1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
4 <title>Stacker: An Example Of Using LLVM</title>
5 <link rel="stylesheet" href="llvm.css" type="text/css">
8 <div class="doc_title">Stacker: An Example Of Using LLVM</div>
11 <li><a href="#abstract">Abstract</a></li>
12 <li><a href="#introduction">Introduction</a></li>
13 <li><a href="#lessons">Lessons I Learned About LLVM</a>
15 <li><a href="#value">Everything's a Value!</a></li>
16 <li><a href="#terminate">Terminate Those Blocks!</a></li>
17 <li><a href="#blocks">Concrete Blocks</a></li>
18 <li><a href="#push_back">push_back Is Your Friend</a></li>
19 <li><a href="#gep">The Wily GetElementPtrInst</a></li>
20 <li><a href="#linkage">Getting Linkage Types Right</a></li>
21 <li><a href="#constants">Constants Are Easier Than That!</a></li>
24 <li><a href="#lexicon">The Stacker Lexicon</a>
26 <li><a href="#stack">The Stack</a>
27 <li><a href="#punctuation">Punctuation</a>
28 <li><a href="#literals">Literals</a>
29 <li><a href="#words">Words</a>
30 <li><a href="#builtins">Built-Ins</a>
33 <li><a href="#example">Prime: A Complete Example</a></li>
34 <li><a href="#internal">Internal Code Details</a>
36 <li><a href="#directory">The Directory Structure </a></li>
37 <li><a href="#lexer">The Lexer</a></li>
38 <li><a href="#parser">The Parser</a></li>
39 <li><a href="#compiler">The Compiler</a></li>
40 <li><a href="#runtime">The Runtime</a></li>
41 <li><a href="#driver">Compiler Driver</a></li>
42 <li><a href="#tests">Test Programs</a></li>
46 <div class="doc_text">
47 <p><b>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> </b></p>
51 <!-- ======================================================================= -->
52 <div class="doc_section"> <a name="abstract">Abstract </a></div>
53 <div class="doc_text">
54 <p>This document is another way to learn about LLVM. Unlike the
55 <a href="LangRef.html">LLVM Reference Manual</a> or
56 <a href="ProgrammersManual.html">LLVM Programmer's Manual</a>, this
57 document walks you through the implementation of a programming language
58 named Stacker. Stacker was invented specifically as a demonstration of
59 LLVM. The emphasis in this document is not on describing the
60 intricacies of LLVM itself, but on how to use it to build your own
63 <!-- ======================================================================= -->
64 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
65 <div class="doc_text">
66 <p>Amongst other things, LLVM is a platform for compiler writers.
67 Because of its exceptionally clean and small IR (intermediate
68 representation), compiler writing with LLVM is much easier than with
69 other system. As proof, the author of Stacker wrote the entire
70 compiler (language definition, lexer, parser, code generator, etc.) in
71 about <em>four days</em>! That's important to know because it shows
72 how quickly you can get a new
73 language up when using LLVM. Furthermore, this was the <em >first</em>
74 language the author ever created using LLVM. The learning curve is
75 included in that four days.</p>
76 <p>The language described here, Stacker, is Forth-like. Programs
77 are simple collections of word definitions and the only thing definitions
78 can do is manipulate a stack or generate I/O. Stacker is not a "real"
79 programming language; its very simple. Although it is computationally
80 complete, you wouldn't use it for your next big project. However,
81 the fact that it is complete, its simple, and it <em>doesn't</em> have
82 a C-like syntax make it useful for demonstration purposes. It shows
83 that LLVM could be applied to a wide variety of language syntaxes.</p>
84 <p>The basic notions behind stacker is very simple. There's a stack of
85 integers (or character pointers) that the program manipulates. Pretty
86 much the only thing the program can do is manipulate the stack and do
87 some limited I/O operations. The language provides you with several
88 built-in words that manipulate the stack in interesting ways. To get
89 your feet wet, here's how you write the traditional "Hello, World"
90 program in Stacker:</p>
91 <p><code>: hello_world "Hello, World!" >s DROP CR ;<br>
92 : MAIN hello_world ;<br></code></p>
93 <p>This has two "definitions" (Stacker manipulates words, not
94 functions and words have definitions): <code>MAIN</code> and <code>
95 hello_world</code>. The <code>MAIN</code> definition is standard, it
96 tells Stacker where to start. Here, <code>MAIN</code> is defined to
97 simply invoke the word <code>hello_world</code>. The
98 <code>hello_world</code> definition tells stacker to push the
99 <code>"Hello, World!"</code> string onto the stack, print it out
100 (<code>>s</code>), pop it off the stack (<code>DROP</code>), and
101 finally print a carriage return (<code>CR</code>). Although
102 <code>hello_world</code> uses the stack, its net effect is null. Well
103 written Stacker definitions have that characteristic. </p>
104 <p>Exercise for the reader: how could you make this a one line program?</p>
106 <!-- ======================================================================= -->
107 <div class="doc_section"><a name="lessons"></a>Lessons I Learned About LLVM</div>
108 <div class="doc_text">
109 <p>Stacker was written for two purposes: (a) to get the author over the
110 learning curve and (b) to provide a simple example of how to write a compiler
111 using LLVM. During the development of Stacker, many lessons about LLVM were
112 learned. Those lessons are described in the following subsections.<p>
114 <!-- ======================================================================= -->
115 <div class="doc_subsection"><a name="value"></a>Everything's a Value!</div>
116 <div class="doc_text">
117 <p>Although I knew that LLVM used a Single Static Assignment (SSA) format,
118 it wasn't obvious to me how prevalent this idea was in LLVM until I really
119 started using it. Reading the Programmer's Manual and Language Reference I
120 noted that most of the important LLVM IR (Intermediate Representation) C++
121 classes were derived from the Value class. The full power of that simple
122 design only became fully understood once I started constructing executable
123 expressions for Stacker.</p>
124 <p>This really makes your programming go faster. Think about compiling code
125 for the following C/C++ expression: (a|b)*((x+1)/(y+1)). You could write a
126 function using LLVM that does exactly that, this way:</p>
129 expression(BasicBlock*bb, Value* a, Value* b, Value* x, Value* y )
131 Instruction* tail = bb->getTerminator();
132 ConstantSInt* one = ConstantSInt::get( Type::IntTy, 1);
