Merge branch 'experiment'

the XE fast-path assumed low byte of ptr never changes
to save a couple cycles

mandel-core.s

@@ -18,6 +18,7 @@ z_buffer_start:  .res 1 ; u8: index into z_buffer
 z_buffer_end:    .res 1 ; u8: index into z_buffer
 iter:            .res 1 ; u8: iteration count
 ptr:             .res 2 ; u16
+pixel_ptr:       .res 2 ; u16
 temp:            .res 2 ; u16
 temp2:           .res 2 ; u16
 
@@ -592,7 +593,7 @@ bank_switch_table:
 .macro imul8 dest, arg1, arg2, xe
     .if xe
         ; using 64KB lookup table
-        ; 51-70 cycles
+        ; 52-69 cycles
         ; clobbers x, y, dest, ptr
         .scope
             output = dest
@@ -636,9 +637,7 @@ bank_switch_table:
             txa          ; 2 cyc
             adc output   ; 3 cyc
             sta output   ; 3 cyc
-            lda #0       ; 2 cyc
-            adc output+1 ; 3 cyc
-            sta output+1 ; 3 cyc
+            add_carry output + 1 ; 2-7 cyc
 
         done:
         .endscope
@@ -775,8 +774,8 @@ inner_loop:
 
 ; min base mem: 81 * 4 + 22 * 2 + 10 + 6 = 384 cyc
 ; max base mem: 92 * 4 + 27 * 2 + 50 + 6 = 478 cyc
-; min ext-mem: 51 * 4 + 22 * 2 + 10 + 6 = 264 cyc
-; max ext-mem: 70 * 4 + 27 * 2 + 50 + 6 = 390 cyc
+; min ext-mem: 52 * 4 + 22 * 2 + 10 + 6 = 268 cyc
+; max ext-mem: 69 * 4 + 27 * 2 + 50 + 6 = 386 cyc
 .macro imul16_impl xe
     .local arg1
     .local arg2
@@ -794,15 +793,15 @@ inner_loop:
     ; h1*256*(h2*256 + l2) + l1*(h2*256 + l2)
     ; h1*h2*256*256 + h1*l2*256 + h2*l1*256 + l1*l2
 
-    imul8 result, arg1, arg2, xe        ; 81-92 or 51-70
+    imul8 result, arg1, arg2, xe        ; 81-92 or 52-69
 
-    imul8 result + 2, arg1 + 1, arg2 + 1, xe ; 81-92 or 51-70
+    imul8 result + 2, arg1 + 1, arg2 + 1, xe ; 81-92 or 52-69
 
-    imul8 inter, arg1 + 1, arg2, xe     ; 81-92 or 51-70
+    imul8 inter, arg1 + 1, arg2, xe     ; 81-92 or 52-69
     add16 result + 1, result + 1, inter ; 20 cyc
     add_carry result + 3                ; 2-7 cyc
 
-    imul8 inter, arg1, arg2 + 1, xe     ; 81-92 or 51-70
+    imul8 inter, arg1, arg2 + 1, xe     ; 81-92 or 52-69
     add16 result + 1, result + 1, inter ; 20 cyc
     add_carry result + 3                ; 2-7 cyc
 
@@ -823,8 +822,8 @@ arg2_pos:
 
 ; min base ram: 5 + 19 * 2 + 81 + 22 * 2 + 6 = 174
 ; max base ram: 23 + 19 * 2 + 92 + 27 * 2 + 6 = 213
-; min ext ram: 5 + 19 * 2 + 51 + 22 * 2 + 6 = 144
-; max ext ram: 23 + 19 * 2 + 70 + 27 * 2 + 6 = 191
+; min ext ram: 5 + 19 * 2 + 52 + 22 * 2 + 6 = 145
+; max ext ram: 23 + 19 * 2 + 69 + 27 * 2 + 6 = 190
 .macro sqr16_impl xe
     .scope
         arg = FR0    ; 16-bit arg (clobbered)
@@ -847,7 +846,7 @@ arg2_pos:
 
         sqr8 result + 2, arg + 1  ; 19 cyc
 
-        imul8 inter, arg + 1, arg, xe        ; 81-92 / 51-70 cyc
+        imul8 inter, arg + 1, arg, xe        ; 81-92 / 52-69 cyc
         add16 result + 1, result + 1, inter  ; 20 cyc
         add_carry result + 3                 ; 2-7 cyc
         add16 result + 1, result + 1, inter  ; 20 cyc
@@ -1267,21 +1266,21 @@ enough:
 negative:
     ; temp1 = top half
     lda #.lobyte(framebuffer_top + stride * half_height)
-    sta ptr
+    sta pixel_ptr
     lda #.hibyte(framebuffer_top + stride * half_height)
-    sta ptr + 1
+    sta pixel_ptr + 1
     jmp point
 
 positive:
 
     lda #.lobyte(framebuffer_bottom)
-    sta ptr
+    sta pixel_ptr
     lda #.hibyte(framebuffer_bottom)
-    sta ptr + 1
+    sta pixel_ptr + 1
 
 point:
 
-    ; ptr += sy * stride
+    ; pixel_ptr += sy * stride
     ;    temp * 40
     ; =  temp * 32  +  temp * 8
     ; = (temp << 5) + (temp << 3)
@@ -1289,10 +1288,10 @@ point:
     shl16 temp
     shl16 temp
     shl16 temp
-    add16 ptr, ptr, temp
+    add16 pixel_ptr, pixel_ptr, temp
     shl16 temp
     shl16 temp
-    add16 ptr, ptr, temp
+    add16 pixel_ptr, pixel_ptr, temp
 
     ; Ok so temp1 points to the start of the line, which is 40 bytes.
     ; Get the byte and bit offsets
@@ -1332,20 +1331,20 @@ shift_done:
 
 draw_pixel:
     ; read, mask, or, write
-    lda (ptr),y
+    lda (pixel_ptr),y
     and pixel_mask
     ora pixel_color
-    sta (ptr),y
+    sta (pixel_ptr),y
 
     dex
     beq done
     clc
     lda #40
-    adc ptr
-    sta ptr
+    adc pixel_ptr
+    sta pixel_ptr
     lda #0
-    adc ptr + 1
-    sta ptr + 1
+    adc pixel_ptr + 1
+    sta pixel_ptr + 1
     jmp draw_pixel
 
 done: