Assembly Language Programming cs.uns.edu

CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 1 of 58 17/11/2005 08:08 p.m. PART One 80x86 Assembly Language Programming PART Two Introduction to MIPS Programming Recommended Texts: Sargent, Murray. - The personal computer from the inside out / Murray Sargent III and Rich. - Rev. ed. - Reading, Mass : Addison-Wesley Pub. Co, 1986. - 0201069180 Waldron, John, 1964-. - Introduction to RISC assembly language programming / John Waldron. - Harlow, England : Addison-Wesley, 1999. - 0201398281 OVERVIEW OF THE 80x86 FAMILY Why Assembly Language ? REPRESENTATION OF NUMBERS IN BINARY REGISTERS General Purpose Regs Index Registers Stack Register SEGMENTS AND OFFSETS THE STACK INTRODUCTION TO ASSEMBLY PUSH AND POP TYPES OF OPERAND SOME USEFUL INSTRUCTIONS MOV INT ADD SUB MUL IMUL DIV IDIV WHAT ARE MEMORY MODELS Tiny Small Medium Large Flat BASIC ASSEMBLY PROGRAM Listing 1:1stProgram.asm COMPILATION INSTRUCTIONS MAKING THINGS EASIER KEYBOARD INPUT PRINTING A CHARACTER Listing 2: DOS Interrupt 21h INTRODUCTION TO PROCEDURES Listing 3: SIMPROC.ASM PROCEDURES THAT PASS PARAMETERS Paramater Passing in Registers Listing 4:Proc1.asm PASSING PARAMETERS THROUGH MEMORY Listing 5: PROC2.ASM CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 2 of 58 17/11/2005 08:08 p.m. PASSING PARAMETERS THROUGH THE STACK Listing 6:Proc3.asm MACROS (in Turbo Assembler) Macros with Parameters FILES AND HOW TO USE THEM Function 3Dh: open file Function 3Eh: close file Function 3Fh: read file/device Listing 7: READFILE.ASM Function 3Ch: Create File OVERVIEW OF THE 80x86 FAMILY -------------------------------------------------- The 80x86 family was first started in 1981 with the8086 and the newest member is the Pentium which was released thirteen years later in 1994. They are all backwards compatible with each other but each new generation has added features and more speed than the previous chip. Today there are very few computers in use that have the 8088 and 8086 chips in them as they are very outdated and slow. There are a few 286's but their numbers are declining as today's software becomes more and more demanding. Even the 386, Intel's first 32-bit CPU, is now declining and it seems that the 486 is nowthe entry level system. Why Assembly Language? An old joke goes something like this: "There are three reasons for using assembly language: speed, speed, and more speed." Even those who absolutely hate assembly language will admit that if speed is your primary concern, assembly language is the way to go. Assembly language has several benefits: Speed. Assembly language programs are generally the fastest programs around. Space. Assembly language programs are often the smallest. Capability. You can do things in assembly which are difficult or impossible in HLLs. Knowledge. Your knowledge of assembly language will help you write better CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 3 of 58 17/11/2005 08:08 p.m. programs, even when using HLLs. Assembly language is the uncontested speed champion among programming languages. An expert assembly language programmer will almost always produce a faster program than an expert C programmer. While certain programs may not benefit much from implementation in assembly, you can speed up many programs by a factor of five or ten over their HLL counterparts by careful coding in assembly language; even greater improvement is possible if you're not using an optimizing compiler. Alas, speedups on the order of five to ten times are generally not achieved by beginning assembly language programmers. However, if you spend the time to learn assembly language really well, you too can achieve these impressive performance gains. Despite some people's claims that programmers no longer have to worry about memory constraints, there are many programmers who need to write smaller programs. Assembly language programs are often less than one-half the size of comparable HLL programs. This is especially impressive when you consider the fact that data items generally consume the same amount of space in both types of programs, and that data is responsible for a good amount of the space used by a typical application. Saving space saves money. Pure and simple. If a program requires 1.5 megabytes, it will not fit on a 1.44 Mbyte floppy. Likewise, if an application requires 2 megabytes RAM, the user will have to install an extra megabyte if there is only one available in the machine. Even on big machines with 32 or more megabytes, writing gigantic applications isn't excusable. Most users put more than eight megabytes in their machines so they can run multiple programs from memory at one time. The bigger a program is, the fewer applications will be able to coexist in memory with it. Virtual memory isn't a particularly attractive solution either. With virtual