Is this going to work?

[palmerj]

With TMS CPU instructions which can contain up to 6 byte instructions?

814520: 0780 09F0 0106           MOVE   A0,@10609F0h,1
814550: 0780 15A0 0106           MOVE   A0,@10615A0h,1
814BB0: 09C1 000A                MOVI   Ah,A1
814BD0: 09E7 5150 FF81           MOVI   FF815150h,A7
814C00: 0D5F 9DD0 FF96           CALLA  FF969DD0h
814C30: 0960                     RETS   
814C40: 4C01                     MOVE   A0,A1
814C50: 0B01 0020                ADDI   20h,A1
814C70: 9220                     MOVE   A1,*A0+,1
814C80: 09C1 0028                MOVI   28h,A1
814CA0: C080 66A0 FF9B           JA     FF9B66A0h
8194A0: B0B0 FF81                MOVE   B5,*B0(FF81h),0

Many thanks in advance.

[pjsoberoi]

No, 6 bytes is just too much data. You can try doing all of the two byte instructions to help you get an initial layout. That might save you some time.

[palmerj]

Thanks for the reply @pjsoberoi. Ok, wow, so that will be heaps of manual work because most instructions I have for this programme are larger than 2 bytes! I have no experience in doing this for Ghidra. So I need to map each unique instruction byte sequence, or are there some instructions (i.e. jump/move with memory addresses, or instructions with different immediates) that I don't need to map?

[pjsoberoi]

I recommend at least disassembling all instructions. If the disassembly engine comes across an unknown instruction everything stops. If the pcode for particular instructions is too much of a pain to implement I suppose leaving some as nops shouldn't be a problem.

I actually ran GPMP on TMS in this thread: NationalSecurityAgency/ghidra#3990. I upload basically what would be the output of running it across all of the two byte instructions already. Maybe it will save you some time. Or maybe you can work with the other user in the thread .

[palmerj]

Thanks @pjsoberoi. I asked the other user again and he's now released what he has. https://github.com/ZPaul2Fresh8/Ghidra_TMS34010 :-) So that's a great start!