Streamlining tile blitter macros

This commit is contained in:
Lucas Scharenbroich 2021-11-19 13:00:59 -06:00
parent 4c31a0d056
commit b4871efe8f
6 changed files with 175 additions and 102 deletions

View File

@ -340,110 +340,66 @@ transparent
next next
eom eom
; Masked renderer for a dynamic tile. What's interesting about this renderer is that the mask
; value is not used directly, but simply indicates if we can use a LDA 0,x / PHA sequence,
; a LDA (00),y / PHA, or a JMP to a blended render
;
; If a dynamic tile is animated, there is the possibility to create a special mask that marks
; words of the tile that a front / back / mixed across all frames.
;
; ]1 : tiledata offset
; ]2 : tilemask offset
; ]3 : code field offset
CopyMaskedDWord MAC
; Need to fill in the first 6 bytes of the JMP handler with the following code sequence
;
; lda (00),y
; and $80,x
; ora $00,x
ldx _X_REG ; Get the addressing offset
ldal JTableOffset,x ; Get the address offset and add to the base address
adc _BASE_ADDR ; of the current code field line
adc #{]1&$F000} ; adjust for the current row offset
sta: ]1+1,y
tax ; This becomes the new address that we use to patch in
lda _X_REG ; Get the offset and render a LDA (dp),y instruction
sep #$20 ; Easier to do 8-bit operations
sta: $0001,x ; Set the LDA (00),y operand
lda #$B1
sta: $0000,x ; Set the LDA (00),y opcode
lda _T_PTR
sta: $0005,x ; Set ORA 00,x operand
ora #$80
sta: $0003,x ; Set AND 00,x operand
lda #$35
sta: $0002,x ; Set AND 00,x opcode
lda #$15
sta: $0004,x ; Set ORA 00,x opcode
rep #$30
lda #$0F80 ; branch to the prologue (BRA *+17)
sta: $0006,x
eom
; Masked renderer for a dynamic tile with sprite data overlaid. ; Masked renderer for a dynamic tile with sprite data overlaid.
; ;
; If a dynamic tile is animated, there is the possibility to create a special mask that marks ; ]1 : sprite plane offset
; words of the tile that a front / back / mixed across all frames. ; ]2 : code field offset
; CopyDynSpriteWord MAC
; ]1 : tiledata offset
; ]2 : tilemask offset
; ]3 : code field offset
CopyMaskedDynSpriteWord MAC
; Need to fill in the first 12 bytes of the JMP handler with the following code sequence ; Need to fill in the first 10 bytes of the JMP handler with the following code sequence where
; the data and mask from from the sprite plane
; ;
; lda (00),y ; lda $00,x
; and $80,x
; ora $00,x
; and #MASK ; and #MASK
; ora #DATA ; ora #DATA
; bra *+11 ; bra *+15
; ;
; If MASK == 0, then we can do a PEA. If MASK == $FFFF, then fall back to the simple Dynamic Masked ; If MASK == 0, then we can do a PEA. If MASK == $FFFF, then fall back to the simple Dynamic Tile
; code. ; code.
ldal spritemask+]1,x ; load the mask value
bne mixed ; a non-zero value may be mixed
; This is a solid word
lda #$00F4 ; PEA instruction
sta: ]2,y
ldal spritedata+]1,x ; load the sprite data
sta: ]2+1,y ; PEA operand
bra next
ldx _X_REG ; Get the addressing offset mixed cmp #$FFFF ; All 1's in the mask is a fully transparent sprite word
ldal JTableOffset,x ; Get the address offset and add to the base address beq transparent
adc _BASE_ADDR ; of the current code field line
lda _JTBL_CACHE ; Get the offset to the exception handler for this column
adc #{]1&$F000} ; adjust for the current row offset adc #{]1&$F000} ; adjust for the current row offset
sta: ]1+1,y sta: ]1+1,y
tax ; This becomes the new address that we use to patch in tay ; This becomes the new address that we use to patch in
lda _X_REG ; Get the offset and render a LDA (dp),y instruction ldx _SPR_X_REG ; Offset into the sprite plane
sep #$20 ; Easier to do 8-bit operations lda _OP_CACHE ; Get the LDA dp,x instruction for this column
sta: $0001,x ; Set the LDA (00),y operand sta: $0000,y
lda #$B1
sta: $0000,x ; Set the LDA (00),y opcode
lda _T_PTR lda #$0029 ; AND #SPRITE_MASK
sta: $0005,x ; Set ORA 00,x operand sta: $0002,y
ora #$80 ldal spritemask+]1,x
sta: $0003,x ; Set AND 00,x operand sta: $0003,y
lda #$35
sta: $0002,x ; Set AND 00,x opcode
lda #$15
sta: $0004,x ; Set ORA 00,x opcode
lda #$29 lda #$0009 ; ORA #SPRITE_DATA
sta: $0006,y ; AND #$0000 opcode sta: $0005,y
lda #$09 ldal spritedata+]1,x
sta: $0009,y ; ORA #$0000 opcode sta: $0006,y
rep #$20
lda ]1+32 ; insert the tile mask and data into the exception lda #$0D80 ; branch to the prologue (BRA *+15)
sta: $0007,y ; handler. sta: $0008,y
lda ]1
sta: $000A,y
lda #$0990 ; BCC *+11 ldy _Y_REG ; restore original y-register value and move on
sta: $000C,x bra next
; This is a transparent word, so just show the dynamic data
transparent
lda #$4800 ; Put the PHA in the third byte
sta: ]2+1,y
lda _OP_CACHE ; Store the LDA dp,x instruction with operand
sta: ]2,y
next
eom eom

