intecode.txt
  interpreter coding prototype (not all is official)

I've made a decision on 2010.12.01 to make each
 instruction an op-word for performance reasons.
(One 2-byte endian conversion across platforms
 versus a 2-8 Big Endian octet read per instruction.)
It's 16-bit now, but you can still carry a second
 operation in the lower byte.


Registers
 F: frame; parameter start and end
 I: instruction pointer
 A: accumulator; result
 X: auxiliary


Opcode bits
 o: operation; see table below
 s: secondary operation, if no value

  oooooaaa abbsssss
   primary addressing modes (a)
   	n form		descr.
	0 A,b		reg
	1 X,b		reg
	2 [X]+,b	ldx
	3 -[S],b	pop
	4 xxx,A/X	ldp, param (S)
	8 *xxx,A/X	rel, instr (I)
	C #xxx,A/X	imm
   secondary addressing modes (b)
	0 A
	1 X
	2 A,[X]+	stx
	3 X,[S]+	psh

Value (xxx): Immediates (imm) and Relatives (rel)
 for the low 8 bits:
  0x81 expands to a half size; 1 (int) or 2 (num) words
  0x80 expands to a full size; 2 (int) or 4 (num) words
  otherwise, a signed byte.
 2/4 bytes integer, sign extended
 4/8 bytes number; zero extended double


Line-zero primary codes
 hex	abbr	description
 0060	lnk	link; pushes reg, inits F pending
 0181	ulk	unlink; pulls reg, exg, resets F pending
 01C0	ret	return; pulls I/F
 01E0	jmp	jump; sets I
 04	bra	branch (regardless of A)
 05	bza	branch if zero A
 06	fma	fused multiply add; [X]+*A+#num -> reg
 08	prl	push relative and reg; reg+*rel
 0A	stp	store parameter (reg)
 0C	str	store relative (reg)
 0E	sti	store immediate ([S])
 0F	phi	push immediate

Line-zero secondary
 lo	abbr	desc.
 0		none
 1	exg	exchange A/X

Operation table
 dec	abbr	desc.
unary
 2	mov	move (can use an immediate)
 3	sqr	square root (can be done 2x in one)

 4	abs	absolute value
 5	neg	negate
 6	not	bin. not (uint)
 7	bno	bool not (uint)
dyadic
 8	ban	bool and (uint)
 9	bor	bool or (uint)
 10	and	bin. and (uint)
 11	ior	bin. or (uint)

 12	eor	exclusive-or (uint)
 13	asr	arithmetic shift right (sint), (int)
 14	shr	shift right
 15	shl	shift left

 16	add	add or append
 17	sub	subtract
 18	mcl	multiply or call (push I/F, set I/F)
 19	pow	raise to power

 20	div	divide
 21	dvs	divide signed
 22	mod	modulus
 23	mds	modulus signed

 24	clt	compare less than
 25	cge	.. greater or equal
 26	cgt	.. greater
 27	cle	.. lesser or equal

 28	ceq	.. equal
 29	cne	.. not equal
 30	cid	.. identical data
 31	cni	.. not identical data


Calling method
 The multiply operation doubles as the call function.
  mcl	X,A	; if X is a func, call; otherwise multiply
  mcl	foo	; foo is relative to an object

 Normal use with parameters
 ...		; parameters
 02 lnk A	; save A and F
 90 mcl	foo	; call foo (relative to ptr)
 08 ulk A	; restore A and F


#f ff(p)
{
 p.1 + p.(2)
}

 ==ff==
 mov	p


Directives
 #cs #ci	case [in]sensitive
 #f [florin]	define function


Mnemonics
 5B00	ior #0	; uin (zero extension)
 6B00	asr #0	; sin (sign extension)
 7B00	add #0	; ref (force reference)
 8B00	sub #0	; nop (no-operation)


Character encodings
 ISO-8859-1's characters from C0-FF, except for
  D7 (times) and F7 (divide), are all Latin letters

 3100	not X	¬


String ops
 `+` appends if both are strings, adds to ptr if num/int
 `-` subtracts length; negative clips length
 (all others would currently force a conversion to number)


Storing
 ->	<identifier>.<number/name>	## plain
	<anything>.+			## incremental


Comments
 ## end-of-line comment
 #[ block
    comment ]#

EOF

e32s1m95 floating point type (128 bits)
 int32 exp
 [sgn.1]
 [man.95]