The middle one is correct. The first form is some...
# verification-be
t
The middle one is correct. The first form is something used internal to the comparison script, and the command does not understand wildcards. I'm not quite following what's going on in the output, though. There is supposed to be a "prematch" stage which does exactly what you are describing, except that it does it automatically (or is supposed to). It should be printing output like "flattening dff_buf_0 in circuit 2 makes a better match". But it would have had to do this earlier, since the transistors are two levels of hieararchy down. One thing I just discovered earlier this evening is that the netlist I have for the SRAM block is using the old s8 names like "nshort", "pshort", etc. That doesn't seem to be the case in what you have, but it means that I'm just getting around to running the same comparison of the SRAM that you are. I should know in a bit what the result is. Otherwise, all I can say is that if netgen is completely confused by something in the input, then manually flattening cells that you know need flattening is a valid way to approach it. But I think there should be a 1:1 matching between subcells in both netlists all the way down the SRAM hierarchy. So somewhere further back in your output, it must have not tried to compare "pk_pk_pdriver_2" to "pdriver_2" but must have just flattened "pk_pk_pdriver_2", suggesting that you are missing an "equate classes" line for that cell (and all the others). What my python script that I wrote to compare the caravel chip from the top level does is to read the netlist, parse out all the ".subckt" statements, find all the prefixes known to be used, and then set up a line in the Tcl script for netgen that automatically tries to equate that cell to the cell without the prefix. I wrap that in a "catch {}" statement because I'm just doing a brute force "equate <cell with prefix> <cell without prefix>" without bothering to check whether or not there is such a cell name in the other netlist. If netgen throws an error when trying to run the "equate" command, then the "catch" statement will cause netgen to ignore it and move on.