<azonenberg>
awesome, but a bit more "short term" than me and awygle's interests
<azonenberg>
we're sitting back waiting for prjxray to progress and starting research into parallel/scalable PnR
<azonenberg>
idk if you've seen some of the discussion here but we want to get an analytic PAR for 7 series that scales to "pretty full, large device in a couple of minutes" on hundreds to thousands of CPUs
<azonenberg>
Even if QoR isnt as nice as vivado or even VPR, we want fast design closure for prototyping
<mithro>
azonenberg: I think the you would be better off trying to plug your parallel/scalable PnR into vpr, but as long as you support the same input / output formats then collaboration is probably possible even if you don't
<azonenberg>
We're planning on using yosys json as input and output
<azonenberg>
and platforms like MPI that are commonly used for high performance / parallel stuff dont seem like they would fit well into a data model like VPR
<azonenberg>
we're talking racks of machines and/or thousands of EC2 cores
<azonenberg>
not just one big multithreaded box
<azonenberg>
i dont know a ton about vpr's internal data model but the impression i got is that it wasnt well suited to that
<awygle>
mithro: what's the high-level status of vpr-for-ice40?
<awygle>
i'm not opposed to integrating if it makes sense. i was planning to play with T-VPACK next week as an initial familiarization project (and because i'm uninterested in spending time on the packer)
<mithro>
azonenberg: I'm all for multiple projects doing the same thing, gcc progress has been a lot better since llvm became viable
<azonenberg>
mithro: i'm not trying to compete like gcc vs llvm, as both essentially do the same thing
<azonenberg>
i see this as more like make vs distcc
<mithro>
awygle: I need to figure out how to convert vpr output to bits and finish up using a real routing graph rather then a virtual generated one
<azonenberg>
Our focus is going to be exclusively on scalability and parallelism
<mithro>
awygle: Any chance you'll be at 34C3?
<awygle>
mithro: nope :(
<mithro>
I would probably have the ice40 stuff working if I hadn't gotten distracted helping CI and stuff setup for Clifford get prjxray stuff
<mithro>
I think I have a pretty good description of all the primitives in a format that vpr accepts
<mithro>
Someone has to write things like READMEs :-P
<awygle>
that sounds great, way too many projects don't have any of those things
<azonenberg>
Nice
<azonenberg>
yeah, i am gonna try and use some of my time over the xmas holiday to write docs for a bunch of my IP cores that i wrote for antikernel
<azonenberg>
as well as TRAGICLASER
<awygle>
also sounds like you've already done almost all of the prelim work i was planning for Tuesday/Wednesday next week, so i'll probably only need to hack together some file conversion stuff
<awygle>
if that
<awygle>
so... thanks! ^_^
<mithro>
awygle: which timezone are you in?
<awygle>
P(S|D)T
<awygle>
i'm in the PNW
<mithro>
PNW?
<mithro>
Well, I'm on whatever Leipzig is on for the next or so - so feel free to poke me on IRC any time
<azonenberg>
mithro: pacific northwest
<awygle>
on the scale of $ENTIRE_WORLD i am basically colocated with azonenberg in seattle
<awygle>
UTC-7 or -8 depending on daylight savings
<awygle>
(sorry for earlier us-centrism lol)
<azonenberg>
awygle: PNW is not US-centric IMO
<azonenberg>
no more than "asia-pacific" or something
<azonenberg>
it encompasses washington, oregon, british columbia, and maybe a bit more
<awygle>
azonenberg: somehow i feel like "pacific northwest" wouldn't be but PNW is
<azonenberg>
Anyway heading out, be back on from the airport later
<awygle>
have a good flight
<rqou>
wtf why do i keep getting ninja'd on projects
<mithro>
awygle: I've tried to make it so that my vpr descriptions also include some useful documentation -- Clifford icestorm documentation is pretty comprehensive but a bit cryptic at times :-P
<awygle>
rqou: grad school
<rqou>
f*ck grad school :P
<awygle>
mithro: _much_ appreciated
<rqou>
btw i got all my grades and they're all "meh"
<rqou>
#berkeley
<awygle>
just get dat paper, nobody cares about your gpa lol
<rqou>
whitequark: btw, i'm subjectively getting great results so far from my "ghetto 3mg melatonin diluted to 0.3mg"
<balrog>
mithro: what are all the null tiles? they don't do anything?
