1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>Stacker: An Example Of Using LLVM</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
8 table, tr, td { border: 2px solid gray }
9 table { border-collapse: collapse; margin-bottom: 2em }
14 <div class="doc_title">Stacker: An Example Of Using LLVM</div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#lessons">Lessons I Learned About LLVM</a>
21 <li><a href="#value">Everything's a Value!</a></li>
22 <li><a href="#terminate">Terminate Those Blocks!</a></li>
23 <li><a href="#blocks">Concrete Blocks</a></li>
24 <li><a href="#push_back">push_back Is Your Friend</a></li>
25 <li><a href="#gep">The Wily GetElementPtrInst</a></li>
26 <li><a href="#linkage">Getting Linkage Types Right</a></li>
27 <li><a href="#constants">Constants Are Easier Than That!</a></li>
29 <li><a href="#lexicon">The Stacker Lexicon</a>
31 <li><a href="#stack">The Stack</a></li>
32 <li><a href="#punctuation">Punctuation</a></li>
33 <li><a href="#comments">Comments</a></li>
34 <li><a href="#literals">Literals</a></li>
35 <li><a href="#words">Words</a></li>
36 <li><a href="style">Standard Style</a></li>
37 <li><a href="#builtins">Built-Ins</a></li>
39 <li><a href="#example">Prime: A Complete Example</a></li>
40 <li><a href="#internal">Internal Code Details</a>
42 <li><a href="#directory">The Directory Structure </a></li>
43 <li><a href="#lexer">The Lexer</a></li>
44 <li><a href="#parser">The Parser</a></li>
45 <li><a href="#compiler">The Compiler</a></li>
46 <li><a href="#runtime">The Runtime</a></li>
47 <li><a href="#driver">Compiler Driver</a></li>
48 <li><a href="#tests">Test Programs</a></li>
49 <li><a href="#exercise">Exercise</a></li>
50 <li><a href="#todo">Things Remaining To Be Done</a></li>
54 <div class="doc_text">
55 <p><b>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a></b></p>
58 <!-- ======================================================================= -->
59 <div class="doc_section"><a name="abstract">Abstract</a></div>
60 <div class="doc_text">
61 <p>This document is another way to learn about LLVM. Unlike the
62 <a href="LangRef.html">LLVM Reference Manual</a> or
63 <a href="ProgrammersManual.html">LLVM Programmer's Manual</a>, here we learn
64 about LLVM through the experience of creating a simple programming language
65 named Stacker. Stacker was invented specifically as a demonstration of
66 LLVM. The emphasis in this document is not on describing the
67 intricacies of LLVM itself but on how to use it to build your own
70 <!-- ======================================================================= -->
71 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
72 <div class="doc_text">
73 <p>Amongst other things, LLVM is a platform for compiler writers.
74 Because of its exceptionally clean and small IR (intermediate
75 representation), compiler writing with LLVM is much easier than with
76 other system. As proof, I wrote the entire compiler (language definition,
77 lexer, parser, code generator, etc.) in about <em>four days</em>!
78 That's important to know because it shows how quickly you can get a new
79 language running when using LLVM. Furthermore, this was the <em >first</em>
80 language the author ever created using LLVM. The learning curve is
81 included in that four days.</p>
82 <p>The language described here, Stacker, is Forth-like. Programs
83 are simple collections of word definitions, and the only thing definitions
84 can do is manipulate a stack or generate I/O. Stacker is not a "real"
85 programming language; it's very simple. Although it is computationally
86 complete, you wouldn't use it for your next big project. However,
87 the fact that it is complete, it's simple, and it <em>doesn't</em> have
88 a C-like syntax make it useful for demonstration purposes. It shows
89 that LLVM could be applied to a wide variety of languages.</p>
90 <p>The basic notions behind stacker is very simple. There's a stack of
91 integers (or character pointers) that the program manipulates. Pretty
92 much the only thing the program can do is manipulate the stack and do
93 some limited I/O operations. The language provides you with several
94 built-in words that manipulate the stack in interesting ways. To get
95 your feet wet, here's how you write the traditional "Hello, World"
96 program in Stacker:</p>
97 <p><code>: hello_world "Hello, World!" >s DROP CR ;<br>
98 : MAIN hello_world ;<br></code></p>
99 <p>This has two "definitions" (Stacker manipulates words, not
100 functions and words have definitions): <code>MAIN</code> and <code>
101 hello_world</code>. The <code>MAIN</code> definition is standard; it
102 tells Stacker where to start. Here, <code>MAIN</code> is defined to
103 simply invoke the word <code>hello_world</code>. The
104 <code>hello_world</code> definition tells stacker to push the
105 <code>"Hello, World!"</code> string on to the stack, print it out
106 (<code>>s</code>), pop it off the stack (<code>DROP</code>), and
107 finally print a carriage return (<code>CR</code>). Although
108 <code>hello_world</code> uses the stack, its net effect is null. Well
109 written Stacker definitions have that characteristic. </p>
110 <p>Exercise for the reader: how could you make this a one line program?</p>
112 <!-- ======================================================================= -->
113 <div class="doc_section"><a name="lessons"></a>Lessons I Learned About LLVM</div>
114 <div class="doc_text">
115 <p>Stacker was written for two purposes: </p>
117 <li>to get the author over the learning curve, and</li>
118 <li>to provide a simple example of how to write a compiler using LLVM.</li>
120 <p>During the development of Stacker, many lessons about LLVM were
121 learned. Those lessons are described in the following subsections.<p>
123 <!-- ======================================================================= -->
124 <div class="doc_subsection"><a name="value"></a>Everything's a Value!</div>
125 <div class="doc_text">
126 <p>Although I knew that LLVM uses a Single Static Assignment (SSA) format,
127 it wasn't obvious to me how prevalent this idea was in LLVM until I really
128 started using it. Reading the <a href="ProgrammersManual.html">
129 Programmer's Manual</a> and <a href="LangRef.html">Language Reference</a>,
130 I noted that most of the important LLVM IR (Intermediate Representation) C++
131 classes were derived from the Value class. The full power of that simple
132 design only became fully understood once I started constructing executable
133 expressions for Stacker.</p>
134 <p>This really makes your programming go faster. Think about compiling code
135 for the following C/C++ expression: <code>(a|b)*((x+1)/(y+1))</code>. Assuming
136 the values are on the stack in the order a, b, x, y, this could be
137 expressed in stacker as: <code>1 + SWAP 1 + / ROT2 OR *</code>.