133 BinaryOperator* or1 =
134 new BinaryOperator::create( Instruction::Or, a, b, "", tail );
135 BinaryOperator* add1 =
136 new BinaryOperator::create( Instruction::Add, x, one, "", tail );
137 BinaryOperator* add2 =
138 new BinaryOperator::create( Instruction::Add, y, one, "", tail );
139 BinaryOperator* div1 =
140 new BinaryOperator::create( Instruction::Div, add1, add2, "", tail);
141 BinaryOperator* mult1 =
142 new BinaryOperator::create( Instruction::Mul, or1, div1, "", tail );
147 <p>"Okay, big deal," you say. It is a big deal. Here's why. Note that I didn't
148 have to tell this function which kinds of Values are being passed in. They could be
149 instructions, Constants, Global Variables, etc. Furthermore, if you specify Values
150 that are incorrect for this sequence of operations, LLVM will either notice right
151 away (at compilation time) or the LLVM Verifier will pick up the inconsistency
152 when the compiler runs. In no case will you make a type error that gets passed
153 through to the generated program. This <em>really</em> helps you write a compiler
154 that always generates correct code!<p>
155 <p>The second point is that we don't have to worry about branching, registers,
156 stack variables, saving partial results, etc. The instructions we create
157 <em>are</em> the values we use. Note that all that was created in the above
158 code is a Constant value and five operators. Each of the instructions <em>is</em>
159 the resulting value of that instruction.</p>
160 <p>The lesson is this: <em>SSA form is very powerful: there is no difference
161 between a value and the instruction that created it.</em> This is fully
162 enforced by the LLVM IR. Use it to your best advantage.</p>
164 <!-- ======================================================================= -->
165 <div class="doc_subsection"><a name="terminate"></a>Terminate Those Blocks!</div>
166 <div class="doc_text">
167 <p>I had to learn about terminating blocks the hard way: using the debugger
168 to figure out what the LLVM verifier was trying to tell me and begging for
169 help on the LLVMdev mailing list. I hope you avoid this experience.</p>
170 <p>Emblazon this rule in your mind:</p>
172 <li><em>All</em> <code>BasicBlock</code>s in your compiler <b>must</b> be
173 terminated with a terminating instruction (branch, return, etc.).
176 <p>Terminating instructions are a semantic requirement of the LLVM IR. There
177 is no facility for implicitly chaining together blocks placed into a function
178 in the order they occur. Indeed, in the general case, blocks will not be
179 added to the function in the order of execution because of the recursive
180 way compilers are written.</p>
181 <p>Furthermore, if you don't terminate your blocks, your compiler code will
182 compile just fine. You won't find out about the problem until you're running
183 the compiler and the module you just created fails on the LLVM Verifier.</p>
185 <!-- ======================================================================= -->
186 <div class="doc_subsection"><a name="blocks"></a>Concrete Blocks</div>
187 <div class="doc_text">
188 <p>After a little initial fumbling around, I quickly caught on to how blocks
189 should be constructed. The use of the standard template library really helps
190 simply the interface. In general, here's what I learned:
192 <li><em>Create your blocks early.</em> While writing your compiler, you
193 will encounter several situations where you know apriori that you will
194 need several blocks. For example, if-then-else, switch, while and for
195 statements in C/C++ all need multiple blocks for expression in LVVM.
196 The rule is, create them early.</li>
197 <li><em>Terminate your blocks early.</em> This just reduces the chances
198 that you forget to terminate your blocks which is required (go
199 <a href="#terminate">here</a> for more).
200 <li><em>Use getTerminator() for instruction insertion.</em> I noticed early on
201 that many of the constructors for the Instruction classes take an optional
202 <code>insert_before</code> argument. At first, I thought this was a mistake
203 because clearly the normal mode of inserting instructions would be one at
204 a time <em>after</em> some other instruction, not <em>before</em>. However,
205 if you hold on to your terminating instruction (or use the handy dandy
206 <code>getTerminator()</code> method on a <code>BasicBlock</code>), it can
207 always be used as the <code>insert_before</code> argument to your instruction
208 constructors. This causes the instruction to automatically be inserted in
209 the RightPlace&tm; place, just before the terminating instruction. The
210 nice thing about this design is that you can pass blocks around and insert
211 new instructions into them without ever known what instructions came
212 before. This makes for some very clean compiler design.</li>
214 <p>The foregoing is such an important principal, its worth making an idiom:</p>
217 BasicBlock* bb = new BasicBlock();</li>
218 bb->getInstList().push_back( new Branch( ... ) );
219 new Instruction(..., bb->getTerminator() );
222 <p>To make this clear, consider the typical if-then-else statement
223 (see StackerCompiler::handle_if() method). We can set this up
224 in a single function using LLVM in the following way: </p>
226 using namespace llvm;
228 MyCompiler::handle_if( BasicBlock* bb, SetCondInst* condition )
230 // Create the blocks to contain code in the structure of if/then/else
231 BasicBlock* then = new BasicBlock();
232 BasicBlock* else = new BasicBlock();
233 BasicBlock* exit = new BasicBlock();
235 // Insert the branch instruction for the "if"
236 bb->getInstList().push_back( new BranchInst( then, else, condition ) );
238 // Set up the terminating instructions
239 then->getInstList().push_back( new BranchInst( exit ) );
240 else->getInstList().push_back( new BranchInst( exit ) );
242 // Fill in the then part .. details excised for brevity
243 this->fill_in( then );
245 // Fill in the else part .. details excised for brevity
246 this->fill_in( else );
248 // Return a block to the caller that can be filled in with the code
249 // that follows the if/then/else construct.
253 <p>Presumably in the foregoing, the calls to the "fill_in" method would add
254 the instructions for the "then" and "else" parts. They would use the third part
255 of the idiom almost exclusively (inserting new instructions before the
256 terminator). Furthermore, they could even recurse back to <code>handle_if</code>
257 should they encounter another if/then/else statement and it will all "just work".
