>
there was a time when i for various reasons was watching crunchyroll
<awygle>
and they have the _worst_ ads
<awygle>
they do like 12 ads per 22-minute episode, and frequently it was _the same ad multiple times_
<awygle>
plus there was this one that was for a snack bar of some sort that still haunts my dreams because it just made _no fucking sense_
<pie_>
awygle, i guess that was the point
<awygle>
but apparently it didn't bother anybody else because searching for "it shows off your long giraffe neck" yields basically no results
<awygle>
pie_: well, yes, but i can't name the company it was an ad for :p
<pie_>
lmao
<pie_>
maybe you hallucinated it after the nth horrible ad
<pie_>
your mind just broke and said "i can do better"
<awygle>
do "BGA", "LFBGA", "TFBGA", "DSBGA", "UFBGA", and "VFBGA" have actual meanings? or are they just marketing things?
<awygle>
like are all TFBGAs 0.8mm pitch or something?
<rqou>
hmm, pretty sure LAB_CLK is a dummy route that doesn't do anything
<rqou>
since they use that term even for IO tiles that aren't LAB tiles
<pie_>
maybe they are all the same thing and naming is arbirtrary? :D
<awygle>
Problem: Naming things is hard. Solution: Only have one thing!
<qu1j0t3>
WUXGA
<rqou>
lol qu1j0t3
<pie_>
awygle, name the same thing multiple things
<rqou>
stahp
<pie_>
internals arent user facing anyway!
<pie_>
rqou, ASSUME NOTHING, DENY EVERYTHING
<rqou>
pie_ i'm already having enough trouble with this already, stahp
<rqou>
also, i love how i can route pin->global net->pin
<cr1901_modern>
Your Best Girl won't love you back, pie_ :(
<cr1901_modern>
Or mine, for that matter
<rqou>
ok in the columns this feeds into the "normal" mux
<pie_>
cr1901_modern, oh no :(
<cr1901_modern>
Or anyone's, really...
<pie_>
cr1901_modern, im not 100% sure how we jumped to this
<cr1901_modern>
>if i ever get into software dev all my flagship products are going to have anime girls
<pie_>
ah ok yeah i follow, i thikn
* pie_
is the one meanwhile writing scifi...I might get a bit closer :c
<pie_>
*attempting to write
<rqou>
oh LAB_CLK is just the buffer in the middle of the column
<rqou>
it's just named weird
<pie_>
<rqou> it's just named weird
<pie_>
semantic gaaaap
<awygle>
semantic gap moe
<awygle>
(that sounds like a wheel of fortune puzzle)
<pie_>
"Moe characters have expanded through Japanese media, and have contributed positively to the Japanese economy"
<pie_>
" have contributed positively to the Japanese economy"
<pie_>
they wake up, go to work, etc
* pie_
had to look moe up because he never quite figured out what it meant
<awygle>
specifically a Before & After puzzle
<awygle>
pie_: that's okay, I had to google semantic gap :-P
<pie_>
awygle, i originally googled it because it sounded like somethng someone would have come up with already xD
<pie_>
still kind of wish it could be my thing :( :p
ZipCPU|ztop has quit [Ping timeout: 265 seconds]
dingbat has joined ##openfpga
<gruetzkopf>
so i have this set of PBXes that's hard to connect to the modern world because germany essentially skipped a step in telco systems most other locations did
<gruetzkopf>
i don't want to build my own E1 card :(
<gruetzkopf>
(germany essentially jumped from a rotary dial system designed in 1955 directly to CCS ISDN, with *very* few exceptions)
<rqou>
so, anybody feel like doing some project chibi work on bits that are much further out ahead?
<rqou>
it'd be nice if somebody could poke at how the JTAG USER0/USER1 functionality works
<rqou>
since this seems to be completely undocumented
pie_ has quit [Ping timeout: 264 seconds]
pie_ has joined ##openfpga
mumptai has quit [Remote host closed the connection]
ZipCPU|ztop has joined ##openfpga
<awygle>
.... do lattice's BGA recommendations look to anyone else like they got SMD and NSMD mixed up?