1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>TableGen Fundamentals</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
10 <div class="doc_title">TableGen Fundamentals</div>
13 <li><a href="#introduction">Introduction</a></li>
15 <li><a href="#concepts">Basic concepts</a></li>
16 <li><a href="#example">An example record</a></li>
17 <li><a href="#running">Running TableGen</a></li>
19 <li><a href="#syntax">TableGen syntax</a></li>
21 <li><a href="#primitives">TableGen primitives</a></li>
23 <li><a href="#comments">TableGen comments</a></li>
24 <li><a href="#types">The TableGen type system</a></li>
25 <li><a href="#values">TableGen values and expressions</a></li>
27 <li><a href="#classesdefs">Classes and definitions</a></li>
29 <li><a href="#valuedef">Value definitions</a></li>
30 <li><a href="#recordlet">'let' expressions</a></li>
31 <li><a href="#templateargs">Class template arguments</a></li>
33 <li><a href="#filescope">File scope entities</a></li>
35 <li><a href="#include">File inclusion</a></li>
36 <li><a href="#globallet">'let' expressions</a></li>
39 <li><a href="#backends">TableGen backends</a></li>
41 <li><a href="#">todo</a></li>
43 <li><a href="#codegenerator">The LLVM code generator</a></li>
45 <li><a href="#">todo</a></li>
49 <!-- *********************************************************************** -->
50 <div class="doc_section"><a name="introduction">Introduction</a></div>
51 <!-- *********************************************************************** -->
53 <div class="doc_text">
55 <p>TableGen's purpose is to help a human develop and maintain records of
56 domain-specific information. Because there may be a large number of these
57 records, it is specifically designed to allow writing flexible descriptions and
58 for common features of these records to be factored out. This reduces the
59 amount of duplication in the description, reduces the chance of error, and
60 makes it easier to structure domain specific information.</p>
62 <p>The core part of TableGen <a href="#syntax">parses a file</a>, instantiates
63 the declarations, and hands the result off to a domain-specific "<a
64 href="#backends">TableGen backend</a>" for processing. The current major user
65 of TableGen is the <a href="#codegenerator">LLVM code generator</a>.
69 Note that if you work on TableGen much, and use emacs or vim, that you can find
70 an emacs "TableGen mode" and a vim language file in <tt>llvm/utils/emacs</tt>
71 and <tt>llvm/utils/vim</tt> directory of your LLVM distribution, respectively.
76 <!-- ======================================================================= -->
77 <div class="doc_subsection">
78 <a name="running">Basic concepts</a>
81 <div class="doc_text">
84 TableGen files consist of two key parts: 'classes' and 'definitions', both of
85 which are considered 'records'.
89 <b>TableGen records</b> have a unique name, a list of values, and a list of
90 superclasses. The list of values is main data that TableGen builds for each
91 record, it is this that holds the domain specific information for the
92 application. The interpretation of this data is left to a specific <a
93 href="#backends">TableGen backend</a>, but the structure and format rules are
94 taken care of and fixed by TableGen.
98 <b>TableGen definitions</b> are the concrete form of 'records'. These generally
99 do not have any undefined values, and are marked with the '<tt>def</tt>'
104 <b>TableGen classes</b> are abstract records that are used to build and describe
105 other records. These 'classes' allow the end-user to build abstractions for
106 either the domain they are targetting (such as "Register", "RegisterClass", and
107 "Instruction" in the LLVM code generator) or for the implementor to help factor
108 out common properties of records (such as "FPInst", which is used to represent
109 floating point instructions in the X86 backend). TableGen keeps track of all of
110 the classes that are used to build up a definition, so the backend can find all
111 definitions of a particular class, such as "Instruction".
