Revision as of 10:26, 16 April 2012

This artikel originally came from Kevin Thackers' archive at http://www.cpctech.org.uk.

Technical information about Locomotive BASIC

The are two versions of Locomotive BASIC used in the Amstrad CPC and CPC+ computers. v1.0 is used by the CPC464, and v1.1 is used by the CPC664, CPC6128, CPC464+ and CPC6128+.

Passing parameters to a RSX (Resident System Extension) command or binary function

A RSX (Resident System Extension) command is accessed through BASIC with a "|" character prefix. e.g. The "DIR" RSX, is accessed by "|DIR". A binary function is accessed through BASIC using the "CALL" command.

Both RSX and CALL commands works (are!) similar from BASIC command line and invokes a machine code - the only difference is: with the help of a RSX command you don't need to know the exact access address. You can pass up to 32 parameters with a CALL or RSX command. Possible parameters could integer, floating points and strings. Just store the parameter into a variable: For integer:

CALL &4000,32,34,&5E,&x10101010
|RSX-command,32,34,&5E,&x10101010

For floating points

@a!=43.375
@b$="test"
CALL or |RSX-command,@a!
CALL or |RSX-command,@b$

The parameters will be stored inside the range of Stack:

     A=n -------> MSB PARAM1
                  LSB PARAM1
                  MSB PARAM2
                  LSB PARAM2
                  ...
                  ...
                  ...
                  MSB PARAMn
     IX+2 ------> LSB PARAMn

with the help of that assembly code is it possible to find the parameters:

ROUT1:  CP   "x"          ;test register A if "x" parameters were hand-overed, e.g. "x" = 2
        RET  NZ         ; if no > failure and return.
        LD   L,(IX+2)
        LD   H,(IX+3)   ; HL = value of last parameter        
        LD   E,(IX+4)
        LD   D,(IX+5)   ; DE = value of next to last parameter
        ...

If you hand-over a string than the first two bytes holds the length of the string.

Hint: the first two bytes at the end of stack shows the internal return address from BASIC interpreter where it was executed.

Passing a string parameter:

BASIC v1.0

a$="A":|DRIVE,@a$

BASIC v1.1:

|DRIVE,"A"

Additional keywords and variables in BASIC v1.1

1. BASIC v1.1 has the following additional keywords:
   • COPYCHR$
   • CURSOR
   • DEC$
   • FILL
   • FRAME
   • GRAPHICS
   • MASK
2. BASIC v1.1 has the following additional predefined variables:
   • DERR

Entering BASIC programs

A maximum of 255 characters can be entered for a single BASIC line. This is converted into the tokenised BASIC program which is more compact than the line entered. The BASIC keywords are converted into "tokens", or 1-2 byte sequences which uniquely identify each keyword.

Structure of a BASIC program

A BASIC program is stored in a tokenised format. Here the keywords are represented by 1 or 2 byte unique sequences. Each line of the BASIC program has the form:

Notes:

1. • This 16-bit value has two functions:
     • if "0" it signals the end of the BASIC program. In this case, there is no furthur BASIC program lines or data.
     • otherwise, this value defines the length of the tokenised BASIC program line in bytes, and includes the 2 bytes defining this value, the 2 bytes defining the program line number, and the end of line marker. This number is stored in little endian notation with low byte followed by high byte.
   • This 16-bit value defines the line number and exists if the length of line data in bytes is not "0". A line number is a integer number in the range 1-65535. This number is stored in little endian notation with low byte followed by high byte.
   • This data defines the tokenised BASIC program line and exists if the length of line data in bytes is not "0". The length is dependant on the BASIC program line contents.
   • This value defines the end of the tokenised BASIC line data and exists if the length of line data in bytes is not "0". The BASIC interpreter looks for this token during the processing of this line, and if found, will stop execution and continue to the next line.

BASIC tokens

This table list the BASIC tokens with no prefix byte.

Notes:

1. • The &ff code is used as a prefix for more keywords. See the table below for the list of keywords using this prefix.
   • &e2,&e8 and &e9 are not used
   • &7c ("|") is a character prefix used to identify a RSX command.

e.g. "|DIR". An RSX is encoded in a BASIC program using the following structure:

1. • This token identifies a integer (16-bit) number. The two bytes following this token is the number, stored low byte then high byte. e.g. &19,&2d,&00 represents the integer number "&002d"="45".
   • This token (&0B) is changed at run-time to &0D
   • This token identifies a integer variable. 02,00,00,var name offset to number in program setup when program RUN. The variable name is stored directly, with bit 7 of the last character set.
   • This token identifies a string variable. offset to string in program string stored AS IS with quotes around it.
   • This token identifies a floating point number. Immediatly following the token are 5 bytes which define the number.

1. • 1. This byte is the BASIC token identifying a floating point number
     2. These 5 bytes define the floating point number.
   • The space symbol is used as a command seperator
   • Variable definitions:
     • The variable definition has the following structure:

The 16-bit byte offset is initially set to 0, but is setup when the program is RUN. The offset is stored in little-endian format, with low byte followed by high byte.

1. • • The token byte defines the variable type (i.e. string, floating point, integer) and the suffix ("$", none and "%").
     • The variable name is stored in the program, with bit 7 of the last character set to "1". e.g. The variable named "abc" will be encoded as 'a','b','c'+&80
   • String values:
     • Each string value is prefixed with &22 token.
     • All strings must be enclosed by double quote symbols (&quot). This symbols define the start and end of the string value.
     • The string value can contain any symbol except the quote ("). i.e. 0-&21, and &23-&ff can be used.
   • After running a program, Tokens of "1d" with 16-bit BASIC program line pointers are changed to token "1e" with the 16-bit memory address of the BASIC line number in memory.