138 You could write a function using LLVM that computes this expression like this: </p>
141 expression(BasicBlock* bb, Value* a, Value* b, Value* x, Value* y )
143 Instruction* tail = bb->getTerminator();
144 ConstantSInt* one = ConstantSInt::get( Type::IntTy, 1);
145 BinaryOperator* or1 =
146 BinaryOperator::create( Instruction::Or, a, b, "", tail );
147 BinaryOperator* add1 =
148 BinaryOperator::create( Instruction::Add, x, one, "", tail );
149 BinaryOperator* add2 =
150 BinaryOperator::create( Instruction::Add, y, one, "", tail );
151 BinaryOperator* div1 =
152 BinaryOperator::create( Instruction::Div, add1, add2, "", tail);
153 BinaryOperator* mult1 =
154 BinaryOperator::create( Instruction::Mul, or1, div1, "", tail );
159 <p>"Okay, big deal," you say? It is a big deal. Here's why. Note that I didn't
160 have to tell this function which kinds of Values are being passed in. They could be
161 <code>Instruction</code>s, <code>Constant</code>s, <code>GlobalVariable</code>s, or
162 any of the other subclasses of <code>Value</code> that LLVM supports.
163 Furthermore, if you specify Values that are incorrect for this sequence of
164 operations, LLVM will either notice right away (at compilation time) or the LLVM
165 Verifier will pick up the inconsistency when the compiler runs. In either case
166 LLVM prevents you from making a type error that gets passed through to the
167 generated program. This <em>really</em> helps you write a compiler that
168 always generates correct code!<p>
169 <p>The second point is that we don't have to worry about branching, registers,
170 stack variables, saving partial results, etc. The instructions we create
171 <em>are</em> the values we use. Note that all that was created in the above
172 code is a Constant value and five operators. Each of the instructions <em>is</em>
173 the resulting value of that instruction. This saves a lot of time.</p>
174 <p>The lesson is this: <em>SSA form is very powerful: there is no difference
175 between a value and the instruction that created it.</em> This is fully
176 enforced by the LLVM IR. Use it to your best advantage.</p>
178 <!-- ======================================================================= -->
179 <div class="doc_subsection"><a name="terminate"></a>Terminate Those Blocks!</div>
180 <div class="doc_text">
181 <p>I had to learn about terminating blocks the hard way: using the debugger
182 to figure out what the LLVM verifier was trying to tell me and begging for
183 help on the LLVMdev mailing list. I hope you avoid this experience.</p>
184 <p>Emblazon this rule in your mind:</p>
186 <li><em>All</em> <code>BasicBlock</code>s in your compiler <b>must</b> be
187 terminated with a terminating instruction (branch, return, etc.).
190 <p>Terminating instructions are a semantic requirement of the LLVM IR. There
191 is no facility for implicitly chaining together blocks placed into a function
192 in the order they occur. Indeed, in the general case, blocks will not be
193 added to the function in the order of execution because of the recursive
194 way compilers are written.</p>
195 <p>Furthermore, if you don't terminate your blocks, your compiler code will
196 compile just fine. You won't find out about the problem until you're running
197 the compiler and the module you just created fails on the LLVM Verifier.</p>
199 <!-- ======================================================================= -->
200 <div class="doc_subsection"><a name="blocks"></a>Concrete Blocks</div>
201 <div class="doc_text">
202 <p>After a little initial fumbling around, I quickly caught on to how blocks
203 should be constructed. In general, here's what I learned:
205 <li><em>Create your blocks early.</em> While writing your compiler, you
206 will encounter several situations where you know apriori that you will
207 need several blocks. For example, if-then-else, switch, while, and for
208 statements in C/C++ all need multiple blocks for expression in LVVM.
209 The rule is, create them early.</li>
210 <li><em>Terminate your blocks early.</em> This just reduces the chances
211 that you forget to terminate your blocks which is required (go
212 <a href="#terminate">here</a> for more).
213 <li><em>Use getTerminator() for instruction insertion.</em> I noticed early on
214 that many of the constructors for the Instruction classes take an optional
215 <code>insert_before</code> argument. At first, I thought this was a mistake
216 because clearly the normal mode of inserting instructions would be one at
217 a time <em>after</em> some other instruction, not <em>before</em>. However,
218 if you hold on to your terminating instruction (or use the handy dandy
219 <code>getTerminator()</code> method on a <code>BasicBlock</code>), it can
220 always be used as the <code>insert_before</code> argument to your instruction
221 constructors. This causes the instruction to automatically be inserted in
222 the RightPlace™ place, just before the terminating instruction. The
223 nice thing about this design is that you can pass blocks around and insert
224 new instructions into them without ever knowing what instructions came
225 before. This makes for some very clean compiler design.</li>
227 <p>The foregoing is such an important principal, its worth making an idiom:</p>
229 BasicBlock* bb = new BasicBlock();
230 bb->getInstList().push_back( new Branch( ... ) );
231 new Instruction(..., bb->getTerminator() );
233 <p>To make this clear, consider the typical if-then-else statement
234 (see StackerCompiler::handle_if() method). We can set this up
235 in a single function using LLVM in the following way: </p>
237 using namespace llvm;
239 MyCompiler::handle_if( BasicBlock* bb, SetCondInst* condition )
241 // Create the blocks to contain code in the structure of if/then/else
242 BasicBlock* then_bb = new BasicBlock();
243 BasicBlock* else_bb = new BasicBlock();
244 BasicBlock* exit_bb = new BasicBlock();
246 // Insert the branch instruction for the "if"
247 bb->getInstList().push_back( new BranchInst( then_bb, else_bb, condition ) );
249 // Set up the terminating instructions
250 then->getInstList().push_back( new BranchInst( exit_bb ) );
251 else->getInstList().push_back( new BranchInst( exit_bb ) );
253 // Fill in the then part .. details excised for brevity
254 this->fill_in( then_bb );
256 // Fill in the else part .. details excised for brevity
257 this->fill_in( else_bb );
259 // Return a block to the caller that can be filled in with the code
260 // that follows the if/then/else construct.
264 <p>Presumably in the foregoing, the calls to the "fill_in" method would add
265 the instructions for the "then" and "else" parts. They would use the third part
266 of the idiom almost exclusively (inserting new instructions before the
267 terminator). Furthermore, they could even recurse back to <code>handle_if</code>
268 should they encounter another if/then/else statement, and it will just work.</p>
269 <p>Note how cleanly this all works out. In particular, the push_back methods on
270 the <code>BasicBlock</code>'s instruction list. These are lists of type
271 <code>Instruction</code> (which is also of type <code>Value</code>). To create
272 the "if" branch we merely instantiate a <code>BranchInst</code> that takes as
273 arguments the blocks to branch to and the condition to branch on. The
274 <code>BasicBlock</code> objects act like branch labels! This new
275 <code>BranchInst</code> terminates the <code>BasicBlock</code> provided
276 as an argument. To give the caller a way to keep inserting after calling
277 <code>handle_if</code>, we create an <code>exit_bb</code> block which is
279 to the caller. Note that the <code>exit_bb</code> block is used as the
280 terminator for both the <code>then_bb</code> and the <code>else_bb</code>
281 blocks. This guarantees that no matter what else <code>handle_if</code>
282 or <code>fill_in</code> does, they end up at the <code>exit_bb</code> block.
