// included by decode.asm
// cycles on extra-padded "Elizabeth Mary Patricia James Robert":
// without any interleaving of instructions and LUT:
// cycles: 7305-55=7250
// instrs: 2665-17=2648
// program size: 0x96   (150, 22 of which are the end-of-string comparisons)
// with JSRs branched-over instead of branched-to-and-back:
// cycles: 7293-55=7238
// instrs: 2649-17=2632
// program size: 0x92   (146, 22 of which are the end-of-string comparisons)
// with BPL instead of CLC+BCC:
// cycles: 7151-55=7096
// instrs: 2578-17=2561
// program size: 0x90   (144, 22 of which are the end-of-string comparisons)
// with interleaved instructions and LUT:
// cycles: 7151-55=7096
// instrs: 2578-17=2561
// program size: 0xC3   (note that this will *always* be 0xC3 with this method)

// so like, same as v3, except
// instead of branching based on the high bits (through asl or ror),
// we mask out the two bits being used, and ORA them in through Absolute,X mode.

// so that means, at these memory positions (possibly offset by X, on another page),
// we need to occupy a byte:
// actually, wait, in v3, the bits we branch on are always the top two.
// so what i could do is use X, both to offset to the end of the page,
// and also for each decoding case.
// X = A & %11000000
// A = %00111100[X]

align(0x100)
decode_ilut: // internal look-up table
db $00,$00,$00

decode_advance:
inc $02 // advance input
beq die // never branch (unless page boundary)
when_to_stop()
ldy #0
lda ($02),y // load from input
tax // stash for after branch
lda #4 // pairs remaining
sta $04 // write pairs remaining
rts

decode_exit:
pla
pla
rts

nops(decode_ilut + 0x40)
db $04,$01,$04

decode_begin_next:
pla
pla

decode:
// NOTE: output/input pointers cannot cross page boundaries.
// that means the effective longest lengths of output/input are 256/192 bytes.

lda #4 // pairs remaining
sta $04 // write pairs remaining

ldy #0
lda ($02),y // load from input
tax // stash for after branch

decode_xx: // decode from offset 0, unknown code length
// two things need to be done here:

// 1. zp[0x05] |= {%0, %100, %1000, %10000}[A >> 6]
txa // TODO: unnecessary?
and #$C0
tay
lda decode_ilut+0,y // would be ORA, but this is our first data point
sta $05

// 2. branch to decode_read4 is both bits were set, decode_read2 otherwise.
tya
eor #$C0
beq decode_read4
bne decode_read2 // always branch

die:
db $F2

decode_nextbytew:
jsr decode_advance
bpl decode_xx // always branch

decode_write:
lda $05 // TODO: unnecessary?
// decode_common stuff:
tay
lda decode_lut0xxx,y
ldy #0
sta ($00),y // write to output
inc $00 // advance output
beq die // never branch (unless page boundary)
txa
asl
asl
tax
dec $04 // decrement pairs remaining
bne decode_xx // branch if we're good, otherwise...
beq decode_nextbytew // (always) branch if we need more pairs

nops(decode_ilut + 0x80)
db $08,$02,$08

decode_read2:
txa
asl
asl
tax

dec $04 // decrement pairs remaining
bne +
jsr decode_advance
+;

decode_read2_again:
// we have at least one pair left to read from X

// zp[0x05] |= {%0, %1, %10, %11}[A >> 6]
txa // TODO: unnecessary?
and #$C0
tay
lda $05
ora decode_ilut+1,y
sta $05

bpl decode_write // always branch

decode_read4:
txa
asl
asl
tax

dec $04 // decrement pairs remaining
bne +
jsr decode_advance
+;
// we have at least one pair left to read from X

// zp[0x05] |= {%0, %1, %10, %11}[A >> 6]
txa // TODO: unnecessary?
and #$C0
tay
lda $05
ora decode_ilut+2,y
sta $05

bpl decode_read2 // always branch

nops(decode_ilut + 0xC0)
db $10,$03,$0C

done:
db $F2

if 0 {
    align(0x100)
    decode_ilut: // internal look-up table
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $04,$01, $04,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $08,$02, $08,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $10,$03, $0C,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
    db $00,$00, $00,$00,  $00,$00, $00,$00,   $00,$00, $00,$00,  $00,$00, $00,$00
}

// decode_ilut[0b00000000] = 0b00000000
// decode_ilut[0b01000000] = 0b00000100
// decode_ilut[0b10000000] = 0b00001000
// decode_ilut[0b11000000] = 0b00010000

// vim:ft=snes_bass