Art of Assembly/Win32 Edition is now available. Let me read that version.
PLEASE: Before emailing me asking how to get a hard copy of this text, read this.
Important Notice: As you have probably discovered by now, I am no longer updating this document. The reason is quite simple: I'm working on a Windows version of "The Art of Assembly Language Programming". In the past I have encouraged individuals to send me corrections to this text. However, as I am no longer updating this material. However, I am collecting errata that I will post to Webster someday, so feel free to continue sending corrections to AoA/DOS (16-bit) to rhyde@cs.ucr.edu. If you're more interested in leading edge material, please see the information about the Win/32 edition, above.
include directives, and initializes necessary Library routines for you. You should not attempt to create a new program from scratch unless you are very familiar with the internal operation of the Standard Library.call instruction for invocation. You cannot, for example, directly call the putc routine. Instead, you invoke the putc macro that includes a call to the sl_putc procedure ("SL" stands for "Standard Library").
meminit routine initializes the memory manager and you must call it before any routine that uses the memory manager. Since many Standard Library routines use the memory manager, you should call this procedure early in the program. The "SHELL.ASM" file makes this call for you.malloc routine allocates storage on the heap and returns a pointer to the block it allocates in the es:di registers. Before calling malloc you need to load the size of the block (in bytes) into the cx register. On return, malloc sets the carry flag if an error occurs (insufficient memory). If the carry is clear, es:di points at a block of bytes the size you've specified:
mov cx, 1024 ;Grab 1024 bytes on the heap
malloc ;Call MALLOC
jc MallocError ;If memory error.
mov word ptr PNTR, DI ;Save away pointer to block.
mov word ptr PNTR+2, ES
When you call malloc, the memory manager promises that the block it gives you is free and clear and it will not reallocate that block until you explicitly free it. To return a block of memory back to the memory manager so you can (possibly) re-use that block of memory in the future, use the free Library routine. free expects you to pass the pointer returned by malloc:
les di, PNTR ;Get pointer to free
free ;Free that block
jc BadFree
As usual for most Standard Library routines, if the free routine has some sort of difficulty it will return the carry flag set to denote an error.
getc (get a character), gets (get a string), and getsm (get a malloc'd string).Getc reads a single character from the keyboard and returns that character in the al register. It returns end of file (EOF ) status in the ah register (zero means EOF did not occur, one means EOF did occur). It does not modify any other registers. As usual, the carry flag returns the error status. You do not need to pass getc any values in the registers. Getc does not echo the input character to the display screen. You must explicitly print the character if you want it to appear on the output monitor.
; Note: "CR" is a symbol that appears in the "consts.a"
; header file. It is the value 13 which is the ASCII code
; for the carriage return character
Wait4Enter: getc
cmp al, cr
jne Wait4Enter
The gets routine reads an entire line of text from the keyboard. It stores each successive character of the input line into a byte array whose base address you pass in the es:di register pair. This array must have room for at least 128 bytes. The gets routine will read each character and place it in the array except for the carriage return character. Gets terminates the input line with a zero byte (which is compatible with the Standard Library string handling routines). Gets echoes each character you type to the display device, it also handles simple line editing functions such as backspace. As usual, gets returns the carry set if an error occurs. The following example reads a line of text from the standard input device and then counts the number of characters typed. This code is tricky, note that it initializes the count and pointer to -1 prior to entering the loop and then immediately increments them by one. This sets the count to zero and adjusts the pointer so that it points at the first character in the string. This simplification produces slightly more efficient code than the straightforward solution would produce:
DSEG segment
MyArray byte 128 dup (?)
DSEG ends
CSEG segment
.
.
.
; Note: LESI is a macro (found in consts.a) that loads
; ES:DI with the address of its operand. It expands to the
; code:
;
; mov di, seg operand
; mov es, di
; mov di, offset operand
;
; You will use the macro quite a bit before many Standard
; Library calls.
lesi MyArray ;Get address of inp buf.
gets ;Read a line of text.
mov ah, -1 ;Save count here.
lea bx, -1[di] ;Point just before string.