View File

@ -412,7 +412,7 @@ ReadControl ENT
put blitter/Tiles00011.s ; dynamic masked tiles put blitter/Tiles00011.s ; dynamic masked tiles
put blitter/Tiles10000.s ; normal tiles + sprites put blitter/Tiles10000.s ; normal tiles + sprites
; put blitter/Tiles10001.s put blitter/Tiles10001.s
put blitter/Tiles10010.s ; normal masked tiles + sprites put blitter/Tiles10010.s ; normal masked tiles + sprites
; put blitter/Tiles10011.s ; put blitter/Tiles10011.s
put blitter/Tiles11000.s ; normal high priority tile + sprites put blitter/Tiles11000.s ; normal high priority tile + sprites

View File

@ -48,6 +48,9 @@ _Y_REG equ tiletmp+2
_T_PTR equ tiletmp+4 ; Copy of the tile address pointer _T_PTR equ tiletmp+4 ; Copy of the tile address pointer
_BASE_ADDR equ tiletmp+6 ; Copy of BTableLow for this tile _BASE_ADDR equ tiletmp+6 ; Copy of BTableLow for this tile
_SPR_X_REG equ tiletmp+8 ; Cache address of sprite plane source for a tile _SPR_X_REG equ tiletmp+8 ; Cache address of sprite plane source for a tile
_JTBL_CACHE equ tiletmp+10 ; Cache the offset to the exception handler for a column
_OP_CACHE equ tiletmp+12 ; Cache of a relevant operand / oeprator
_TILE_ID equ tiletmp+14 ; Copy of the tile descriptor
; Low-level function to take a tile descriptor and return the address in the tiledata ; Low-level function to take a tile descriptor and return the address in the tiledata
; bank. This is not too useful in the fast-path because the fast-path does more ; bank. This is not too useful in the fast-path because the fast-path does more
@ -119,7 +122,7 @@ _RenderTile2
stx _SPR_X_REG stx _SPR_X_REG
:nosprite :nosprite
sta tmp0 ; Some tile blitters need to get the tile descriptor sta _TILE_ID ; Some tile blitters need to get the tile descriptor
and #TILE_CTRL_MASK and #TILE_CTRL_MASK
xba xba
tax tax
@ -180,10 +183,11 @@ TileProcs dw _TBSolidTile_00,_TBSolidTile_0H,_TBSolidTile_V0,_TBSolid
dw _TBDynamicMaskTile_00,_TBDynamicMaskTile_00 dw _TBDynamicMaskTile_00,_TBDynamicMaskTile_00
; Here are all the sprite variants of the tiles ; Here are all the sprite variants of the tiles
dw _TBSolidSpriteTile_00,_TBSolidSpriteTile_0H, dw _TBSolidSpriteTile_00,_TBSolidSpriteTile_0H
dw _TBSolidSpriteTile_V0,_TBSolidSpriteTile_VH ; 10000 : normal tiles w/sprite dw _TBSolidSpriteTile_V0,_TBSolidSpriteTile_VH ; 10000 : normal tiles w/sprite
dw _TBSolidTile_00,_TBSolidTile_0H,_TBSolidTile_V0,_TBSolidTile_VH ; 10001 : dynamic tiles w/sprite dw _TBDynamicSpriteTile_00,_TBDynamicSpriteTile_00
dw _TBMaskedSpriteTile_00,_TBMaskedSpriteTile_0H, dw _TBDynamicSpriteTile_00,_TBDynamicSpriteTile_00 ; 10001 : dynamic tiles w/sprite
dw _TBMaskedSpriteTile_00,_TBMaskedSpriteTile_0H
dw _TBMaskedSpriteTile_V0,_TBMaskedSpriteTile_VH ; 10010 : masked normal tiles w/sprite dw _TBMaskedSpriteTile_V0,_TBMaskedSpriteTile_VH ; 10010 : masked normal tiles w/sprite
dw _TBSolidTile_00,_TBSolidTile_0H,_TBSolidTile_V0,_TBSolidTile_VH ; 10011 : masked dynamic tiles w/sprite dw _TBSolidTile_00,_TBSolidTile_0H,_TBSolidTile_V0,_TBSolidTile_VH ; 10011 : masked dynamic tiles w/sprite

