prjtrellis documentation shows that EBR clock inputs have optional inverters.
The bram techmap outputs those parameters, and nextpnr consumes them. But for
whatever reason, Diamond doesn't include those parameters in its blackbox
models. This makes synth_lattice fail when targeting ECP5 with a design that
maps block RAMs if you include any pass that needs cells_bb_ecp5.v's definitions.
This change fixes up the ECP5 bram blackbox models at generation time, by
adding the missing parameters back in.
Signed-off-by: David Anderson <dave@natulte.net>
Mostly memory_libmap arg checks; puts the checks into an else block on the `if (help_mode)` check to avoid cases like `synth_ice40` listing `-no-auto-huge [-no-auto-huge]`.
Also fix `map_iopad` section being empty in `synth_fabulous`.
After `memory_map` maps memories to flip-flops we need to let `opt`
remove undef muxes, otherwise we block enable/reset signal inference by
`opt_dff` which is in detriment to QoR.
`dfflegalize` will emit NOT gates to drive control signals on flip-flops
when mapping to supported flip-flop polarities. Typically in a design
this will produce a number of NOT gates driven by the same signal. For
one reason or another ABC doesn't fully cancel this redundancy during
LUT mapping. Insert an explicit `opt_merge` pass to improve synthesis
QoR.