116 <!-- ======================================================================= -->
117 <div class="doc_subsection">
118 <a name="example">An example record</a>
121 <div class="doc_text">
124 With no other arguments, TableGen parses the specified file and prints out all
125 of the classes, then all of the definitions. This is a good way to see what the
126 various definitions expand to fully. Running this on the <tt>X86.td</tt> file
127 prints this (at the time of this writing):
133 <b>def</b> ADDrr8 { <i>// Instruction X86Inst I2A8 Pattern</i>
134 <b>string</b> Name = "add";
135 <b>string</b> Namespace = "X86";
136 <b>list</b><Register> Uses = [];
137 <b>list</b><Register> Defs = [];
138 <b>bit</b> isReturn = 0;
139 <b>bit</b> isBranch = 0;
140 <b>bit</b> isCall = 0;
141 <b>bit</b> isTwoAddress = 1;
142 <b>bit</b> isTerminator = 0;
143 <b>dag</b> Pattern = (set R8, (plus R8, R8));
144 <b>bits</b><8> Opcode = { 0, 0, 0, 0, 0, 0, 0, 0 };
145 Format Form = MRMDestReg;
146 <b>bits</b><5> FormBits = { 0, 0, 0, 1, 1 };
148 <b>bits</b><3> TypeBits = { 0, 0, 1 };
149 <b>bit</b> hasOpSizePrefix = 0;
150 <b>bit</b> printImplicitUses = 0;
151 <b>bits</b><4> Prefix = { 0, 0, 0, 0 };
153 <b>bits</b><3> FPFormBits = { 0, 0, 0 };
159 This definition corresponds to an 8-bit register-register add instruction in the
160 X86. The string after the '<tt>def</tt>' string indicates the name of the
161 record ("<tt>ADDrr8</tt>" in this case), and the comment at the end of the line
162 indicates the superclasses of the definition. The body of the record contains
163 all of the data that TableGen assembled for the record, indicating that the
164 instruction is part of the "X86" namespace, should be printed as "<tt>add</tt>"
165 in the assembly file, it is a two-address instruction, has a particular
166 encoding, etc. The contents and semantics of the information in the record is
167 specific to the needs of the X86 backend, and is only shown as an example.
171 As you can see, a lot of information is needed for every instruction supported
172 by the code generator, and specifying it all manually would be unmaintainble,
173 prone to bugs, and tiring to do in the first place. Because we are using
174 TableGen, all of the information was derived from the following definition:
178 <b>def</b> ADDrr8 : I2A8<"add", 0x00, MRMDestReg>,
179 Pattern<(set R8, (plus R8, R8))>;
183 This definition makes use of the custom I2A8 (two address instruction with 8-bit
184 operand) class, which is defined in the X86-specific TableGen file to factor out
185 the common features that instructions of its class share. A key feature of
186 TableGen is that it allows the end-user to define the abstractions they prefer
187 to use when describing their information.
192 <!-- ======================================================================= -->
193 <div class="doc_subsection">
194 <a name="running">Running TableGen</a>
197 <div class="doc_text">
200 TableGen runs just like any other LLVM tool. The first (optional) argument
201 specifies the file to read. If a filename is not specified, <tt>tblgen</tt>
202 reads from standard input.
206 To be useful, one of the <a href="#backends">TableGen backends</a> must be used.
207 These backends are selectable on the command line (type '<tt>tblgen --help</tt>'
208 for a list). For example, to get a list of all of the definitions that subclass
209 a particular type (which can be useful for building up an enum list of these
210 records), use the <tt>--print-enums</tt> option:
214 $ tblgen X86.td -print-enums -class=Register
215 AH, AL, AX, BH, BL, BP, BX, CH, CL, CX, DH, DI, DL, DX,
216 EAX, EBP, EBX, ECX, EDI, EDX, ESI, ESP, FP0, FP1, FP2, FP3, FP4, FP5, FP6,
217 SI, SP, ST0, ST1, ST2, ST3, ST4, ST5, ST6, ST7,
219 $ tblgen X86.td -print-enums -class=Instruction
220 ADCrr32, ADDri16, ADDri16b, ADDri32, ADDri32b, ADDri8, ADDrr16, ADDrr32,
221 ADDrr8, ADJCALLSTACKDOWN, ADJCALLSTACKUP, ANDri16, ANDri16b, ANDri32, ANDri32b,
222 ANDri8, ANDrr16, ANDrr32, ANDrr8, BSWAPr32, CALLm32, CALLpcrel32, ...
226 The default backend prints out all of the records, as described <a
227 href="#example">above</a>.
231 If you plan to use TableGen for some purpose, you will most likely have to <a
232 href="#backends">write a backend</a> that extracts the information specific to
233 what you need and formats it in the appropriate way.
