preparing to optimize monogfx.fill()

compiler/res/prog8lib/cx16/monogfx.p8

@@ -789,31 +789,14 @@ invert:
         while cx16.r12L!=0 {
             pop_stack()
             xx = x1
-            while xx >= 0 {
-                if pget(xx as uword, yy as uword) as ubyte != cx16.r11L
-                    break
-                xx--
-            }
-            if x1!=xx
-                horizontal_line(xx as uword+1, yy as uword, x1-xx as uword, cx16.r10L as bool)
-            else
-                goto skip
-
+            if fill_scanline_left() goto skip
             left = xx + 1
             if left < x1
                 push_stack(left, x1 - 1, yy, -dy)
             xx = x1 + 1
 
             do {
-                cx16.r9s = xx
-                while xx <= width-1 {
-                    if pget(xx as uword, yy as uword) as ubyte != cx16.r11L
-                        break
-                    xx++
-                }
-                if cx16.r9s!=xx
-                    horizontal_line(cx16.r9, yy as uword, xx-cx16.r9s as uword, cx16.r10L as bool)
-
+                fill_scanline_right()
                 push_stack(left, xx - 1, yy, dy)
                 if xx > x2 + 1
                     push_stack(x2 + 1, xx - 1, yy, -dy)
@@ -827,6 +810,33 @@ skip:
                 left = xx
             } until xx>x2
         }
+
+        sub fill_scanline_left() -> bool {
+            ; TODO optimize this to use vera auto-decrements, but requires masking etc because of 8 pixels per byte...
+            cx16.r9s = xx
+            while xx >= 0 {
+                if pget(xx as uword, yy as uword) as ubyte != cx16.r11L
+                    break
+                xx--
+            }
+            if xx!=cx16.r9s {
+                horizontal_line(xx+1 as uword, yy as uword, cx16.r9s-xx as uword, cx16.r10L as bool)
+                return false
+            }
+            return true
+        }
+
+        sub fill_scanline_right() {
+            ; TODO optimize this to use vera auto-increments, but requires masking etc because of 8 pixels per byte...
+            cx16.r9s = xx
+            while xx <= width-1 {
+                if pget(xx as uword, yy as uword) as ubyte != cx16.r11L
+                    break
+                xx++
+            }
+            if xx!=cx16.r9s
+                horizontal_line(cx16.r9, yy as uword, xx-cx16.r9s as uword, cx16.r10L as bool)
+        }
     }
 
     sub position(uword @zp xx, uword yy) {

docs/source/todo.rst

@@ -1,6 +1,9 @@
 TODO
 ====
 
+Optimize monogfx fill_scanline_left() and fill_scanline_right().
+
+
 Improve register load order in subroutine call args assignments:
 in certain situations, the "wrong" order of evaluation of function call arguments is done which results
 in overwriting registers that already got their value, which requires a lot of stack juggling (especially on plain 6502 cpu!)

examples/test.p8

@@ -1,4 +1,4 @@
-%import gfx2
+%import monogfx
 %import textio
 %import math
 
@@ -9,31 +9,31 @@
 main {
 
     sub start() {
-        gfx2.screen_mode(2)
+        monogfx.lores()
         demofill()
     }
 
     sub demofill() {
-        gfx2.circle(160, 120, 110, 1)
-        gfx2.rect(180, 5, 25, 190, 2)
-        gfx2.line(100, 150, 240, 10, 2)
-        gfx2.rect(150, 130, 10, 100, 3)
+        monogfx.circle(160, 120, 110, true)
+        monogfx.rect(180, 5, 25, 190, true)
+        monogfx.line(100, 150, 240, 10, true)
+        monogfx.line(101, 150, 241, 10, true)
+        monogfx.rect(150, 130, 10, 100, true)
 
         sys.wait(30)
 
         cbm.SETTIM(0,0,0)
-        gfx2.fill(100,100,3)
-        gfx2.fill(100,100,2)
-        gfx2.fill(100,100,0)
+        monogfx.fill(100,100,true)
+        monogfx.fill(100,100,false)
         uword duration = cbm.RDTIM16()
         sys.wait(30)
 
-        gfx2.screen_mode(0)
+        monogfx.textmode()
         txt.nl()
         txt.print_uw(duration)
         txt.print(" jiffies\n")
 
-        ; hires 4c before optimizations: ~345 jiffies
+        ; before optimizations: ~166 jiffies
 
     }
 }