tweaks:

mandel.s

@@ -435,9 +435,15 @@ fill_masks:
 ;.align 256 ; warning - if not aligned this will cost an extra cycle sometimes
 bank_switch_table:
     .repeat 256, i
-        .byte ((i & $c0) >> 5) | $c1
+        .byte ((i & $c0) >> 4) | $d1
     .endrepeat
 
+.macro bank_switch bank
+    lda #((bank << 2) | $d1)
+    sta PORTB
+.endmacro
+
+
 ; 58-77 cycles
 ; clobbers x, y, dest to dest + 3
 .macro imul8xe dest, arg1, arg2
@@ -497,11 +503,6 @@ bank_switch_table:
 done:
 .endmacro
 
-.macro bank_switch bank
-    lda #((bank << 1) | $c1)
-    sta PORTB
-.endmacro
-
 .proc imul8xe_init
 
     ; go through the input set, in four 16KB chunks
@@ -576,13 +577,14 @@ inner_loop:
     sta result
     lda #0
     adc result + 1
-    sta result
+    sta result + 1
+    clc
     lda arg2
     adc result
     sta result
     lda #0
     adc result + 1
-    sta result
+    sta result + 1
 
     ; inner loop check
     inc arg1