step 2: break out the multiplication hotspot

mandelbrot_hotspot_impl macro calls through to the
sqr16_impl and imul16_impl macros with the xe mode

this is backed by two realized functions, as
mandelbrot_hotspot and mandelbrot_hotspot_xe

these are also called via direct jmp instead of jsr/rts
because there's only one call site so we can save 6 cycles
per iteration by jmp/jmp

could save 6 more cycles per iter by specializing all of
the mandelbrot proc but there isn't room in ram right now

sqr16_func is removed as it is unused and we ran out of
code space adding the hotspot's extra implementations

imul16_func is kept, as it's called via zoom_factor
in a couple of places. this forwards to the xe version
at a cost of 3 cycles, as the call sites aren't patched

mandel-core.s

@@ -558,8 +558,6 @@ input_max:
 
 ; input: arg1, arg2 as fixed4.12
 ; output: dest as fixed8.24
-; patch point jsr at 16 bytes in
-imul16_patch_offset = 16
 .macro imul16 dest, arg1, arg2
     copy16 FR0, arg1  ; 12 cyc
     copy16 FR1, arg2  ; 12 cyc
@@ -569,13 +567,11 @@ imul16_patch_offset = 16
 
 ; input: arg as fixed4.12
 ; output: dest as fixed8.24
-; patch point jsr at 8 bytes in
-sqr16_patch_offset = 8
-.macro sqr16 dest, arg
-    copy16 FR0, arg   ; 12 cyc
-    jsr sqr16_func    ; ? cyc
-    copy32 dest, FR2  ; 24 cyc
-.endmacro
+;.macro sqr16 dest, arg
+;    copy16 FR0, arg   ; 12 cyc
+;    jsr sqr16_func    ; ? cyc
+;    copy32 dest, FR2  ; 24 cyc
+;.endmacro
 
 ; input: arg as u8
 ; output: dest as u16
@@ -799,15 +795,15 @@ inner_loop:
 
 .endproc
 
-; 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: 50 * 4 + 22 * 2 + 10 + 6 = 260 cyc
-; max ext-mem: 55 * 4 + 27 * 2 + 50 + 6 = 330 cyc
-.macro imul16_impl xe
+; min base mem: 81 * 4 + 22 * 2 + 10 = 378 cyc
+; max base mem: 92 * 4 + 27 * 2 + 50 = 472 cyc
+; min ext-mem: 50 * 4 + 22 * 2 + 10 = 254 cyc
+; max ext-mem: 55 * 4 + 27 * 2 + 50 = 324 cyc
+.macro imul16_impl result, arg1, arg2, xe
     .scope
-        arg1 = FR0    ; 16-bit arg
-        arg2 = FR1    ; 16-bit arg
-        result = FR2  ; 32-bit result (output)
+        ;arg1         ; 16-bit arg
+        ;arg2         ; 16-bit arg
+        ;result       ; 32-bit result (output)
         inter = temp2 ; 16-bit temporary (clobbered)
 
         ; h1l1 * h2l2
@@ -843,8 +839,6 @@ inner_loop:
         bpl arg2_pos                       ; 2 cyc
         sub16 result + 2, result + 2, arg1 ; 20 cyc
     arg2_pos:
-
-        rts ; 6 cyc
     .endscope
 .endmacro
 
@@ -877,15 +871,15 @@ inner_loop:
 .endmacro
 
 
-; min base ram: 5 + 163 + 6 = 174
-; max base ram: 7 + 184 + 18 + 6 = 215
-; min ext ram: 5 + 132 + 6 = 143
-; max ext ram: 7 + 147 + 18 + 6 = 178
-.macro sqr16_impl xe
+; min base ram: 5 + 163 + 3 = 171
+; max base ram: 7 + 184 + 18 = 209
+; min ext ram: 5 + 132 + 3 = 140
+; max ext ram: 7 + 147 + 18 = 172
+.macro sqr16_impl result, arg, xe
     .scope
-        arg = FR0    ; 16-bit arg
+        ; arg        ; 16-bit arg
         negated = FR1
-        result = FR2 ; 32-bit result
+        ; result     ; 32-bit result
 
