diff --git a/mandel-core.s b/mandel-core.s
index 34dff8b..5080d7e 100644
--- a/mandel-core.s
+++ b/mandel-core.s
@@ -563,12 +563,13 @@ sqr16_patch_offset = 8
 ; input: arg as u8
 ; output: dest as u16
 ; clobbers a, x
+; 19 cyc
 .macro sqr8 dest, arg
-    ldx arg
-    lda sqr_lobyte,x
-    sta dest
-    lda sqr_hibyte,x
-    sta dest + 1
+    ldx arg           ; 3 cyc
+    lda sqr_lobyte,x  ; 5 cyc
+    sta dest          ; 3 cyc
+    lda sqr_hibyte,x  ; 5 cyc
+    sta dest + 1      ; 3 cyc
 .endmacro
 
 .segment "TABLES"
@@ -770,6 +771,10 @@ inner_loop:
 
 .endproc
 
+; min low-mem: 81 * 4 + 28 * 2 + 10 + 6 = 396 cyc
+; max low-mem: 92 * 4 + 28 * 2 + 50 + 6 = 480 cyc
+; min ext-mem: 51 * 4 + 28 * 2 + 10 + 6 = 276 cyc
+; max low-mem: 70 * 4 + 28 * 2 + 50 + 6 = 392 cyc
 .macro imul16_impl xe
     .local arg1
     .local arg2
@@ -787,32 +792,37 @@ 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
+    imul8 result, arg1, arg2, xe        ; 81-92 or 51-70
 
-    imul8 result + 2, arg1 + 1, arg2 + 1, xe
+    imul8 result + 2, arg1 + 1, arg2 + 1, xe ; 81-92 or 51-70
 
-    imul8 inter, arg1 + 1, arg2, xe
-    add16 result + 1, result + 1, inter
-    add_carry result + 3
+    imul8 inter, arg1 + 1, arg2, xe     ; 81-92 or 51-70
+    add16 result + 1, result + 1, inter ; 20 cyc
+    add_carry result + 3                ; 8 cyc
 
-    imul8 inter, arg1, arg2 + 1, xe
-    add16 result + 1, result + 1, inter
-    add_carry result + 3
+    imul8 inter, arg1, arg2 + 1, xe     ; 81-92 or 51-70
+    add16 result + 1, result + 1, inter ; 20 cyc
+    add_carry result + 3                ; 8 cyc
 
     ; In case of negative inputs, adjust high word
     ; https://stackoverflow.com/a/28827013
-    lda arg1 + 1
-    bpl arg1_pos
-    sub16 result + 2, result + 2, arg2
+    ; 10-50 cycles
+    lda arg1 + 1                       ; 3 cyc
+    bpl arg1_pos                       ; 2 cyc
+    sub16 result + 2, result + 2, arg2 ; 20 cyc
 arg1_pos:
-    lda arg2 + 1
-    bpl arg2_pos
-    sub16 result + 2, result + 2, arg1
+    lda arg2 + 1                       ; 3 cyc
+    bpl arg2_pos                       ; 2 cyc
+    sub16 result + 2, result + 2, arg1 ; 20 cyc
 arg2_pos:
 
     rts ; 6 cyc
 .endmacro
 
+; min base ram: 19 * 2 + 81 + 28 * 2 + 6 = 181
+; max base ram: 19 * 2 + 92 + 28 * 2 + 6 = 192
+; min ext ram: 19 * 2 + 51 + 28 * 2 + 6 = 151
+; max ext ram: 19 * 2 + 70 + 28 * 2 + 6 = 170
 .macro sqr16_impl xe
     .scope
         arg = FR0    ; 16-bit arg (clobbered)
@@ -820,9 +830,9 @@ arg2_pos:
         ;inter = temp2
         inter = FR1
 
-        lda arg + 1
-        bpl arg_pos
-        neg16 arg
+        lda arg + 1  ; 3 cyc
+        bpl arg_pos  ; 2 cyc
+        neg16 arg    ; 18 cyc
     arg_pos:
 
         ; hl * hl
@@ -830,15 +840,15 @@ arg2_pos:
         ; h*256*(h*256 + l) + l*(h*256 + l)
         ; h*h*256*256 + h*l*256 + h*l*256 + l*l
 
-        sqr8 result, arg
+        sqr8 result, arg          ; 19 cyc
 
-        sqr8 result + 2, arg + 1
+        sqr8 result + 2, arg + 1  ; 19 cyc
 
-        imul8 inter, arg + 1, arg, xe
-        add16 result + 1, result + 1, inter
-        add_carry result + 3
-        add16 result + 1, result + 1, inter
-        add_carry result + 3
+        imul8 inter, arg + 1, arg, xe        ; 81-92 / 51-70 cyc
+        add16 result + 1, result + 1, inter  ; 20 cyc
+        add_carry result + 3                 ; 8 cyc
+        add16 result + 1, result + 1, inter  ; 20 cyc
+        add_carry result + 3                 ; 8 cyc
 
         rts ; 6 cyc
     .endscope