App. A: Machine Code

AMOS Professional Basic includes a range of commands that provide the advanced programmer with total access to the inner workings of the Amiga's hardware. In experienced hands these instructions can be invaluable. But be warned, if you are not completely sure what you are doing, these commands may be lethal for your programs!

All the important hardware routines are built in to AMOS Professional Basic, allowing you to exploit the Amiga's full potential using the simple commands already detailed in this User Guide. In other words, if you prefer to opt for the simple life, you are free to ignore this machine code section altogether. You certainly do not need these functions to create superb computer games.

Converting numbers

Human beings normally count in multiples of ten, based on the number of fingers most of us possess. These "decimal" numbers are expressed by a group of characters from 0 to 9, and the relative position of these "digits" determines whether a character represents the number of ones, or tens, or millions, and so on.

So in the decimal system, the number 1234 is equivalent to:

1*1000 + 2*100 + 3*10 + 4

Computers do not posses ten fingers, but count in a "binary" system instead, which means that each digit can only be in one of two possible states, non-existent or existent. This is represented by either a 0 or a 1.

As with the decimal system, the meaning of a binary digit depends entirely on its position in a binary number. Both systems rely on the fact that the value of digits change depending on their position from right to left in the number. The human system uses a base of ten, but the computer prefers a base of 2. So as you move from right to left, the values of binary numbers increase by a factor of two. In other words, the digit at the extreme right of a binary number represents the number of ones, the next one along represents the number of twos, then fours, eights, and so on.

Examine the number 15. In the decimal system this is depicted as follows:

1*10 + 5

But in binary, the same number is stored like this:

1*8 + 1*4 + 1*2 + 1

Which can be written as follows:



function: convert a decimal value into a string of binary digits

This is the function that converts a decimal number or expression into the equivalent string of binary digits. The binary number that is returned will automatically have a leading % (per cent) character added to it. This character acts as an introduction sign, to indicate that the number which follows it is in binary notation, rather than the standard decimal system.

After the decimal value that is to be converted, an optional number between 1 and 31 can be added which sets the number of digits to be returned in the binary string. If this parameter is omitted, AMOS Professional will express the value in the fewest possible digits, with no leading zeros. Here are a few examples:

Print Bin$(5) Print Bin$(10) Print Bin$(255) X$=Bin$(100) : Print X$

You may enter binary numbers directly, as part of an expression, providing that the % (per cent) character is placed in front of your binary value. Such numbers will be converted into standard decimal notation automatically. For example:

Print %101 Print %1010 Print %11111111 X$=Bin$(100) : Print Val(X$)

Certain functions make use of yet another system of counting. The Hexadecimal system counts in units of 16 rather than ten, so a total of 16 different digits is needed to represent all the different numbers. The digits from 0 to 9 are used as normal, but the digits from 10 to 15 are signified by the letters A to F, as shown in the following table:

Hex digit0123456789A BCDEF


function: convert a decimal value into a string of hexadecimal digits

HEX$ converts numbers from the decimal system into a string of hexadecimal (Hex) digits. The decimal value to be converted is specified in brackets. The hex number that is returned will automatically have a leading $ (dollar) character added to it.

This character acts as an introduction sign, to indicate that the number which follows it is in hexadecimal notation, rather than the standard decimal system.

After the decimal value that is to be converted, an optional number can be added which sets the number of digits to be returned in the hex string. If this parameter is omitted, AMOS Professional will return the value in the fewest possible digits needed to express the hex number. For example:

Print Hex$(100) Print Hex$(100,3)

HEX$ is often used with the COLOUR function to display the precise mixture of Red, Green and Blue components in a particular colour, as follows:

Print Hex$(Colour(2))

Hexadecimal notation is ideal for handling large numbers such as addresses, and it may be entered directly in any AMOS Professional Basic expression. The $ (dollar) character must be placed in front of hex numbers, and they will be converted into standard decimal notation automatically. For example:

Print $64 Print $A

Do not confuse the use of the leading $ character for a hex number with the use of a trailing $ character for a string. $A is a hexadecimal number, but A$ is a variable!