239 <!-- *********************************************************************** -->
240 <div class="doc_section"><a name="syntax">TableGen syntax</a></div>
241 <!-- *********************************************************************** -->
243 <div class="doc_text">
246 TableGen doesn't care about the meaning of data (that is up to the backend to
247 define), but it does care about syntax, and it enforces a simple type system.
248 This section describes the syntax and the constructs allowed in a TableGen file.
253 <!-- ======================================================================= -->
254 <div class="doc_subsection">
255 <a name="primitives">TableGen primitives</tt></a>
258 <!----------------------------------------------------------------------------->
259 <div class="doc_subsubsection">
260 <a name="comments">TableGen comments</tt></a>
263 <div class="doc_text">
265 <p>TableGen supports BCPL style "<tt>//</tt>" comments, which run to the end of
266 the line, and it also supports <b>nestable</b> "<tt>/* */</tt>" comments.</p>
271 <!----------------------------------------------------------------------------->
272 <div class="doc_subsubsection">
273 <a name="types">The TableGen type system</tt></a>
276 <div class="doc_text">
278 TableGen files are strongly typed, in a simple (but complete) type-system.
279 These types are used to perform automatic conversions, check for errors, and to
280 help interface designers constrain the input that they allow. Every <a
281 href="#valuedef">value definition</a> is required to have an associated type.
285 TableGen supports a mixture of very low-level types (such as <tt>bit</tt>) and
286 very high-level types (such as <tt>dag</tt>). This flexibility is what allows
287 it to describe a wide range of information conveniently and compactly. The
293 <li>"<tt><b>bit</b></tt>" - A 'bit' is a boolean value that can hold either 0 or
296 <li>"<tt><b>int</b></tt>" - The 'int' type represents a simple 32-bit integer
297 value, such as 5.</li>
299 <li>"<tt><b>string</b></tt>" - The 'string' type represents an ordered sequence
300 of characters of arbitrary length.</li>
302 <li>"<tt><b>bits</b><n></tt>" - A 'bits' type is a arbitrary, but fixed,
303 size integer that is broken up into individual bits. This type is useful
304 because it can handle some bits being defined while others are undefined.</li>
306 <li>"<tt><b>list</b><ty></tt>" - This type represents a list whose
307 elements are some other type. The contained type is arbitrary: it can even be
308 another list type.</li>
310 <li>Class type - Specifying a class name in a type context means that the
311 defined value must be a subclass of the specified class. This is useful in
312 conjunction with the "list" type, for example, to constrain the elements of the
313 list to a common base class (e.g., a <tt><b>list</b><Register></tt> can
314 only contain definitions derived from the "<tt>Register</tt>" class).</li>
316 <li>"<tt><b>code</b></tt>" - This represents a big hunk of text. NOTE: I don't
317 remember why this is distinct from string!</li>
319 <li>"<tt><b>dag</b></tt>" - This type represents a nestable directed graph of
325 To date, these types have been sufficient for describing things that TableGen
326 has been used for, but it is straight-forward to extend this list if needed.
331 <!----------------------------------------------------------------------------->
332 <div class="doc_subsubsection">
333 <a name="values">TableGen values and expressions</tt></a>
338 TableGen allows for a pretty reasonable number of different expression forms
339 when building up values. These forms allow the TableGen file to be written in a
340 natural syntax and flavor for the application. The current expression forms
345 <li>? - Uninitialized field.</li>
346 <li>0b1001011 - Binary integer value.</li>
347 <li>07654321 - Octal integer value (indicated by a leading 0).</li>
348 <li>7 - Decimal integer value.</li>
349 <li>0x7F - Hexadecimal integer value.</li>
350 <li>"foo" - String value.</li>
351 <li>[{ .... }] - Code fragment.</li>
352 <li>[ X, Y, Z ] - List value.</li>
353 <li>{ a, b, c } - Initializer for a "bits<3>" value.</li>
354 <li>value - Value reference.</li>
355 <li>value{17} - Access to one or more bits of a value.</li>
356 <li>DEF - Reference to a record definition.</li>
357 <li>X.Y - Reference to the subfield of a value.</li>
359 <li>(DEF a, b) - A dag value. The first element is required to be a record
360 definition, the remaining elements in the list may be arbitrary other values,
361 including nested 'dag' values.</li>
366 Note that all of the values have rules specifying how they convert to to values
367 for different types. These rules allow you to assign a value like "7" to a
368 "bits<4>" value, for example.
