diff --git a/mandel.s b/mandel.s
index 39571b3..b706dc5 100644
--- a/mandel.s
+++ b/mandel.s
@@ -115,7 +115,7 @@ plus:
     shr 4, arg
 .endmacro
 
-; 24 to 49 cycles
+; 24 to 44 cycles
 .macro bitmul16 arg1, arg2, result, bitnum
     .local next
 
@@ -137,12 +137,18 @@ plus:
     beq next ; 2 cyc
 
     ; 16-bit add on the top bits
-    add16 result + 2, arg2 ; 20 cyc
+    lda result + 2 ; 3 cyc
+    adc arg2       ; 3 cyc
+    sta result + 2 ; 3 cyc
+    lda result + 3 ; 3 cyc
+    adc arg2 + 1   ; 3 cyc
+    ror a          ; 2 cyc - get a jump on the shift
+    sta result + 3 ; 3 cyc
 
-next:
     ; Shift the 32-bit result down by one bit,
     ; saving the previous carry.
     ror result + 3 ; 5 cyc
+next:
     ror result + 2 ; 5 cyc
     ror result + 1 ; 5 cyc
     .if bitnum >= 8
@@ -164,7 +170,8 @@ next:
 positive:
 .endmacro
 
-; 579 to 725 cycles
+; min 454 cycles
+; max 756 cycles
 .proc imul16
     arg1 = FR0   ; 16-bit arg (clobbered)
     arg2 = FR1   ; 16-bit arg (clobbered)
@@ -183,9 +190,10 @@ positive:
 
     ; unrolled loop for maximum speed, at the cost
     ; of a larger routine
+    ; 424 to 672 cycles
     .repeat 16, bitnum
-        ; first half: 24 to 45 cycles
-        ; second half: 29 to 49 cycles
+        ; first half: 24 to 40 cycles
+        ; second half: 29 to 44 cycles
         bitmul16 arg1, arg2, result, bitnum
     .endrepeat