Manipulating memory

A "byte" is a single unit of data, resident at an address in memory. If no unit of data exists at a particular address, the address remains empty. Each byte can hold a number between 0 and 255. Each byte is made up of eight smaller units of memory called "bits", and a bit is the smallest unit of data that can be represented in a computer's memory by a 1 or a 0.


function: read a byte from an address

The PEEK function returns a single 8-bit byte from an address in memory.


instruction: change a byte at an address
Poke address,number

The POKE command moves a number from 0 to 255 into the memory location at the specified address. Take great care with this instruction! Only POKE addresses that you completely understand, such as the contents of an AMOS Professional memory bank. Random poking will provoke your Amiga into taking horrible reprisals!


function: read two bytes from an even address

DEEK reads a two-byte "word" at a specified address. This address must be even, or an address error will be generated.


instruction: change a two-byte word at an even address
Doke address,number

Use the DOKE command to copy a two-byte number between 0 and 65535 into the memory location at a specified even address. Only DOKE into places where you are certain of safety, because indiscriminate use of this command will almost certainly crash your Amiga!


function: read four bytes from an even address

The LEEK function returns a four-byte "long word" stored at the specified even address. The result will be in exactly the same format as a standard AMOS Professional integer. This may result in negative values being returned in certain circumstances, such as if bit 31 of your number is set to 1. The correct value can be calculated by making use of DEEK, then loading the result into a floating point number, like this:

A=$FFFFFFFE Print Leek(Varptr(A)) A#=Deek(Varptr(A))*65535.0+Deek(Varptr(A)+2) Print A#


instruction: change a four-byte word at an even address
Loke address,number

LOKE copies a four-byte number into the memory location at a specified address. As before, the address must be an even location, and this command must be used with the greatest of care to avoid crashing the computer.


instruction: poke a string of characters into memory
Poke$ address, string$

Use the POKE$ command to take a source string and copy it directly to a chosen memory location, one character at a time. The address parameter holds the address of the first byte to be loaded with the new string data.

The copying operation will continue until the last character of the source string is reached, and the end address will be as follows:

Reserve As Data 10,1000: Rem Reserve a memory bank NAME=Start(10)-8 : Rem Get the address of the name T$="Testbank" : Rem Choose a new name of up to 8 characters Poke$ NAME,Left$(T$,8) : Rem Poke the first 8 characters into the name


function: read a string of characters from memory

PEEK$ reads the maximum number of characters specified in the length parameter, into a new string. If this is not a sensible value, the length is assumed to be 65500. The address parameter is the location of the first character to be read.

There is an optional stop$ parameter, and if this is included, AMOS Professional will stop in its tracks the moment a specified stop$ character is encountered in the memory area. You will then be left with a string of characters up to the final stop$. Here is an example using PEEKS:

Reserve As Data 10,1000 : Rem Reserve a memory bank NAME=Start(10)-8 : Rem Get the address of the name Print Peek$(NAME,8)


instruction: copy a memory block
Copy start,finish To destination

The COPY command is used to move large sections of the Amiga's memory rapidly from one place to another. Specify the start and finish locations of the data to be moved, then give the destination of the position of memory area which is to be loaded with the data. Addresses may be odd or even, and special care should be taken to ensure that the destination area points to somewhere safe!


instruction: fill memory block with the contents of a variable
Fill start To finish,pattern

The FILL instruction packs an area of memory specified from start to finish. This area is filled with multiple copies of a specified four-byte pattern. The addresses of the start and finish determine the size and position of the memory block, and they must both be even.

The fill pattern is a standard four-byte integer, but if you need to fill the area with multiple copies of a single byte, this value can be calculated as follows, where V will contain the required value of your fill command:

BYTE=255 : Rem This can be any number from 0 to 255 V=0 Poke Varptr(V),BYTE : Poke Varptr(V)+1,BYTE Poke Varptr(V)+2,BYTE : Poke Varptr(V)+3,BYTE Print V


function: find a string of characters in memory
first=Hunt(start To finish,s$)

HUNT is really a low level version of the familiar INSTR$ command. It searches the memory area defined by the given start and finish addresses, looking for the first occurrence of the characters held in your specified string.

If the search is successful, the position of the first character in memory is returned, otherwise a value of zero will be given. When using this function, take great care in selecting the start and finish points for the search.

Direct access to variables


function: read the address of a variable

This useful function returns the location of any AMOS Professional variable in the Amiga's memory. Programmers familiar with C should find it very similar to the & (ampersand) operator in that language.

VARPTR provides back-door access to your variables, and with careful use you are able to get to them directly, without having to rely on standard routines. This is especially valuable with procedures, because if procedures are loaded with the address of a variable instead of its actual value, that variable can be changed from inside the procedure. For example:

TEST=0 Rem Correct use of square brackets ANSWER[Varptr(Test)] : Rem Load ADDRESS of variable into AD parameter Print TEST Procedure ANSWER[AD] Loke AD,42 : Rem Copy new value into variable, by back door! End Proc

There should be few problems encountered when reading the variables, but changing them is a very hazardous process!

