google
Pin Yosys back to 0.29 as workaround for Nexus flow (Yosys #4081).
Workaround for https://github.com/YosysHQ/yosys/issues/4081.
This issue is specific to the Lattice Nexus flow.
With current Yosys, the synthesized netlist from synth_nexus contains an explicit 1'b0 driving the low order CIN of a chain of CCU2 (carry chain) primitives. This causes a nextpnr-nexus error. nextpnr-nexus expects this input to be undriven which I assume implies an implicit 1'b0 input.
Hi @PiotrZierhoffer! I still seem to be seeing intermittent certificate issues with dl.antmicro.com. I could probably get it to green if I could selectively rerun the failing test(s), but unfortunately I don't have the permission to do so.
This causes a nextpnr-nexus error. nextpnr-nexus expects this input to be undriven which I assume implies an implicit 1'b0 input.
I'm curious why this error is not showing up in the CI, are we missing some tests? https://github.com/google/CFU-Playground/actions/runs/7270368055
I'm not sure where that link is from, but test-projects.yml is just one of the CI tests; it only runs Renode simulation and doesn't run synth / place / route for any platforms, so it doesn't encounter the issue here. Only the oxide-ci test in oxide.yml and other tests building bitstreams for the Nexus FPGA will encounter the issue.
Looks like https://github.com/google/CFU-Playground/actions/workflows/oxide.yml?query=branch%3Amain++ was never ran on main, do we have a existing run of this job failing? (or can you share instruction to reproduce this locally? is that similar to #808 and #803?)
Here is a run from the semi-automatic Renode update, from within the last 24hrs -- it is failing on the issue that's fixed by this PR (CCU2):
https://github.com/google/CFU-Playground/actions/runs/8368656346/job/22913100078
But I agree, the CI runs could use a review and reorg, to make it easier to find the ongoing failures.
But I agree, the CI runs could use a review and reorg, to make it easier to find the ongoing failures.
I wasn't planning on doing that in this PR though, @proppy.
Makes sense, I just wanted a way to check that it's indeed fixing the issue:
- 🔴 before https://github.com/google/CFU-Playground/actions/runs/8368656346/job/22913100078
- 🟢 after https://github.com/google/CFU-Playground/actions/runs/7493242461/job/20398519775?pr=810