Reverse Code Engineering RCE CD +sandman 2000

home *** CD-ROM | disk | FTP | other *** search

/ Reverse Code Engineering RCE CD +sandman 2000 / ReverseCodeEngineeringRceCdsandman2000.iso / RCE / Library / Manuels & Misc / Assembly / ASM-TUT.ZIP / CHAP20.DOC < prev next >

Wrap

Text File | 1990-07-26 | 13.2 KB | 447 lines

212 CHAPTER 20 - CONTROL STRUCTURES Control structures are things like 'for', 'do', 'while' and 'if' that you have in high level languages. It is possible to do these things in assembler language if you want. They are included here so if you ever want to use them at this level, you will have a template on how to set them up. All the examples will use integers only. If you have real numbers: if ( variable < 3.891 ) x = 4 * y ; it is too difficult to implement on the 8086. Of course, if you have an 8087, it's a piece of cake. IF IF is the easiest and it will illustrate how we are going to set up our labels. Let's take that first example. Since it is a waste of space to define variables over and over, we will assume that all variables are words, not bytes. The examples will all be written in C. if (variable < 7) { x = 4 + y ; } The assembler code looks like this: if1: ;---------------- cmp variable, 7 jge end_if1 mov x, 4 mov ax, y add x, ax end_if1: ;---------------- All labels will look like variations of this. The control words will be the labels.{1} They will have a unique number for each block so you can write as many different blocks as you want; the label that signals the end of the block will be the word 'end_' followed by the control word. The semicolon with the line is a ____________________ 1. The words 'if' and 'endif' are Microsoft macro directives, so if you use them without a tag number, you will get some bizarre code indeed. ______________________ The PC Assembler Tutor - Copyright (C) 1989 Chuck Nelson Chapter 20 - Control Structures 213 _______________________________ comment, and it is used for visual separation. By using a different number each time, you may nest as deeply as you want, though the code will be pretty unreadable: if (variable < 7 ) { if ( x > 9 ) { if ( z == 0 ) { y = 12 ; } q = 5 ; } r = 7 ; } becomes: if1: ;--------------- cmp variable, 7 jge end_if1 if2: ;--------------- cmp x, 9 jle end_if2 if3: ;--------------- cmp z, 0 jne end_if3 mov y, 12 end_if3: ;----------- mov q, 5 end_if2: ;----------- mov r, 7 end_if1: ;----------- With the Microsoft assembler, you don't need to have the labels start at the beginning of the line, but they MUST be the first thing on the line. You can indent them: if1: ; -------------- cmp variable, 7 jge end_if1 if2: ;------------- cmp x, 9 jle end_if2 if3: ;------------ cmp z, 0 jne end_if3 mov y, 12 The PC Assembler Tutor 214 ______________________ end_if3: ;-------- mov q, 5 end_if2: ;--------- mov r, 7 end_if1: ;--------- You will notice that they are all compares and jumps using reverse logic. If the condition is NOT true, then we jump. If it is true, we just go on. What about IF ELSE? It looks almost the same. if ( j < 12 ) { y = 19 ; } else { z = 25 ; } We get: if16: ;---------------- cmp j, 12 jge else16 mov y, 19 jmp end_if16 else16: mov z, 25 end_else16: end_if16: ;--------------- We jump to another part inside the block, and throw in a JMP after the IF part. WHILE WHILE is the same as IF except that at the end of the block, we jump back to the beginning. while ( j < 20 ) { j = j + 1 ; } while25: ;------------ cmp j, 20 jge end_while25 inc j jmp while25 end_while25: ;----------- Chapter 20 - Control Structures 215 _______________________________ DO WHILE DO WHILE always goes through the code once. The decision process is at the end. do { k = k + 7 ; } while (k < 0) ; do_while74: ;------------ add k, 7 cmp k, 0 jl do_while74 end_do_while74: ;------------ Notice that here the jump is a positive decision. If the condition is fulfilled, we jump back to the