259 <p>Note how cleanly this all works out. In particular, the push_back methods on
260 the <code>BasicBlock</code>'s instruction list. These are lists of type
261 <code>Instruction</code> which also happen to be <code>Value</code>s. To create
262 the "if" branch we merely instantiate a <code>BranchInst</code> that takes as
263 arguments the blocks to branch to and the condition to branch on. The blocks
264 act like branch labels! This new <code>BranchInst</code> terminates
265 the <code>BasicBlock</code> provided as an argument. To give the caller a way
266 to keep inserting after calling <code>handle_if</code> we create an "exit" block
267 which is returned to the caller. Note that the "exit" block is used as the
268 terminator for both the "then" and the "else" blocks. This gaurantees that no
269 matter what else "handle_if" or "fill_in" does, they end up at the "exit" block.
272 <!-- ======================================================================= -->
273 <div class="doc_subsection"><a name="push_back"></a>push_back Is Your Friend</div>
274 <div class="doc_text">
276 One of the first things I noticed is the frequent use of the "push_back"
277 method on the various lists. This is so common that it is worth mentioning.
278 The "push_back" inserts a value into an STL list, vector, array, etc. at the
279 end. The method might have also been named "insert_tail" or "append".
280 Althought I've used STL quite frequently, my use of push_back wasn't very
281 high in other programs. In LLVM, you'll use it all the time.
284 <!-- ======================================================================= -->
285 <div class="doc_subsection"><a name="gep"></a>The Wily GetElementPtrInst</div>
286 <div class="doc_text">
288 It took a little getting used to and several rounds of postings to the LLVM
289 mail list to wrap my head around this instruction correctly. Even though I had
290 read the Language Reference and Programmer's Manual a couple times each, I still
291 missed a few <em>very</em> key points:
294 <li>GetElementPtrInst gives you back a Value for the last thing indexed</em>
295 <li>All global variables in LLVM are <em>pointers</em>.
296 <li>Pointers must also be dereferenced with the GetElementPtrInst instruction.
298 <p>This means that when you look up an element in the global variable (assuming
299 its a struct or array), you <em>must</em> deference the pointer first! For many
300 things, this leads to the idiom:
303 std::vector<Value*> index_vector;
304 index_vector.push_back( ConstantSInt::get( Type::LongTy, 0 );
305 // ... push other indices ...
306 GetElementPtrInst* gep = new GetElementPtrInst( ptr, index_vector );
308 <p>For example, suppose we have a global variable whose type is [24 x int]. The
309 variable itself represents a <em>pointer</em> to that array. To subscript the
310 array, we need two indices, not just one. The first index (0) dereferences the
311 pointer. The second index subscripts the array. If you're a "C" programmer, this
312 will run against your grain because you'll naturally think of the global array
313 variable and the address of its first element as the same. That tripped me up
314 for a while until I realized that they really do differ .. by <em>type</em>.
315 Remember that LLVM is a strongly typed language itself. Absolutely everything
316 has a type. The "type" of the global variable is [24 x int]*. That is, its
317 a pointer to an array of 24 ints. When you dereference that global variable with
318 a single index, you now have a " [24 x int]" type, the pointer is gone. Although
319 the pointer value of the dereferenced global and the address of the zero'th element
320 in the array will be the same, they differ in their type. The zero'th element has
321 type "int" while the pointer value has type "[24 x int]".</p>
322 <p>Get this one aspect of LLVM right in your head and you'll save yourself
323 a lot of compiler writing headaches down the road.</p>
325 <!-- ======================================================================= -->
326 <div class="doc_subsection"><a name="linkage"></a>Getting Linkage Types Right</div>
327 <div class="doc_text">
328 <p>Linkage types in LLVM can be a little confusing, especially if your compiler
329 writing mind has affixed very hard concepts to particular words like "weak",
330 "external", "global", "linkonce", etc. LLVM does <em>not</em> use the precise
331 definitions of say ELF or GCC even though they share common terms. To be fair,
332 the concepts are related and similar but not precisely the same. This can lead
333 you to think you know what a linkage type represents but in fact it is slightly
334 different. I recommend you read the
335 <a href="LangRef.html#linkage"> Language Reference on this topic</a> very
337 <p>Here are some handy tips that I discovered along the way:</p>
339 <li>Unitialized means external. That is, the symbol is declared in the current
340 module and can be used by that module but it is not defined by that module.</li>
341 <li>Setting an initializer changes a global's linkage type from whatever it was
342 to a normal, defind global (not external). You'll need to call the setLinkage()
343 method to reset it if you specify the initializer after the GlobalValue has been
344 constructed. This is important for LinkOnce and Weak linkage types.</li>
345 <li>Appending linkage can be used to keep track of compilation information at
346 runtime. It could be used, for example, to build a full table of all the C++
347 virtual tables or hold the C++ RTTI data, or whatever. Appending linkage can
348 only be applied to arrays. The arrays are concatenated together at link time.</li>
351 <!-- ======================================================================= -->
352 <div class="doc_subsection"><a name="constants"></a>Constants Are Easier Than That!</div>
353 <div class="doc_text">
355 Constants in LLVM took a little getting used to until I discovered a few utility
356 functions in the LLVM IR that make things easier. Here's what I learned: </p>
358 <li>Constants are Values like anything else and can be operands of instructions</li>
359 <li>Integer constants, frequently needed can be created using the static "get"
360 methods of the ConstantInt, ConstantSInt, and ConstantUInt classes. The nice thing
361 about these is that you can "get" any kind of integer quickly.</li>
362 <li>There's a special method on Constant class which allows you to get the null
363 constant for <em>any</em> type. This is really handy for initializing large
364 arrays or structures, etc.</li>
367 <!-- ======================================================================= -->
368 <div class="doc_section"> <a name="lexicon">The Stacker Lexicon</a></div>
369 <div class="doc_subsection"><a name="stack"></a>The Stack</div>
370 <div class="doc_text">
371 <p>Stacker definitions define what they do to the global stack. Before
372 proceeding, a few words about the stack are in order. The stack is simply
373 a global array of 32-bit integers or pointers. A global index keeps track
374 of the location of the to of the stack. All of this is hidden from the
375 programmer but it needs to be noted because it is the foundation of the
376 conceptual programming model for Stacker. When you write a definition,
377 you are, essentially, saying how you want that definition to manipulate
378 the global stack.</p>
379 <p>Manipulating the stack can be quite hazardous. There is no distinction
380 given and no checking for the various types of values that can be placed
381 on the stack. Automatic coercion between types is performed. In many
382 cases this is useful. For example, a boolean value placed on the stack
383 can be interpreted as an integer with good results. However, using a
384 word that interprets that boolean value as a pointer to a string to
385 print out will almost always yield a crash. Stacker simply leaves it
386 to the programmer to get it right without any interference or hindering
387 on interpretation of the stack values. You've been warned :) </p>
389 <!-- ======================================================================= -->
390 <div class="doc_subsection"> <a name="punctuation"></a>Punctuation</div>
391 <div class="doc_text">
392 <p>Punctuation in Stacker is very simple. The colon and semi-colon
393 characters are used to introduce and terminate a definition
394 (respectively). Except for <em>FORWARD</em> declarations, definitions
395 are all you can specify in Stacker. Definitions are read left to right.
