[rtl] Move prim_generic dependency to ibex_top

[SamuelRiedel]

The prim_generic fileset dependency was declared in ibex_top_tracing.core rather than ibex_top.core, causing fusesoc to fail when targeting ibex_top directly. This PR fixes this by moving the prim_generic dependency to ibex_top.core.

Fixes #2396.

[rswarbrick]

Looks good to me. One thing that confuses me is that the dependency was added in commit d1e3125, which sounds like it's about tooling versions.

How did this work at the start of November?

[rswarbrick]

Ohh! Is it just that #2339 has some intermediate broken commits?

[SamuelRiedel]

I think this worked, because we never only use ibex_top, but always ibex_top_tracing. E.g., for all the CI jobs. Afaik, the only exception is the demo system, so we should actually check whether that one still works with those changes. I suspect that the generics are usually included (and Ibex includes the ones it requires) but the mapping is ambiguous. For example, linting just the top (no tracing) will give the following fusesoc warnings:

WARNING: Non-deterministic selection of virtual core lowrisc:prim:and2:0 selected lowrisc:prim_generic:and2:0
WARNING: Non-deterministic selection of virtual core lowrisc:prim:buf:0 selected lowrisc:prim_generic:buf:0
WARNING: Non-deterministic selection of virtual core lowrisc:prim:flop:0 selected lowrisc:prim_generic:flop:0
WARNING: Non-deterministic selection of virtual core lowrisc:prim:clock_gating:0 selected lowrisc:prim_xilinx:clock_gating:0
WARNING: Non-deterministic selection of virtual core lowrisc:prim:clock_mux2:0 selected lowrisc:prim_xilinx:clock_mux2:0
WARNING: Non-deterministic selection of virtual core lowrisc:prim:ram_1p:0 selected lowrisc:prim_generic:ram_1p:0

And the mix between FPGA and generic prims leads to an error.

But yes, this has to do with the updated version of fusesoc and the virtual cores feature.

[SamuelRiedel]

After quickly discussing this with @marnovandermaas offline, we agreed to not get blocked by the demo system, since that one anyway points to a fixed commit of Ibex.

[rswarbrick]

Looks good to me, and thanks for the answers for the gently confused questions :-)