;
; Hi James,
;
; I suddenly saw how to use eor.l to shave a few more cycles.  Now it's
; 67.0 cycles/pixel with no nasty tricks.  (Use a7 as another bitplane
; pointer to get 66.5 cycles/pixel.)  It's still untested.
;
; I think I've about reached my limit, unless someone gives me some more
; clues.
;
; Regards, Peter.

		xdef	_chunky2planar

;-----------------------------------------------------------------------------
; chunky2planar:	(new Motorola syntax)
;  a0 -> chunky pixels
;  a1 -> plane0 (assume other 7 planes are allocated contiguously)
; d0-d1/a0-a1 are trashed


width		equ	320		; must be a multiple of 8
height		equ	200
plsiz		equ	(width/8)*height

_chunky2planar:

		movem.l	d2-d7/a2-a6,-(sp)

; set up register constants

		move.l	#$0f0f0f0f,d5	; d5 = constant $0f0f0f0f
		move.l	#$55555555,d6	; d6 = constant $55555555
		move.l	#$3333cccc,d7	; d7 = constant $3333cccc
		lea	(plsiz,a1),a2	; a2 -> plane1 (end of plane0)

; load up (otherwise) unused address registers with bitplane ptrs

		movea.l	a2,a3		; a3 -> plane1
		lea	(2*plsiz,a1),a4	; a4 -> plane2
		lea	(2*plsiz,a4),a5	; a5 -> plane4
		lea	(2*plsiz,a5),a6	; a6 -> plane6

; main loop (starts here) processes 8 chunky pixels at a time

mainloop:

; d0 = a7a6a5a4a3a2a1a0 b7b6b5b4b3b2b1b0 c7c6c5c4c3c2c1c0 d7d6d5d4d3d2d1d0

		move.l	(a0)+,d0	; 12 get next 4 chunky pixels in d0

; d1 = e7e6e5e4e3e2e1e0 f7f6f5f4f3f2f1f0 g7g6g5g4g3g2g1g0 h7h6h5h4h3h2h1h0

		move.l	(a0)+,d1	; 12 get next 4 chunky pixels in d1

; d2 = d0 & 0f0f0f0f
; d2 = ........a3a2a1a0 ........b3b2b1b0 ........c3c2c1c0 ........d3d2d1d0

		move.l	d0,d2		;  4
		and.l	d5,d2		;  8 d5=$0f0f0f0f

; d0 ^= d2
; d0 = a7a6a5a4........ b7b6b5b4........ c7c6c5c4........ d7d6d5d4........

		eor.l	d2,d0		;  8

; d3 = d1 & 0f0f0f0f
; d3 = ........e3e2e1e0 ........f3f2f1f0 ........g3g2g1g0 ........h3h2h1h0

		move.l	d1,d3		;  4
		and.l	d5,d3		;  8 d5=$0f0f0f0f

; d1 ^= d3
; d1 = e7e6e5e4........ f7f6f5f4........ g7g6g5g4........ h7h6h5h4........

		eor.l	d3,d1		;  8

; d2 = (d2 << 4) | d3
; d2 = a3a2a1a0e3e2e1e0 b3b2b1b0f3f2f1f0 c3c2c1c0g3g2g1g0 d3d2d1d0h3h2h1h0

		lsl.l	#4,d2		; 16
		or.l	d3,d2		;  8

; d0 = d0 | (d1 >> 4)
; d0 = a7a6a5a4e7e6e5e4 b7b6b5b4f7f6f5f4 c7c6c5c4g7g6g5g4 d7d6d5d4h7h6h5h4

		lsr.l	#4,d1		; 16
		or.l	d1,d0		;  8

; d3 = ((d2 & 33330000) << 2) | (swap(d2) & 3333cccc) | ((d2 & 0000cccc) >> 2)
; d3 = a1a0c1c0e1e0g1g0 b1b0d1d0f1f0h1h0 a3a2c3c2e3e2g3g2 b3b2d3d2f3f2h3h2

		move.l	d2,d3		;  4
		and.l	d7,d3		;  8 d7=$3333cccc
		move.w	d3,d1		;  4
		clr.w	d3		;  4
		lsl.l	#2,d3		; 12
		lsr.w	#2,d1		; 10
		or.w	d1,d3		;  4
		swap	d2		;  4
		and.l	d7,d2		;  8 d7=$3333cccc
		or.l	d2,d3		;  8

