acme/docs/Illegals.txt
marcobaye 64a29e4504 amended docs on illegals
git-svn-id: https://svn.code.sf.net/p/acme-crossass/code-0/trunk@204 4df02467-bbd4-4a76-a152-e7ce94205b78
2020-05-31 13:04:21 +00:00

140 lines
6.7 KiB
Plaintext

ACME
...the ACME Crossassembler for Multiple Environments
--- Undocumented ("illegal") opcodes ---
In release 0.87, support for some of the undocumented opcodes of the
6502 processor was added.
In release 0.89, some more were added.
In release 0.94.8, another one was added (lxa).
In release 0.95.3, C64DTV2 support was added, which includes these
opcodes as well.
In release 0.95.4, the remaining seven were added.
In release 0.95.6, "ANC" was removed from C64DTV2 mode.
Here are the new mnemonics, possible addressing modes and generated
opcodes (mnemonics in parentheses are used by other sources):
| addressing mode |
mnemonic | 8 8,x 8,y 16 16,x 16,y (8,x) (8),y | performs:
----------------+--------------------------------------+-----------
slo (aso) | 07 17 0f 1f 1b 03 13 | asl + ora
rla (rln) | 27 37 2f 3f 3b 23 33 | rol + and
sre (lse) | 47 57 4f 5f 5b 43 53 | lsr + eor
rra (rrd) | 67 77 6f 7f 7b 63 73 | ror + adc
sax (axs, aax) | 87 97 8f 83 | stx + sta
lax | a7 b7 af bf a3 b3 | ldx + lda
dcp (dcm) | c7 d7 cf df db c3 d3 | dec + cmp
isc (isb, ins) | e7 f7 ef ff fb e3 f3 | inc + sbc
las (lar, lae) | bb | A,X,S = {addr} & S
These five are said to be unstable:
tas (shs, xas) | 9b | S = A & X {addr} = A&X& {H+1}
sha (axa, ahx) | 9f 93 | {addr} = A & X & {H+1}
shx (xas, sxa) | 9e | {addr} = X & {H+1}
shy (say, sya) | 9c | {addr} = Y & {H+1}
| addressing mode |
mnemonic | implied #8 8 8,x 16 16,x | performs:
----------------+---------------------------------+-----------------------
anc (ana, anb) | 0b* | A = A & arg, then C=N
asr (alr) | 4b | A = A & arg, then lsr
arr | 6b | A = A & arg, then ror
sbx (axs, sax) | cb | X = (A & X) - arg
dop (nop, skb) | 80** 80 04 14 | skips next byte
top (nop, skw) | 0c** 0c 1c | skips next two bytes
jam (kil, hlt) | 02 | crash (wait for reset)
These two are somewhat unstable, because they involve an arbitrary value:
ane (xaa, axm) | 8b*** | A = (A | ??) & X & arg
lxa (lax, atx) | ab*** | A,X = (A | ??) & arg
Example:
!cpu 6510 ; activate additional mnemonics...
lax (some_zp_label,x) ; ...and use them. No, this
dcp (other_zp_label),y ; example does not make sense.
*) Up until ACME version 0.95.1, anc#8 generated opcode 0x2b. Since
ACME version 0.95.2, anc#8 generates opcode 0x0b. Both opcodes work
the same way on a real 6510 CPU, but they do not work on the C64DTV2.
**) Note that DOP ("double nop") and TOP ("triple nop") can be used
with implied addressing, but the generated opcodes are those for
immediate and 16-bit absolute addressing, respectively. Using dop/top
with x-indexed addressing might have its uses when timing is critical
(crossing a page border adds a penalty cycle).
***) ANE and LXA first perform an ORA with an arbitrary(!) value and
then perform an AND with the given argument. So they are unstable and
therefore useless - unless the given argument is zero:
ANE #0 reliably clears A - which is still useless; just use LDA #0.
LXA #0 reliably clears both A and X.
ACME will output a warning if these opcodes get assembled with a
nonzero argument.
There is no guarantee that these opcodes actually work on a given 6502
(or 6510, or 8500, or 8502) CPU. But as far as I know, nobody ever
found an unmodified C64/C128 where these illegals didn't work. That's
why I used "6510" as the CPU keyword instead of "6502illegal" or
something like that.
These illegals will definitely *not* work on 65c02 and 65816 CPUs. But
I really should not have to tell you that ;)
Because there are no official mnemonics for these opcodes, different
people use different names for them. I hope my choices are not too
exotic for your taste.
Just for the sake of completeness: Here are all the remaining opcodes
(the ones ACME won't generate even with "6510" cpu):
Opcode| Description C64DTV2
------+--------------------------------------------------------------
12 | same as 02 and others jam CRASH bra rel
1a | same as (*legal*) ea nop
22 | same as 02 and others jam CRASH
2b | same as 0b anc #8 dop
32 | same as 02 and others jam CRASH sac #8
34 | same as 14 and others dop 8,x
3a | same as (*legal*) ea nop
3c | same as 1c and others top 16,x
42 | same as 02 and others jam CRASH sir #8
44 | same as 04 dop 8
52 | same as 02 and others jam CRASH
54 | same as 14 and others dop 8,x
5a | same as (*legal*) ea nop
5c | same as 1c and others top 16,x
62 | same as 02 and others jam CRASH
64 | same as 04 dop 8
72 | same as 02 and others jam CRASH
74 | same as 14 and others dop 8,x
7a | same as (*legal*) ea nop
7c | same as 1c and others top 16,x
82 | same as c2/e2 dop #8, but said to CRASH sometimes
89 | same as 80 dop #8
92 | same as 02 and others jam CRASH
b2 | same as 02 and others jam CRASH
c2 | same as 82/e2 dop #8, but said to CRASH sometimes
d2 | same as 02 and others jam CRASH
d4 | same as 14 and others dop 8,x
da | same as (*legal*) ea nop
dc | same as 1c and others top 16,x
e2 | same as 82/c2 dop #8, but said to CRASH sometimes
eb | same as (*legal*) e9 sbc #8
f2 | same as 02 and others jam CRASH
f4 | same as 14 and others dop 8,x
fa | same as (*legal*) ea nop
fc | same as 1c and others top 16,x
For more information about what these opcodes do, see these documents:
John West, Marko Mäkelä. '64doc' file, 1994/06/03.
Extra Instructions Of The 65XX Series CPU, Adam Vardy, 27 Sept. 1996
6502 Undocumented Opcodes, by Freddy Offenga, 5/17/1997
AAY64 (All About Your 64)
and the most comprehensive work is:
"No More Secrets - NMOS 6510 Unintended Opcodes"
Download it from https://csdb.dk/release/?id=185341
or ask google for the latest version.