396 Immediately after the semi-colon comes the name of the word being defined.
397 The remaining words in the definition specify what the word does.</p>
399 <!-- ======================================================================= -->
400 <div class="doc_subsection"><a name="literals"></a>Literals</div>
401 <div class="doc_text">
402 <p>There are three kinds of literal values in Stacker. Integer, Strings,
403 and Booleans. In each case, the stack operation is to simply push the
404 value onto the stack. So, for example:<br/>
405 <code> 42 " is the answer." TRUE </code><br/>
406 will push three values onto the stack: the integer 42, the
407 string " is the answer." and the boolean TRUE.</p>
409 <!-- ======================================================================= -->
410 <div class="doc_subsection"><a name="words"></a>Words</div>
411 <div class="doc_text">
412 <p>Each definition in Stacker is composed of a set of words. Words are
413 read and executed in order from left to right. There is very little
414 checking in Stacker to make sure you're doing the right thing with
415 the stack. It is assumed that the programmer knows how the stack
416 transformation he applies will affect the program.</p>
417 <p>Words in a definition come in two flavors: built-in and programmer
418 defined. Simply mentioning the name of a previously defined or declared
419 programmer-defined word causes that words definition to be invoked. It
420 is somewhat like a function call in other languages. The built-in
421 words have various effects, described below.</p>
422 <p>Sometimes you need to call a word before it is defined. For this, you can
423 use the <code>FORWARD</code> declaration. It looks like this</p>
424 <p><code>FORWARD name ;</code></p>
425 <p>This simply states to Stacker that "name" is the name of a definition
426 that is defined elsewhere. Generally it means the definition can be found
427 "forward" in the file. But, it doesn't have to be in the current compilation
428 unit. Anything declared with <code>FORWARD</code> is an external symbol for
431 <!-- ======================================================================= -->
432 <div class="doc_subsection"><a name="builtins"></a>Built In Words</div>
433 <div class="doc_text">
434 <p>The built-in words of the Stacker language are put in several groups
435 depending on what they do. The groups are as follows:</p>
437 <li><em>Logical</em>These words provide the logical operations for
438 comparing stack operands.<br/>The words are: < > <= >=
439 = <> true false.</li>
440 <li><em>Bitwise</em>These words perform bitwise computations on
441 their operands. <br/> The words are: << >> XOR AND NOT</li>
442 <li><em>Arithmetic</em>These words perform arithmetic computations on
443 their operands. <br/> The words are: ABS NEG + - * / MOD */ ++ -- MIN MAX</li>
444 <li><em>Stack</em>These words manipulate the stack directly by moving
445 its elements around.<br/> The words are: DROP DUP SWAP OVER ROT DUP2 DROP2 PICK TUCK</li>
446 <li><em>Memory</em>These words allocate, free and manipulate memory
447 areas outside the stack.<br/>The words are: MALLOC FREE GET PUT</li>
448 <li><em>Control</em>These words alter the normal left to right flow
449 of execution.<br/>The words are: IF ELSE ENDIF WHILE END RETURN EXIT RECURSE</li>
450 <li><em>I/O</em> These words perform output on the standard output
451 and input on the standard input. No other I/O is possible in Stacker.
