revert restriction on certain associative operator reshuffling

it caused larger generated code

compiler/src/prog8/compiler/astprocessing/StatementReorderer.kt

@@ -307,13 +307,12 @@ internal class StatementReorderer(val program: Program,
                         val newValue = BinaryExpression(leftBinExpr.left, binExpr.operator, newRight, binExpr.position)
                         listOf(IAstModification.ReplaceNode(binExpr, newValue, assignment))
                     }
-                    else if(leftBinExpr.left.constValue(program)!=null && binExpr.right.constValue(program)!=null) {
+                    else {
                         // A = (x <associative-operator> A) <same-operator> y ==> A = A <associative-operator> (x <same-operator> y)
                         val newRight = BinaryExpression(leftBinExpr.left, binExpr.operator, binExpr.right, binExpr.position)
                         val newValue = BinaryExpression(leftBinExpr.right, binExpr.operator, newRight, binExpr.position)
                         listOf(IAstModification.ReplaceNode(binExpr, newValue, assignment))
                     }
-                    else noModifications
                 }
                 val rightBinExpr = binExpr.right as? BinaryExpression
                 if(rightBinExpr?.operator == binExpr.operator) {

docs/source/todo.rst

@@ -3,22 +3,12 @@ TODO
 
 For next release
 ^^^^^^^^^^^^^^^^
-- petaxian roller.p8 line 49  (also see test.p8) generates large code compared to 8.2
+- logical.p8 is huge compared to 8.2
 
-- code gen for if statements has become inefficient? vm/6502?
-        if not a or not b
-            return 0
-- code gen for while loops has become inefficient:  (until loops probably as well)
-  (maybe solved when if statements code has been fixed)
-             while c64.CHRIN()!='\"'  {
-                if c64.READST()
-                    goto close_end
-            }
+- compiling logical.p8 to virtual with optimization generates code that is smaller,
+  but takes many more vm instructions to execute than not-optimized code!?
 
-- petaxian.prg became quite a bit (200 bytes) larger, why!? because of the above?
-
-- compiling logical.p8 to virtual with optimization generates a lot larger code as without optimizations.
-  this is not the case for the 6502 codegen.
+- petaxian.prg became quite a bit (150 bytes) larger since 8.2, what causes that
 
 - add some more optimizations in vmPeepholeOptimizer
 - vm Instruction needs to know what the read-registers/memory are, and what the write-register/memory is.

examples/test.p8

@@ -7,90 +7,22 @@ main  {
 
     sub start() {
 
-        ubyte a1 = 0
-        uword w1 = 0
-        ubyte vv
+        ; a "pixelshader":
+        sys.gfx_enable(0)       ; enable lo res screen
+        ubyte shifter
 
-        if not a1
-            txt.print("ok ")
-        else
-            txt.print("fail ")
-        vv = not a1
-        txt.print_ub(vv)
-        txt.nl()
-        a1=128
-        if not a1
-            txt.print("fail ")
-        else
-            txt.print("ok ")
-        vv = not a1
-        txt.print_ub(vv)
-        txt.nl()
-        if not w1
-            txt.print("ok ")
-        else
-            txt.print("fail ")
-        vv = not w1
-        txt.print_ub(vv)
-        txt.nl()
-        w1 = 4096
-        if not w1
-            txt.print("fail ")
-        else
-            txt.print("ok ")
-        vv = not w1
-        txt.print_ub(vv)
-        txt.nl()
-
-
-        ; petaxian roller.p8 line 49
-        ; super large in new version, ok in 8.2....
-        ; TODO cx16.r0 = a2 + 25 + (a1/40)
-        ; txt.setcc( a1, a2 + 25 + (a1/40), 11,22)
-
-;        if a1 and a2 {
-;            a1++
-;        }
-;
-;        if not a1 or not a2 {
-;            a1++
-;        }
-;
-;        if a1!=99 and not a2 {
-;            a1++
-;        }
-
-;        while a1 != a2 {
-;            a1++
-;        }
-;
-;        while a1!='\"'  {
-;            a1++
-;        }
-;        do {
-;            a1++
-;        } until a1==99
-;
-;close_end:
-;        a1++
-
-
-;        ; a "pixelshader":
-;        sys.gfx_enable(0)       ; enable lo res screen
-;        ubyte shifter
-;
-;        repeat {
-;            uword xx
-;            uword yy = 0
-;            repeat 240 {
-;                xx = 0
-;                repeat 320 {
-;                    sys.gfx_plot(xx, yy, xx*yy + shifter as ubyte)
-;                    xx++
-;                }
-;                yy++
-;            }
-;            shifter+=4
-;        }
+        repeat {
+            uword xx
+            uword yy = 0
+            repeat 240 {
+                xx = 0
+                repeat 320 {
+                    sys.gfx_plot(xx, yy, xx*yy + shifter as ubyte)
+                    xx++
+                }
+                yy++
+            }
+            shifter+=4
+        }
      }
 }