         ; 5-7 cycles
         lda arg + 1                ; 3 cyc
@@ -894,30 +888,35 @@ inner_loop:
 
     arg_pos:
         sqr16_impl_inner result, arg, xe      ; 163-184 / 132-147 cyc
-        rts                                   ; 6 cyc
+        jmp done                              ; 3 cyc
 
     arg_neg:
         copy_neg16 negated, arg               ; 18 cyc
         sqr16_impl_inner result, negated, xe  ; 163-184 / 132-147 cyc
-        rts                                   ; 6 cyc
+
+    done:
     .endscope
 .endmacro
 
 .proc imul16_func
-    imul16_impl 0
+    imul16_impl FR2, FR0, FR1, 0
+    rts
 .endproc
 
 .proc imul16xe_func
-    imul16_impl 1
+    imul16_impl FR2, FR0, FR1, 1
+    rts
 .endproc
 
-.proc sqr16_func
-    sqr16_impl 0
-.endproc
+;.proc sqr16_func
+;    sqr16_impl FR2, FR0, 0
+;    rts
+;.endproc
 
-.proc sqr16xe_func
-    sqr16_impl 1
-.endproc
+;.proc sqr16xe_func
+;    sqr16_impl FR2, FR0, 1
+;    rts
+;.endproc
 
 ; 11-27 cycles
 .macro round16 arg
@@ -1157,16 +1156,11 @@ keep_going:
     shift_round_16 zy, 3
 
     ; zx_2 = zx * zx
-fixup_sqr16_1:
-    sqr16 zx_2, zx + 2
-
     ; zy_2 = zy * zy
-fixup_sqr16_2:
-    sqr16 zy_2, zy + 2
-
     ; zx_zy = zx * zy
-fixup_imul16_1:
-    imul16 zx_zy, zx + 2, zy + 2
+fixup_mandelbrot_hotspot:
+    jmp mandelbrot_hotspot
+after_mandelbrot_hotspot:
 
     ; dist = zx_2 + zy_2
     add32 dist, zx_2, zy_2
@@ -1262,6 +1256,27 @@ next:
 
 .endproc
 
+.macro mandelbrot_hotspot_impl xe 
+    ; zx_2 = zx * zx
+    sqr16_impl zx_2, zx + 2, xe
+
+    ; zy_2 = zy * zy
+    sqr16_impl zy_2, zy + 2, xe
+
+    ; zx_zy = zx * zy
+    imul16_impl zx_zy, zx + 2, zy + 2, xe
+
+    jmp mandelbrot::after_mandelbrot_hotspot
+.endmacro
+
+.proc mandelbrot_hotspot
+    mandelbrot_hotspot_impl 0
+.endproc
+
+.proc mandelbrot_hotspot_xe
+    mandelbrot_hotspot_impl 1
+.endproc
+
 .macro scale_zoom dest
     ; clobbers X, flags
     .local cont
@@ -2189,22 +2204,21 @@ init:
     sta imul16_func
     lda #.lobyte(imul16xe_func)
     sta imul16_func + 1
-    sta mandelbrot::fixup_imul16_1 + imul16_patch_offset + 1
     lda #.hibyte(imul16xe_func)
     sta imul16_func + 2
-    sta mandelbrot::fixup_imul16_1 + imul16_patch_offset + 2
 
     ; ditto for sqr16_func -> sqr16xe_func
-    lda #$4c ; 'jmp' opcode
-    sta sqr16_func
-    lda #.lobyte(sqr16xe_func)
-    sta sqr16_func + 1
-    sta mandelbrot::fixup_sqr16_1 + sqr16_patch_offset + 1
-    sta mandelbrot::fixup_sqr16_2 + sqr16_patch_offset + 1
-    lda #.hibyte(sqr16xe_func)
-    sta sqr16_func + 2
-    sta mandelbrot::fixup_sqr16_1 + sqr16_patch_offset + 2
-    sta mandelbrot::fixup_sqr16_2 + sqr16_patch_offset + 2
+    ;lda #$4c ; 'jmp' opcode
+    ;sta sqr16_func
+    ;lda #.lobyte(sqr16xe_func)
+    ;sta sqr16_func + 1
+    ;lda #.hibyte(sqr16xe_func)
+    ;sta sqr16_func + 2
+
+    lda #.lobyte(mandelbrot_hotspot_xe)
+    sta mandelbrot::fixup_mandelbrot_hotspot + 1
+    lda #.hibyte(mandelbrot_hotspot_xe)
+    sta mandelbrot::fixup_mandelbrot_hotspot + 2
 
 
     ; create the lookup table