-
-
Save asiekierka/6862636 to your computer and use it in GitHub Desktop.
| AREIA-1 | |
| ------------ | |
| 16 registers, 16 bits each, | |
| 0 = always 0 | |
| 15 = SP | |
| Memory is treated as little-endian. | |
| Opcode byte: oooooott (o - opcode, t - type of opcode) | |
| l is .b or .w | |
| Opcode formats: | |
| OP0 0000oooo | |
| OP1.1 010loooo xxxxyyyy ( op(@x,@y) -> @x ) | |
| OP1.2 011loooo xxxxyyyy iiiiiiii [iiiiiiii] ( op(@y,#i) -> @x ) | |
| OP2.1 1100oooo xxxxbbbb aaaaaaaa aaaaaaaa (jmp $baaaa + @x) | |
| OP2.2 110loooo aaaaaaaa aaaaaaaa (jmp $Saaaa (S taken from upper 4 bits of PC)) | |
| OP2.3 111loooo aaaaaaaa [aaaaaaaa] (jmp PC_AT_END_OF_OPCODE + (signed)$aa[aa] - CAN CROSS SEGMENT BOUNDARIES.) | |
| OP3.1 1000olss xxxxyyyy (ld @x, $S0000 + (unsigned int16)@y - segment picked by ss) | |
| OP3.2. 1001olss xxxxbbbb aaaaaaaa aaaaaaaa (ld @x, $baaaa + $S0000) (if ss is set to 3, force to $00) | |
| OP3.3. 1010olss xxxxyyyy aaaaaaaa (ld @x, $S0000 + (signed int8 converted to unsigned int16)$aa + @y) | |
| OP3.4. 1011olss xxxxyyyy aaaaaaaa aaaaaaaa (ld @x, $S0000 + (unsigned)$aaaa + @y) | |
| (12 more modes left for ld/st) | |
| (The ordering of the 4 type-defining bits is going to stay. Makes it saner to hexedit. --asie) | |
| Opcodes: (OP0 - 64 opcodes total (00??oooo), OP1-2 - 16 opcodes each, OP3 - full!) | |
| OP0 nop, ret, popf, pushf, cli, sei, hlt, (unused), ss0, ss1, ss2, ss3, dbg (prints out debug information in console) | |
| OP1 move, cmp, add, sub, xor, or, and, (unused), asl, asr, lsr, rol, ror, rcl, rcr, (unused) | |
| OP2 all the jumps | |
| OP3 ld, st | |
| Notes: | |
| move with @0 as destination is: | |
| for l=0, move.b SEGMENT, @y | |
| for l=1, move.b @y, SEGMENT | |
| SS0/SS1/SS2/SS3 is "set segment 0/1/2/3", which selects the segment used for memory access. | |
| By convention: | |
| SS0 is for general access. | |
| SS1 is for streaming reads. | |
| SS2 is for streaming writes. | |
| SS3 is for stack accesses and is explicitly the one used by PUSHF/POPF/RET. | |
| PC is a full 20 bits, and doesn't use the segment registers. | |
| FLAGS: | |
| bit description | |
| --------------------------------------- | |
| 0 ZERO (CMP equality) | |
| 1 CARRY (whether carrying a bit, for subtract follow Z80 impl.) | |
| 2 OVERFLOW (add/sub/cmp: if signed overflow occurs; xor/and/or: sum of all bits over GF(2) (parity)) | |
| 3 SIGNED (whether >= 0x8000 or >= 0x80 for b) | |
| 4 INTERRUPT (set to 1 to enable) | |
| 5 SEGB0 (1 uses segments 1/3, 0 uses segments 0/2) | |
| 6 SEGB1 (1 uses segments 2/3, 0 uses segments 0/1) | |
| 7-15 Reserved. |
OP3.2. 1001olss xxxxbbbb aaaaaaaa aaaaaaaa (ld @x, $baaaa + $S0000) (if ss is set to 3, ignore its existence - debate this one plz)
Yeah I would have to agree with the "SS3 is actually zero" thing, although I also feel that the segment selector is kinda unnecessary for this op. I'm not sure what we should reassign those bits to, though.
I'm not fond of there being an op for debugging - it really should be more of an MMIO feature. $FFF80-$FFFFF bank, perhaps?
While we're at it, let's reorder OP1. (It'll make it easier to interpret. Maybe.)
Old:
OP1 move, cmp, add, sub, xor, or, and, asl, asr, lsr, rol, ror, rcl, rcr
New:
OP1 move, add, cmp, sub, xor, or, and, (unused), rol, ror, rcl, rcr, (unused), asl, asr, lsr
Yes. The segment selector is good for easy relocation of code larger than 64KB and similar things.
OP1 reorder is fine. The OP for debugging is because we have a lot of free space and because it is WAY easier to just inject a few dbgs in the code than stores and stuff
Just don't make the PCs look like sh*t and we're good. I get that it's MC, but a box with a screen a computer does not make.
No. If we support indirect addressing, it will complicate things, and I believe it will bottleneck our optimisations. (issue dealt to --GM)