This wires up enough of the Apicula timing data to give an Fmax for most designs. There are gaps in the implementation, but those gaps also existed in nextpnr-gowin, so it's not much of a loss.
The Fmax numbers do appear to be optimistic though - compared to the vendor tooling the numbers are about 20% higher and I don't entirely trust them. However, if the optimistic numbers say that a design doesn't meet timing, it definitely doesn't.
It would be nice to get BRAM timings added at some point, depending on the clock-to-out time it's not uncommon for it to actually be the critical path.
This wires up enough of the Apicula timing data to give an Fmax for most designs. There are gaps in the implementation, but those gaps also existed in nextpnr-gowin, so it's not much of a loss.
The Fmax numbers do appear to be optimistic though - compared to the vendor tooling the numbers are about 20% higher and I don't entirely trust them. However, if the optimistic numbers say that a design doesn't meet timing, it definitely doesn't.