What is the fastest "packed to planar" algorithm?

[psycopathicteen]

I managed to get down to 102 cycles to convert 8 4bit pixels, with a long complicated LUT. I'm not done typing out the LUT, so the LUT isn't posted. I can't seem to find an algorithm faster than this.

Code: Select all

ldx !input		;3
ldy !input+1		;3	6
lda LUT+$0000,x		;4	10
ora LUT+$0400,y		;4	14
xba			;2	16
lda LUT+$0100,x		;4	20
ora LUT+$0500,y		;4	24
ldx !input+2		;3	27
ldy !input+3		;3	30
ora LUT+$0900,x		;4	34
ora LUT+$0d00,y		;4	38
sta !output+1		;3	41
xba			;2	43
ora LUT+$0800,x		;4	47
ora LUT+$0c00,y		;4	51
sta !output		;3	54
lda LUT+$0a00,x		;4	58
ora LUT+$0e00,y		;4	62
xba			;2	64
lda LUT+$0b00,x		;4	68
ora LUT+$0f00,y		;4	72
ldx !input		;3	75
ldy !input+1		;3	78
ora LUT+$0300,x		;4	82
ora LUT+$0700,y		;4	86
sta !output+3		;3	89
xba			;2	91
ora LUT+$0200,x		;4	95
ora LUT+$0600,y		;4	99
sta !output+2		;3	102

Does anybody know of a way I can bring this down even more?

[tepples]

What might be even faster than that is shortcutting around the conversion entirely. Where are these 4-bit pixels coming from?