<geertu>
Where are the days I would loved to have an SS20...
keesj has joined #litex
<keesj>
lo
<Melkhior>
geertu: yeah wanted one back in the day, learned Unix on a SS1+ running SunOS 4
<Melkhior>
Got myself a Ultra 1 Creator in the late 90s, it was nice - still have it, though with a LCD as the CRT power supply died a long time ago
alainlou has joined #litex
<Melkhior>
Only got SS20s about 6 years ago I think ; recovered then from a DEC maintenance shop that didn't know what to do with them...
<Melkhior>
they're in a bad shape (except the one given to me by my current CEO, he had kept it in great shape!) but still, they are SS20s :-)
FabM has quit [Quit: Leaving]
<alainlou>
what's the best way to debug if it seems whenever a subroutine is called (__udivsi3) it causes the system to halt? running with `--cpu-type=vexriscv --cpu-variant=minimal`
<leons>
Hm, can you reproduce it in litex_sim? I wanted to upstream my RVFI tracer implementation for too long now…
<alainlou>
leons I'll give that a shot!
<leons>
Other than that, it does seem like maybe your mul/div intrinsics generate proper multiplication or division instructions, which won’t work without HW multiplication support. Maybe you’re linking to a library which wasn’t compiled for the `rv32i` architecture (i.e. without extensions)?
<leons>
Be aware IIRC minimal really supports only `rv32i`…
<alainlou>
lol this is when printf'ing the CRC hash of the BIOS on bootup in what I think is a totally vanilla setup (just tried to follow installation instructions with the sifive gcc and what not) - maybe I should look at the assembly for `__udivsi3` :D
<gatecat>
you should also try cleaning out the build directory
<gatecat>
I've seen problems where a library build from a different CPU variant breaks the current build
<leons>
gatecat: absolutely. The incremental builds have occasionally cost me a lot of headaches. :)