i went down a very deep rabbit hole, and this is what followed me home.

working on a 6502 emulator, i wanted to add support for the undocumented opcodes.

one of these is ARR #imm.  (x6B)

i spent hours trying to figure out how this opcode works.

i even found a wonderful set of test vectors at
https://github.com/SingleStepTests/65x02

long story short, it is unstable, and i was unable to find any solid implementations that matched the test vectors.

also, the test vectors are incomplete.
they contain only 10,000 tests each, when the entire search space is 2^18.
(2^8 values for A * 2^8 values for #imm * 2 values for C * 2 values for D).

not enough for me to fully test any implementation i might come up with.

i also found perfect6502 at
https://github.com/mist64/perfect6502
which allowed me to generate a complete set of tests.

curiosly, it did not match well with ANY of the SingleStepTests.

as i was giving up (it's unstable, there is no single solution),
i realized that this could be leveraged to generate a unique fingerprint for different silicon versions.

i analyzed the test vectors for differences, and found 4 tests that together would allow me to differentiate between:
perfect6502
SingleStepTests/6502
SingleStepTests/nes6502
SingleStepTests/rockwell65c02
SingleStepTests/synertek65c02
SingleStepTests/wdc65c02

now, this doesn't mean i know what the fingerprints are;
i would need real silicon or a complete set of 2^18 for each to calculate that.

but i was able to calculate it for perfect6502, obvs.

maybe this will be useful to someone in some way.

maybe people will come in after me and add some fingerprints to the registry.

who knows?