shave 60481 cycles from all precomputed 1-bit HGR transitions (~3% faster)

src/fx/fx.hgr.precomputed.1bit.a

@@ -1,17 +1,6 @@
 ;license:MIT
 ;(c) 2019 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..6 but in high 3 bits, so $00, $20, $40, $60, $80, $A0, $C0
 mirror_copymasks = $0201
 hgrlo          = $0300    ; $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   -
@@ -65,53 +54,53 @@ hgr1himirror   = $BF40    ; $C0 bytes
 }
 
 !macro ROW_X_TO_BASE_ADDRESSES {
-         lda   hgrlo,x
+         lda   hgrlo, x
-         sta   dest1
+         sta   <dest1
-         sta   src1
+         sta   <src1
-         lda   hgr1hi,x
+         lda   hgr1hi, x
-         sta   dest1+1
+         sta   <dest1+1
          eor   #$60
-         sta   src1+1
+         sta   <src1+1
-         lda   hgrlo+1,x
+         lda   hgrlo+1, x
-         sta   dest2
+         sta   <dest2
-         sta   src2
+         sta   <src2
-         lda   hgr1hi+1,x
+         lda   hgr1hi+1, x
-         sta   dest2+1
+         sta   <dest2+1
          eor   #$60
-         sta   src2+1
+         sta   <src2+1
 }
 
 !macro ROW_X_TO_MIRROR_ADDRESSES {
-         lda   hgrlomirror,x
+         lda   hgrlomirror, x
-         sta   mirror_dest1
+         sta   <mirror_dest1
-         sta   mirror_src1
+         sta   <mirror_src1
-         lda   hgr1himirror,x
+         lda   hgr1himirror, x
-         sta   mirror_dest1+1
+         sta   <mirror_dest1+1
          eor   #$60
-         sta   mirror_src1+1
+         sta   <mirror_src1+1
-         lda   hgrlomirror+1,x
+         lda   hgrlomirror+1, x
-         sta   mirror_dest2
+         sta   <mirror_dest2
-         sta   mirror_src2
+         sta   <mirror_src2
-         lda   hgr1himirror+1,x
+         lda   hgr1himirror+1, x
-         sta   mirror_dest2+1
+         sta   <mirror_dest2+1
          eor   #$60
-         sta   mirror_src2+1
+         sta   <mirror_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
+         inc   <input
          beq   +
          jmp   .loop
 +        bit   $c000
          bmi   +
-         inc   input+1
+         inc   <input+1
          jmp   .loop
 +        rts
 }
@@ -136,44 +125,77 @@ hgr1himirror   = $BF40    ; $C0 bytes
          +BUILD_HGR_MIRROR_LOOKUP_TABLES hgrlomirror, hgr1himirror
          +BUILD_MIRROR_COLS mirror_cols
          +BUILD_SPARSE_BITMASKS copymasks, mirror_copymasks
-         +LDADDR .coords
+         ldx   #(end-start)          ; copy InputLoop code to zero page
-         +ST16 input
+-        lda   start-1, x
-         jmp   InputLoop
+         sta   $FF, x
+         dex
+         bne   -
+         txa
+         pha
+         pha
+start
+!pseudopc 0 {
 Exit1Bit rts
 InputLoop
          ldy   #0
-         lda   (input),y             ; first value: HGR row (only 0..95 will be in input array)
+input=*+1
+         lda   .coords, y            ; first value: HGR row (only 0..95 will be in input array)
          bmi   Exit1Bit              ; if > 127 then we're done
          tax
          +ROW_X_TO_BASE_ADDRESSES
          +ROW_X_TO_MIRROR_ADDRESSES
 
          inc   input
-         lda   (input),y
+         lda   (input), y
          +HIGH_3_LOW_5 input
 
          ; main 1x2 block in top-left quadrant
-         +COPY_BIT src1, dest1, copymasks
+src1=*+1
-         +COPY_BIT src2, dest2, copymasks
+         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 1x2 block in top-right quadrant (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
 
          ; corresponding 1x2 block in bottom-right quadrant (opposite row, opposite column)
-         +COPY_BIT mirror_src1, mirror_dest1, mirror_copymasks
+mirror_src1=*+1
-         +COPY_BIT mirror_src2, mirror_dest2, mirror_copymasks
+         lda   $FDFD, y
+         eor   (<mirror_dest1), y
+         and   mirror_copymasks, x
+         eor   (<mirror_dest1), y
+mirror_dest1=*+1
+         sta   $FDFD, y
+mirror_src2=*+1
+         lda   $FDFD, y
+         eor   (<mirror_dest2), y
+         and   mirror_copymasks, x
+         eor   (<mirror_dest2), y
+mirror_dest2=*+1
+         sta   $FDFD, y
 
          ; corresponding 1x2 block in bottom-left quadrant (opposite row, original column)
-         lda   mirror_cols,y
+         lda   mirror_cols, y
          tay
          +COPY_BIT mirror_src1, mirror_dest1, copymasks
          +COPY_BIT mirror_src2, mirror_dest2, copymasks
 
          +INC_INPUT_AND_LOOP InputLoop
-         rts
+}
+end
 !if * and 1 {
          !byte 0 ;align 2 but avoids the fake allocation bug if it was aligned already
 }

src/fx/fx.hgr.radial.a

@@ -5,6 +5,16 @@
 !to "build/FX/RADIAL",plain
 *=$6000
 
+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
+input          = $FE      ; word
+
          !source "src/fx/fx.hgr.precomputed.1bit.a"
 
          +BUILD_HGR_LOOKUP_TABLES hgrlo, hgr1hi

src/fx/fx.hgr.radial.common.a

@@ -1,3 +1,14 @@
+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
+
          +BUILD_HGR_LOOKUP_TABLES hgrlo, hgr1hi
          +BUILD_HGR_MIRROR_LOOKUP_TABLES hgrlomirror, hgr1himirror
          +BUILD_MIRROR_COLS mirror_cols

src/fx/fx.hgr.radial3.a

@@ -5,6 +5,17 @@
 !to "build/FX/RADIAL3",plain
 *=$6000
 
+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
+
          !source "src/fx/fx.hgr.precomputed.1bit.a"
 
          +BUILD_HGR_LOOKUP_TABLES hgrlo, hgr1hi