<daveshah>
I was assuming this was ECP5 for some reason
<pepijndevos>
ah no, it's apicula, sorry
<pepijndevos>
Does that mean it *is* in fact my own fault?
<daveshah>
Probably, unless something else has broken with the pybind11 move
<daveshah>
I know the nextpnr-generic example was still working as that was one of our main regression tests for that
<mwk>
oh nice, a yosys/nextpnr regression that wasn't me
<pepijndevos>
lol
<pepijndevos>
So, uh... what does this error indicate? Not really sure where to start looking. Apicula doesn't support ALU or carry at all
<daveshah>
That name will just be from Yosys' internal mapping of an adder
<daveshah>
The problem is that it has failed to find a route for some reason
<daveshah>
--debug --verbose will enable more info on the failed route
<daveshah>
there are some ifdefs that can be enabled in the router for even more debugging
<pepijndevos>
Ok ill try thanks :)
* pepijndevos
waiting 10 minutes for Python to load the wires
<pepijndevos>
So at least it looks like looking at the input JSON files it's just a normal wire from a LUT output to another LUT input. Not much special.
<pepijndevos>
So it looks like it can't find any route from the LUT output at column 8 to the LUT input in column 9... extremely weird
<daveshah>
Are other LUTs routing fine?
<daveshah>
Or is this just the first LUT it finds and so the first one that fails?
<pepijndevos>
If I understand the output correctly, the first one succeeds, and then the second one fails
<pepijndevos>
but the first one is within the same tile
<daveshah>
Yep, seems like some routing has gone missing
<pepijndevos>
ugh...
emeb_mac has joined #yosys
<pepijndevos>
I will be sad if it's the PR that worked around the assert problem
<daveshah>
Can you create a nextpnr issue for that btw?
<pepijndevos>
Oh yea will do
<daveshah>
thx
<pepijndevos>
... while I wait for python to load all my wires haha
<daveshah>
incidentally, that patch seems like it might not be helping performance wise?
<daveshah>
added_wires should be a set not a list, probably
craigo has joined #yosys
<pepijndevos>
probably
<kbeckmann>
ah good catch..
<pepijndevos>
I was thingking... I don't remember it being THAT slow
<pepijndevos>
lol, it's instant now
<kbeckmann>
just as a datapoint in case there is a regression in nextpnr, i was building a design that worked a few days ago, but when i rebuilt it with a brand new yosys/nextpnr it breaks down after a few seconds of runtime. worked fine earlier.
<kbeckmann>
cool
<kbeckmann>
sorry for the bad code :p
<daveshah>
I think this might be something broken in the new pybind11 stuff then
<daveshah>
Not sure what though
<kbeckmann>
unfortunately i don't have the git hash of the working revision :/
<daveshah>
was it before 10 days ago?
<kbeckmann>
it was probably around 10 days old give or take.
<pepijndevos>
printing all the pips that match the broken net, and I'm too sleepy to see what's going on https://bpa.st/AO5Q
<tpb>
Title: View paste AO5Q (at bpa.st)
<pepijndevos>
A4 seems to be connected topip R2C9_E251_A4 E251 R2C8_E25 A4 R2C9_A4