<marcan>
whitequark: what was the nextpnr (?) option to make stuff fast?
<whitequark>
fast to build? --placer heap
<marcan>
and which version did that need?
Shiz has quit [Ping timeout: 246 seconds]
<marcan>
ah, yeah, I have it already, cool
<marcan>
glasgow run selftest 7.678 user 0.363 system 90% cpu (8.935 wasted time).
<marcan>
just to compare that with some xilinx shit:
<marcan>
make 30.260 user 1.208 system 70% cpu (44.730 wasted time).
<marcan>
(both should be trivial designs)
<marcan>
and the glasgow one includes python time obviously
<whitequark>
yup
<marcan>
what's the current bottleneck? I see about 1 second in python before it starts calling yosys, then maybe 4 seconds of no output, then all the nextpnr stuff
<marcan>
I guess the 4 seconds is yosys?
<whitequark>
yep
<marcan>
no low hanging fruit to improve there?
<whitequark>
i'm currnetly making it emit less awful code for processes
<whitequark>
because right now it emits a *fuckton* of logic for trivial things
<marcan>
heh
<whitequark>
and expects later passes to clean up
anuejn has quit [Quit: No Ping reply in 180 seconds.]