Recently, with the advent of console verification of TASes, there seems to be an increasing concern about the accuracy of emulators. In view of this, I ran some test ROMs with the GB emulators accepted for submission here. Anyone can do this, a similar test is available in GBdev for general purpose emulators, but I think this is unknown to most here. I'm posting the results here to show that there are some accuracy problems with these emulators, hoping that they'll be fixed if it's easy enough or to help people make decisions of which emulator version to choose if they want their runs to be more likely to sync on real hardware in case GB verification becomes a reality.
The test ROMs are due to Blargg, except the last one for SGB (which I don't know what it tests) you can google them if you want to download and do this. The screenshot to the left is VBA-rr v23.5 and the one to the right is VBAM-rr v24.
Notice that passing these tests doesn't mean the emulator is accurate, but failing them doesn't say anything good about it.
cgb_sound (the answer is the same both in CGB and DMG mode, a real CGB would pass and a real DMG would fail 8,9,10,11,12 with the values 1,1,1,4,2 respectively)
dmg_sound (v2)
cpu_instructions
instruction_timing
memory_timing (v2)
oam_bug (v2)
SGB Pack