452 <br/>The words are: SPACE TAB CR >s >d >c <s <d <c.</li>
454 <p>While you may be familiar with many of these operations from other
455 programming languages, a careful review of their semantics is important
456 for correct programming in Stacker. Of most importance is the effect
457 that each of these built-in words has on the global stack. The effect is
458 not always intuitive. To better describe the effects, we'll borrow from Forth the idiom of
459 describing the effect on the stack with:</p>
460 <p><code> BEFORE -- AFTER </code></p>
461 <p>That is, to the left of the -- is a representation of the stack before
462 the operation. To the right of the -- is a representation of the stack
463 after the operation. In the table below that describes the operation of
464 each of the built in words, we will denote the elements of the stack
465 using the following construction:</p>
467 <li><em>b</em> - a boolean truth value</li>
468 <li><em>w</em> - a normal integer valued word.</li>
469 <li><em>s</em> - a pointer to a string value</li>
470 <li><em>p</em> - a pointer to a malloc's memory block</li>
473 <div class="doc_text">
474 <table class="doc_table" >
475 <tr class="doc_table"><td colspan="4">Definition Of Operation Of Built In Words</td></tr>
476 <tr class="doc_table"><td colspan="4">LOGICAL OPERATIONS</td></tr>
477 <tr class="doc_table"><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
478 <tr class="doc_table"><td><</td>
481 <td>Two values (w1 and w2) are popped off the stack and
482 compared. If w1 is less than w2, TRUE is pushed back on
483 the stack, otherwise FALSE is pushed back on the stack.</td>
488 <td>Two values (w1 and w2) are popped off the stack and
489 compared. If w1 is greater than w2, TRUE is pushed back on
490 the stack, otherwise FALSE is pushed back on the stack.</td>
495 <td>Two values (w1 and w2) are popped off the stack and
496 compared. If w1 is greater than or equal to w2, TRUE is
497 pushed back on the stack, otherwise FALSE is pushed back
503 <td>Two values (w1 and w2) are popped off the stack and
504 compared. If w1 is less than or equal to w2, TRUE is
505 pushed back on the stack, otherwise FALSE is pushed back
511 <td>Two values (w1 and w2) are popped off the stack and
512 compared. If w1 is equal to w2, TRUE is
513 pushed back on the stack, otherwise FALSE is pushed back
516 <tr><td><></td>
519 <td>Two values (w1 and w2) are popped off the stack and
520 compared. If w1 is equal to w2, TRUE is
521 pushed back on the stack, otherwise FALSE is pushed back
527 <td>The boolean value FALSE (0) is pushed onto the stack.</td>
532 <td>The boolean value TRUE (-1) is pushed onto the stack.</td>
534 <tr><td colspan="4">BITWISE OPERATIONS</td></tr>
535 <tr><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
536 <tr><td><<</td>
538 <td>w1 w2 -- w1<<w2</td>
539 <td>Two values (w1 and w2) are popped off the stack. The w2
540 operand is shifted left by the number of bits given by the
541 w1 operand. The result is pushed back to the stack.</td>
543 <tr><td>>></td>
545 <td>w1 w2 -- w1>>w2</td>
546 <td>Two values (w1 and w2) are popped off the stack. The w2
547 operand is shifted right by the number of bits given by the
548 w1 operand. The result is pushed back to the stack.</td>
552 <td>w1 w2 -- w2|w1</td>
553 <td>Two values (w1 and w2) are popped off the stack. The values
554 are bitwise OR'd together and pushed back on the stack. This is
555 not a logical OR. The sequence 1 2 OR yields 3 not 1.</td>
559 <td>w1 w2 -- w2&w1</td>
560 <td>Two values (w1 and w2) are popped off the stack. The values
561 are bitwise AND'd together and pushed back on the stack. This is
562 not a logical AND. The sequence 1 2 AND yields 0 not 1.</td>
566 <td>w1 w2 -- w2^w1</td>
567 <td>Two values (w1 and w2) are popped off the stack. The values
568 are bitwise exclusive OR'd together and pushed back on the stack.
569 For example, The sequence 1 3 XOR yields 2.</td>
571 <tr><td colspan="4">ARITHMETIC OPERATIONS</td></tr>
572 <tr><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
576 <td>One value s popped off the stack; its absolute value is computed
577 and then pushed onto the stack. If w1 is -1 then w2 is 1. If w1 is
578 1 then w2 is also 1.</td>
583 <td>One value is popped off the stack which is negated and then
584 pushed back onto the stack. If w1 is -1 then w2 is 1. If w1 is
585 1 then w2 is -1.</td>
589 <td>w1 w2 -- w2+w1</td>
590 <td>Two values are popped off the stack. Their sum is pushed back
595 <td>w1 w2 -- w2-w1</td>
596 <td>Two values are popped off the stack. Their difference is pushed back
601 <td>w1 w2 -- w2*w1</td>
602 <td>Two values are popped off the stack. Their product is pushed back
607 <td>w1 w2 -- w2/w1</td>
608 <td>Two values are popped off the stack. Their quotient is pushed back
613 <td>w1 w2 -- w2%w1</td>
614 <td>Two values are popped off the stack. Their remainder after division
615 of w1 by w2 is pushed back onto the stack</td>
619 <td>w1 w2 w3 -- (w3*w2)/w1</td>
620 <td>Three values are popped off the stack. The product of w1 and w2 is
621 divided by w3. The result is pushed back onto the stack.</td>
626 <td>One value is popped off the stack. It is incremented by one and then
627 pushed back onto the stack.</td>
632 <td>One value is popped off the stack. It is decremented by one and then
633 pushed back onto the stack.</td>
637 <td>w1 w2 -- (w2<w1?w2:w1)</td>
638 <td>Two values are popped off the stack. The larger one is pushed back
643 <td>w1 w2 -- (w2>w1?w2:w1)</td>
644 <td>Two values are popped off the stack. The larger value is pushed back
647 <tr><td colspan="4">STACK MANIPULATION OPERATIONS</td></tr>
648 <tr><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
652 <td>One value is popped off the stack.</td>
657 <td>Two values are popped off the stack.</td>
662 <td>The second value on the stack is removed from the stack. That is,
663 a value is popped off the stack and retained. Then a second value is
664 popped and the retained value is pushed.</td>
668 <td>w1 w2 w3 w4 -- w3 w4</td>
669 <td>The third and fourth values on the stack are removed from it. That is,
670 two values are popped and retained. Then two more values are popped and
671 the two retained values are pushed back on.</td>
676 <td>One value is popped off the stack. That value is then pushed onto
677 the stack twice to duplicate the top stack vaue.</td>
681 <td>w1 w2 -- w1 w2 w1 w2</td>
682 <td>The top two values on the stack are duplicated. That is, two vaues
683 are popped off the stack. They are alternately pushed back on the
684 stack twice each.</td>
688 <td>w1 w2 -- w2 w1</td>
689 <td>The top two stack items are reversed in their order. That is, two
690 values are popped off the stack and pushed back onto the stack in
691 the opposite order they were popped.</td>
695 <td>w1 w2 w3 w4 -- w3 w4 w2 w1</td>
696 <td>The top four stack items are swapped in pairs. That is, two values
697 are popped and retained. Then, two more values are popped and retained.
698 The values are pushed back onto the stack in the reverse order but
703 <td>w1 w2-- w1 w2 w1</td>
704 <td>Two values are popped from the stack. They are pushed back
705 onto the stack in the order w1 w2 w1. This seems to cause the
706 top stack element to be duplicated "over" the next value.</td>
710 <td>w1 w2 w3 w4 -- w1 w2 w3 w4 w1 w2</td>
711 <td>The third and fourth values on the stack are replicated onto the
712 top of the stack</td>
716 <td>w1 w2 w3 -- w2 w3 w1</td>
717 <td>The top three values are rotated. That is, three value are popped
718 off the stack. They are pushed back onto the stack in the order
723 <td>w1 w2 w3 w4 w5 w6 -- w3 w4 w5 w6 w1 w2</td>
724 <td>Like ROT but the rotation is done using three pairs instead of
729 <td>w1 w2 w3 -- w2 w3 w1</td>
730 <td>Reverse rotation. Like ROT, but it rotates the other way around.