The slightest error made in your address calculations will crash AMOS Professional, so it is vital to save your programs before attempting to change your variables in this way.

With machine code programs, VARPTR can also be used to manipulate entire strings or arrays directly. Each type of variable is stored in its own individual format, as listed below.

Integers are held as a simple group of four bytes. They can be read from your Basic program using LEEK, and altered by LOKE. Here is an example of this (dangerous) method:

ANSWER=43 : Rem Load a variable AN=Varptr(ANSWER) : Rem Find address of variable Loke AN,LEEK(AN)-1 : Rem Equivalent to ANSWER=ANSWER-1 Print ANSWER

Floating point numbers are stored as four bytes, using the special Fast-Floating point format. However, if DOUBLE PRECISION is being used, floating point numbers are held as a group of eight bytes in IEEE double precision format.

Strings are stored' as a series of characters in standard Ascii format. The address given by VARPTR points to the first character in the string, and this can be examined with PEEK or replaced using POKE. Note that the length of the string is contained in two bytes immediately before the string. This means that it can be loaded into Basic using a line like this:

Print Deek(Varptr(A$)-2) : Rem Equivalent to Print Len(A$)

One application of this function is to return the Ascii value of a single character in an AMOS Professional string. The standard method is to make use of the ASC and M1D$ functions, like this:

A=Asc(Mid$(A$),C,1) : Rem Return Ascii code of character C in A$

Using VARPTR, that could be replaced by the following line:

A=Asc(Mid$(A$),C,1) : Rem Return Ascii code of character C in A$

To avoid danger, special precautions must be taken before new values are poked into a string. During the course of a program, the address of a string may change many times, so it is vital to load the current address of a string using VARPTR immediately before that string is used.

AMOS Professional regularly reorganises all strings in memory, using a "garbage collection" process. This frees valuable space needed for variables, and is essential for the smooth running of the system. But if you wish to pass the address of a string as a procedure garbage collection can play havoc. The obvious solution is to collect the garbage before the address is calculated, using a simple line like this:


The address of the string can now be established, and passed to the procedure. So providing strings themselves are not used in the procedure, you should remain safe.

Another hazard can be encountered if you try to POKE values straight into a string.

A$="123456789": Rem Define a string of characters For C=0 To Len(A$)-1 AD=Varptr(A$) : Rem Get the address V=Peek(AD+C) : Rem Get Ascii value of current element Poke AD+C,V+1 : Rem Add 1 to it Next C Print A$

When you return to the Editor, you will discover that your listing has been changed!- AMOS Professional is trying to save space by storing your string in the program listing, rather than the standard variable "buffer". This problem can be solved by loading the first character into the start of the string, then adding the remaining characters later. Here is how:

A$="1" : Rem Set up the first character A$=A$+"23456789" : Rem Now add remaining characters

This fools AMOS Professional into creating a separate copy of the string in the variable buffer.

Numerical arrays are stored as a simple list of values, with each dimension stored in turn. Look at the following array:

Dim TEST(3,3)

That array is held in the following order:

0,0 0,1 0,2 0,3
1,0 1,1 1,2 1,3
2,0 2,1 2,2 2,3
3,0 3,1 3,2 3,3

So to return the address of the first value of the array, you would use this:

Varptr TEST(0,0)

String arrays are more complex, because their length needs to change whenever one of their elements is assigned to a new value. The only way of ensuring total safety is to avoid them altogether! If you ignore this advice and try to access them using VARPTR, you are risking real danger.

Manipulating bits


instruction: rotate left
Rol.B number,bin value
Rol.W number,bin value
Rol.L number,bin value

ROL is the AMOS Professional Basic version of the ROL command available from 68000 assembly language. It takes the given binary value, and rotates it the specified number of places to the left. The value can be a normal variable, or an expression. Expressions will be treated as a memory location, and AMOS Professional will change the value at the address of the result.

This command allows instant rotation of any part of the Amiga's memory, and it must be used with extreme caution! If variables are confused with bit numbers, your machine will crash. Take heed of the next lines:

A=1 Rol.l 1,A : Rem This is fine Rol.l A,1 : Rem This is lethal. DO NOT DO IT!

