<mithro>
Finding contractors to work on nextpnr related stuff is also very hard -- daveshah only has so much time to go around
<whitequark>
can other people contribute to nextpnr? why not?
<mwk>
... hmm, I thought there was some company that has people with experience working on nextpnr
<whitequark>
also that
<whitequark>
if i didn't have my hands full i'd totally consider working on nextpnr
<whitequark>
i mean i'll work on nextpnr, just not on an architecture
<mithro>
Yes there is, but business to business relationships are complicated
<whitequark>
then fund these people as individuals
<whitequark>
anyway, as long as the result is good it doesn't matter very much
<whitequark>
my concern here is along the lines of "if I can't easily cross-compile the toolchain I wouldn't bother using it in any place where I currently use nextpnr"
<whitequark>
the point of having a FOSS toolchain, for me, is it being head and shoulders above the proprietary ones, not being basically as much pain to use but OSS
<mithro>
Impact: Able to achieve 12-18 minutes using Vivado -- Contrast 42-160 minutes no PRflow -- Plausible to achieve 2-5 minutes with open source Symbiflow
<mithro>
I actually /think/ that was the first academic paper to reference symbiflow and kind of came out of nowhere
<mithro>
> First considering the VPR 8 and VPR 7+ timing-driven routers (Timing + WL), we observe the VPR 8 router is substantially more run-time efficient (6.2× faster) while also improving QoR (18% wirelength and 16% critical path delay reductions).
<mithro>
> From a run-time perspective, VPR 8 HE runs 1.2× faster than Quartus and has a 4% smaller memory footprint. VPR 8’s packer runs 18% slower than Quartus’, but is much more general purpose. For these large benchmarks, placement dominates run-time (78% of total time), and VPR 8’s placer run-time is comparable to Quartus. The VPR 8 router is more efficient and runs 2.9× faster than Quartus even in this high effort
<mithro>
mode. Compared to VPR 8, Quartus spends approximately 3.3× more run-time on STA.
<mithro>
The largest benchmark in VPR 1.8 million LUTs too
<TD-Linux>
in particular there's some shenanigans with the central gnd balls
indy has joined ##openfpga
<tnt>
That PDF only has 343 pages ...
hitomi2500 has joined ##openfpga
keesj_ has quit [Quit: leaving]
keesj has joined ##openfpga
<tnt>
mithro: well, it's possible something is wrong but I would have no clue where to even begin looking ... the bitstream it generate work but the 'route' phase takes forever.
<tnt>
My laptop is by no mean recent ( i7-4600U dual core 2GHz ) but building a full soc with usb filling up 80% of a up5k is in the 2m total build time on ice40. A trivial 400 LC design driving a mipi screen takes 3s.
<tnt>
whitequark: it'd still be good to have bitstream caching :p When you constantly swtich between two applets, it's a pain even if it only take 10-20s ... (and yeah I know I can manually 'cache' it)
<daveshah>
I think there is something in VPR called lookahead that needs to be set up correctly or routing is terrible
<daveshah>
idk how it is set up or if that js actually the problem though
<whitequark>
tnt: yeah that will happen sometime
<tnt>
daveshah: at least in the log there is a "# Computing router lookahead map took 32.86 seconds (max_rss 622.5 MiB, delta_rss +0.0 MiB)"
<daveshah>
I think for the Xilinx arches this is computed during build
OmniMancer1 has joined ##openfpga
OmniMancer has quit [Ping timeout: 260 seconds]
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
OmniMancer has joined ##openfpga
OmniMancer1 has quit [Ping timeout: 260 seconds]
Asu has joined ##openfpga
jfcaron has joined ##openfpga
mumptai has joined ##openfpga
Bike has joined ##openfpga
<tnt>
Anyone know if Olof Kindgren hangs around here ?
<daveshah>
Probably you would have more luck on the librecores gitter
<tnt>
There is a librecores gitter ? ... -ETOOMANYPLATFORMS :/
<daveshah>
Yes
<tnt>
But tx :)
indy has quit [Read error: Connection reset by peer]
X-Scale` has joined ##openfpga
X-Scale has quit [Ping timeout: 264 seconds]
X-Scale` is now known as X-Scale
indy has joined ##openfpga
genii has joined ##openfpga
OmniMancer has quit [Read error: Connection reset by peer]
OmniMancer has joined ##openfpga
OmniMancer has quit [*.net *.split]
mumptai has quit [*.net *.split]
pepijndevos has quit [*.net *.split]
promach3 has quit [*.net *.split]
jhol has quit [*.net *.split]
filt3r has quit [*.net *.split]
OmniMancer has joined ##openfpga
pepijndevos has joined ##openfpga
mumptai has joined ##openfpga
filt3r has joined ##openfpga
jhol has joined ##openfpga
futarisIRCcloud has joined ##openfpga
swedishhat[m]1 has quit [Ping timeout: 246 seconds]
omnitechnomancer has quit [Ping timeout: 246 seconds]
eddyb has quit [Ping timeout: 260 seconds]
xobs has quit [Ping timeout: 260 seconds]
emily has quit [Ping timeout: 260 seconds]
henriknj has quit [Ping timeout: 260 seconds]
john_k[m]1 has quit [Ping timeout: 240 seconds]
hitomi2500 has quit [Quit: Nettalk6 - www.ntalk.de]
OmniMancer has quit [Quit: Leaving.]
mossmann_ is now known as mossmann
promach3 has joined ##openfpga
X-Scale` has joined ##openfpga
X-Scale has quit [Ping timeout: 260 seconds]
xobs has joined ##openfpga
eddyb has joined ##openfpga
swedishhat[m] has joined ##openfpga
Guest95944 has joined ##openfpga
henriknj has joined ##openfpga
john_k[m] has joined ##openfpga
indefini[m] has joined ##openfpga
omnitechnomancer has joined ##openfpga
emeb has joined ##openfpga
X-Scale` is now known as X-Scale
nurelin has joined ##openfpga
jemk has quit [Remote host closed the connection]
jemk has joined ##openfpga
<mithro>
tnt: upload an example and people can look into it