This table list the BASIC tokens with a &ff prefix byte.

NOTES:

1. • These codes are prefixed by &FF. e.g. The keyword "ABS" is stored as the following sequence:

     &FF,&00

   • Codes &1e...&3f inclusive are not used
   • Codes &4a...&6f inclusive are not used
   • Codes &80...&ff inclusive are not used

Floating Point data definition

A floating point number represents a number with both an integer and fractional part.

In Amstrad BASIC, a floating point number is stored in base-2 in a normalized form: 1 x 2^<exponent> The representation uses 5 bytes and is stored using the following structure:

Sign

The sign is represented by bit 7 of byte 3. The value of the sign bit indicates the sign of the floating point number.

1. • If the sign bit is "1", then the number is negative.
   • If the sign bit is "0", then the number is positive.

Mantissa

The mantissa holds the normalized number. The exponent is manipulated so that the most significant bit is always 1. Since it is always 1, it is not stored. It's value is implied by the representation.

Exponent

The exponent is 8-bit and is stored in byte 4. It is stored with a bias of 128.

1. • 128-255 are positive exponents,
   • 0-127 are negative exponents.

To obtain the signed exponent, you must subtract 128 from the stored exponent. The minimum exponent is 0 and this describes a number of 2^-127. The maximum exponent is 255 and this describes a number of 2^128.

floating point example in RAM by MaV

The initial figure = 43.375. An example in BASIC:

a!=43.375
READY
PRINT a!
 43.375
PRINT @a!
 374
FOR I=0 TO 4:PRINT HEX$(PEEK(374+I),2);:NEXT I
0000802D86

the result for better readability:
00 00 80 2D 86
The first byte contains the least significant bits of the mantissa, and every following byte up to the fourth byte contain the bits following the first in significance and in exactly that order!

That means that you have to turn around the first four bytes if you want to read the mantissa correctly: 2D800000 or 0010.1101.1000.0000.0000.0000.0000.0000 in binary (the stops are inserted for better readability).

Now the highest significant bit here is 0, so the number is positive. What's more, you need to clear that bit and substitute the implied highest bit (see description in the link), which is always 1 (regardless of the deleted sign bit!!!). Thus you get: AD800000 or 1010.1101.1000.0000.0000.0000.0000.0000 in binary.

The mantissa has an implied decimal point before the number. Since the number is 2^32 bits (i.e. 4 bytes) long you have to divide by 2^32 to get the proper decimal representation.

AD800000 (hex) = 2910846976 (dec)

2910846976 / 2^32 = 0.677734375

The mantissa in a decimal representation is: 0.677734375

Now, the last byte contains the exponent: 86 (hex) or 134 (dec). You need to subtract 128 to get the real exponent, the highest bit tells you that the exponent is positive (1) or negative (0). Your exponent is 6.

0.677734375 * 2^6 = 43.375

Another example:
PI is represented in memory as:
A2 DA 0F 49 82
The mantissa is
490FDAA2 (highest significant bit is zero => number is positive) The exponent is
82 (=positive mantissa, value=2 thus multiply mantissa with 2^2)

Delete the sign bit (which is 0 anyway), and then add the implied 1 bit:
C90FDAA2 (hex) = 3373259426 (dec) = 1100.1001.0000.1111.1101.1010.1010.0010 (bin)

Calculate the decimal representation of the mantissa:
3373259426 / 2^32 = 0,7853981633670628070831298828125

Then add the exponent (82h - 80h = 2 thus 2^2)
0,7853981633670628070831298828125 * 2^2 = 3,14159265346825122833251953125

Voilà! The internal representation of the number PI on the CPC is: 3,14159265346825122833251953125

If you compare to a (more) correct representation of PI, you'll see that the CPC is correct up to the 9th decimal place:

CPC:
3.14159265346825122833251953125
more correctly:
3.1415926535897932384626433832795028841971693993751058209749445923078164062862089986...

BASIC floating-point/real variables

A floating point (real) variable describes a number with integer and fractional parts.

1. • A integer variable is defined with a "!" character postfix.

e.g.

a! = 3.141592654

1. • A real number uses 5 bytes of memory as storage. The format is described above.
   • The address of a real variable can be found in BASIC by:

     PRINT @a!

Where "a" should be replaced with the name of the variable.

BASIC integer variables

A integer variable describes a whole number. i.e. a number without any fractional part.

NOTES:

1. • A integer variable is defined with a "%" character postfix.

e.g.

a% = 3

1. • A integer variable has a range of -32768 to 32767
   • A integer variable always uses two bytes of memory as storage with the following format:

1. • The address of the integer variable can be found in BASIC by:

     PRINT @a%

where "a" should be replaced with the name of the variable.

BASIC string variables

A string is a variable which contains a group of characters.

NOTES:

1. • • A string variable is defined in BASIC with a "$" character postfix.

e.g.

a$ = "hello"

1. • • A string variable is described internally by a "string descriptor block" which has the following structure:

1. • • The address of the "string descriptor block" for the string variable can be found in BASIC by typing:

       PRINT @a$

replacing "a$" with the name of the string variable.

1. • • A string uses a minimum of 3 bytes of storage (a 0 character length string) and a maximum of 258 bytes of storage. 3 bytes are used by the "string descriptor block")
     • A string variable can contain any ASCII character code in the range 0-255.
     • A string variable can store a minimum of 0 characters and a maximum of 255 characters.
     • The length of the string in characters is defined in the string descriptor block. The string does not have a termination character.