There are three forms of the ROL instruction:

ROL.B rotates the first eight bits of the value
ROL.W rotates the bottom 16 bits of the value
ROL.L rotates the entire number

The ROL command is invaluable as a rapid method of multiplying and positive number by a power of two, like this:

B=1 Rol.l 2,B Print B

Here is an example routine:

Curs Off : Locate 0,20 : Centre "Press a key to ROL the number" Locate 0,0 : Print "Binary version" Locate 0,4 : Print "Decimal version" B=1 : Rem Set initial value Do Locate 0,2: Print Bin$(B,32) : Rem Display number in binary Locate 0,6: Print B;" "; : Rem Nine spaces Wait Key Rol.l 1,B : Rem Try ROL.W and ROL.B too Loop


instruction: rotate right
Ror.B number,bin value
Ror.W number,bin value
Ror.L number,bin value

The ROR commands are similar to ROL, except they rotate numbers from left to right. As before, the number of places to be moved must be set, followed by a variable or an expression. If an expression is used, it will be treated as the address of your value. ROR can be used as a fast way of dividing any positive number by a power of two, like this:

A=8 Ror.l 1,A Print A


function: test a bit

The BTST function tests a single binary bit in a given value. Specify the number of the bit to be tested, from 0 to 31, then give the chosen variable or expression. If the given value is an expression, it will be used as an address, so the bit will then be checked at LEEK(value) instead. Note that only bits 0 to 7 can be tested by this system, and that AMOS Professional will take your bit number and perform an automatic AND operation with 7, to ensure that it lies in the correct range.

If the test is successful, a value of -1 (True) is returned, otherwise a zero (False) is given. For example:

B=%1010 Print Btst(3,B) Print Btst(2,B)


instruction: set a bit to 1
Bset position,value

The BSET command sets a bit to 1. Specify the bit by giving its position in a variable or an expression. If an expression is used, it will be treated as an address of a value in the Amiga's memory. If the bit number and the variable are given in the wrong order, your computer will crash!


instruction: toggle a bit
Bchg position,value

This instruction flips a binary bit from 0 to 1, or from 1 to 0, as appropriate.

As usual, the bit is specified by giving the number of its position, followed by either a variable or expression. If the value is an expression, it is assumed to be an address, and great care should be taken.


instruction: clear a bit
Bclr number,value

The BCLR command clears a bit by setting it to zero. The bit number can be from 0 to 31, and pinpoints the single binary digit to be cleared. If the value is an expression, it will be used as an address in memory.

Using assembly language

AMOS Professional exploits the most useful machine code routines and transforms them into simple Basic commands. Even if you are an experienced machine code programmer, you will have to work very hard to better the speed and range of AMOS Professional.

Even though assembly language is hazardous and you are best advised to avoid it, there are a few routines which could be improved by its use. So for this reason, you are provided with several methods of accessing machine code directly from AMOS Professional Basic. These features are strictly for experts, and should be ignored by anyone not familiar with assembly language.

Machine code procedures

The easiest option for exploiting assembly language is to install machine code directly into an AMOS Professional procedure. These procedures can be saved and loaded using standard commands, and then executed from the Basic program simply by typing their name! Apart from the fact that it cannot be listed on screen, the only effective difference between a machine code procedure and its Basic equivalent is speed.

Machine code procedures are completely compatible with the AMOS Compiler, which will not only run most of your programs at double speed, but also compact them to a fraction of their original size. This means that if you decide to compile your programs at a later date, you will not need to alter your assembly language at all. The following points should be noted before using an assembler:

Creating a machine code language procedure

Machine code procedures are installed using the [Inset Program] option from the [Editor/Procedures] menu. The following steps should be followed:

Procedure _MACHINE[A,A$] End Proc

Existing closed procedures may also be used for this purpose, and it is perfectly legal to update a routine after the machine code program has been re-assembled.

Once the machine code is installed in this manner, it will be called automatically whenever the new procedure is run from AMOS Professional Basic.

Communicating with a machine code procedure

There are two methods of exchanging information with a machine code procedure.

With the first method, values are loaded into the appropriate Address and Data registers, before the procedure is called using the AREG and DREG functions. For example:

Dreg(0)=1 : Dreg(1)=Varptr(A$) : _MACHINE Procedure _MACHINE

Note that AREG(3) to AREG(6) will not be transferred to the routine. These registers cannot be changed, as they are used to store important system information. To return values to AMOS Professional Basic, AREG and DREG can be used to read the contents of the Address and Data registers, after the procedure has been called.

The second alternative method is much neater. Values are entered using normal parameters. As usual, a list of parameters is specified as part of the procedure definition, like this:

Rem Use no parameters but take info directly from AREG and DREG values Procedure _MACHINE° Procedure _MACHINE1[A] : Rem Enter one integer into procedure Procedure _MACH1NE2[A,B,C$] : Rem Get two integers and one string

The values of the parameters are pushed onto the Parameter stack, pointed to by A3. These parameters are stored in reverse order, and are four bytes in length.