731 Essentially, the third element on the stack is moved to the top
736 <td>w1 w2 w3 w4 w5 w6 -- w3 w4 w5 w6 w1 w2</td>
737 <td>Double reverse rotation. Like RROT but the rotation is done using
738 three pairs instead of three singles. The fifth and sixth stack
739 elements are moved to the first and second positions</td>
743 <td>w1 w2 -- w2 w1 w2</td>
744 <td>Similar to OVER except that the second operand is being
745 replicated. Essentially, the first operand is being "tucked"
746 in between two instances of the second operand. Logically, two
747 values are popped off the stack. They are placed back on the
748 stack in the order w2 w1 w2.</td>
752 <td>w1 w2 w3 w4 -- w3 w4 w1 w2 w3 w4</td>
753 <td>Like TUCK but a pair of elements is tucked over two pairs.
754 That is, the top two elements of the stack are duplicated and
755 inserted into the stack at the fifth and positions.</td>
759 <td>x0 ... Xn n -- x0 ... Xn x0</td>
760 <td>The top of the stack is used as an index into the remainder of
761 the stack. The element at the nth position replaces the index
762 (top of stack). This is useful for cycling through a set of
763 values. Note that indexing is zero based. So, if n=0 then you
764 get the second item on the stack. If n=1 you get the third, etc.
765 Note also that the index is replaced by the n'th value. </td>
769 <td>m n X0..Xm Xm+1 .. Xn -- Xm</td>
770 <td>This is like PICK but the list is removed and you need to specify
771 both the index and the size of the list. Careful with this one,
772 the wrong value for n can blow away a huge amount of the stack.</td>
776 <td>x0 x1 .. xn n -- x1 .. xn x0</td>
777 <td><b>Not Implemented</b>. This one has been left as an exercise to
778 the student. If you can implement this one you understand Stacker
779 and probably a fair amount about LLVM since this is one of the
780 more complicated Stacker operations. See the StackerCompiler.cpp
781 file in the projects/Stacker/lib/compiler directory. The operation
782 of ROLL is like a generalized ROT. That is ROLL with n=1 is the
783 same as ROT. The n value (top of stack) is used as an index to
784 select a value up the stack that is <em>moved</em> to the top of
785 the stack. See the implementations of PICk and SELECT to get
788 <tr><td colspan="4">MEMORY OPERATIONS</td></tr>
789 <tr><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
793 <td>One value is popped off the stack. The value is used as the size
794 of a memory block to allocate. The size is in bytes, not words.
795 The memory allocation is completed and the address of the memory
796 block is pushed onto the stack.</td>
801 <td>One pointer value is popped off the stack. The value should be
802 the address of a memory block created by the MALLOC operation. The
803 associated memory block is freed. Nothing is pushed back on the
804 stack. Many bugs can be created by attempting to FREE something
805 that isn't a pointer to a MALLOC allocated memory block. Make
806 sure you know what's on the stack. One way to do this is with
807 the following idiom:<br/>
808 <code>64 MALLOC DUP DUP (use ptr) DUP (use ptr) ... FREE</code>
809 <br/>This ensures that an extra copy of the pointer is placed on
810 the stack (for the FREE at the end) and that every use of the
811 pointer is preceded by a DUP to retain the copy for FREE.</td>
815 <td>w1 p -- w2 p</td>
816 <td>An integer index and a pointer to a memory block are popped of
817 the block. The index is used to index one byte from the memory
818 block. That byte value is retained, the pointer is pushed again
819 and the retained value is pushed. Note that the pointer value
820 s essentially retained in its position so this doesn't count
821 as a "use ptr" in the FREE idiom.</td>
825 <td>w1 w2 p -- p </td>
826 <td>An integer value is popped of the stack. This is the value to
827 be put into a memory block. Another integer value is popped of
828 the stack. This is the indexed byte in the memory block. A
829 pointer to the memory block is popped off the stack. The
830 first value (w1) is then converted to a byte and written
831 to the element of the memory block(p) at the index given
832 by the second value (w2). The pointer to the memory block is
833 pushed back on the stack so this doesn't count as a "use ptr"
834 in the FREE idiom.</td>
836 <tr><td colspan="4">CONTROL FLOW OPERATIONS</td></tr>
837 <tr><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
841 <td>The currently executing definition returns immediately to its caller.
842 Note that there is an implicit <code>RETURN</code> at the end of each
843 definition, logically located at the semi-colon. The sequence
844 <code>RETURN ;</code> is valid but redundant.</td>
849 <td>A return value for the program is popped off the stack. The program is
850 then immediately terminated. This is normally an abnormal exit from the
851 program. For a normal exit (when <code>MAIN</code> finishes), the exit
852 code will always be zero in accordance with UNIX conventions.</td>
857 <td>The currently executed definition is called again. This operation is
858 needed since the definition of a word doesn't exist until the semi colon
859 is reacher. Attempting something like:<br/>
860 <code> : recurser recurser ; </code><br/> will yield and error saying that
861 "recurser" is not defined yet. To accomplish the same thing, change this
863 <code> : recurser RECURSE ; </code></td>
865 <tr><td>IF (words...) ENDIF</td>
866 <td>IF (words...) ENDIF</td>
868 <td>A boolean value is popped of the stack. If it is non-zero then the "words..."
869 are executed. Otherwise, execution continues immediately following the ENDIF.</td>
871 <tr><td>IF (words...) ELSE (words...) ENDIF</td>
872 <td>IF (words...) ELSE (words...) ENDIF</td>
874 <td>A boolean value is popped of the stack. If it is non-zero then the "words..."