start, otherwise we fall through. What do you do with 'break' and 'continue'? {2} Let's put them in: do { k = k + 7 ; if ( y < 9 ) { continue ; } j = j - 6 ; if ( z == 5 ) { break ; } y = y * 2 ; } while ( j < 17) ; In assembler language, this becomes: do_while143: ;---------- add k, 7 if144: ;----------- cmp y, 9 jge end_if144 jmp continue143 end_if144: ;----------- sub j, 6 ____________________ 2. For those of you who are not C people, BREAK breaks you out of the innermost loop. CONTINUE skips all the code till the end of the loop. The PC Assembler Tutor 216 ______________________ if145: ;----------- cmp z, 5 jne end_if145 jmp break143 end_if145: ;----------- sal y, 1 ; shift left 1 = multiply by 2 continue143: cmp j, 17 jl do_while143 end_do_while143: break143: In DO WHILE, CONTINUE is the label just before the decision process. BREAK is always the first label after the block. We can just use the label that marks the end of our block for the break. Instead of: jmp break143 we can have: jmp end_do_while143 From now on, we'll put in the continue and break, even if they aren't used. Once you are used to it, you can drop the break label and use the end of block label. Going back to the WHILE statement, we have: while25: ;------------ cmp j, 20 jge end_while25 inc j continue25: jmp while25 end_while25: ;----------- break25: FOR The FOR statement in C is more sophisticated than in other languages, but we will keep the example simple. It has 3 parts: for ( i = 11 ; i < 20 ; i = i + 1 ) can be looked at as: for ( initialize ; test ; update ) The reason that it is sophisticated is that it can have any number of things in the first and third parts. Chapter 20 - Control Structures 217 _______________________________ for ( i = 11, j = 27 z = 4 ; i < 20 ; k = k + 1, j = j/2) is legitimate. Therefore we need to set aside a block for the initialize part, a block for the test part, and a block for the update part. One question is whether we want the initialization inside the FOR label. I'm not doing it, but it is a possibility. Here's the C code. for ( i = 11 ; i < 20 ; i = i + 1 ) { k = k + 9 } And its corresponding assembler code: init217: ;--------------- mov i, 11 end_init217: ;------ for217: ; --------------- test217: cmp i, 20 jge end_for217 end_test217: ;----- add k,9 continue217: update217: inc i end_update217: ; ---- jmp test217 end_for217: ;------------- break217: My that's a lot of labels. In this example we would get rid of most of them because there is so little code, but if the block contained a lot of code they would be a help, not a hindrance. Remember, a label generates no code, so this makes no difference in the size of the object file. Some of the labels which are not targets of jumps could be made into comments. end_update217: ;---- can become: ; end_update217 ---- Whichever way looks better to you. Using comments instead of labels gives you a slightly faster compile time. Just make sure you don't change one that IS a target of a jump. SWITCH Finally there is the switch statement. If you don't know what it is, skip this section because it will probably be confusing. The PC Assembler Tutor 218 ______________________ switch ( k ) { case 'A': x = x + 9 ; break ; case '&': y = y * 2 ; break ; case 'j': z = z - 7 ; break ; } We get the following: switch82: ;----------------- cmp k, 'A' jne test82_2 jmp case82_1 test82_2 cmp k, '&' jne test82_3 jmp case82_2 test82_3 cmp k, 'j' jne end_test82 jmp case82_3 end_test82: jmp deafult82 ; ---------- case82_1: add x, 9 jmp break82 case82_2: sal y, 1 ; multiply by 2 jmp break82 case82_3: sub z, 7 jmp break82 default82: jmp break82 end_switch82: ;------------ break82: It may look like there are unnecessary jumps at the beginning, but it is likely that in a real program some of the case statements would be more than +127 bytes away from the conditional jumps, so you need JMP, which can go anywhere in the segment.