<mithro>
balrog: Not sure yet
<balrog>
ahh, okau
<balrog>
okay*
carl0s has joined ##openfpga
digshadow has quit [Ping timeout: 248 seconds]
m_t has joined ##openfpga
digshadow has joined ##openfpga
<awygle>
Horizon EDA is disappointing currently
<jn>
in which way does it disappoint? (not arguing, just trying to understand)
<awygle>
UX primarily
<awygle>
It feels like a mobile or tablet app
<awygle>
Graphics performance (on the machine it ran on) seems much better than kicad though
<awygle>
It's possible the ux issues are down to the newness of it, but it feels like a deliberate design strategy that I disagree with
<awygle>
I'll keep an eye on it and see how it evolves
soylentyellow has quit [Ping timeout: 256 seconds]
<digshadow>
awygle: who is backing horizon? what is the motivation vs say kicad?
<digshadow>
"carrotIndustries"
<awygle>
digshadow: no idea. "carrot industries" whoever that is
<awygle>
The motivation vs kicad is probably "kicad is pretty bad" lol
<awygle>
jn: was that what you were looking for? If you're involved in the project in some way I can be more specific
<jn>
i posted the link here, but i'm not involved (i just found the project in the list of fosdem talks)
<jn>
i thought that their claim of "sane libray management" sounded noteworthy
<qu1j0t3>
awygle: Well they do claim graphics acceleration as an exciting reason developers would want to work on it
<qu1j0t3>
i don't think UX was mentioned ODDLY
<balrog>
Horizon seems to be a one person project
<qu1j0t3>
I did rather snerk at 'Written in modern C++, legacy-free codebase!
<qu1j0t3>
but i wouldn't kinkshame anyone
<balrog>
qu1j0t3: I think that means no legacy cruft from when reimplementing STL features was commonplace
<balrog>
(pre-C++11)
<awygle>
It just means they use auto and shared_ptr
<awygle>
Re: library management, it felt like they had a good *library management* system but that *part creation* was very bad
<awygle>
I would love an EDA tool that has a toolbar button in the schematic editor for "create new part"
<qu1j0t3>
haha
<awygle>
That's where you are in the design flow when you want a new part. I don't know about you but I don't correctly guess every part I'm going to need ahead of time.
soylentyellow has joined ##openfpga
pie_ has quit [Ping timeout: 248 seconds]
TDelectro has quit [Quit: East bound and down, loaded up and truckin']
TDelectron has joined ##openfpga
TDelectron is now known as TDelectro_
TDelectro_ is now known as TDelectron
TDelectron has quit [Quit: I used to think I was indecisive, but now I'm not too sure.]
TDelectron has joined ##openfpga
nrossi has quit [Quit: Connection closed for inactivity]
carl0s has quit [Quit: Leaving]
soylentyellow has quit [Ping timeout: 265 seconds]
gnufan has joined ##openfpga
pie_ has joined ##openfpga
soylentyellow has joined ##openfpga
m_t has quit [Quit: Leaving]
soylentyellow has quit [Ping timeout: 265 seconds]
<cr1901_modern>
My other concern at the moment is how do I know my work won't conflict with the up-to-date DB, if each person is working on a separate ROI/ROI_FRAMES?
Bike has joined ##openfpga
<digshadow>
cr1901_modern: to be clear, you'd contribute a fuzzer, not to the actual DB file
<digshadow>
mithro is generally in charge of actually updating the official DB
<digshadow>
so as long as your coordinated with us that you'd be working on say IOB or DSP, it wouldn't conflict with what we are doing
* mithro
is a manual build bot :-P
<cr1901_modern>
digshadow: Okay then, I'll focus on IOB. Let's see how much of a dent I can make by Sunday
<digshadow>
cr1901_modern: btw please use Vivado 2017.2 if you can
<digshadow>
I think at least for short term we are going to standardize on that
<cr1901_modern>
Can I install multiple Vivados side-by-side?
<digshadow>
yes, if your hard drive doesn't cry itself to death
<cr1901_modern>
^mithro... made sure that I'm able to do that
<digshadow>
cr1901_modern: to be clear, are you dev on windows or linux
<digshadow>
everything we are doing today assumes linux I think
<cr1901_modern>
digshadow: That's a complicated question, but for the purposes of this project I'm running Vivado "headless" on a Linux box
<digshadow>
heh okay
<digshadow>
you mean no x server?
<cr1901_modern>
If I use the GUI it'll be through xpra
<digshadow>
that may make some things harder
<digshadow>
but its not a strict requirement for the project
<cr1901_modern>
Meaning I won't be physically present at the machine, but I do have access to X, even on my Windows machine