#define lcd_delay()  ex (sp),hl \ ex (sp),hl

GraphToLCD_:
  ld hl,(BufPtr)
  ld ix,(GrayBufPtr)

BufferToLCD_:
;Input:
;     HL points to the buffer to copy to the LCD
;     IX points to a secondary buffer to use in grayscale
;     (graymask) has the gray mask
;     (graymask+1) is the method
;     expects the LCD to be in X-increment mode
;       (If you don't know what this means, you probably don't have to worry!)
;Outputs:
;     HL is incremented by 12
;     DE is incremented by 12
;Destroys:
;     AF,BC,IX
;       This might change in the future, but A and C are the gray mask, B is 0,
;       and IX points to the routine that mixes the pixels from two layers

; Don't want interrupts interfering
  di

; Expect that the LCD increment mode is already set to increment down
;
; We shouldn't need a delay here!
;  in a,(16) \ rla \ jr c,$-3

; Set the LCD pointer to the top row
  ld a,$80
  out (16),a

  lcd_delay()   ;new calcs seem to ignore the LCD busy bit

;we'll need to increment HL by 12 each iteration anyways, so may as well make
;DE useful
  ld de,12

  ld a,(graymode)
  or a
  jr z,gray2
  dec a
  jp z,gray3
  dec a
  jp z,gray4
  sub 14
  jp z,LCD_buf_OR
  dec a
  jp z,mask_buffer_to_LCD

gray2:
  ld c,16

; Set the LCD pointer to the left-most column of 8 pixels
  ld a,$20

gray2_col:
  lcd_delay()   ;new calcs seem to ignore the LCD busy bit

;Wait for the LCD to be ready for new data
  in f,(c) \ jp m,$-2

;Set the column
  out (10h),a

;Save the column number
  push af

;Set B to 64, we'll be writing a whole column, so 64 writes.
  ld b,64
_:
;Now wait for the LCD to be ready
  ld a,(hl)
  add hl,de
  bit InvertLCDFlag,(iy+userflags)
  jr z,$+3
  cpl
  in f,(c) \ jp m,$-2
  out (17),a
  djnz -_
  dec h
  dec h
  dec h
  inc hl

;Restore A to the column number, then increment. If it hits $2C, we've drawn all
;12 columns, so no more looping. Otherwise, draw the next column.
  pop af
  inc a
  cp $2C
  jp nz,gray2_col
  ret

gray3:
;We have to wait for the LCD anyways, so may as well do something useful.
;Get the gray mask and put it in C
  ld a,(graymask)
  ld c,a

;Wait for the LCD to be ready for new data
  in a,(16) \ rla \ jr c,$-3

; Set the LCD pointer to the left-most column of 8 pixels
  ld a,$20

gray3_col:
;Set the column
  out (10h),a

;Save the column number
  push af

;Set B to 64, we'll be writing a whole column, so 64 writes.
  ld b,64
_:

;Toggles the mask between %10101010 and %01010101.
  rlc c

;Now wait for the LCD to be ready
  in a,(16) \ rla \ jr c,$-3

; Now we use C to select pixels from either (IX), or (HL).
; If a bit in C is 0, the corresponding bit in the output is from (HL),
; else it is from (IX). There is an explaination of this at the bottom.
  ld a,(ix)
  xor (hl)
  and c
  xor (hl)

  add hl,de
  add ix,de

  bit InvertLCDFlag,(iy+userflags)
  jr z,$+3
  cpl
  out (17),a

  djnz -_

;Now point to HL to the start of the next column.
;If we subtract 768, it is where it started, so we'd need to increment 1 more.
;Basically, HL-(3*256)+1
  dec h
  dec h
  dec h
  inc hl

;Same with IX
  dec ixh
  dec ixh
  dec ixh
  inc ix

;Now wait for the LCD to be ready
  in a,(16) \ rla \ jr c,$-3

;Restore A to the column number, then increment. If it hits $2C, we've drawn all
;12 columns, so no more looping. Otherwise, draw the next column.
  pop af
  inc a
  cp $2C
  jp nz,gray3_col

;Rotate the mask one final time. Our masks are either 2- or 3-cycles, and we
;want to avoid a static image if possible!
  ld a,c
  rlca
  ld (graymask),a
  ret

gray4:
;We have to wait for the LCD anyways, so may as well do something useful.
;Get the gray mask and put it in C
  ld a,(graymask)
  ld c,a

;Wait for the LCD to be ready for new data
  in a,(16) \ rla \ jr c,$-3

; Set the LCD pointer to the left-most column of 8 pixels
  ld a,$20

gray4_col:
;Set the column
  out (10h),a

;Save the column number
  push af

;Set B to 64, we'll be writing a whole column, so 64 writes.
  ld b,64
_:

;Now wait for the LCD to be ready
  in a,(16) \ rla \ jr c,$-3

  ld a,c  ;\
  cp $C0  ; | Rotate the graymask
  rra     ; | Can be efficiently done by rotating left while
  ld c,a  ;/  shifting in the XOR of the top two bits.