_MACHINE1 will grab the parameters like this:

Move.l (a3)+,d0

_MACHINE2 will grab the parameters as follows:

; Grab the string. Each string is stored at an EVEN address,
; starting with the length of the string, and then the string itself

Move.l (a3)+,a2             * Address of the string
Move.w (a2)+,d2             * Length of the string

; A2 now points to the first character

; Grab the two integers

Move.l (a3)+,d1             * Grab "B"
Move.l (a3)+,d0             * Grab "A"

The AMOS Professional stack works in the same way as a conventional stack, so although anything below can be changed, do not touch any values above the base address contained in A3. Also note that the space available for the routine depends on the level of the procedure, so if it is called from the main program approximately 3k is available. This can be increased by a call to the SET STACK command from AMOS Professional Basic.

To return values to AMOS Professional Basic, the value of DO is available from the PARAM function automatically.

Calling machine code from an address or bank

There is another option for calling machine code directly from a memory bank or an address.


instruction: load machine code directly into memory
Pload "filename",bank number

The PLOAD command reserves a memory bank and loads some machine code into it from disc. Specify the filename that contains the machine code file on disc, followed by the number of a new memory bank to be reserved for the program. If the bank number is negative, the number will be multiplied by -1, and the bank will be allocated using Chip memory.

Once machine code is loaded in this way, it is installed as a permanent memory bank, so whenever the current program is saved, the machine code is stored too. Also note that the machine code file can be saved onto disc as a standard ".Abk" file, then loaded directly into AMOS Professional Basic. After PLOAD has performed its work, the memory bank can be executed immediately! The following factors should be noted:


instruction: execute a machine code program from memory
Call address
Call address,parameters
Call bank
Call bank,parameters

The CALL instruction is used to run a machine code program straight from the Amiga's memory. You can specify either an absolute memory location or the number of a memory bank, previously installed using the PLOAD command.

On entry to the program, registers D0 to D7 and A0 to A2 will be loaded from values stored in the DREG and AREG functions. The assembly language program can change any 68000 registers it chooses. At the start of the routine, register A3 will point to the optional parameter list, which is explained next, and A5 will contain the address of the AMOS Professional data zone. When the routine has completed its task, you can return to Basic with a RTS.

After the memory location or bank number, a list of optional parameters may be given in the form of a list of values. These values will be taken from the AMOS Professional Basic program and pushed onto the A3 stack by the CALL command. They must be removed in reverse order, so the last value in the list will be the first on the stack. The format of a parameter depends on what type of variable they are, as follows:

Integers. The parameter holds a long word, containing a normal AMOS Professional number. It can be grabbed with a line such as this:

Move.l (a3)+,d0

Single precision numbers. These are stored in Fast Floating Point format, and are held in one long word. To load such a number into register d0, use the following:

Move.l (a3)+,d0

Double precision numbers. These are stored in IEEE double precision format, and are held as two long words. To load a double precision variable into registers d0 and dl, you could use this:

Move.l (a3)+,d0   * Top half
Move.l (a3)+,d1   * Bottom half

Strings. The stack contains the Address of the string in memory. All strings begin with a single word that holds their length. For example:

; Grab the string. Each string is stored at an EVEN address,
; starting with the length of the string, and then the string itself

Move.l (a3)+,a2    * Address of the string
Move.w (a2)+,d2    * Length of the string


reserved variable: pass values to and from 68000 address register

AREG is a special array which is used to pass values to and from any of the 68000 processor's address registers. Specify the number of the register from 0 to 6, selected from either of the following two groups:

A0, Al, A2. These registers can be read from AMOS Professional Basic, and changed at will. Whenever a machine code program is run, any new values will be transferred straight into the relevant address register. For example:

Areg(0)=Varptr(A$) : Rem Load the address of A$ into A0 Areg(1)=Varptr(B(0,0)) : Rem Load the address of B(0,0) into Al

A3, A4, A5, A6. These are read-only registers. Any attempt to change their current contents will generate an "illegal function call" error message.


reserved variable: pass value into 68000 data register

DREG can be used to move values back and forth between AMOS Professional Basic and the 68000's Data registers, by specifying the number of a data register from 0 to 7.

This function can be thought of as an array which holds an exact copy of registers DO to D7. This array is automatically moved into the Data registers, either by the CALL command or whenever a machine code procedure is run. Once the routine has ended, the new contents of DO to D7 is copied directly into the array, so that the results may be read directly from AMOS Professional programs. For example:

Dreg(0)=10 : Rem Save 10 into DO Print Dreg(0) : Rem Print the contents of DO