proteusguy has quit [Remote host closed the connection]
proteusguy has joined #yosys
notafile has joined #yosys
jryans has joined #yosys
wiizzard has joined #yosys
promach3 has joined #yosys
gkelly has joined #yosys
vidbina has joined #yosys
FFY00 has quit [Ping timeout: 240 seconds]
FFY00 has joined #yosys
kraiskil has joined #yosys
_whitelogger has joined #yosys
jryans has joined #yosys
notafile has joined #yosys
peepsalot has quit [Ping timeout: 265 seconds]
kmehall_ has joined #yosys
peepsalot has joined #yosys
kraiskil has quit [Ping timeout: 260 seconds]
ejcspii has joined #yosys
ejcspii has quit [Ping timeout: 272 seconds]
<pepijndevos>
uh... nextpnr claims to use 100% CPU but my fans are idle and routing is making close to zero progress
<pepijndevos>
It's taking 2 minutes per batch
<daveshah>
try running with --debug --verbose to see what it is stuck on
<pepijndevos>
and abort 3467 seconds of progress? XD okay let's see
<daveshah>
are your delays messed up by any chance?
az0re has quit [Remote host closed the connection]
<daveshah>
either that or unresolveably conflicting signals are usually the cause of weird router behaviour
<pepijndevos>
uh, this is Apicula, so delays could be... interesting
<daveshah>
as we've discussed before, they don't need to be accurate
<daveshah>
they just need to be consistent with your delay estimate
<tnt>
Mmm, I usually cherry pick " ice40: Pack LUT and carry in more cases" and "ice40: Use correct delays for LUT-permutation pips" from your branches, I'm wondering if that could explain why some seed yield router ending in a infinite loop.
<pepijndevos>
yea I think we changed them and that seemed to resolve the problem back then
<pepijndevos>
But this design uses a significant part of the FPGA, so without clock routing it could be that it just struggles a lot with timing
<pepijndevos>
running...
<pepijndevos>
I have no idea what I'm looking at. How much of this output do you want?
<pepijndevos>
It seems to be doing a fair amount of unbinding
<daveshah>
About 20-30 nets worth of output should be enough to get a rough idea
<daveshah>
The visited node count is quite high so I suspect it is just the delay estimate needs tuning
<daveshah>
probably increasing
<pepijndevos>
ctx.setDelayScaling(0.05, 0.5) this one right?
<daveshah>
Yeah
<pepijndevos>
pick a number haha
<daveshah>
Try doubling both to start with
<pepijndevos>
woooo that's going so much faster
<pepijndevos>
before it was showing 100ms+ pauses in the output, now it's racing by faster than I can read
<pepijndevos>
I think the number of visited nodes is much lower indeed now
<pepijndevos>
some output: https://bpa.st/76OQ I'll run it without debug now to see if it finishes
<tpb>
Title: View paste 76OQ (at bpa.st)
<daveshah>
If the estimate is now too high (I suspect the second number in particular might be) it will go much faster but also take more iterations
<daveshah>
This is a picorv32 demo?
<pepijndevos>
I see... I'll try tweaking it a bit more
<pepijndevos>
yea
<daveshah>
do you have any RAM primitives yet?
<pepijndevos>
no haha
<daveshah>
for some reason I've found bitblasted RAMs tend to be quite slow to route
<pepijndevos>
so it's taking up most of the 9k luts with registers
ktemkin has quit [Ping timeout: 260 seconds]
ktemkin has joined #yosys
<pepijndevos>
it's actually only marginally worse in terms of itterations than before. I recall at 31000 it was at 8k unrouted before, while now it's at 9k
<pepijndevos>
yay, I'm counting primes ;)
X-Scale` has joined #yosys
X-Scale has quit [Ping timeout: 265 seconds]
X-Scale` is now known as X-Scale
vidbina has quit [Ping timeout: 246 seconds]
alexhw has quit [Ping timeout: 244 seconds]
alexhw has joined #yosys
ejcspii has joined #yosys
ejcspii has quit [Ping timeout: 272 seconds]
vidbina has joined #yosys
kristianpaul has quit [Read error: Connection reset by peer]