Cycle timings are sometimes wrong

[maxim-zhao]

Some of the timings seem to be wrong, for example ld a,ixh. Possibly due to Marat's core being based on MSX timings. MAME timings are probably correct.

[ocornut]

I will eventually look into it.
If you can list exact instructions or provide a PR obviously it'd be easier :)
Do you know of games that are visibly affected?

[maxim-zhao]

Not sure, it was affecting a sample player where I needed to count cycles and I came across differences when using CLOCK to verify my counts due to mistakes in the CB prefix. There are equivalent tables for cycle counts in MAME, which I have copied over. PR will come soon.

[ocornut]

Looking at it in details now, can't just nilly willy merge that. If you have details/verifications somewhere I would take care.

Now that MAME is free-software we can also use its Z80 emulator, the only problem is that I thnik it needs a few extra dozen bytes of save-state so I'd have the break Meka state-state format. Perhaps it's just a better solution to do that.

[maxim-zhao]

I guess I need to write some code to give a pass/fail per instruction, which I think will be quite tricky... then compare to hardware.