monogfx and gfx2: flood fill uses optimized horizontal line drawing

compiler/res/prog8lib/cx16/gfx2.p8

@@ -17,7 +17,7 @@
 ;   mode 2 = bitmap 640 x 480 x 4c (2 bpp. there's not enough vram for more colors in hires mode.)
 ;   mode 3 = bitmap 320 x 240 x 16c (not yet implemented: just use 256c, there's enough vram for that)
 ;   mode 4 = bitmap 320 x 240 x 4c (not yet implemented: just use 256c, there's enough vram for that)
-;   higher color dephts in highres are not supported due to lack of VRAM
+;   higher color depths in highres are not supported due to lack of VRAM
 
 gfx2 {
 
@@ -481,11 +481,20 @@ gfx2 {
                     lda  cx16.r0L
                     sta  cx16.VERA_ADDR_L
                     lda  cx16.VERA_DATA0
-                    sta  cx16.r0L
+                    pha
+                    lda  xx
+                    and  #3
+                    tay
+                    lda  gfx2.plot.shift4c,y
+                    tay
+                    pla
+                    cpy  #0
+                    beq  +
+-                   lsr  a
+                    dey
+                    bne  -
++                   and #3
                 }}
-                cx16.r1L = lsb(xx) & 3
-                cx16.r0L >>= gfx2.plot.shift4c[cx16.r1L]
-                return cx16.r0L & 3
             }
             else -> return 0
         }
@@ -574,10 +583,12 @@ gfx2 {
         while cx16.r12L {
             pop_stack()
             xx = x1
-            while xx >= 0 and pget(xx as uword, yy as uword) == cx16.r11L {
+            cx16.r9 = xx
-                plot(xx as uword, yy as uword, cx16.r10L)           ; TODO use vera auto decrement
+            while xx >= 0 and pget(xx as uword, yy as uword) == cx16.r11L
                 xx--
-            }
+            if cx16.r9!=xx
+                horizontal_line(xx as uword+1, yy as uword, cx16.r9-(xx as uword), cx16.r10L)
+
             if xx >= x1
                 goto skip
 
@@ -587,10 +598,12 @@ gfx2 {
             xx = x1 + 1
 
             do {
-                while xx <= width-1 and pget(xx as uword, yy as uword) == cx16.r11L {
+                cx16.r9 = xx
-                    plot(xx as uword, yy as uword, cx16.r10L)       ; TODO use vera auto increment
+                while xx <= width-1 and pget(xx as uword, yy as uword) == cx16.r11L
                     xx++
-                }
+                if cx16.r9!=xx
+                    horizontal_line(cx16.r9, yy as uword, (xx as uword)-cx16.r9, cx16.r10L)
+
                 push_stack(left, xx - 1, yy, dy)
                 if xx > x2 + 1
                     push_stack(x2 + 1, xx - 1, yy, -dy)

compiler/res/prog8lib/cx16/monogfx.p8

@@ -608,10 +608,13 @@ _done
         while cx16.r12L {
             pop_stack()
             xx = x1
-            while xx >= 0 and pget(xx as uword, yy as uword) == cx16.r11L {
+
-                plot(xx as uword, yy as uword, cx16.r10L)
+            cx16.r9 = xx
+            while xx >= 0 and pget(xx as uword, yy as uword) == cx16.r11L
                 xx--
-            }
+            if cx16.r9!=xx
+                horizontal_line(xx as uword+1, yy as uword, cx16.r9-(xx as uword), cx16.r10L)
+
             if xx >= x1
                 goto skip
 
@@ -621,10 +624,12 @@ _done
             xx = x1 + 1
 
             do {
-                while xx <= width-1 and pget(xx as uword, yy as uword) == cx16.r11L {
+                cx16.r9 = xx
-                    plot(xx as uword, yy as uword, cx16.r10L)
+                while xx <= width-1 and pget(xx as uword, yy as uword) == cx16.r11L
                     xx++
-                }
+                if cx16.r9!=xx
+                    horizontal_line(cx16.r9, yy as uword, (xx as uword)-cx16.r9, cx16.r10L)
+
                 push_stack(left, xx - 1, yy, dy)
                 if xx > x2 + 1
                     push_stack(x2 + 1, xx - 1, yy, -dy)

docs/source/todo.rst

@@ -1,8 +1,11 @@
 TODO
 ====
 
-- gfx2: use vera auto in/decrement in the flood fill routine
+- codegen for cx16.r2L >>= gfx2.plot.shift4c[cx16.r1L] is wrong (gfx2.p8 older version fill hires 4c)
-- add an option to inject an actual instruction (STP on 65c02, BRK on 6502) for breakpoints, instead of only listing them in the vice symbol file.
+- why does cx16.r1L = lsb(xx) & 3  generate 5 asm instructions? (noopt)
+- optimize asm:  sta VAR / <something with A> / sta VAR
+- gfx2/monogfx: use vera auto in/decrement in the flood fill routine (getpixels)
+- gfx2: use vera fx fill for horizontal_line of sufficient length?
 
 - [on branch: shortcircuit] investigate McCarthy evaluation again? this may also reduce code size perhaps for things like if a>4 or a<2 ....
 - [on branch: ir-less-branch-opcodes] IR: reduce the number of branch instructions such as BEQ, BEQR, etc (gradually), replace with CMP(I) + status branch instruction

examples/test.p8

@@ -1,49 +1,27 @@
+%import gfx2
 %import textio
+%import math
+
+%option no_sysinit
 %zeropage basicsafe
 
+
 main {
+
     sub start() {
-        ubyte[] b_array = [11,22,33,44]
+        gfx2.screen_mode(2)
-        uword[] w_array = [1111,2222,3333,4444]
+        demofill()
-        uword[] @split split_array = [1111,2222,3333,4444]
+        repeat {
-        ubyte index = 2
-
-        uword value = 9999
-        w_array[index] = value
-        split_array[index] = value
-
-        uword pstep = w_array[index]
-        uword psteps = split_array[index]
-
-        txt.print_uw(pstep)
-        txt.nl()
-        txt.print_uw(psteps)
-        txt.nl()
-
-        uword @zp ptr = &w_array[index]
-        txt.print_uw(peekw(ptr))
-        txt.nl()
-        txt.print_uwhex(&w_array, true)
-        txt.spc()
-        txt.print_uwhex(ptr, true)
-        txt.nl()
-        %breakpoint
-        ptr = &split_array
-        txt.print_uw(peekw(ptr))
-        txt.nl()
-        txt.print_uwhex(&w_array, true)
-        txt.spc()
-        txt.print_uwhex(ptr, true)
-        txt.nl()
-
-        ptr = &b_array[index]
-        txt.print_ub(peek(ptr))
-        txt.nl()
-        txt.print_uwhex(&b_array, true)
-        txt.spc()
-        txt.print_uwhex(ptr, true)
-        txt.nl()
-
         }
     }
 
+    sub demofill() {
+        gfx2.circle(160, 120, 110, 1)
+        gfx2.rect(180, 5, 25, 190, 1)
+        gfx2.line(100, 150, 240, 10, 1)
+        gfx2.line(101, 150, 241, 10, 1)
+        gfx2.fill(100,100,2)
+        gfx2.fill(182,140,3)
+        gfx2.fill(182,40,1)
+    }
+}