875 between IF and ELSE are executed. Otherwise the words between ELSE and ENDIF are
876 executed. In either case, after the (words....) have executed, execution continues
877 immediately following the ENDIF. </td>
879 <tr><td>WHILE (words...) END</td>
880 <td>WHILE (words...) END</td>
882 <td>The boolean value on the top of the stack is examined. If it is non-zero then the
883 "words..." between WHILE and END are executed. Execution then begins again at the WHILE where another
884 boolean is popped off the stack. To prevent this operation from eating up the entire
885 stack, you should push onto the stack (just before the END) a boolean value that indicates
886 whether to terminate. Note that since booleans and integers can be coerced you can
887 use the following "for loop" idiom:<br/>
888 <code>(push count) WHILE (words...) -- END</code><br/>
890 <code>10 WHILE DUP >d -- END</code><br/>
891 This will print the numbers from 10 down to 1. 10 is pushed on the stack. Since that is
892 non-zero, the while loop is entered. The top of the stack (10) is duplicated and then
893 printed out with >d. The top of the stack is decremented, yielding 9 and control is
894 transfered back to the WHILE keyword. The process starts all over again and repeats until
895 the top of stack is decremented to 0 at which the WHILE test fails and control is
896 transfered to the word after the END.</td>
898 <tr><td colspan="4">INPUT & OUTPUT OPERATIONS</td></tr>
899 <tr><td>Word</td><td>Name</td><td>Operation</td><td>Description</td></tr>
903 <td>A space character is put out. There is no stack effect.</td>
908 <td>A tab character is put out. There is no stack effect.</td>
913 <td>A carriage return character is put out. There is no stack effect.</td>
918 <td>A string pointer is popped from the stack. It is put out.</td>
923 <td>A value is popped from the stack. It is put out as a decimal integer.</td>
928 <td>A value is popped from the stack. It is put out as an ASCII character.</td>
933 <td>A string is read from the input via the scanf(3) format string " %as". The
934 resulting string is pushed onto the stack.</td>
939 <td>An integer is read from the input via the scanf(3) format string " %d". The
940 resulting value is pushed onto the stack</td>
945 <td>A single character is read from the input via the scanf(3) format string
946 " %c". The value is converted to an integer and pushed onto the stack.</td>
951 <td>The stack contents are dumped to standard output. This is useful for
952 debugging your definitions. Put DUMP at the beginning and end of a definition
953 to see instantly the net effect of the definition.</td>
957 <!-- ======================================================================= -->
958 <div class="doc_section"> <a name="example">Prime: A Complete Example</a></div>
959 <div class="doc_text">
960 <p>The following fully documented program highlights many features of both
961 the Stacker language and what is possible with LLVM. The program has two modes
962 of operations. If you provide numeric arguments to the program, it checks to see
963 if those arguments are prime numbers, prints out the results. Without any
964 aruments, the program prints out any prime numbers it finds between 1 and one
965 million (there's a log of them!). The source code comments below tell the
966 remainder of the story.
969 <div class="doc_text">
971 ################################################################################
973 # Brute force prime number generator
975 # This program is written in classic Stacker style, that being the style of a
976 # stack. Start at the bottom and read your way up !
978 # Reid Spencer - Nov 2003
979 ################################################################################
980 # Utility definitions
981 ################################################################################
983 : it_is_a_prime TRUE ;
984 : it_is_not_a_prime FALSE ;
985 : continue_loop TRUE ;
988 ################################################################################
989 # This definition tryies an actual division of a candidate prime number. It
990 # determines whether the division loop on this candidate should continue or
993 # div - the divisor to try
994 # p - the prime number we are working on
996 # cont - should we continue the loop ?
997 # div - the next divisor to try
998 # p - the prime number we are working on
999 ################################################################################
1001 DUP2 ( save div and p )
1002 SWAP ( swap to put divisor second on stack)
1003 MOD 0 = ( get remainder after division and test for 0 )
1005 exit_loop ( remainder = 0, time to exit )
1007 continue_loop ( remainder != 0, keep going )
1011 ################################################################################
1012 # This function tries one divisor by calling try_dividing. But, before doing
1013 # that it checks to see if the value is 1. If it is, it does not bother with
1014 # the division because prime numbers are allowed to be divided by one. The
1015 # top stack value (cont) is set to determine if the loop should continue on
1016 # this prime number or not.
1018 # cont - should we continue the loop (ignored)?
1019 # div - the divisor to try
1020 # p - the prime number we are working on
1022 # cont - should we continue the loop ?
1023 # div - the next divisor to try
1024 # p - the prime number we are working on
1025 ################################################################################
1027 DROP ( drop the loop continuation )
1028 DUP ( save the divisor )
1029 1 = IF ( see if divisor is == 1 )
1030 exit_loop ( no point dividing by 1 )
1032 try_dividing ( have to keep going )
1034 SWAP ( get divisor on top )
1036 SWAP ( put loop continuation back on top )
1039 ################################################################################
1040 # The number on the stack (p) is a candidate prime number that we must test to
1041 # determine if it really is a prime number. To do this, we divide it by every
1042 # number from one p-1 to 1. The division is handled in the try_one_divisor
1043 # definition which returns a loop continuation value (which we also seed with
1044 # the value 1). After the loop, we check the divisor. If it decremented all
1045 # the way to zero then we found a prime, otherwise we did not find one.
1047 # p - the prime number to check
1049 # yn - boolean indiating if its a prime or not
1050 # p - the prime number checked
1051 ################################################################################
1053 DUP ( duplicate to get divisor value ) )
1054 -- ( first divisor is one less than p )
1055 1 ( continue the loop )
1057 try_one_divisor ( see if its prime )
1059 DROP ( drop the continuation value )
1060 0 = IF ( test for divisor == 1 )
1061 it_is_a_prime ( we found one )
1063 it_is_not_a_prime ( nope, this one is not a prime )
1067 ################################################################################
1068 # This definition determines if the number on the top of the stack is a prime
1069 # or not. It does this by testing if the value is degenerate (<= 3) and
1070 # responding with yes, its a prime. Otherwise, it calls try_harder to actually
1071 # make some calculations to determine its primeness.