CountLoop: inc ah ;Bump count by one.
inc bx ;Point at next char in str.
cmp byte ptr es:[bx], 0
jne CoutLoop
; Now AH contains the number of chars in the string.
.
.
.
The getsm routine also reads a string from the keyboard and returns a pointer to that string in es:di. The difference between gets and getsm is that you do not have to pass the address of an input buffer in es:di. Getsm automatically allocates storage on the heap with a call to malloc and returns a pointer to the buffer in es:di. Don't forget that you must call meminit at the beginning of your program if you use this routine. The SHELL.ASM skeleton file calls meminit for you. Also, don't forget to call free to return the storage to the heap when you're done with the input line.
getsm ;Returns pointer in ES:DI
.
.
.
free ;Return storageto heap.
Putc outputs a single character to the display device. It outputs the character appearing in the al register. It does not affect any registers unless there is an error on output (the carry flag denotes error/no error, as usual). See the Standard Library documentation for more details.Putcr outputs a "newline" (carriage return/line feed combination) to the standard output. It is completely equivalent to the code:
mov al, cr ;CR and LF are constants
putc ; appearing in the consts.a
mov al, lf ; header file.
putc
The puts (put a string) routine prints the zero terminated string at which es:di points. Note that puts does not automatically output a newline after printing the string. You must either put the carriage return/line feed characters at the end of the string or call putcr after calling puts if you want to print a newline after the string. Puts does not affect any registers (unless there is an error). In particular, it does not change the value of the es:di registers. The following code sequence uses this fact:
getsm ;Read a string
puts ;Print it
putcr ;Print a new line
free ;Free the memory for string.
Since the routines above preserve es:di (except, of course, getsm), the call to free deallocates the memory allocated by the call to getsm.
The puth routine prints the value in the al register as exactly two hexadecimal digits, including a leading zero byte if the value is in the range 0..Fh. The following loop reads a sequence of keys from the keyboard and prints their ASCII values until the user presses the Enter key:
KeyLoop: getc
cmp al, cr
je Done
puth
putcr
jmp KeyLoop
Done:
The puti routine prints the value in ax as a signed 16 bit integer. The following code fragment prints the sum of I and J to the display:
mov ax, I
add ax, J
puti
putcr
Putu is similar to puti except it outputs unsigned integer values rather than signed integers.
Routines like puti and putu always output numbers using the minimum number of possible print positions. For example, puti uses three print positions on the string to print the value 123. Sometimes, you may want to force these output routines to print their values using a fixed number of print positions, padding any extra positions with spaces. The putisize and putusize routines provide this capability. These routines expect a numeric value in ax and a field width specification in cx. They will print the number in a field width of at least cx positions. If the value in cx is larger than the number of print position the value requires, these routines will right justify the number in a field of cx print positions. If the value in cx is less than the number of print positions the value requires, these routines ignore the value in cx and use however many print positions the number requires.
; The following loop prints out the values of a 3x3 matrix in matrix form:
; On entry, bx points at element [0,0] of a row column matrix.
mov dx, 3 ;Repeat for each row.
PrtMatrix: mov ax, [bx] ;Get first element in this row.
mov cx, 7 ;Use seven print positions.
putisize ;Print this value.
mov ax, 2[bx] ;Get the second element.
putisize ;CX is still seven.
mov ax, 4[bx] ;Get the third element.
putisize
putcr ;Output a new line.
add bx, 6 ;Move on to next row.
dec dx ;Repeat for each row.
jne PrtMatrix
The print routine is one of the most-often called procedures in the library. It prints the zero terminated string that immediately follows the call to print:
print
byte "Print this string to the display",cr,lf,0
The example above prints the string "Print this string to the display" followed by a new line. Note that print will print whatever characters immediately follow the call to print, up to the first zero byte it encounters. In particular, you can print the newline sequence and any other control characters as shown above. Also note that you are not limited to printing one line of text with the print routine:
print
byte "This example of the PRINT routine",cr,lf
byte "prints several lines of text.",cr,lf
byte "Also note,",cr,lf,"that the source lines "
byte "do not have to correspond to the output."
byte cr,lf
byte 0
The above displays:
This example of the PRINT routine prints several lines of text. Also note, that the source lines do not have to correspond to the output.