285 <!-- ======================================================================= -->
286 <div class="doc_subsection"><a name="push_back"></a>push_back Is Your Friend</div>
287 <div class="doc_text">
289 One of the first things I noticed is the frequent use of the "push_back"
290 method on the various lists. This is so common that it is worth mentioning.
291 The "push_back" inserts a value into an STL list, vector, array, etc. at the
292 end. The method might have also been named "insert_tail" or "append".
293 Although I've used STL quite frequently, my use of push_back wasn't very
294 high in other programs. In LLVM, you'll use it all the time.
297 <!-- ======================================================================= -->
298 <div class="doc_subsection"><a name="gep"></a>The Wily GetElementPtrInst</div>
299 <div class="doc_text">
301 It took a little getting used to and several rounds of postings to the LLVM
302 mailing list to wrap my head around this instruction correctly. Even though I had
303 read the Language Reference and Programmer's Manual a couple times each, I still
304 missed a few <em>very</em> key points:
307 <li>GetElementPtrInst gives you back a Value for the last thing indexed.</li>
308 <li>All global variables in LLVM are <em>pointers</em>.</li>
309 <li>Pointers must also be dereferenced with the GetElementPtrInst
312 <p>This means that when you look up an element in the global variable (assuming
313 it's a struct or array), you <em>must</em> deference the pointer first! For many
314 things, this leads to the idiom:
317 std::vector<Value*> index_vector;
318 index_vector.push_back( ConstantSInt::get( Type::LongTy, 0 );
319 // ... push other indices ...
320 GetElementPtrInst* gep = new GetElementPtrInst( ptr, index_vector );
322 <p>For example, suppose we have a global variable whose type is [24 x int]. The
323 variable itself represents a <em>pointer</em> to that array. To subscript the
324 array, we need two indices, not just one. The first index (0) dereferences the
325 pointer. The second index subscripts the array. If you're a "C" programmer, this
326 will run against your grain because you'll naturally think of the global array
327 variable and the address of its first element as the same. That tripped me up
328 for a while until I realized that they really do differ .. by <em>type</em>.
329 Remember that LLVM is strongly typed. Everything has a type.
330 The "type" of the global variable is [24 x int]*. That is, it's
331 a pointer to an array of 24 ints. When you dereference that global variable with
332 a single (0) index, you now have a "[24 x int]" type. Although
333 the pointer value of the dereferenced global and the address of the zero'th element
334 in the array will be the same, they differ in their type. The zero'th element has
335 type "int" while the pointer value has type "[24 x int]".</p>
336 <p>Get this one aspect of LLVM right in your head, and you'll save yourself
337 a lot of compiler writing headaches down the road.</p>
339 <!-- ======================================================================= -->
340 <div class="doc_subsection"><a name="linkage"></a>Getting Linkage Types Right</div>
341 <div class="doc_text">
342 <p>Linkage types in LLVM can be a little confusing, especially if your compiler
343 writing mind has affixed firm concepts to particular words like "weak",
344 "external", "global", "linkonce", etc. LLVM does <em>not</em> use the precise
345 definitions of, say, ELF or GCC, even though they share common terms. To be fair,
346 the concepts are related and similar but not precisely the same. This can lead
347 you to think you know what a linkage type represents but in fact it is slightly
348 different. I recommend you read the
349 <a href="LangRef.html#linkage"> Language Reference on this topic</a> very
350 carefully. Then, read it again.<p>
351 <p>Here are some handy tips that I discovered along the way:</p>
353 <li><em>Uninitialized means external.</em> That is, the symbol is declared in the current
354 module and can be used by that module, but it is not defined by that module.</li>
355 <li><em>Setting an initializer changes a global' linkage type.</em> Setting an
356 initializer changes a global's linkage type from whatever it was to a normal,
357 defined global (not external). You'll need to call the setLinkage() method to
358 reset it if you specify the initializer after the GlobalValue has been constructed.
359 This is important for LinkOnce and Weak linkage types.</li>
360 <li><em>Appending linkage can keep track of things.</em> Appending linkage can
361 be used to keep track of compilation information at runtime. It could be used,
362 for example, to build a full table of all the C++ virtual tables or hold the
363 C++ RTTI data, or whatever. Appending linkage can only be applied to arrays.
364 All arrays with the same name in each module are concatenated together at link
368 <!-- ======================================================================= -->
369 <div class="doc_subsection"><a name="constants"></a>Constants Are Easier Than That!</div>
370 <div class="doc_text">
372 Constants in LLVM took a little getting used to until I discovered a few utility
373 functions in the LLVM IR that make things easier. Here's what I learned: </p>
375 <li>Constants are Values like anything else and can be operands of instructions</li>
376 <li>Integer constants, frequently needed, can be created using the static "get"
377 methods of the ConstantInt, ConstantSInt, and ConstantUInt classes. The nice thing
378 about these is that you can "get" any kind of integer quickly.</li>
379 <li>There's a special method on Constant class which allows you to get the null
380 constant for <em>any</em> type. This is really handy for initializing large
381 arrays or structures, etc.</li>
384 <!-- ======================================================================= -->
385 <div class="doc_section"> <a name="lexicon">The Stacker Lexicon</a></div>
386 <div class="doc_text"><p>This section describes the Stacker language</p></div>
387 <div class="doc_subsection"><a name="stack"></a>The Stack</div>
388 <div class="doc_text">
389 <p>Stacker definitions define what they do to the global stack. Before
390 proceeding, a few words about the stack are in order. The stack is simply
391 a global array of 32-bit integers or pointers. A global index keeps track
392 of the location of the top of the stack. All of this is hidden from the
393 programmer, but it needs to be noted because it is the foundation of the
394 conceptual programming model for Stacker. When you write a definition,
395 you are, essentially, saying how you want that definition to manipulate
396 the global stack.</p>
397 <p>Manipulating the stack can be quite hazardous. There is no distinction
398 given and no checking for the various types of values that can be placed
399 on the stack. Automatic coercion between types is performed. In many
400 cases, this is useful. For example, a boolean value placed on the stack
401 can be interpreted as an integer with good results. However, using a
402 word that interprets that boolean value as a pointer to a string to
403 print out will almost always yield a crash. Stacker simply leaves it
404 to the programmer to get it right without any interference or hindering
405 on interpretation of the stack values. You've been warned. :) </p>
407 <!-- ======================================================================= -->
408 <div class="doc_subsection"> <a name="punctuation"></a>Punctuation</div>
409 <div class="doc_text">
410 <p>Punctuation in Stacker is very simple. The colon and semi-colon
411 characters are used to introduce and terminate a definition
412 (respectively). Except for <em>FORWARD</em> declarations, definitions
413 are all you can specify in Stacker. Definitions are read left to right.