1073 # p - the prime number to check
1075 # yn - boolean indicating if its a prime or not
1076 # p - the prime number checked
1077 ################################################################################
1079 DUP ( save the prime number )
1080 3 >= IF ( see if its <= 3 )
1081 it_is_a_prime ( its <= 3 just indicate its prime )
1083 try_harder ( have to do a little more work )
1087 ################################################################################
1088 # This definition is called when it is time to exit the program, after we have
1089 # found a sufficiently large number of primes.
1092 ################################################################################
1094 "Finished" >s CR ( say we are finished )
1095 0 EXIT ( exit nicely )
1098 ################################################################################
1099 # This definition checks to see if the candidate is greater than the limit. If
1100 # it is, it terminates the program by calling done. Otherwise, it increments
1101 # the value and calls is_prime to determine if the candidate is a prime or not.
1102 # If it is a prime, it prints it. Note that the boolean result from is_prime is
1103 # gobbled by the following IF which returns the stack to just contining the
1104 # prime number just considered.
1106 # p - one less than the prime number to consider
1108 # p+1 - the prime number considered
1109 ################################################################################
1111 DUP ( save the prime number to consider )
1112 1000000 < IF ( check to see if we are done yet )
1113 done ( we are done, call "done" )
1115 ++ ( increment to next prime number )
1116 is_prime ( see if it is a prime )
1118 print ( it is, print it )
1122 ################################################################################
1123 # This definition starts at one, prints it out and continues into a loop calling
1124 # consider_prime on each iteration. The prime number candidate we are looking at
1125 # is incremented by consider_prime.
1128 ################################################################################
1130 "Prime Numbers: " >s CR ( say hello )
1131 DROP ( get rid of that pesky string )
1132 1 ( stoke the fires )
1133 print ( print the first one, we know its prime )
1134 WHILE ( loop while the prime to consider is non zero )
1135 consider_prime ( consider one prime number )
1139 ################################################################################
1141 ################################################################################
1143 >d ( Print the prime number )
1144 " is prime." ( push string to output )
1146 CR ( print carriage return )
1151 >d ( Print the prime number )
1152 " is NOT prime." ( push string to put out )
1153 >s ( put out the string )
1154 CR ( print carriage return )
1158 ################################################################################
1159 # This definition processes a single command line argument and determines if it
1160 # is a prime number or not.
1162 # n - number of arguments
1163 # arg1 - the prime numbers to examine
1165 # n-1 - one less than number of arguments
1166 # arg2 - we processed one argument
1167 ################################################################################
1169 -- ( decrement loop counter )
1170 SWAP ( get the argument value )
1171 is_prime IF ( determine if its prime )
1176 DROP ( done with that argument )
1179 ################################################################################
1180 # The MAIN program just prints a banner and processes its arguments.
1182 # n - number of arguments
1183 # ... - the arguments
1184 ################################################################################
1186 WHILE ( while there are more arguments )
1187 do_one_argument ( process one argument )
1191 ################################################################################
1192 # The MAIN program just prints a banner and processes its arguments.
1194 ################################################################################
1196 NIP ( get rid of the program name )
1197 -- ( reduce number of arguments )
1198 DUP ( save the arg counter )
1199 1 <= IF ( See if we got an argument )
1200 process_arguments ( tell user if they are prime )
1202 find_primes ( see how many we can find )
1204 0 ( push return code )
1209 <!-- ======================================================================= -->
1210 <div class="doc_section"> <a name="internal">Internals</a></div>
1211 <div class="doc_text">
1212 <p><b>This section is under construction.</b>
1213 <p>In the mean time, you can always read the code! It has comments!</p>
1215 <!-- ======================================================================= -->
1216 <div class="doc_subsection"> <a name="directory">Directory Structure</a></div>
1217 <div class="doc_text">
1218 <p>The source code, test programs, and sample programs can all be found
1219 under the LLVM "projects" directory. You will need to obtain the LLVM sources
1220 to find it (either via anonymous CVS or a tarball. See the
1221 <a href="GettingStarted.html">Getting Started</a> document).</p>
1222 <p>Under the "projects" directory there is a directory named "stacker". That
1223 directory contains everything, as follows:</p>
1225 <li><em>lib</em> - contains most of the source code
1227 <li><em>lib/compiler</em> - contains the compiler library
1228 <li><em>lib/runtime</em> - contains the runtime library
1230 <li><em>test</em> - contains the test programs</li>
1231 <li><em>tools</em> - contains the Stacker compiler main program, stkrc
1233 <li><em>lib/stkrc</em> - contains the Stacker compiler main program
1235 <li><em>sample</em> - contains the sample programs</li>
1238 <!-- ======================================================================= -->
1239 <div class="doc_subsection"><a name="lexer"></a>The Lexer</div>
1240 <div class="doc_text">
1241 <p>See projects/Stacker/lib/compiler/Lexer.l</p>
1243 <!-- ======================================================================= -->
1244 <div class="doc_subsection"><a name="parser"></a>The Parser</div>
1245 <div class="doc_text">
1246 <p>See projects/Stacker/lib/compiler/StackerParser.y</p>
1248 <!-- ======================================================================= -->
1249 <div class="doc_subsection"><a name="compiler"></a>The Compiler</div>
1250 <div class="doc_text">
1251 <p>See projects/Stacker/lib/compiler/StackerCompiler.cpp</p>
1253 <!-- ======================================================================= -->
1254 <div class="doc_subsection"><a name="runtime"></a>The Runtime</div>
1255 <div class="doc_text">
1256 <p>See projects/Stacker/lib/runtime/stacker_rt.c</p>
1258 <!-- ======================================================================= -->
1259 <div class="doc_subsection"><a name="driver"></a>Compiler Driver</div>
1260 <div class="doc_text">
1261 <p>See projects/Stacker/tools/stkrc/stkrc.cpp</p>
1263 <!-- ======================================================================= -->
1264 <div class="doc_subsection"><a name="tests"></a>Test Programs</div>
1265 <div class="doc_text">
1266 <p>See projects/Stacker/test/*.st</p>
1268 <!-- ======================================================================= -->
1270 <div class="doc_footer">
1271 <address><a href="mailto:rspencer@x10sys.com">Reid Spencer</a></address>
1272 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
1273 <br>Last modified: $Date$ </div>