It is very important that you not forget about that zero terminating byte. The print routine begins executing the first 80x86 machine language instruction following that zero terminating byte. If you forget to put the zero terminating byte after your string, the print routine will gladly eat up the instruction bytes following your string (printing them) until it finds a zero byte (zero bytes are common in assembly language programs). This will cause your program to misbehave and is a big source of errors beginning programmers have when they use the print routine. Always keep this in mind.
Printf, like its "C" namesake, provides formatted output capabilities for the Standard Library package. A typical call to printf always takes the following form:
printf
byte "format string",0
dword operand1, operand2, ..., operandn
The format string is comparable to the one provided in the "C" programming language. For most characters, printf simply prints the characters in the format string up to the terminating zero byte. The two exceptions are characters prefixed by a backslash ("\") and characters prefixed by a percent sign ("%"). Like C's printf, the Standard Library's printf uses the backslash as an escape character and the percent sign as a lead-in to a format string.
Printf uses the escape character ("\") to print special characters in a fashion similar to, but not identical to C's printf. The Standard Library's printf routine supports the following special characters:
C users should note a couple of differences between Standard Library's escape sequences and C's. First, use "\%" to print a percent sign within a format string, not "%%". C doesn't allow the use of "\%" because the C compiler processes "\%" at compile time (leaving a single "%" in the object code) whereas printf processes the format string at run-time. It would see a single "%" and treat it as a format lead-in character. The Standard Library's printf, on the other hand, processes both the "\" and "%" at run-time, therefore it can distinguish "\%".
Strings of the form "\0xhh" must contain exactly two hex digits. The current printf routine isn't robust enough to handle sequences of the form "\0xh" which contain only a single hex digit. Keep this in mind if you find printf chopping off characters after you print a value.
There is absolutely no reason to use any hexadecimal escape character sequence except "\0x00". Printf grabs all characters following the call to printf up to the terminating zero byte (which is why you'd need to use "\0x00" if you want to print the null character, printf will not print such values). The Standard Library's printf routine doesn't care how those characters got there. In particular, you are not limited to using a single string after the printf call. The following is perfectly legal:
printf
byte "This is a string",13,10
byte "This is on a new line",13,10
byte "Print a backspace at the end of this line:"
byte 8,13,10,0
Your code will run a tiny amount faster if you avoid the use of the escape character sequences. More importantly, the escape character sequences take at least two bytes. You can encode most of them as a single byte by simply embedding the ASCII code for that byte directly into the code stream. Don't forget, you cannot embed a zero byte into the code stream. A zero byte terminates the format string. Instead, use the "\0x00" escape sequence.
Format sequences always begin with "%". For each format sequence, you must provide a far pointer to the associated data immediately following the format string, e.g.,
printf
byte "%i %i",0
dword i,j
Format sequences take the general form "%s\cn^f" where:
The "s", "\c", "n", and "^" items are optional, the "%" and "f" items must be present. Furthermore, the order of these items in the format item is very important. The "\c" entry, for example, cannot precede the "s" entry. Likewise, the "^" character, if present, must follow everything except the "f" character(s).
The format characters i, d, x, h, u, c, s, ld, li, lx, and lu control the output format for the data. The i and d format characters perform identical functions, they tell printf to print the following value as a 16 bit signed decimal integer. The x and h format characters instruct printf to print the specified value as a 16 bit or 8-bit hexadecimal value (respectively). If you specify u, printf prints the value as a 16-bit unsigned decimal integer. Using c tells printf to print the value as a single character. S tells printf that you're supplying the address of a zero-terminated character string, printf prints that string. The ld, li, lx, and lu entries are long (32-bit) versions of d/i, x, and u. The corresponding address points at a 32-bit value that printf will format and print to the standard output.