414 Immediately after the colon comes the name of the word being defined.
415 The remaining words in the definition specify what the word does. The definition
416 is terminated by a semi-colon.</p>
417 <p>So, your typical definition will have the form:</p>
418 <pre><code>: name ... ;</code></pre>
419 <p>The <code>name</code> is up to you but it must start with a letter and contain
420 only letters, numbers, and underscore. Names are case sensitive and must not be
421 the same as the name of a built-in word. The <code>...</code> is replaced by
422 the stack manipulating words that you wish to define <code>name</code> as. <p>
424 <!-- ======================================================================= -->
425 <div class="doc_subsection"><a name="comments"></a>Comments</div>
426 <div class="doc_text">
427 <p>Stacker supports two types of comments. A hash mark (#) starts a comment
428 that extends to the end of the line. It is identical to the kind of comments
429 commonly used in shell scripts. A pair of parentheses also surround a comment.
430 In both cases, the content of the comment is ignored by the Stacker compiler. The
431 following does nothing in Stacker.
434 # This is a comment to end of line
435 ( This is an enclosed comment )
437 <p>See the <a href="#example">example</a> program to see comments in use in
440 <!-- ======================================================================= -->
441 <div class="doc_subsection"><a name="literals"></a>Literals</div>
442 <div class="doc_text">
443 <p>There are three kinds of literal values in Stacker: Integers, Strings,
444 and Booleans. In each case, the stack operation is to simply push the
445 value on to the stack. So, for example:<br/>
446 <code> 42 " is the answer." TRUE </code><br/>
447 will push three values on to the stack: the integer 42, the
448 string " is the answer.", and the boolean TRUE.</p>
450 <!-- ======================================================================= -->
451 <div class="doc_subsection"><a name="words"></a>Words</div>
452 <div class="doc_text">
453 <p>Each definition in Stacker is composed of a set of words. Words are
454 read and executed in order from left to right. There is very little
455 checking in Stacker to make sure you're doing the right thing with
456 the stack. It is assumed that the programmer knows how the stack
457 transformation he applies will affect the program.</p>
458 <p>Words in a definition come in two flavors: built-in and programmer
459 defined. Simply mentioning the name of a previously defined or declared
460 programmer-defined word causes that word's stack actions to be invoked. It
461 is somewhat like a function call in other languages. The built-in
462 words have various effects, described <a href="#builtins">below</a>.</p>
463 <p>Sometimes you need to call a word before it is defined. For this, you can
464 use the <code>FORWARD</code> declaration. It looks like this:</p>
465 <p><code>FORWARD name ;</code></p>
466 <p>This simply states to Stacker that "name" is the name of a definition
467 that is defined elsewhere. Generally it means the definition can be found
468 "forward" in the file. But, it doesn't have to be in the current compilation
469 unit. Anything declared with <code>FORWARD</code> is an external symbol for
472 <!-- ======================================================================= -->
473 <div class="doc_subsection"><a name="builtins"></a>Built In Words</div>
474 <div class="doc_text">
475 <p>The built-in words of the Stacker language are put in several groups
476 depending on what they do. The groups are as follows:</p>
478 <li><em>Logical</em>: These words provide the logical operations for
479 comparing stack operands.<br/>The words are: < > <= >=
480 = <> true false.</li>
481 <li><em>Bitwise</em>: These words perform bitwise computations on
482 their operands. <br/> The words are: << >> XOR AND NOT</li>
483 <li><em>Arithmetic</em>: These words perform arithmetic computations on
484 their operands. <br/> The words are: ABS NEG + - * / MOD */ ++ -- MIN MAX</li>
485 <li><em>Stack</em>These words manipulate the stack directly by moving
486 its elements around.<br/> The words are: DROP DROP2 NIP NIP2 DUP DUP2
487 SWAP SWAP2 OVER OVER2 ROT ROT2 RROT RROT2 TUCK TUCK2 PICK SELECT ROLL</li>
488 <li><em>Memory</em>These words allocate, free, and manipulate memory
489 areas outside the stack.<br/>The words are: MALLOC FREE GET PUT</li>
490 <li><em>Control</em>: These words alter the normal left to right flow
491 of execution.<br/>The words are: IF ELSE ENDIF WHILE END RETURN EXIT RECURSE</li>
492 <li><em>I/O</em>: These words perform output on the standard output
493 and input on the standard input. No other I/O is possible in Stacker.
494 <br/>The words are: SPACE TAB CR >s >d >c <s <d <c.</li>
496 <p>While you may be familiar with many of these operations from other
497 programming languages, a careful review of their semantics is important
498 for correct programming in Stacker. Of most importance is the effect
499 that each of these built-in words has on the global stack. The effect is
500 not always intuitive. To better describe the effects, we'll borrow from Forth the idiom of
501 describing the effect on the stack with:</p>
502 <p><code> BEFORE -- AFTER </code></p>
503 <p>That is, to the left of the -- is a representation of the stack before
504 the operation. To the right of the -- is a representation of the stack
505 after the operation. In the table below that describes the operation of
506 each of the built in words, we will denote the elements of the stack
507 using the following construction:</p>
509 <li><em>b</em> - a boolean truth value</li>
510 <li><em>w</em> - a normal integer valued word.</li>
511 <li><em>s</em> - a pointer to a string value</li>
512 <li><em>p</em> - a pointer to a malloc'd memory block</li>
515 <div class="doc_text" >
516 <table class="doc_table">
517 <tr class="doc_table"><td colspan="4">Definition Of Operation Of Built In Words</td></tr>
518 <tr class="doc_table"><td colspan="4"><b>LOGICAL OPERATIONS</b></td></tr>
519 <tr class="doc_table">
525 <tr class="doc_table">
529 <td>Two values (w1 and w2) are popped off the stack and
530 compared. If w1 is less than w2, TRUE is pushed back on
531 the stack, otherwise FALSE is pushed back on the stack.</td>
536 <td>Two values (w1 and w2) are popped off the stack and
537 compared. If w1 is greater than w2, TRUE is pushed back on
538 the stack, otherwise FALSE is pushed back on the stack.</td>
543 <td>Two values (w1 and w2) are popped off the stack and
544 compared. If w1 is greater than or equal to w2, TRUE is
545 pushed back on the stack, otherwise FALSE is pushed back
551 <td>Two values (w1 and w2) are popped off the stack and
552 compared. If w1 is less than or equal to w2, TRUE is
553 pushed back on the stack, otherwise FALSE is pushed back
559 <td>Two values (w1 and w2) are popped off the stack and
560 compared. If w1 is equal to w2, TRUE is
561 pushed back on the stack, otherwise FALSE is pushed back
564 <tr><td><></td>
567 <td>Two values (w1 and w2) are popped off the stack and
568 compared. If w1 is equal to w2, TRUE is
569 pushed back on the stack, otherwise FALSE is pushed back
575 <td>The boolean value FALSE (0) is pushed on to the stack.</td>
580 <td>The boolean value TRUE (-1) is pushed on to the stack.</td>
582 <tr><td colspan="4"><b>BITWISE OPERATORS</b></td></tr>
589 <tr><td><<</td>
591 <td>w1 w2 -- w1<<w2</td>
592 <td>Two values (w1 and w2) are popped off the stack. The w2
593 operand is shifted left by the number of bits given by the
594 w1 operand. The result is pushed back to the stack.</td>
596 <tr><td>>></td>
598 <td>w1 w2 -- w1>>w2</td>
599 <td>Two values (w1 and w2) are popped off the stack. The w2
600 operand is shifted right by the number of bits given by the
601 w1 operand. The result is pushed back to the stack.</td>
605 <td>w1 w2 -- w2|w1</td>
606 <td>Two values (w1 and w2) are popped off the stack. The values
607 are bitwise OR'd together and pushed back on the stack. This is
608 not a logical OR. The sequence 1 2 OR yields 3 not 1.</td>
612 <td>w1 w2 -- w2&w1</td>
613 <td>Two values (w1 and w2) are popped off the stack. The values
614 are bitwise AND'd together and pushed back on the stack. This is
615 not a logical AND. The sequence 1 2 AND yields 0 not 1.</td>
619 <td>w1 w2 -- w2^w1</td>
620 <td>Two values (w1 and w2) are popped off the stack. The values
621 are bitwise exclusive OR'd together and pushed back on the stack.
