What does the Commodore PET BASIC assemble starting from $00C2?

Question

Zero-page memory maps of the PET that I've found claim that the zero page address range$00C2..$00D9 are used for static data, e.g. http://www.classiccmp.org/dunfield/pet/petmem.txt says:

 RIDATA 00C2        Cassette Temp (64#00AA) read flags: 0=scan,
                    1-15=count, $40=load, $80=end of tape marker
 RIPRTY 00C3        Cassette Short Cnt (64#00AB): counter of seconds
                    before tape write / checksum
 PNT    00C4-00C5   Pointer: Current Screen Line Address
 PNTR   00C6        Cursor Column on Current Line
 SAL    00C7-00C8   Pointer: Tape Buffer/ Screen Scrolling
 EAL    00C9-00CA   Tape End Addresses/End of Program
 CMP0   00CB-00CC   Tape Timing Constants
 QTSW   00CD        Flag: Editor in Quote Mode, $00 = NO
 BITTS  00CE        Cassette Temp (64#00B4): Tape read timer flag
                    =IRQ enabled for Timer 1
        00CF        End of tape read
        00D0        Read character error
 FNLEN  00D1        Length of Current File Name
 LA     00D2        Current Logical File Number
 SA     00D3        Current Secondary Address
 FA     00D4        Current Device Number
 LNMX   00D5        Physical Screen Line Length
        00D5        4.80: right side of window
 TAPE1  00D6-00D7   Pointer: Start of Tape Buffer
 TBLX   00D8        Current Cursor Physical Line Number
 DATAX  00D9        Current Character to Print

However, looking at the ROM disassembly, one can find places where the address $00C2 is jumped to, e.g. http://www.zimmers.net/anonftp/pub/cbm/firmware/computers/pet/d/rom-1.html#C70A :

 C70A  4C C2 00             JMP iC2       

Looking at a disassembly starting at $00C2 after booting up a PET, I can see reasonable-looking code:

.C:00c2  E6 C9       INC $C9
.C:00c4  D0 02       BNE $00C8
.C:00c6  E6 CA       INC $CA
.C:00c8  AD 00 04    LDA $0400
.C:00cb  C9 3A       CMP #$3A
.C:00cd  B0 0A       BCS $00D9
.C:00cf  C9 20       CMP #$20
.C:00d1  F0 EF       BEQ $00C2
.C:00d3  38          SEC
.C:00d4  E9 30       SBC #$30
.C:00d6  38          SEC
.C:00d7  E9 D0       SBC #$D0
.C:00d9  60          RTS

What is this area used for? Where is the code that assembles this program into this area? What's this code supposed to do? (It seems to be scanning the area starting at $0400 for : and characters?)

Answer 1

It's part of the BASIC interpreter loop. It reads one byte of the tokenized BASIC program, setting zero flag if it's a colon or a zero byte and clearing the carry if it's a number. You can see it used in the main part of the interpreter loop at address C6B5.

I'm not sure why this routine was placed in zero page. It's a cycle (or rarely two) faster to use LDA $0400 over LDA ($C9),Y, but I can't see it actually making much difference.

I should also note that the ROM disassembly you're looking at appears to be for a BASIC 1.0 ROM, while the memory map you've referenced is for versions 2.0 and 4.0.

Here's what Mapping the Commodore 64 by Sheldon Leemon says about the equivalent C64 routine:

115-138   $73-$8A   CHRGET
Subroutine: Get Next BASIC Text Character

...

CHRGET is a crucial routine which BASIC uses to read text characters, such as the text of the BASIC program which is being interpreted. It is placed on zero page to make the routine run faster. Since it keeps track of the address of the character being read within the routine itself, the routine must be in RAM in order to update that pointer. The pointer to the address of the byte currently being read is really the operand of a LDA instruction. When entered from CHRGET, the routine increments the pointer by modifying the operand at TXTPTR (122, $7A), thus allowing the next character to be read.

Entry at CHRGOT (121, $79) allows the current character to be read again. The CHRGET routine skips spaces, sets the various flags or the status register (.P) to indicate whether the character read was a digit, statement terminator, or other type of character, and returns with the retrieved character in the Accumulator (.A).

...

As this is such a central routine, a disassembly listing is given below to provide a better understanding of how it works.

115 $73   CHRGET  INC TXTPTR   ; increment low byte of TXTPTR
117 $75           BNE CHRGOT   ; if low byte isn't 0, skip next
119 $77           INC TXTPTR+1 ; increment high byte of TXTPTR
121 $79   CHRGOT  LDA          ; load byte from where TXTPTR points
                               ; entry here does not update TXTPTR,
                               ; allowing you to readl the old byte again
122 $7A   TXTPTR  $0207        ; pointer is really the LDA operand
                               ; TXTPTR+1 points to 512-580 ($200-$250)
                               ; when reading from the input buffer
                               ; in direct mode
124 $7C   POINTB  CMP #$3A     ; carry flag set if > ASCII numeral 9
126 $7E           BCS EXIT     ; character is not a numeral--exit
128 $80           CMP #$20     ; if it is an ASCII space...
130 $82           BEQ CHRGET   ; ignore it and get next character
132 $84           SEC          ; prepare to subtract
133 $85           SBC #$30     ; ASCII 0-9 are between 48-57 ($30-$39)
135 $87           SEC          ; prepare to subtract again
136 $88           SBC #$D0     ; if < ASCII 0 (57, $39) then carry is set
138 $8A   EXIT    RTS          ; carry is clear only for numeral on return

The Accumulator (.A register) holds the character that was read on exit from the routine. Status register (.P) bits which can be tested for on exit are:

Carry Clear if the character was an ASCII digit 0-9. Carry Set, otherwise. Zero Set only if the character was a statement terminator 0 or an ASCII colon, 58 ($3A). Otherwise, Zero Clear.

Answer 2

It's self modifying code, notice the embedded pointer at $C9. The source for it is at $E0B4, and it's copied to the zero page by code at $E0E5.

Looks like it is indeed scanning text, and classifying the current character. It will set ZF if the character is a space, and CF will be clear if it is a digit. As such it seems useful for converting string to number, but that's just a guess.

Update: An example routine that uses this code is at $C863. It is certainly doing some string-to-number conversion, you can recognize the pattern that calculates result = result * 10 + (current_char - '0') in a loop.