The following example demonstrates these format items:
printf
byte "I= %i, U= %u, HexC= %h, HexI= %x, C= %c, "
dbyte "S= %s",13,10
byte "L= %ld",13,10,0
dword i,u,c,i,c,s,l
The number of far addresses (specified by operands to the "dd" pseudo-opcode) must match the number of "%" format items in the format string. Printf counts the number of "%" format items in the format string and skips over this many far addresses following the format string. If the number of items do not match, the return address for printf will be incorrect and the program will probably hang or otherwise malfunction. Likewise (as for the print routine), the format string must end with a zero byte. The addresses of the items following the format string must point directly at the memory locations where the specified data lies.
When used in the format above, printf always prints the values using the minimum number of print positions for each operand. If you want to specify a minimum field width, you can do so using the "n" format option. A format item of the format "%10d" prints a decimal integer using at least ten print positions. Likewise, "%16s" prints a string using at least 16 print positions. If the value to print requires more than the specified number of print positions, printf will use however many are necessary. If the value to print requires fewer, printf will always print the specified number, padding the value with blanks. Printf will print the value right justified in the print field (regardless of the data's type). If you want to print the value left justified in the output file, use the "-" format character as a prefix to the field width, e.g.,
printf
byte "%-17s",0
dword string
In this example, printf prints the string using a 17 character long field with the string left justified in the output field.
By default, printf blank fills the output field if the value to print requires fewer print positions than specified by the format item. The "\c" format item allows you to change the padding character. For example, to print a value, right justified, using "*" as the padding character you would use the format item "%\*10d". To print it left justified you would use the format item "%-\*10d". Note that the "-" must precede the "\*". This is a limitation of the current version of the software. The operands must appear in this order. Normally, the address(es) following the printf format string must be far pointers to the actual data to print.
On occasion, especially when allocating storage on the heap (using malloc), you may not know (at assembly time) the address of the object you want to print. You may have only a pointer to the data you want to print. The "^" format option tells printf that the far pointer following the format string is the address of a pointer to the data rather than the address of the data itself. This option lets you access the data indirectly.
Note: unlike C, Standard Library's printf routine does not support floating point output. Putting floating point into printf would increase the size of this routine a tremendous amount. Since most people don't need the floating point output facilities, it doesn't appear here. There is a separate routine, printff, that includes floating point output.
The Standard Library printf routine is a complex beast. However, it is very flexible and extremely useful. You should spend the time to master its major functions. You will be using this routine quite a bit in your assembly language programs.
The standard output package provides many additional routines besides those mentioned here. There simply isn't enough room to go into all of them in this chapter. For more details, please consult the Standard Library documentation.
puti, putu, and putl routines output the numeric strings using the minimum number of print positions necessary. For example, puti uses three character positions to print the value -12. On occasion, you may need to specify a different field width so you can line up columns of numbers or achieve other formatting tasks. Although you can use printf to accomplish this goal, printf has two major drawbacks - it only prints values in memory (i.e., it cannot print values in registers) and the field width you specify for printf must be a constant. The putisize, putusize, and putlsize routines overcome these limitations.ax register (putisize and putusize) or the dx:ax register pair (putlsize). They also expect a minimum field width in the cx register. As with printf, if the value in the cx register is smaller than the number of print positions that the number actually needs to print, putisize, putusize, and putlsize will ignore the value in cx and print the value using the minimum necessary number of print positions.
isize, usize, and lsize routines do this for you.isize routine expects a signed integer in the ax register. It returns the minimum field width of that value (including a position for the minus sign, if necessary) in the ax register. Usize computes the size of the unsigned integer in ax and returns the minimum field width in the ax register. Lsize computes the minimum width of the signed integer in dx:ax (including a position for the minus sign, if necessary) and returns this width in the ax register.