622 For example, The sequence 1 3 XOR yields 2.</td>
624 <tr><td colspan="4"><b>ARITHMETIC OPERATORS</b></td></tr>
634 <td>One value s popped off the stack; its absolute value is computed
635 and then pushed on to the stack. If w1 is -1 then w2 is 1. If w1 is
636 1 then w2 is also 1.</td>
641 <td>One value is popped off the stack which is negated and then
642 pushed back on to the stack. If w1 is -1 then w2 is 1. If w1 is
643 1 then w2 is -1.</td>
647 <td>w1 w2 -- w2+w1</td>
648 <td>Two values are popped off the stack. Their sum is pushed back
653 <td>w1 w2 -- w2-w1</td>
654 <td>Two values are popped off the stack. Their difference is pushed back
659 <td>w1 w2 -- w2*w1</td>
660 <td>Two values are popped off the stack. Their product is pushed back
665 <td>w1 w2 -- w2/w1</td>
666 <td>Two values are popped off the stack. Their quotient is pushed back
671 <td>w1 w2 -- w2%w1</td>
672 <td>Two values are popped off the stack. Their remainder after division
673 of w1 by w2 is pushed back on to the stack</td>
677 <td>w1 w2 w3 -- (w3*w2)/w1</td>
678 <td>Three values are popped off the stack. The product of w1 and w2 is
679 divided by w3. The result is pushed back on to the stack.</td>
684 <td>One value is popped off the stack. It is incremented by one and then
685 pushed back on to the stack.</td>
690 <td>One value is popped off the stack. It is decremented by one and then
691 pushed back on to the stack.</td>
695 <td>w1 w2 -- (w2<w1?w2:w1)</td>
696 <td>Two values are popped off the stack. The larger one is pushed back
697 on to the stack.</td>
701 <td>w1 w2 -- (w2>w1?w2:w1)</td>
702 <td>Two values are popped off the stack. The larger value is pushed back
703 on to the stack.</td>
705 <tr><td colspan="4"><b>STACK MANIPULATION OPERATORS</b></td></tr>
715 <td>One value is popped off the stack.</td>
720 <td>Two values are popped off the stack.</td>
725 <td>The second value on the stack is removed from the stack. That is,
726 a value is popped off the stack and retained. Then a second value is
727 popped and the retained value is pushed.</td>
731 <td>w1 w2 w3 w4 -- w3 w4</td>
732 <td>The third and fourth values on the stack are removed from it. That is,
733 two values are popped and retained. Then two more values are popped and
734 the two retained values are pushed back on.</td>
739 <td>One value is popped off the stack. That value is then pushed on to
740 the stack twice to duplicate the top stack vaue.</td>
744 <td>w1 w2 -- w1 w2 w1 w2</td>
745 <td>The top two values on the stack are duplicated. That is, two vaues
746 are popped off the stack. They are alternately pushed back on the
747 stack twice each.</td>
751 <td>w1 w2 -- w2 w1</td>
752 <td>The top two stack items are reversed in their order. That is, two
753 values are popped off the stack and pushed back on to the stack in
754 the opposite order they were popped.</td>
758 <td>w1 w2 w3 w4 -- w3 w4 w2 w1</td>
759 <td>The top four stack items are swapped in pairs. That is, two values
760 are popped and retained. Then, two more values are popped and retained.
761 The values are pushed back on to the stack in the reverse order but
766 <td>w1 w2-- w1 w2 w1</td>
767 <td>Two values are popped from the stack. They are pushed back
768 on to the stack in the order w1 w2 w1. This seems to cause the
769 top stack element to be duplicated "over" the next value.</td>
773 <td>w1 w2 w3 w4 -- w1 w2 w3 w4 w1 w2</td>
774 <td>The third and fourth values on the stack are replicated on to the
775 top of the stack</td>
779 <td>w1 w2 w3 -- w2 w3 w1</td>
780 <td>The top three values are rotated. That is, three value are popped
781 off the stack. They are pushed back on to the stack in the order
786 <td>w1 w2 w3 w4 w5 w6 -- w3 w4 w5 w6 w1 w2</td>
787 <td>Like ROT but the rotation is done using three pairs instead of
792 <td>w1 w2 w3 -- w2 w3 w1</td>
793 <td>Reverse rotation. Like ROT, but it rotates the other way around.
794 Essentially, the third element on the stack is moved to the top
799 <td>w1 w2 w3 w4 w5 w6 -- w3 w4 w5 w6 w1 w2</td>
800 <td>Double reverse rotation. Like RROT but the rotation is done using
801 three pairs instead of three singles. The fifth and sixth stack
802 elements are moved to the first and second positions</td>
806 <td>w1 w2 -- w2 w1 w2</td>
807 <td>Similar to OVER except that the second operand is being
808 replicated. Essentially, the first operand is being "tucked"
809 in between two instances of the second operand. Logically, two
810 values are popped off the stack. They are placed back on the
811 stack in the order w2 w1 w2.</td>
815 <td>w1 w2 w3 w4 -- w3 w4 w1 w2 w3 w4</td>
816 <td>Like TUCK but a pair of elements is tucked over two pairs.
