implemented ROUND math routine, corrected some others

STATUS.md

@@ -196,6 +196,7 @@ Floating point mathematical routines - not official, but well known and broadly
 | `$BBD4`   | `MOVMF`      | NOT DONE | copy FAC1 to memory location, rounds if asked      |
 | `$BBFC`   | `MOVEF`      | NOT DONE | copy FAC2 to FAC1                                  |
 | `$BC0F`   | `MOVFA`      | NOT DONE | copy FAC1 to FAC2, skips rounding                  |
+| `$BC1B`   | `ROUND`      | PARTIAL  | not tested                                         |
 | `$BC2B`   | `SIGN`       | NOT DONE | evaluate FAC1 sign                                 |
 | `$BC39`   | `SGN`        | NOT DONE | evaluate FAC1 sign, to FAC1                        |
 | `$BC58`   | `ABS`        | NOT DONE | absolute value                                     |

c64/basic/,stubs_math/bb12.interop

@@ -1,3 +1,4 @@
 This is an important floating point math routine, documented in many places.
 
-XXX - try to find real-life example
+Example code utilizing the routine:
+- https://www.lemon64.com/forum/viewtopic.php?t=67576&sid=d41e77cdddd28d6f9c9d52ce5b4e8dc3

c64/basic/,stubs_math/bcf3.interop

@@ -1,3 +1,4 @@
 This is an important floating point math routine, documented in many places.
 
-XXX - try to find real-life example
+Example code utilizing the routine:
+- https://www.lemon64.com/forum/viewtopic.php?t=67576&sid=d41e77cdddd28d6f9c9d52ce5b4e8dc3

c64/basic/,stubs_math/bddd.interop

@@ -1,3 +1,4 @@
 This is an important floating point math routine, documented in many places.
 
-XXX - try to find real-life example
+Example code utilizing the routine:
+- https://www.lemon64.com/forum/viewtopic.php?t=67576&sid=d41e77cdddd28d6f9c9d52ce5b4e8dc3

c64/basic/math/CONUPK.s

@@ -38,7 +38,7 @@ CONUPK:
 	// 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
 
-	// Copy the mantissa - reverse byte order
+	// Copy the mantissa
 
 	ldy #$04
 	lda (INDEX), y

c64/basic/math/MOVFM.s

@@ -35,7 +35,7 @@ MOVMF:
 	// 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
 
-	// Copy the mantissa - reverse byte order
+	// Copy the mantissa
 
 	ldy #$04
 	lda (INDEX), y

c64/basic/math/NORMAL.s

@@ -44,10 +44,10 @@ normal_by_bit:
 
 	clc
 	rol FACOV
-	rol FAC1_mantissa+0
-	rol FAC1_mantissa+1
-	rol FAC1_mantissa+2
 	rol FAC1_mantissa+3
+	rol FAC1_mantissa+2
+	rol FAC1_mantissa+1
+	rol FAC1_mantissa+0
 	
 	bcc NORMAL                         // branch if next iteration needed
 	rts

c64/basic/math/ROUND.s

@@ -0,0 +1,61 @@
+// #LAYOUT# STD *       #TAKE
+// #LAYOUT# *   BASIC_0 #TAKE
+// #LAYOUT# *   *       #IGNORE
+
+//
+// Math package - rounds FAC1 before copying
+//
+// See also:
+// - [CM64] Computes Mapping the Commodore 64 - page 115
+// - https://codebase64.org/doku.php?id=base:kernal_floating_point_mathematics
+//
+
+// XXX test this
+
+ROUND:
+
+	// If number is 0, do not do anything
+
+	lda FAC1_exponent
+	beq round_done
+
+	lda FACOV
+	bpl round_zero_FACOV                 // branch if it is enough to clear FACOV
+
+	// Increment the 4 bytes of matissa
+
+	// Curse you, who invented this format! Had the mantissa been little endian,
+	// we could have used INW on 65C02 and above, the code would have been smaller!
+
+	inc FAC1_mantissa+3
+	bne round_zero_FACOV
+	inc FAC1_mantissa+2
+	bne round_zero_FACOV
+	inc FAC1_mantissa+1
+	bne round_zero_FACOV
+	inc FAC1_mantissa+0
+	bne round_zero_FACOV
+
+	// If we are here, it means mantissa is 0; put it to minimum allowed and increment exponent
+
+	lda #$80
+	sta FAC1_mantissa+0
+	inc FAC1_exponent
+	bne round_zero_FACOV
+
+	// If w are here, it means the exponent wrapped around to $00; put it back to $FF
+
+	dec FAC1_exponent
+
+	// FALLTROUGH
+
+round_zero_FACOV:
+
+	lda #$00
+	sta FACOV
+
+	// FALLTROUGH
+
+round_done:
+
+	rts

c64/basic/math/bc1b.ROUND.jmp.s

@@ -0,0 +1,7 @@
+// #LAYOUT# STD *       #TAKE
+// #LAYOUT# *   BASIC_0 #TAKE
+// #LAYOUT# *   *       #IGNORE
+
+// Just a compatibility jump - our routine is longer than original one
+
+	jmp ROUND

c64/basic/math/bc1b.interop

@@ -0,0 +1,4 @@
+This is an important floating point math routine, documented in many places.
+
+Example code utilizing the routine:
+- https://www.lemon64.com/forum/viewtopic.php?t=67576&sid=d41e77cdddd28d6f9c9d52ce5b4e8dc3