atoi, atoh, atou, itoa, htoa, wtoa, and utoa (plus others). The ATOx routines convert an ASCII string in the appropriate format to a numeric value and leave that value in ax or al. The ITOx routines convert the value in al/ax to a string of digits and store this string in the buffer whose address is in es:di. There are several variations on each routine that handle different cases. The following paragraphs describe each routine.atoi routine assumes that es:di points at a string containing integer digits (and, perhaps, a leading minus sign). They convert this string to an integer value and return the integer in ax. On return, es:di still points at the beginning of the string. If es:di does not point at a string of digits upon entry or if an overflow occurs, atoi returns the carry flag set. Atoi preserves the value of the es:di register pair. A variant of atoi, atoi2, also converts an ASCII string to an integer except it does not preserve the value in the di register. The atoi2 routine is particularly useful if you need to convert a sequence of numbers appearing in the same string. Each call to atoi2 leaves the di register pointing at the first character beyond the string of digits. You can easily skip over any spaces, commas, or other delimiter characters until you reach the next number in the string; then you can call atoi2 to convert that string to a number. You can repeat this process for each number on the line.Atoh works like the atoi routine, except it expects the string to contain hexadecimal digits (no leading minus sign). On return, ax contains the converted 16 bit value and the carry flag denotes error/no error. Like atoi, the atoh routine preserves the values in the es:di register pair. You can call atoh2 if you want the routine to leave the di register pointing at the first character beyond the end of the string of hexadecimal digits.Atou converts an ASCII string of decimal digits in the range 0..65535 to an integer value and returns this value in ax. Except that the minus sign is not allowed, this routine behaves just like atoi. There is also an atou2 routine that does not preserve the value of the di register; it leaves di pointing at the first character beyond the string of decimal digits.geti, geth, or getu routines available in the Standard Library, you will have to construct these yourself. The following code demonstrates how to read an integer from the keyboard:
print
byte "Enter an integer value:",0
getsm
atoi ;Convert string to an integer in AX
free ;Return storage allocated by getsm
print
byte "You entered ",0
puti ;Print value returned by ATOI.
putcr
The itoa, utoa, htoa, and wtoa routines are the logical inverse to the atox routines. They convert numeric values to their integer, unsigned, and hexadecimal string representations. There are several variations of these routines depending upon whether you want them to automatically allocate storage for the string or if you want them to preserve the di register.
Itoa converts the 16 bit signed integer in ax to a string and stores the characters of this string starting at location es:di. When you call itoa, you must ensure that es:di points at a character array large enough to hold the resulting string. Itoa requires a maximum of seven bytes for the conversion: five numeric digits, a sign, and a zero terminating byte. Itoa preserves the values in the es:di register pair, so upon return es:di points at the beginning of the string produced by itoa.
Occasionally, you may not want to preserve the value in the di register when calling the itoa routine. For example, if you want to create a single string containing several converted values, it would be nice if itoa would leave di pointing at the end of the string rather than at the beginning of the string. The itoa2 routine does this for you; it will leave the di register pointing at the zero terminating byte at the end of the string. Consider the following code segment that will produce a string containing the ASCII representations for three integer variables, Int1, Int2, and Int3:
; Assume es:di already points at the starting location to store the converted
; integer values
mov ax, Int1
itoa2 ;Convert Int1 to a string.
; Okay, output a space between the numbers and bump di so that it points
; at the next available position in the string.
mov byte ptr es:[di], ' '
inc di
; Convert the second value.
mov ax, Int2
itoa2
mov byte ptr es:[di], ' '
inc di
; Convert the third value.
mov ax, Int3
itoa2
; At this point, di points at the end of the string containing the
; converted values. Hopefully you still know where the start of the
; string is so you can manipulate it!
Another variant of the itoa routine, itoam, does not require you to initialize the es:di register pair. This routine calls malloc to automatically allocate the storage for you. It returns a pointer to the converted string on the heap in the es:di register pair. When you are done with the string, you should call free to return its storage to the heap.
; The following code fragment converts the integer in AX to a string and prints
; this string. Of course, you could do this same operation with PUTI, but this
; code does demonstrate how to call itoam.
itoam ;Convert integer to string.
puts ;Print the string.
free ;Return storage to the heap.
The utoa, utoa2, and utoam routines work just like itoa, itoa2, and itoam, except they convert the unsigned integer value in ax to a string. Note that utoa and utoa2 require, at most, six bytes since they never output a sign character.