; d1 = ((d0 & 33330000) << 2) | (swap(d0) & 3333cccc) | ((d0 & 0000cccc) >> 2)
; d1 = a5a4c5c4e5e4g5g4 b5b4d5d4f5f4h5h4 a7a6c7c6e7e6g7g6 b7b6d7d6f7f6h7h6

		move.l	d0,d1		;  4
		and.l	d7,d1		;  8 d7=$3333cccc
		move.w	d1,d2		;  4
		clr.w	d1		;  4
		lsl.l	#2,d1		; 12
		lsr.w	#2,d2		; 10
		or.w	d2,d1		;  4
		swap	d0		;  4
		and.l	d7,d0		;  8 d7=$3333cccc
		or.l	d0,d1		;  8

; d2 = d1 >> 7
; d2 = ..............a5 a4c5c4e5e4g5g4b5 b4d5d4f5f4h5h4a7 a6c7c6e7e6g7g6..

		move.l	d1,d2		;  4
		lsr.l	#7,d2		; 22

; d0 = d1 & 55555555
; d0 = ..a4..c4..e4..g4 ..b4..d4..f4..h4 ..a6..c6..e6..g6 ..b6..d6..f6..h6

		move.l	d1,d0		;  4
		and.l	d6,d0		;  8 d6=$55555555

; d1 ^= d0
; d1 = a5..c5..e5..g5.. b5..d5..f5..h5.. a7..c7..e7..g7.. b7..d7..f7..h7..

		eor.l	d0,d1		;  8

; d4 = d2 & 55555555
; d4 = ..............a5 ..c5..e5..g5..b5 ..d5..f5..h5..a7 ..c7..e7..g7....

		move.l	d2,d4		;  4
		and.l	d6,d4		;  8 d6=$55555555

; d2 ^= d4
; d2 = ................ a4..c4..e4..g4.. b4..d4..f4..h4.. a6..c6..e6..g6..

		eor.l	d4,d2		;  8

; d1 = (d1 | d4) >> 1
; d1 = ................ a5b5c5d5e5f5g5h5 ................ a7b7c7d7e7f7g7h7

		or.l	d4,d1		;  8
		lsr.l	#1,d1		; 10

		move.b	d1,(plsiz,a6)	; 12 plane 7
		swap	d1		;  4
		move.b	d1,(plsiz,a5)	; 12 plane 5

; d2 |= d0
; d2 = ................ a4b4c4d4e4f4g4h4 ................ a6b6c6d6e6f6g6h6

		or.l	d0,d2		;  8

		move.b	d2,(a6)+	;  8 plane 6
		swap	d2		;  4
		move.b	d2,(a5)+	;  8 plane 4

; d2 = d3 >> 7
; d2 = ..............a1 a0c1c0e1e0g1g0b1 b0d1d0f1f0h1h0a3 a2c3c2e3e2g3g2..

		move.l	d3,d2		;  4
		lsr.l	#7,d2		; 22

; d0 = d3 & 55555555
; d0 = ..a0..c0..e0..g0 ..b0..d0..f0..h0 ..a2..c2..e2..g2 ..b2..d2..f2..h2

		move.l	d3,d0		;  4
		and.l	d6,d0		;  8 d6=$55555555

; d3 ^= d0
; d3 = a1..c1..e1..g1.. b1..d1..f1..h1.. a3..c3..e3..g3.. b3..d3..f3..h3..

		eor.l	d0,d3		;  8

; d4 = d2 & 55555555
; d4 = ..............a1 ..c1..e1..g1..b1 ..d1..f1..h1..a3 ..c3..e3..g3....

		move.l	d2,d4		;  4
		and.l	d6,d4		;  8 d6=$55555555

; d2 ^= d4
; d2 = ................ a0..c0..e0..g0.. b0..d0..f0..h0.. a2..c2..e2..g2..

		eor.l	d4,d2		;  8

; d3 = (d3 | d4) >> 1
; d3 = ................ a1b1c1d1e1f1g1h1 ................ a3b3c3d3e3f3g3h3

		or.l	d4,d3		;  8
		lsr.l	#1,d3		; 10

		move.b	d3,(plsiz,a4)	; 12 plane 3
		swap	d3		;  4
		move.b	d3,(a3)+	;  8 plane 1

; d2 = d2 | d0
; d2 = ................ a0b0c0d0e0f0g0h0 ................ a2b2c2d2e2f2g2h2

		or.l	d0,d2		;  8

		move.b	d2,(a4)+	;  8 plane 2
		swap	d2		;  4
		move.b	d2,(a1)+	;  8 plane 0

; test if finished

		cmpa.l	a1,a2		;  6
		bne.w	mainloop	; 10	total=536 (67.0 cycles/pixel)

		movem.l	(sp)+,d2-d7/a2-a6

		rts