litghost
litghost
> Ok, we should then enable that in archdefs, as it is not currently set. From what I have seen, this will likely result in a worsening of QoR and...
Good catch, let me review the CI logs.
> Actually I think we might have missed to upgrade also the device_define parameters, hence the behaviour I have seen: > https://github.com/SymbiFlow/symbiflow-arch-defs/blob/f0e7b4212544e1d55da776fb7a2ff79117e01454/xc/common/cmake/device_define.cmake#L357 > > > GitHub**[SymbiFlow/symbiflow-arch-defs](https://github.com/SymbiFlow/symbiflow-arch-defs)**FOSS architecture definitions of FPGA...
This one can be removed when check-route isn't so slow.
> so the main part to implement here is the test script part in nextpnr, right Yes
> This seems like a good idea and something I'm happy to look into. > > BTW, for generic issues like this, can you create them on the YosysHQ repo...
> As for the specific MSAN issue here, unfortunately the [first example](https://www.boost.org/doc/libs/1_75_0/libs/program_options/example/first.cpp) of boost program_options is failing with MSAN in exactly the same way for me, so this is definitely...
Looks like the initial DB push had a bad IOB result for at least one of the pads. I'll dig and see if can find a fix.
I would propose we continue to emit the bits for unused LUT's within a used CLE, but totally empty CLE's we can safely drop. There would need to be a...
> SGTM - so long as explicitly setting LUT init to 0, even if the site is otherwise empty, doesn't still set the default 1s as this may be a...