Wtoa, wtoa2, and wtoam convert the 16 bit value in ax to a string of exactly four hexadecimal characters plus a zero terminating byte. Otherwise, they behave exactly like itoa, itoa2, and itoam. Note that these routines output leading zeros so the value is always four digits long.
The htoa, htoa2, and htoam routines are similar to the wtoa, wtoa2, and wtoam routines. However, the htox routines convert the eight bit value in al to a string of two hexadecimal characters plus a zero terminating byte.
The Standard Library provides several other conversion routines as well as the ones mentioned in this section. See the Standard Library documentation in the appendices for more details.
al register to see if it falls within a certain set of characters. These routines all return the status in the zero flag. If the condition is true, they return the zero flag set (so you can test the condition with a je instruction). If the condition is false, they clear the zero flag (test this condition with jne). These routines are
al contains an alphanumeric character.
al to see if it contains a hexadecimal digit character.
al to see if it contains a decimal digit character.
al to see if it contains an alphabetic character.
al to see if it contains a lower case alpha character.
al to see if it contains an upper case alpha character.
al register. They will convert the character in al to the appropriate alphabetic case.al contains a lower case alphabetic character, ToUpper will convert it to the equivalent upper case character. If al contains any other character, ToUpper will return it unchanged.al contains an upper case alphabetic character, ToLower will convert it to the equivalent lower case character. If the value is not an upper case alphabetic character ToLower will return it unchanged.
Random routine generates a sequence of pseudo-random numbers. It returns a random value in the ax register on each call. You can treat this value as a signed or unsigned value since Random manipulates all 16 bits of the ax register.div and idiv instructions to force the output of random to a specific range. Just divide the value random returns by some number n and the remainder of this division will be a value in the range 0..n-1. For example, to compute a random number in the range 1..10, you could use code like the following:
random ;Get a random number in range 0..65535.
sub dx, dx ;Zero extend to 16 bits.
mov bx, 10 ;Want value in the range 1..10.
div bx ;Remainder goes to dx!
inc dx ;Convert 0..9 to 1..10.
; At this point, a random number in the range 1..10 is in the dx register.
The random routine always returns the same sequence of values when a program loads from disk and executes. Random uses an internal table of seed values that it stores as part of its code. Since these values are fixed, and always load into memory with the program, the algorithm that random uses will always produce the same sequence of values when a program containing it loads from the disk and begins running. This might not seem very "random" but, in fact, this is a nice feature since it is very difficult to test a program that uses truly random values. If a random number generator always produces the same sequence of numbers, any tests you run on that program will be repeatable.
Unfortunately, there are many examples of programs that you may want to write (e.g., games) where having repeatable results is not acceptable. For these applications you can call the randomize routine. Randomize uses the current value of the time of day clock to generate a nearly random starting sequence. So if you need a (nearly) unique sequence of random numbers each time your program begins execution, call the randomize routine once before ever calling the random routine. Note that there is little benefit to calling the randomize routine more than once in your program. Once random establishes a random starting point, further calls to randomize will not improve the quality (randomness) of the numbers it generates.
NULL = 0 ;Some common ASCII codes BELL = 07 ;Bell character bs = 08 ;Backspace character tab = 09 ;Tab character lf = 0ah ;Line feed character cr = 0dh ;Carriage return
In addition to the constants above, "stdlib.a" also defines some useful macros including ExitPgm, lesi, and ldxi. These macros contain the following instructions:
; ExitPgm- Returns control to MS-DOS
ExitPgm macro
mov ah, 4ch ;DOS terminate program opcode
int 21h ;DOS call.
endm
; LESI ADRS-
; Loads ES:DI with the address of the specified operand.
lesi macro adrs
mov di, seg adrs
mov es, di
mov di, offset adrs
endm
; LDXI ADRS-
; Loads DX:SI with the address of the specified operand.
ldxi macro adrs
mov dx, seg adrs
mov si, offset adrs
endm
The lesi and ldxi macros are especially useful for load addresses into es:di or dx:si before calling various standard library routines (see Chapter Eight for details about macros).