View File

@ -17,7 +17,7 @@ _TBDynamicTile_00
; LDA 00,x / PHA where the operand is fixed when the tile is rendered ; LDA 00,x / PHA where the operand is fixed when the tile is rendered
; $B5 $00 $48 ; $B5 $00 $48
_TBDynamicData _TBDynamicData
lda tmp0 ; Get the original tile descriptor lda _TILE_ID ; Get the original tile descriptor
and #$007F ; clamp to < (32 * 4) and #$007F ; clamp to < (32 * 4)
ora #$4800 ; insert the PHA instruction ora #$4800 ; insert the PHA instruction

View File

@ -8,11 +8,26 @@ _TBDynamicMaskTile_00
; A = dynamic tile id (must be <32) ; A = dynamic tile id (must be <32)
_TBDynamicDataAndMask _TBDynamicDataAndMask
sta _X_REG sta _X_REG ; Cache some column values derived from _X_REG
tax
ora #$B100 ; Pre-calc the LDA (dp),y opcode + operand
xba
sta _OP_CACHE
lda tmp0 ; get the original tile descriptor clc
ldal JTableOffset,x ; Get the address offset and add to the base address
adc _BASE_ADDR ; of the current code field line
sta _JTBL_CACHE
; We need to do an AND dp|$80,x / ORA dp,x. The opcode values are $35 and $15, respectively.
; We pre-calculate the AND opcode with the high bit of the operand set and then, in the macro
; perform and EOR #$2080 to covert the opcode and operand in one instruction
lda _TILE_ID ; Get the original tile descriptor
and #$007F ; clamp to < (32 * 4) and #$007F ; clamp to < (32 * 4)
sta _T_PTR ora #$3580 ; Pre-calc the AND $80,x opcode + operand
xba
sta _T_PTR ; This is an op to load the dynamic tile data
CopyMaskedDWord $0003 CopyMaskedDWord $0003
CopyMaskedDWord $1003 CopyMaskedDWord $1003
@ -23,10 +38,20 @@ _TBDynamicDataAndMask
CopyMaskedDWord $6003 CopyMaskedDWord $6003
CopyMaskedDWord $7003 CopyMaskedDWord $7003
inc _T_PTR ; Move to the next column ldx _X_REG
inc _T_PTR inx
inc _X_REG ; Move to the next column inx
inc _X_REG clc
ldal JTableOffset,x ; Get the address offset and add to the base address
adc _BASE_ADDR ; of the current code field line
sta _JTBL_CACHE
lda _OP_CACHE
adc #$0200
sta _OP_CACHE
lda _T_PTR
adc #$0200
sta _T_PTR
CopyMaskedDWord $0000 CopyMaskedDWord $0000
CopyMaskedDWord $1000 CopyMaskedDWord $1000
@ -61,3 +86,39 @@ _TBFillJMPOpcode
sta $7003,y sta $7003,y
rep #$20 rep #$20
rts rts
; Masked renderer for a dynamic tile. What's interesting about this renderer is that the mask
; value is not used directly, but simply indicates if we can use a LDA 0,x / PHA sequence,
; a LDA (00),y / PHA, or a JMP to a blended render
;
; If a dynamic tile is animated, there is the possibility to create a special mask that marks
; words of the tile that a front / back / mixed across all frames.
;
; ]1 : code field offset
;
; This macro does not set the opcode since they will all be JMP instructions, they can be
; filled more efficiently in a separate routine.
CopyMaskedDWord MAC
; Need to fill in the first 6 bytes of the JMP handler with the following code sequence
;
; lda (00),y
; and $80,x
; ora $00,x
; bra *+17
lda _JTBL_CACHE
ora #{]1&$F000} ; adjust for the current row offset
sta: ]1+1,y
tax ; This becomes the new address that we use to patch in
lda _OP_CACHE
sta: $0000,x ; LDA (00),y
lda _T_PTR
sta: $0002,x ; AND $80,x
eor #$2080 ; Switch the opcode to an ORA and remove the high bit of the operand
sta: $0004,x ; ORA $00,x
lda #$0F80 ; branch to the prologue (BRA *+17)
sta: $0006,x
eom

