Weird instructions

[absence]

During disassembly I've gotten some invalid opcodes that seem to be BTST or BSET with memory address operand and bit number bigger than 7. Can anyone confirm that the bit number is simply anded by 7 or something?

There's also a read from address $bfe0ff. That's the CIA address range, but those registers end in $000 or $001. Are the extra bits ignored, causing the instruction to read from $bfe001?

[hitchhikr]

During disassembly I've gotten some invalid opcodes that seem to be BTST or BSET with memory address operand and bit number bigger than 7. Can anyone confirm that the bit number is simply anded by 7 or something?

Bit instructions can access to 32 bits ([0..31]) with memory operands so if you got some thing like "btst #????14, $dff002" with resource, it's a bug from resource.

There's also a read from address $bfe0ff. That's the CIA address range, but those registers end in $000 or $001. Are the extra bits ignored, causing the instruction to read from $bfe001?

$bfe0ff = bfe001. You're probably getting this from an older program where some wise guy thought it'll be fun to use uncommon addresses range.

[Leffmann]

Not sure about the CIA address range, but for the single-bit BSET, BCLR, BCHG and BTST with a memory address as destination you are correct.

[hitchhikr]

Yes, internally it's "anded" by 7.

[Photon]

Originally Posted by hitchhikr

Bit instructions can access to 32 bits ([0..31]) with memory operands so if you got some thing like "btst #????14, $dff002" with resource, it's a bug from resource.



$bfe0ff = bfe001. You're probably getting this from an older program where some wise guy thought it'll be fun to use uncommon addresses range.

It's the completely other way around. When accessing a bit in a register, it's modulo 32 (bit number is ANDed with 31), but all memory access is byte-size (bit number is ANDed with 7).

So when a source says "btst #14,$dff002", the coder thinks he's accessing bit 14 of 16 in a word-size custom address. Which in fact he is, as he's accessing bit 6 in the byte at $dff002. Luckily, the bit number was >8, or it wouldn't work

There are actually a whopping 8 bits to store the bit number in the instruction word, so even btst #254,$dff002 should work (if your assembler doesn't say "You shall not pass!") So actually, any disassembler that shows higher numbers than 7 or 31 is showing the truth and hasn't manipulated the instruction words.


This means you can store data in the higher bits, if you feel like a cheeky devil


About the CIA registers, well without looking in HRM, in OCS there's 16 odd and 16 even registers spaced out $100 apart, but I guess any address that fits that description should make the custom chips happy.