817 That is, the top two elements of the stack are duplicated and
818 inserted into the stack at the fifth and positions.</td>
822 <td>x0 ... Xn n -- x0 ... Xn x0</td>
823 <td>The top of the stack is used as an index into the remainder of
824 the stack. The element at the nth position replaces the index
825 (top of stack). This is useful for cycling through a set of
826 values. Note that indexing is zero based. So, if n=0 then you
827 get the second item on the stack. If n=1 you get the third, etc.
828 Note also that the index is replaced by the n'th value. </td>
832 <td>m n X0..Xm Xm+1 .. Xn -- Xm</td>
833 <td>This is like PICK but the list is removed and you need to specify
834 both the index and the size of the list. Careful with this one,
835 the wrong value for n can blow away a huge amount of the stack.</td>
839 <td>x0 x1 .. xn n -- x1 .. xn x0</td>
840 <td><b>Not Implemented</b>. This one has been left as an exercise to
841 the student. See <a href="#exercise">Exercise</a>. ROLL requires
842 a value, "n", to be on the top of the stack. This value specifies how
843 far into the stack to "roll". The n'th value is <em>moved</em> (not
844 copied) from its location and replaces the "n" value on the top of the
845 stack. In this way, all the values between "n" and x0 roll up the stack.
846 The operation of ROLL is a generalized ROT. The "n" value specifies
847 how much to rotate. That is, ROLL with n=1 is the same as ROT and
848 ROLL with n=2 is the same as ROT2.</td>
850 <tr><td colspan="4"><b>MEMORY OPERATORS</b></td></tr>
860 <td>One value is popped off the stack. The value is used as the size
861 of a memory block to allocate. The size is in bytes, not words.
862 The memory allocation is completed and the address of the memory
863 block is pushed on to the stack.</td>
868 <td>One pointer value is popped off the stack. The value should be
869 the address of a memory block created by the MALLOC operation. The
870 associated memory block is freed. Nothing is pushed back on the
871 stack. Many bugs can be created by attempting to FREE something
872 that isn't a pointer to a MALLOC allocated memory block. Make
873 sure you know what's on the stack. One way to do this is with
874 the following idiom:<br/>
875 <code>64 MALLOC DUP DUP (use ptr) DUP (use ptr) ... FREE</code>
876 <br/>This ensures that an extra copy of the pointer is placed on
877 the stack (for the FREE at the end) and that every use of the
878 pointer is preceded by a DUP to retain the copy for FREE.</td>
882 <td>w1 p -- w2 p</td>
883 <td>An integer index and a pointer to a memory block are popped of
884 the block. The index is used to index one byte from the memory
885 block. That byte value is retained, the pointer is pushed again
886 and the retained value is pushed. Note that the pointer value
887 s essentially retained in its position so this doesn't count
888 as a "use ptr" in the FREE idiom.</td>
892 <td>w1 w2 p -- p </td>
893 <td>An integer value is popped of the stack. This is the value to
894 be put into a memory block. Another integer value is popped of
895 the stack. This is the indexed byte in the memory block. A
896 pointer to the memory block is popped off the stack. The
897 first value (w1) is then converted to a byte and written
898 to the element of the memory block(p) at the index given
899 by the second value (w2). The pointer to the memory block is
900 pushed back on the stack so this doesn't count as a "use ptr"
901 in the FREE idiom.</td>
903 <tr><td colspan="4"><b>CONTROL FLOW OPERATORS</b></td></tr>
913 <td>The currently executing definition returns immediately to its caller.
914 Note that there is an implicit <code>RETURN</code> at the end of each
915 definition, logically located at the semi-colon. The sequence
916 <code>RETURN ;</code> is valid but redundant.</td>
921 <td>A return value for the program is popped off the stack. The program is
922 then immediately terminated. This is normally an abnormal exit from the
923 program. For a normal exit (when <code>MAIN</code> finishes), the exit
924 code will always be zero in accordance with UNIX conventions.</td>
929 <td>The currently executed definition is called again. This operation is
930 needed since the definition of a word doesn't exist until the semi colon
931 is reacher. Attempting something like:<br/>
932 <code> : recurser recurser ; </code><br/> will yield and error saying that
933 "recurser" is not defined yet. To accomplish the same thing, change this
935 <code> : recurser RECURSE ; </code></td>
937 <tr><td>IF (words...) ENDIF</td>
938 <td>IF (words...) ENDIF</td>
940 <td>A boolean value is popped of the stack. If it is non-zero then the "words..."
941 are executed. Otherwise, execution continues immediately following the ENDIF.</td>
943 <tr><td>IF (words...) ELSE (words...) ENDIF</td>
944 <td>IF (words...) ELSE (words...) ENDIF</td>
946 <td>A boolean value is popped of the stack. If it is non-zero then the "words..."
947 between IF and ELSE are executed. Otherwise the words between ELSE and ENDIF are
948 executed. In either case, after the (words....) have executed, execution continues
949 immediately following the ENDIF. </td>
951 <tr><td>WHILE (words...) END</td>
952 <td>WHILE (words...) END</td>
954 <td>The boolean value on the top of the stack is examined. If it is non-zero then the
955 "words..." between WHILE and END are executed. Execution then begins again at the WHILE where another
956 boolean is popped off the stack. To prevent this operation from eating up the entire
957 stack, you should push on to the stack (just before the END) a boolean value that indicates
958 whether to terminate. Note that since booleans and integers can be coerced you can
959 use the following "for loop" idiom:<br/>
960 <code>(push count) WHILE (words...) -- END</code><br/>
962 <code>10 WHILE DUP >d -- END</code><br/>
963 This will print the numbers from 10 down to 1. 10 is pushed on the stack. Since that is
964 non-zero, the while loop is entered. The top of the stack (10) is duplicated and then
965 printed out with >d. The top of the stack is decremented, yielding 9 and control is
966 transfered back to the WHILE keyword. The process starts all over again and repeats until
967 the top of stack is decremented to 0 at which the WHILE test fails and control is
968 transfered to the word after the END.</td>
970 <tr><td colspan="4"><b>INPUT & OUTPUT OPERATORS</b></td></tr>
980 <td>A space character is put out. There is no stack effect.</td>
985 <td>A tab character is put out. There is no stack effect.</td>
990 <td>A carriage return character is put out. There is no stack effect.</td>
995 <td>A string pointer is popped from the stack. It is put out.</td>
1000 <td>A value is popped from the stack. It is put out as a decimal
1006 <td>A value is popped from the stack. It is put out as an ASCII
1012 <td>A string is read from the input via the scanf(3) format string " %as".
1013 The resulting string is pushed on to the stack.</td>
1018 <td>An integer is read from the input via the scanf(3) format string " %d".