52
src/blitter/Tiles10001.s Normal file
View File

@ -0,0 +1,52 @@
; _TBDynamicSpriteTile
;
; This tile type does not explicitly support horizontal or vertical flipping. An appropriate tile
; descriptor should be passed into CopyTileToDyn to put the horizontally or vertically flipped source
; data into the dynamic tile buffer
_TBDynamicSpriteTile_00
sty _Y_REG ; This is restored in the macro
sta _X_REG ; Cache some column values derived from _X_REG
tax
clc
ldal JTableOffset,x ; Get the address offset and add to the base address
adc _BASE_ADDR ; of the current code field line
sta _JTBL_CACHE
lda _TILE_ID ; Get the original tile descriptor
and #$007F ; clamp to < (32 * 4)
ora #$B500
xba
sta _OP_CACHE ; This is the 2-byte opcode for to load the data
CopyDynSpriteWord {0*SPRITE_PLANE_SPAN};$0003
CopyDynSpriteWord {1*SPRITE_PLANE_SPAN};$1003
CopyDynSpriteWord {2*SPRITE_PLANE_SPAN};$2003
CopyDynSpriteWord {3*SPRITE_PLANE_SPAN};$3003
CopyDynSpriteWord {4*SPRITE_PLANE_SPAN};$4003
CopyDynSpriteWord {5*SPRITE_PLANE_SPAN};$5003
CopyDynSpriteWord {6*SPRITE_PLANE_SPAN};$6003
CopyDynSpriteWord {7*SPRITE_PLANE_SPAN};$7003
ldx _X_REG
inx
inx
clc
ldal JTableOffset,x ; Get the address offset and add to the base address
adc _BASE_ADDR ; of the current code field line
sta _JTBL_CACHE
lda _OP_CACHE
adc #$0200
sta _OP_CACHE
CopyDynSpriteWord {0*SPRITE_PLANE_SPAN}+2;$0000
CopyDynSpriteWord {1*SPRITE_PLANE_SPAN}+2;$1000
CopyDynSpriteWord {2*SPRITE_PLANE_SPAN}+2;$2000
CopyDynSpriteWord {3*SPRITE_PLANE_SPAN}+2;$3000
CopyDynSpriteWord {4*SPRITE_PLANE_SPAN}+2;$4000
CopyDynSpriteWord {5*SPRITE_PLANE_SPAN}+2;$5000
CopyDynSpriteWord {6*SPRITE_PLANE_SPAN}+2;$6000
CopyDynSpriteWord {7*SPRITE_PLANE_SPAN}+2;$7000
rts