Specifically, the SDP configurations for RBRAM (ignoring 2048x09 because it makes the memlib format unhappy).
Drop the unused defines from the synth pass.
Remove comments from the lutram files referencing xilinx.
UG303-1.0E_Arora Ⅴ Configurable Function Unit (CFU) User Guide.pdf
specifies that the only flip-flop types supported in GW5 are DFFSE,
DFFRE, DFFPE, and DFFCE.
However, the bit streams generated by the vendor IDE also contain DFF
flip-flops, which are probably the result of optimisation, so we leave
them in the list of permitted items, but add a flag that will allow the
generation of completely correct output files, acceptable for further P&
R using vendor tools (they will not allow the use of flip-flops other
than the four specified in the netlist).
In the GW5 SemiDual Port BSRAM series, the primitive does not have
RESETA and RESETB ports—they are replaced by the RESET port, so we
separate the files for BSRAM generation, especially since in the future
we may have to take into account other, as yet unexplored, differences
in BSRAM.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
We were performing the helper passes `abc9_ops -replace_zbufs` and
`abc9_ops -restore_zbufs` for every module, but those passes act on the
full design (and can't be applied entirely selectively due to entering
and leaving bufnorm).
This lead to an explosive creation of a lot of redundant bufnorm helper
cells that would have been cleaned up by `clean` but that never ran.
Instead we now run each helper pass once, one before and one after
iterating over the selected modules. This limits the number of bufnorm
helper cells.
For consistency.
Also trying a new thing: only rebuilding objects that use the pybind11 library. The idea is these are the only objects that include the Python/pybind headers and thus the only ones that depend on the Python ABI in any capacity, so other objects can be reused across wheel builds. This has the potential to cut down build times.