memory, the bigger an application is, the slower the system will run as a result of that program's size. Capability is another reason people resort to assembly language. HLLs are an abstraction of a typical machine architecture. They are designed to be independent of the particular machine architecture. As a result, they rarely take into account any special features of the machine, features which are available to assembly language programmers. If you want to use such features, you will need to use assembly language. A really good example is the input/output instructions available on the 80x86 microprocessors. These instructions let you directly access certain I/O devices on the computer. In general, such access is not part of any high level language. Indeed, some languages like C pride themselves on not supporting any specific I/O operations. In assembly language you have no such restrictions. Anything you can do on the machine you can do in assembly language. This is definitely not the case with most HLLs. CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 4 of 58 17/11/2005 08:08 p.m. Of course, another reason for learning assembly language is just for the knowledge. Now some of you may be thinking, "Gee, that would be wonderful, but I've got lots to do. My time would be better spent writing code than learning assembly language." There are some practical reasons for learning assembly, even if you never intend to write a single line of assembly code. If you know assembly language well, you'll have an appreciation for the compiler, and you'll know exactly what the compiler is doing with all those HLL statements. Once you see how compilers translate seemingly innocuous statements into a ton of machine code, you'll want to search for better ways to accomplish the same thing. Good assembly language programmers make better HLL programmers because they understand the limitations of the compiler and they know what it's doing with their code. Those who don't know assembly language will accept the poor performance their compiler produces and simply shrug it off. Representation of numbers in binary --------------------------------------------------------------------- Before we begin to understand how to program in assembly it is best to try to understand how numbers are represented in computers. Numbers are stored in binary, base two. There are several terms which are used to describe different size numbers and I will describe what these mean. 1 BIT: 0 One bit is the simplest piece of data that exists. Its either a one or a zero. 1 NIBBLE: 0000 4 BITS The nibble is four bits or half a byte. Note that it has a maximum value of 15 (1111 = 15). This is the basis for the hexadecimal (base 16) number system which is used as it is far easier to understand. Hexadecimal numbers go from 1 to F and are followed by a h to state that the are in hex. i.e. Fh = 15 decimal. Hexadecimal numbers that begin with a letter are prefixed with a 0 (zero). 1 BYTE 00000000 2 NIBBLES 8 BITS A byte is 8 bits or 2 nibbles. A byte has a maximum value of FFh (255 decimal). CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 5 of 58 17/11/2005 08:08 p.m. Because a byte is 2 nibbles the hexadecimal representation is two hex digits in a row i.e. 3Dh. The byte is also that size of the 8-bit registers which we will be covering later. 1 WORD 0000000000000000 2 BYTES 4 NIBBLES 16 BITS A word is two bytes that are stuck together. A word has a maximum value of FFFFh (65,536). Since a word is four nibbles, it is represented by four hex digits. This is the size of the 16-bit registers. Registers --------------------------------------------------------------------- Registers are a place in the CPU where a number can be stored and manipulated. There are three sizes of registers: 8-bit, 16-bit and on 386 and above 32-bit. There are four different types of registers; general purpose registers, segment egisters, index registers and stack registers. Firstly here are descriptions of the main registers. Stack registers and segment registers will be covered later. General Purpose Registers --------------------------------------------------------------------- These are 16-bit registers. There are four general purpose registers; AX, BX, CX and DX. They are split up into 8-bit registers. AX is split up into AH which contains the high byte and AL which contains the low byte. On 386's and above there are also 32-bit registers, these have the same names as the 16-bit registers but with an 'E' in front i.e. EAX. You can use AL, AH, AX and EAX separatly and treat them as separate registers for some tasks. CPU registers are very special memory locations constructed from flip-flops. They are not part of main memory; the CPU implements them on-chip. Various members of the 80x86 family have different register sizes. The 886, 8286, 8486, and 