; Now we use C to select pixels from either (IX), or (HL).
; If a bit in C is 0, the corresponding bit in the output is from (HL),
; else it is from (IX). There is an explaination of this at the end.
  ld a,(ix)
  xor (hl)
  and c
  xor (hl)

  add hl,de
  add ix,de

  bit InvertLCDFlag,(iy+userflags)
  jr z,$+3
  cpl
  out (17),a

  djnz -_

;Now point to HL to the start of the next column.
;If we subtract 768, it is where it started, so we'd need to increment 1 more.
;Basically, HL-(3*256)+1
  dec h
  dec h
  dec h
  inc hl

;Same with IX
  dec ixh
  dec ixh
  dec ixh
  inc ix

;Now wait for the LCD to be ready
  in a,(16) \ rla \ jr c,$-3

;Restore A to the column number, then increment. If it hits $2C, we've drawn all
;12 columns, so no more looping. Otherwise, draw the next column.
  pop af
  inc a
  cp $2C
  jp nz,gray4_col

;Rotate the mask one final time. Our masks are either 2- or 3-cycles, and we
;want to avoid a static image if possible!
  ld a,c
  cp $C0
  rra
  ld (graymask),a
  ret


LCD_buf_OR:
  ld c,16

; Set the LCD pointer to the left-most column of 8 pixels
  ld a,$20

LCD_buf_OR_col:
;Wait for the LCD to be ready for new data
  in f,(c) \ jp m,$-2

;Set the column
  out (10h),a

;Save the column number
  push af

;Set B to 64, we'll be writing a whole column, so 64 writes.
  ld b,64
_:

;Now wait for the LCD to be ready
  ld a,(hl)
  or (ix)

  add hl,de
  add ix,de

  bit InvertLCDFlag,(iy+userflags)
  jr z,$+3
  cpl
  in f,(c) \ jp m,$-2
  out (17),a
  djnz -_
  dec h
  dec h
  dec h
  inc hl

  dec ixh
  dec ixh
  dec ixh
  inc ix

  in f,(c) \ jp m,$-2

;Restore A to the column number, then increment. If it hits $2C, we've drawn all
;12 columns, so no more looping. Otherwise, draw the next column.
  pop af
  inc a
  cp $2C
  jp nz,LCD_buf_OR_col
  ret

mask_buffer_to_LCD:
;This will AND the secondary buf on to the tilemap_buf, then OR the primary buf
;on to that.
  ld de,(tilemap_buf)

; Expect that the LCD increment mode is already set to increment down
;
; We shouldn't need a delay here!
;  in a,(16) \ rla \ jr c,$-3

; ; Set the LCD pointer to the top row
;   ld a,$80
;   out (16),a

;Wait for the LCD to be ready for new data
  in a,(16) \ rla \ jr c,$-3

; Set the LCD pointer to the left-most column of 8 pixels
  ld a,$20

maskcol:
;Set the column
  out (10h),a

;Save the column number
  push af

;Set B to 64, we'll be writing a whole column, so 64 writes.
  ld bc,$400C
_:
  push bc
  ld a,(de)
  and (ix)
  xor (hl)
  bit InvertLCDFlag,(iy+userflags)
  jr z,$+3
  cpl
  push af

  ld b,0
  add hl,bc
  ex de,hl
  add hl,bc
  ex de,hl
  add ix,bc

;Now wait for the LCD to be ready
  in a,(16) \ rla \ jr c,$-3
  pop af
  out (17),a
  pop bc
  djnz -_

;Now point to HL to the start of the next column.
;If we subtract 768, it is where it started, so we'd need to increment 1 more.
;Basically, HL-(3*256)+1
  dec h
  dec h
  dec h
  inc hl

;Same with DE
  dec d
  dec d
  dec d
  inc de

;Same with IX
  dec ixh
  dec ixh
  dec ixh
  inc ix

;Restore A to the column number, then increment. If it hits $2C, we've drawn all
;12 columns, so no more looping. Otherwise, draw the next column.
  pop af
  inc a
  cp $2C
  jp nz,maskcol
  ret






; How C can select from two different buffers
;
;
;We'll use this magical formula:
; ((x^y)&m)^y
;
;When m is 0:
;   ((x^y)&m)^y
; = ((x^y)&0)^y
; = (0)^y
; = y
;
;When m is 1:
;   ((x^y)&m)^y
; = ((x^y)&1)^y
; = (x^y)^y
; = x^(y^y)     ;We can do this when all operations are XOR
; = x
;
;So a 0 in the mask will choose the pixel from (HL) (primary buffer),
;and a 1 will choose a pixel from (IX) (gray/back/secondary buffer).