implemented CONUPK math routine

c64/basic/,stubs_math/ba8c.CONUPK.s

@@ -10,7 +10,11 @@
 // - .Y - address high byte
 //
 // Output:
-// - .A - FAC1 exponent (XXX really FAC1?)
+// - .A - FAC1 exponent (yes, really - Codebase64 is right)
+// - .Y - 0 (tested with original ROMs)
+//
+// Preserves:
+// - .X (tested with original ROMs)
 //
 // See also:
 // - [CM64] Computes Mapping the Commodore 64 - page 114
@@ -18,7 +22,61 @@
 // - https://codebase64.org/doku.php?id=base:kernal_floating_point_mathematics
 //
 
-// XXX provide implementation
+// XXX test this
 
 CONUPK:
-	STUB_IMPLEMENTATION()
+
+	// Noticed in the avilable materials:
+	// - https://www.c64-wiki.com/wiki/Floating_point_arithmetic
+	// - https://www.lemon64.com/forum/viewtopic.php?t=67576&sid=d41e77cdddd28d6f9c9d52ce5b4e8dc3
+	// that at least some math routines use $22/$23 (INDEX+0 and INDEX+1) for pointer storage,
+	// confirmed that this routine is not an exception
+
+	sty INDEX+1
+	sta INDEX+0
+
+	// Now copy the data from RAM to FAC2, for the format description see:
+	// - https://www.c64-wiki.com/wiki/Floating_point_arithmetic#Representation_in_the_C-64
+
+	ldy #$04
+
+	// Copy the mantissa - reverse byte order
+
+	ldy #$04
+	lda (INDEX), y
+	sta FAC2_mantissa+3
+
+	dey
+	lda (INDEX), y
+	sta FAC2_mantissa+2
+
+	dey
+	lda (INDEX), y
+	sta FAC2_mantissa+1
+
+	dey
+	lda (INDEX), y
+	ora #$80                           // clear the sign bit
+	sta FAC2_mantissa+0
+
+	// Copy the sign
+
+	lda (INDEX), y
+	bpl !+                             // assumption: non-negative numbers are more frequent
+	lda #$FF
+	skip_2_bytes_trash_nvz
+!:
+	lda $00
+
+	sta FAC2_sign
+
+	// Copy the exponent
+
+	dey
+	lda (INDEX), y
+	sta FAC2_exponent
+
+	// Return the FAC1 exponent
+
+	lda FAC1_exponent
+	rts

c64/basic/math/ba8c.CONUPK.s

@@ -0,0 +1,82 @@
+// #LAYOUT# STD *       #TAKE
+// #LAYOUT# *   BASIC_0 #TAKE
+// #LAYOUT# *   *       #IGNORE
+
+//
+// Math package - fetch FAC2 from RAM location
+//
+// Input:
+// - .A - address low byte
+// - .Y - address high byte
+//
+// Output:
+// - .A - FAC1 exponent (yes, really - Codebase64 is right)
+// - .Y - XXX 0?
+//
+// Preserves:
+// - .X - (experimentation with original routine)
+//
+// See also:
+// - [CM64] Computes Mapping the Commodore 64 - page 114
+// - https://www.c64-wiki.com/wiki/Floating_point_arithmetic
+// - https://codebase64.org/doku.php?id=base:kernal_floating_point_mathematics
+//
+
+// XXX test this
+
+CONUPK:
+
+	// Not sure which temporary variables are used for storage by the original ROMs,
+	// but $22/$23 (INDEX+0 and INDEX+1) seem to be good candidates, at least some
+	// math routines use them as pointer storage, see:
+	// - https://www.c64-wiki.com/wiki/Floating_point_arithmetic
+	// - https://www.lemon64.com/forum/viewtopic.php?t=67576&sid=d41e77cdddd28d6f9c9d52ce5b4e8dc3
+
+	sty INDEX+1
+	sta INDEX+0
+
+	// Now copy the data from RAM to FAC2, for the format description see:
+	// - https://www.c64-wiki.com/wiki/Floating_point_arithmetic#Representation_in_the_C-64
+
+	ldy #$04
+
+	// Copy the mantissa - reverse byte order
+
+	ldy #$04
+	lda (INDEX), y
+	sta FAC2_mantissa+3
+
+	dey
+	lda (INDEX), y
+	sta FAC2_mantissa+2
+
+	dey
+	lda (INDEX), y
+	sta FAC2_mantissa+1
+
+	dey
+	lda (INDEX), y
+	ora #$80                           // clear the sign bit
+	sta FAC2_mantissa+0
+
+	// Copy the sign
+
+	lda (INDEX), y
+	bpl !+                             // assumption: non-negative numbers are more frequent
+	lda #$FF
+	skip_2_bytes_trash_nvz
+!:
+	lda $00
+
+	sta FAC2_sign
+
+	// Copy the exponent
+
+	dey
+	lda (INDEX), y
+	sta FAC2_exponent
+
+	// Return the FAC1 exponent
+
+	lda FAC1_exponent
+	rts