8686 (x86 from now on) CPUs have exactly four registers, all 16 bits wide. All arithmetic and location operations occur in the CPU registers. CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 6 of 58 17/11/2005 08:08 p.m. Because the x86 processor has so few registers, we'll give each register its own name and refer to it by that name rather than its address. The names for the x86 registers are AX -The accumulator register BX -The base address register CX -The count register DX -The data register Besides the above registers, which are visible to the programmer, the x86 processors also have an instruction pointer register which contains the address of the next instruction to execute. There is also a flags register that holds the result of a comparison. The flags register remembers if one value was less than, equal to, or greater than another value. Because registers are on-chip and handled specially by the CPU, they are much faster than memory. Accessing a memory location requires one or more clock cycles. Accessing data in a register usually takes zero clock cycles. Therefore, you should try to keep variables in the registers. Register sets are very small and most registers have special purposes which limit their use as variables, but they are still an excellent place to store temporary data. If AX contained 24689 decimal: AH AL 01100000 01110001 AH would be 96 and AL would be 113. If you added one to AL it would be 114 and AH would be unchanged. SI, DI, SP and BP can also be used as general purpose registers but have more specific uses. They are not split into two halves. CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 7 of 58 17/11/2005 08:08 p.m. Index Registers --------------------------------------------------------------------- These are sometimes called pointer registers and they are 16-bit registers. They are mainly used for string instructions. There are three index registers SI (source index), DI (destination index) and IP (instruction pointer). On 386's and above there are also 32-bit index registers: EDI and ESI. You can also use BX to index strings. IP is a index register but it can't be manipulated directly as it stores the address of the next instruction. Stack registers --------------------------------------------------------------------- BP and SP are stack registers and are used when dealing with the stack. They will be covered when we talk about the stack later on. Segments and offsets --------------------------------------------------------------------- You cannot discuss memory addressing on the 80x86 processor family without first discussing segmentation. Among other things, segmentation provides a powerful memory management mechanism. It allows programmers to partition their programs into modules that operate independently of one another. Segments provide a way to easily implement object-oriented programs. Segments allow two processes to easily CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 8 of 58 17/11/2005 08:08 p.m. share data. All in all, segmentation is a really neat feature. On the other hand, if you ask ten programmers what they think of segmentation, at least nine of the ten will claim it's terrible. Why such a response? Well, it turns out that segmentation provides one other nifty feature: it allows you to extend the addressability of a processor. In the case of the 8086, segmentation let Intel's designers extend the maximum addressable memory from 64K to one megabyte. Gee, that sounds good. Why is everyone complaining? Well, a little history lesson is in order to understand what went wrong. In 1976, when Intel began designing the 8086 processor, memory was very expensive. Personal computers, such that they were at the time, typically had four thousand bytes of memory. Even when IBM introduced the PC five years later, 64K was still quite a bit of memory, one megabyte was a tremendous amount. Intel's designers felt that 64K memory would remain a large amount throughout the lifetime of the 8086. The only mistake they made was completely underestimating the lifetime of the 8086. They figured it would last about five years, like their earlier 8080 processor. They had plans for lots of other processors at the time, and "86" was not a suffix on the names of any of those. Intel figured they were set. Surely one megabyte would be more than enough to last until they came out with something better. Unfortunately, Intel didn't count on the IBM PC and the massive amount of software to appear for it. By 1983, it was very clear that Intel could not abandon the 80x86 architecture. They were stuck with it, but by then people were running up against the one megabyte limit of 8086. So Intel gave us the 80286. This processor could address up to 16 megabytes of memory. Surely more than enough. The only problem was that all that wonderful software