1019 The resulting value is pushed on to the stack</td>
1024 <td>A single character is read from the input via the scanf(3) format string
1025 " %c". The value is converted to an integer and pushed on to the stack.</td>
1030 <td>The stack contents are dumped to standard output. This is useful for
1031 debugging your definitions. Put DUMP at the beginning and end of a definition
1032 to see instantly the net effect of the definition.</td>
1037 <!-- ======================================================================= -->
1038 <div class="doc_section"> <a name="example">Prime: A Complete Example</a></div>
1039 <div class="doc_text">
1040 <p>The following fully documented program highlights many features of both
1041 the Stacker language and what is possible with LLVM. The program has two modes
1042 of operation. If you provide numeric arguments to the program, it checks to see
1043 if those arguments are prime numbers and prints out the results. Without any
1044 arguments, the program prints out any prime numbers it finds between 1 and one
1045 million (there's a lot of them!). The source code comments below tell the
1046 remainder of the story.
1049 <div class="doc_text">
1051 ################################################################################
1053 # Brute force prime number generator
1055 # This program is written in classic Stacker style, that being the style of a
1056 # stack. Start at the bottom and read your way up !
1058 # Reid Spencer - Nov 2003
1059 ################################################################################
1060 # Utility definitions
1061 ################################################################################
1063 : it_is_a_prime TRUE ;
1064 : it_is_not_a_prime FALSE ;
1065 : continue_loop TRUE ;
1068 ################################################################################
1069 # This definition tries an actual division of a candidate prime number. It
1070 # determines whether the division loop on this candidate should continue or
1073 # div - the divisor to try
1074 # p - the prime number we are working on
1076 # cont - should we continue the loop ?
1077 # div - the next divisor to try
1078 # p - the prime number we are working on
1079 ################################################################################
1081 DUP2 ( save div and p )
1082 SWAP ( swap to put divisor second on stack)
1083 MOD 0 = ( get remainder after division and test for 0 )
1085 exit_loop ( remainder = 0, time to exit )
1087 continue_loop ( remainder != 0, keep going )
1091 ################################################################################
1092 # This function tries one divisor by calling try_dividing. But, before doing
1093 # that it checks to see if the value is 1. If it is, it does not bother with
1094 # the division because prime numbers are allowed to be divided by one. The
1095 # top stack value (cont) is set to determine if the loop should continue on
1096 # this prime number or not.
1098 # cont - should we continue the loop (ignored)?
1099 # div - the divisor to try
1100 # p - the prime number we are working on
1102 # cont - should we continue the loop ?
1103 # div - the next divisor to try
1104 # p - the prime number we are working on
1105 ################################################################################
1107 DROP ( drop the loop continuation )
1108 DUP ( save the divisor )
1109 1 = IF ( see if divisor is == 1 )
1110 exit_loop ( no point dividing by 1 )
1112 try_dividing ( have to keep going )
1114 SWAP ( get divisor on top )
1116 SWAP ( put loop continuation back on top )
1119 ################################################################################
1120 # The number on the stack (p) is a candidate prime number that we must test to
1121 # determine if it really is a prime number. To do this, we divide it by every
1122 # number from one p-1 to 1. The division is handled in the try_one_divisor
1123 # definition which returns a loop continuation value (which we also seed with
1124 # the value 1). After the loop, we check the divisor. If it decremented all
1125 # the way to zero then we found a prime, otherwise we did not find one.
1127 # p - the prime number to check
1129 # yn - boolean indicating if its a prime or not
1130 # p - the prime number checked
1131 ################################################################################
1133 DUP ( duplicate to get divisor value ) )
1134 -- ( first divisor is one less than p )
1135 1 ( continue the loop )
1137 try_one_divisor ( see if its prime )
1139 DROP ( drop the continuation value )
1140 0 = IF ( test for divisor == 1 )
1141 it_is_a_prime ( we found one )
1143 it_is_not_a_prime ( nope, this one is not a prime )
1147 ################################################################################
1148 # This definition determines if the number on the top of the stack is a prime
1149 # or not. It does this by testing if the value is degenerate (<= 3) and
1150 # responding with yes, its a prime. Otherwise, it calls try_harder to actually
1151 # make some calculations to determine its primeness.
1153 # p - the prime number to check
1155 # yn - boolean indicating if its a prime or not
1156 # p - the prime number checked
1157 ################################################################################
1159 DUP ( save the prime number )
1160 3 >= IF ( see if its <= 3 )
1161 it_is_a_prime ( its <= 3 just indicate its prime )
1163 try_harder ( have to do a little more work )
1167 ################################################################################
1168 # This definition is called when it is time to exit the program, after we have
1169 # found a sufficiently large number of primes.
1170 # STACK<: ignored
1172 ################################################################################
1174 "Finished" >s CR ( say we are finished )
1175 0 EXIT ( exit nicely )
1178 ################################################################################
1179 # This definition checks to see if the candidate is greater than the limit. If
1180 # it is, it terminates the program by calling done. Otherwise, it increments
1181 # the value and calls is_prime to determine if the candidate is a prime or not.
1182 # If it is a prime, it prints it. Note that the boolean result from is_prime is
1183 # gobbled by the following IF which returns the stack to just contining the
1184 # prime number just considered.
1186 # p - one less than the prime number to consider
1188 # p+1 - the prime number considered
1189 ################################################################################
1191 DUP ( save the prime number to consider )
1192 1000000 < IF ( check to see if we are done yet )
1193 done ( we are done, call "done" )
1195 ++ ( increment to next prime number )
1196 is_prime ( see if it is a prime )
1198 print ( it is, print it )
1202 ################################################################################
1203 # This definition starts at one, prints it out and continues into a loop calling
1204 # consider_prime on each iteration. The prime number candidate we are looking at
1205 # is incremented by consider_prime.
1208 ################################################################################
1210 "Prime Numbers: " >s CR ( say hello )
1211 DROP ( get rid of that pesky string )
1212 1 ( stoke the fires )
1213 print ( print the first one, we know its prime )
1214 WHILE ( loop while the prime to consider is non zero )
1215 consider_prime ( consider one prime number )
1219 ################################################################################
1221 ################################################################################
1223 >d ( Print the prime number )
1224 " is prime." ( push string to output )
1226 CR ( print carriage return )
1231 >d ( Print the prime number )
1232 " is NOT prime." ( push string to put out )
1233 >s ( put out the string )
1234 CR ( print carriage return )
1238 ################################################################################
1239 # This definition processes a single command line argument and determines if it
1240 # is a prime number or not.
1242 # n - number of arguments
1243 # arg1 - the prime numbers to examine
1245 # n-1 - one less than number of arguments
1246 # arg2 - we processed one argument
1247 ################################################################################
1249 -- ( decrement loop counter )
1250 SWAP ( get the argument value )
1251 is_prime IF ( determine if its prime )
1256 DROP ( done with that argument )
1259 ################################################################################
1260 # The MAIN program just prints a banner and processes its arguments.
