Apple-2-Tools/4cade
1
0
Fork 0
mirror of https://github.com/a2-4am/4cade.git synced 2025-03-02 19:29:40 +00:00
Code Issues Projects Releases Wiki Activity

shave 61381 cycles from all precomputed 2-bit HGR transitions (~4% faster)

Browse Source
This commit is contained in:
4am 2020-09-05 22:06:35 -04:00
parent 9bb4524402
commit f081f94756
1 changed files with 55 additions and 48 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

103
src/fx/fx.hgr.precomputed.2bit.a
View File

@ -1,20 +1,9 @@
;license:MIT
;(c) 2019 by 4am
;(c) 2019-2020 by 4am
;
mirror_src1 = $E8 ; word
mirror_dest1 = $EA ; word
mirror_src2 = $EC ; word
mirror_dest2 = $EE ; word
src1 = $F0 ; word
dest1 = $F2 ; word
src2 = $F4 ; word
dest2 = $F6 ; word
reverse_input = $FC ; word
input = $FE ; word
copymasks = $0200 ; $100 bytes but sparse, index is 0..4 but in high 3 bits, so $00, $20, $40, $60, $80
mirror_copymasks = $0201
hgrlo = $0300 ; $C0 bytes
hgrlo = $0301 ; $C0 bytes
hgrlomirror = $BD40 ; $C0 bytes
mirror_cols = $BE00 ; $28 bytes
hgr1hi = $BE40 ; $C0 bytes
@ -27,7 +16,7 @@ hgr1himirror = $BF40 ; $C0 bytes
ldx #$27
ldy #$00
- tya
sta .mirror_cols,x
sta .mirror_cols, x
iny
dex
bpl -
@ -70,71 +59,89 @@ hgr1himirror = $BF40 ; $C0 bytes
!macro ROW_X_TO_2BIT_BASE_ADDRESSES {
; X = $01..$C0, mapping to row 0..191
lda hgrlo-1,x
sta dest1
sta src1
lda hgr1hi-1,x
sta dest1+1
lda hgrlo-1, x
sta <dest1
sta <src1
lda hgr1hi-1, x
sta <dest1+1
eor #$60
sta src1+1
lda hgrlo,x
sta dest2
sta src2
lda hgr1hi,x
sta dest2+1
sta <src1+1
lda hgrlo, x
sta <dest2
sta <src2
lda hgr1hi, x
sta <dest2+1
eor #$60
sta src2+1
sta <src2+1
}
!macro HIGH_3_LOW_5 .input {
and #%11100000 ; second value: high 3 bits = index into tables to find bitmasks
tax
eor (.input),y ; second value: low 5 bits = byte offset within the row (implicitly "and #%00011111")
eor (.input), y ; second value: low 5 bits = byte offset within the row (implicitly "and #%00011111")
tay
}
!macro INC_INPUT_AND_LOOP .loop {
inc input
beq +
jmp .loop
+ bit $c000
bmi +
inc input+1
jmp .loop
+ rts
}
!macro FX_PRECOMPUTED_2BIT .coords {
+BUILD_HGR_LOOKUP_TABLES hgrlo, hgr1hi
+BUILD_MIRROR_COLS mirror_cols
+BUILD_SPARSE_BITMASKS_2BIT copymasks, mirror_copymasks
ldx #(end-start) ; copy InputLoop code to zero page
- lda start-1, x
sta $FF, x
dex
bne -
+LDADDR .coords
+ST16 input
jmp InputLoop
sta <input
sty <input+1
lda #0
pha
pha
start
!pseudopc 0 {
Exit2Bit rts
InputLoop
ldy #0
lda (input),y ; first value: HGR row + 1
input=*+1
lda $FDFD, y ; first value: HGR row + 1
beq Exit2Bit ; if 0 then we're done
tax
+ROW_X_TO_2BIT_BASE_ADDRESSES
inc input
lda (input),y
inc <input
lda (<input), y
+HIGH_3_LOW_5 input
; main 2x2 block in left half
+COPY_BIT src1, dest1, copymasks
+COPY_BIT src2, dest2, copymasks
src1=*+1
lda $FDFD, y
eor (<dest1), y
and copymasks, x
eor (<dest1), y
dest1=*+1
sta $FDFD, y
src2=*+1
lda $FDFD, y
eor (<dest2), y
and copymasks, x
eor (<dest2), y
dest2=*+1
sta $FDFD, y
; corresponding 2x2 block in right half (same row, opposite column)
lda mirror_cols,y
lda mirror_cols, y
tay
+COPY_BIT src1, dest1, mirror_copymasks
+COPY_BIT src2, dest2, mirror_copymasks
+INC_INPUT_AND_LOOP InputLoop
rts
inc <input
bne InputLoop
bit $c000
bmi Exit2Bit
inc <input+1
bne InputLoop ; always branches
}
end
!if * and 1 {
!byte 0 ;align 2 but avoids the fake allocation bug if it was aligned already
}