written for the IBM PC was written in such a way that it couldn't take advantage of any memory beyond one megabyte. It turns out that the maximum amount of addressable memory is not everyone's main complaint. The real problem is that the 8086 was a 16 bit processor, with 16 bit registers and 16 bit addresses. This limited the processor to addressing 64K chunks of memory. Intel's clever use of segmentation extended this to one megabyte, but addressing more than 64K at one time takes some effort. Addressing more than 256K at one time takes a lot of effort. Despite what you might have heard, segmentation is not bad. In fact, it is a really great memory management scheme. What is bad is Intel's 1976 implementation of segmentation still in use today. You can't blame Intel for this - they fixed the problem in the 80's with the release of the 80386. The real culprit is MS-DOS that forces programmers to continue to use 1976 style segmentation. Fortunately, newer operating systems such as Linux, UNIX, Windows 9x, Windows NT, and CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 9 of 58 17/11/2005 08:08 p.m. OS/2 don't suffer from the same problems as MS-DOS. Furthermore, users finally seem to be more willing to switch to these newer operating systems so programmers can take advantage of the new features of the 80x86 family. With the history lesson aside, it's probably a good idea to figure out what segmentation is all about. Consider the current view of memory: it looks like a linear array of bytes. A single index (address) selects some particular byte from that array. Let's call this type of addressing linear or flat addressing. Segmented addressing uses two components to specify a memory location: a segment value and an offset within that segment. Ideally, the segment and offset values are independent of one another. The best way to describe segmented addressing is with a two-dimensional array. The segment provides one of the indices into the array, the offset provides the other: Now you may be wondering, "Why make this process more complex?" Linear addresses seem to work fine, why bother with this two dimensional addressing scheme? Well, let's consider the way you typically write a program. If you were to write, say, a SIN(X) routine and you needed some temporary variables, you probably would not use global variables. Instead, you would use local variables inside the SIN(X) function. In a broad sense, this is one of the features that segmentation offers - the ability to attach blocks of variables (a segment) to a particular piece of code. You could, for xample, have a segment containing local variables for SIN, a segment for SQRT, a segment for DRAWWindow, etc. Since the variables for SIN appear in the segment for SIN, it's less likely your SIN routine will affect the variables belonging to the SQRT routine. Indeed, on the 80286 and later operating in protected mode, the CPU can prevent one routine from accidentally modifying the variables in a different segment. A full segmented address contains a segment component and an offset component. This text will write segmented addresses as segment:offset. On the 8086 through the 80286, these two values are 16 bit constants. On the 80386 and later, the offset can be a 16 bit constant or a 32 bit constant. CA225 Assembly Language Programming http://www.computing.dcu.ie/%7Eray/CA225.html 10 of 58 17/11/2005 08:08 p.m. The size of the offset limits the maximum size of a segment. On the 8086 with 16 bit offsets, a segment may be no longer than 64K; it could be smaller (and most segments are), but never larger. The 80386 and later processors allow 32 bit offsets with segments as large as four gigabytes. The segment portion is 16 bits on all 80x86 processors. This lets a single program have up to 65,536 different segments in the program. Most programs have less than 16 segments (or thereabouts) so this isn't a practical limitation. Of course, despite the fact that the 80x86 family uses segmented addressing, the actual (physical) memory connected to the CPU is still a linear array of bytes. There is a function that converts the segment value to a physical memory address. The processor then adds the offset to this physical address to obtain the actual address of the data in memory. This text will refer to addresses in your programs as segmented addresses or logical addresses. The actual linear address that appears on the address bus is the physical address : On the 8086, 8088, 80186, and 80188 (and other processors operating in real mode), the function that maps a segment to a physical address is very simple. The CPU multiplies the segment value by sixteen (10h) and adds the offset portion. For example, consider the segmented address: 1000:1F00. To convert this to a physical address you multiply the segment value (1000h) by sixteen. Multiplying by the