Rewriting count_neighbourgs in assembly

Makefile

@@ -5,7 +5,7 @@ LINUX_OUT	:= gol.out
 
 APPLE2_CL	:= $(CC65_HOME)/bin/cl65
 APPLE2_CC	:= $(CC65_HOME)/bin/cc65
-APPLE2_SRC	:= gol_apple2.c
+APPLE2_SRC	:= gol_apple2.c gol_apple2_optimized.asm
 APPLE2_MAP  := gol_apple2.map
 APPLE2_CFLAGS	:= -Oirs -v -t apple2
 APPLE2_OUT	:= gol.a2

benchs.txt

@@ -0,0 +1,5 @@
+Time to 50 iterations
+
+50ef808 : 61:25 
+        : 38:51
+

gol_apple2.c

@@ -17,7 +17,7 @@ uint8_t toggle_cell( const uint8_t x, const uint8_t y ); /* toggles the cell at
 
 void run( void );                           /* runs the simulation */
 void update( void );                        /* updates the simulation */
-void count_neighbours( uint8_t* cell, uint8_t* count ); /* counts nb neighbours of the cell */
+uint8_t __fastcall__  count_neighbours( uint8_t* cell ); /* counts nb neighbours of the cell */
 
 void quit( void );
 
@@ -224,7 +224,6 @@ void run( void  )
 
 void update( void )
 {
-    uint8_t nb_neighbours;
     uint8_t x, y;
     for( y = 1u; y < NB_LINES-1; ++y )
     {
@@ -232,7 +231,7 @@ void update( void )
         {
           register uint8_t nb_neighbours;
           uint8_t* cell = &Cells[x-1u][y-1u];
-          count_neighbours( cell, &nb_neighbours );
+          nb_neighbours = count_neighbours( cell );
           if( Cells[x][y] == ALIVE && \
               (nb_neighbours < 2u || nb_neighbours > 3u )
           ) {
@@ -248,12 +247,3 @@ void update( void )
     memcpy( Cells, Cells_Future, sizeof(Cells) );
 }
 
-
-void count_neighbours( uint8_t* cell, uint8_t* count )
-{
-    *count = *cell++; *count += *cell++; *count += *cell;
-    cell += JUMP_BEGINNING_NEXT_LINE;
-    *count += *cell; cell+=2 ; *count += *cell;
-    cell += JUMP_BEGINNING_NEXT_LINE;
-    *count += *cell++; *count += *cell++; *count += *cell;
-}

gol_apple2_optimized.asm

@@ -0,0 +1,56 @@
+
+    .include        "apple2.inc"
+    .include        "zeropage.inc"
+    
+    .export _count_neighbours
+
+; ******************
+;uint8_t __fastcall__ count_neighbours( uint8_t* cell )
+;param: cell is in AX
+_count_neighbours:
+
+    ;ASSUMPTIONS:
+    ; -> A and Y (offset to starting ptr) won't overflow!
+
+    ;init
+    STA ptr1
+    STX ptr1+1
+    LDA #0
+    LDY #0
+    CLC
+    
+    ;acc 1st row
+    ADC (ptr1),Y
+    INY
+    ADC (ptr1),Y
+    INY
+    ADC (ptr1),Y
+    
+    ;next row    
+    STA tmp1
+    TYA
+    ADC #$15
+    TAY
+    LDA tmp1
+    ADC (ptr1),Y
+    INY
+    INY
+    ADC (ptr1),Y
+    
+    ;next row    
+    STA tmp1
+    TYA
+    ADC #$15
+    TAY
+    LDA tmp1
+    ADC (ptr1),Y
+    INY
+    ADC (ptr1),Y
+    INY
+    ADC (ptr1),Y
+    
+    ;return
+    TAX
+    LDX #0
+    RTS
+