1262 # n - number of arguments
1263 # ... - the arguments
1264 ################################################################################
1266 WHILE ( while there are more arguments )
1267 do_one_argument ( process one argument )
1271 ################################################################################
1272 # The MAIN program just prints a banner and processes its arguments.
1273 # STACK<: arguments
1274 ################################################################################
1276 NIP ( get rid of the program name )
1277 -- ( reduce number of arguments )
1278 DUP ( save the arg counter )
1279 1 <= IF ( See if we got an argument )
1280 process_arguments ( tell user if they are prime )
1282 find_primes ( see how many we can find )
1284 0 ( push return code )
1289 <!-- ======================================================================= -->
1290 <div class="doc_section"> <a name="internal">Internals</a></div>
1291 <div class="doc_text">
1292 <p><b>This section is under construction.</b>
1293 <p>In the mean time, you can always read the code! It has comments!</p>
1295 <!-- ======================================================================= -->
1296 <div class="doc_subsection"> <a name="directory">Directory Structure</a></div>
1297 <div class="doc_text">
1298 <p>The source code, test programs, and sample programs can all be found
1299 under the LLVM "projects" directory. You will need to obtain the LLVM sources
1300 to find it (either via anonymous CVS or a tarball. See the
1301 <a href="GettingStarted.html">Getting Started</a> document).</p>
1302 <p>Under the "projects" directory there is a directory named "Stacker". That
1303 directory contains everything, as follows:</p>
1305 <li><em>lib</em> - contains most of the source code
1307 <li><em>lib/compiler</em> - contains the compiler library
1308 <li><em>lib/runtime</em> - contains the runtime library
1310 <li><em>test</em> - contains the test programs</li>
1311 <li><em>tools</em> - contains the Stacker compiler main program, stkrc
1313 <li><em>lib/stkrc</em> - contains the Stacker compiler main program
1315 <li><em>sample</em> - contains the sample programs</li>
1318 <!-- ======================================================================= -->
1319 <div class="doc_subsection"><a name="lexer"></a>The Lexer</div>
1320 <div class="doc_text">
1321 <p>See projects/Stacker/lib/compiler/Lexer.l</p>
1323 <!-- ======================================================================= -->
1324 <div class="doc_subsection"><a name="parser"></a>The Parser</div>
1325 <div class="doc_text">
1326 <p>See projects/Stacker/lib/compiler/StackerParser.y</p>
1328 <!-- ======================================================================= -->
1329 <div class="doc_subsection"><a name="compiler"></a>The Compiler</div>
1330 <div class="doc_text">
1331 <p>See projects/Stacker/lib/compiler/StackerCompiler.cpp</p>
1333 <!-- ======================================================================= -->
1334 <div class="doc_subsection"><a name="runtime"></a>The Runtime</div>
1335 <div class="doc_text">
1336 <p>See projects/Stacker/lib/runtime/stacker_rt.c</p>
1338 <!-- ======================================================================= -->
1339 <div class="doc_subsection"><a name="driver"></a>Compiler Driver</div>
1340 <div class="doc_text">
1341 <p>See projects/Stacker/tools/stkrc/stkrc.cpp</p>
1343 <!-- ======================================================================= -->
1344 <div class="doc_subsection"><a name="tests"></a>Test Programs</div>
1345 <div class="doc_text">
1346 <p>See projects/Stacker/test/*.st</p>
1348 <!-- ======================================================================= -->
1349 <div class="doc_subsection"> <a name="exercise">Exercise</a></div>
1350 <div class="doc_text">
1351 <p>As you may have noted from a careful inspection of the Built-In word
1352 definitions, the ROLL word is not implemented. This word was left out of
1353 Stacker on purpose so that it can be an exercise for the student. The exercise
1354 is to implement the ROLL functionality (in your own workspace) and build a test
1355 program for it. If you can implement ROLL, you understand Stacker and probably
1356 a fair amount about LLVM since this is one of the more complicated Stacker
1357 operations. The work will almost be completely limited to the
1358 <a href="#compiler">compiler</a>.
1359 <p>The ROLL word is already recognized by both the lexer and parser but ignored
1360 by the compiler. That means you don't have to futz around with figuring out how
1361 to get the keyword recognized. It already is. The part of the compiler that
1362 you need to implement is the <code>ROLL</code> case in the
1363 <code>StackerCompiler::handle_word(int)</code> method.</p> See the
1364 implementations of PICK and SELECT in the same method to get some hints about
1365 how to complete this exercise.<p>
1368 <!-- ======================================================================= -->
1369 <div class="doc_subsection"><a name="todo">Things Remaining To Be Done</a></div>
1370 <div class="doc_text">
1371 <p>The initial implementation of Stacker has several deficiencies. If you're
1372 interested, here are some things that could be implemented better:</p>
1374 <li>Write an LLVM pass to compute the correct stack depth needed by the
1375 program. Currently the stack is set to a fixed number which means programs
1376 with large numbers of definitions might fail.</li>
1377 <li>Enhance to run on 64-bit platforms like SPARC. Right now the size of a
1378 pointer on 64-bit machines will cause incorrect results because of the
1379 32-bit size of a stack element currently supported. This feature was not
1380 implemented because LLVM needs a union type to be able to support the
1381 different sizes correctly (portably and efficiently).</li>
1382 <li>Write an LLVM pass to optimize the use of the global stack. The code
1383 emitted currently is somewhat wasteful. It gets cleaned up a lot by existing
1384 passes but more could be done.</li>
1385 <li>Add -O -O1 -O2 and -O3 optimization switches to the compiler driver to
1386 allow LLVM optimization without using "opt."</li>
1387 <li>Make the compiler driver use the LLVM linking facilities (with IPO)
1388 before depending on GCC to do the final link.</li>
1389 <li>Clean up parsing. It doesn't handle errors very well.</li>
1390 <li>Rearrange the StackerCompiler.cpp code to make better use of inserting
1391 instructions before a block's terminating instruction. I didn't figure this
1392 technique out until I was nearly done with LLVM. As it is, its a bad example
1393 of how to insert instructions!</li>
1394 <li>Provide for I/O to arbitrary files instead of just stdin/stdout.</li>
1395 <li>Write additional built-in words; with inspiration from FORTH</li>
1396 <li>Write additional sample Stacker programs.</li>
1397 <li>Add your own compiler writing experiences and tips in the
1398 <a href="#lessons">Lessons I Learned About LLVM</a> section.</li>
1402 <!-- *********************************************************************** -->
1406 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1407 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1408 <a href="http://validator.w3.org/check/referer"><img
1409 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
1411 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
1412 <a href="http://llvm.cs.uiuc.edu">LLVM Compiler Infrastructure</a><br>
1413 Last modified: $Date$