376 <!-- ======================================================================= -->
377 <div class="doc_subsection">
378 <a name="classesdefs">Classes and definitions</tt></a>
383 As mentioned in the <a href="#concepts">intro</a>, classes and definitions
384 (collectively known as 'records') in TableGen are the main high-level unit of
385 information that TableGen collects. Records are defined with a <tt>def</tt> or
386 <tt>class</tt> keyword, the record name, and an optional list of "<a
387 href="templateargs">template arguments</a>". If the record has superclasses,
388 they are specified as a comma seperated list that starts with a colon character
389 (":"). If <a href="#valuedef">value definitions</a> or <a href="#recordlet">let
390 expressions</a> are needed for the class they are enclosed in curly braces
391 ("{}"), otherwise the record ends with a semicolon. Here is a simple TableGen
396 <b>class</b> C { <b>bit</b> V = 1; }
399 <b>string</b> Greeting = "hello";
404 This example defines two definitions, <tt>X</tt> and <tt>Y</tt>, both of which
405 derive from the <tt>C</tt> class. Because of this, they both get the <tt>V</tt>
406 bit value. The <tt>Y</tt> definition also gets the Greeting member as well.
411 <!----------------------------------------------------------------------------->
412 <div class="doc_subsubsection">
413 <a name="valuedef">Value definitions</tt></a>
416 <div class="doc_text">
418 Value definitions define named entries in records. A value must be defined
419 before it can be referred to as the operand for another value definition, or
420 before the value is reset with a <a href="#recordlet">let expression</a>. A
421 value is defined by specifying a <a href="#types">TableGen type</a> and a name.
422 If an initial value is available, it may be specified after the type with an
423 equal sign. Value definitions require terminating semicolons.
426 <!----------------------------------------------------------------------------->
427 <div class="doc_subsubsection">
428 <a name="recordlet">'let' expressions</tt></a>
431 <div class="doc_text">
433 A record-level let expression is used to change the value of a value definition
434 in a record. This is primarily useful when a superclass defines a value that a
435 derived class or definitions wants to override. Let expressions consist of the
436 '<tt>let</tt>' keyword, followed by a value name, an equal sign ("="), and a new
437 value for example, a new class could be added to the example above, redefining
438 the <tt>V</tt> field for all of its subclasses:</p>
441 <b>class</b> D : C { let V = 0; }
446 In this case, the <tt>Z</tt> definition will have a zero value for its "V"
447 value, despite the fact that it derives (indirectly) from the <tt>C</tt> class,
448 because the <tt>D</tt> class overrode its value.
453 <!----------------------------------------------------------------------------->
454 <div class="doc_subsubsection">
455 <a name="templateargs">Class template arguments</tt></a>
458 <div class="doc_text">
459 and default values...
464 <!-- ======================================================================= -->
465 <div class="doc_subsection">
466 <a name="filescope">File scope entities</tt></a>
469 <!----------------------------------------------------------------------------->
470 <div class="doc_subsubsection">
471 <a name="include">File inclusion</tt></a>
474 <div class="doc_text">
476 TableGen supports the '<tt>include</tt>' token, which textually substitutes the
477 specified file in place of the include directive. The filename should be
478 specified as a double quoted string immediately after the '<tt>include</tt>'
482 <b>include</b> "foo.td"
487 <!----------------------------------------------------------------------------->
488 <div class="doc_subsubsection">
489 <a name="globallet">'let' expressions</tt></a>
492 <div class="doc_text">
494 "let" expressions at file scope are similar to <a href="#recordlet">"let"
495 expressions within a record</a>, except they can specify a value binding for
496 multiple records at a time, and may be useful in certain other cases.
497 File-scope let expressions are really just another way that TableGen allows the
498 end-user to factor out commonality from the records.
502 File-scope "let" expressions take a comma-seperated list of bindings to apply,
503 and one of more records to bind the values in. Here are some examples:
507 <b>let</b> isTerminator = 1, isReturn = 1 <b>in</b>
508 <b>def</b> RET : X86Inst<"ret", 0xC3, RawFrm, NoArg>;
510 <b>let</b> isCall = 1 <b>in</b>
511 <i>// All calls clobber the non-callee saved registers...</i>
512 <b>let</b> Defs = [EAX, ECX, EDX, FP0, FP1, FP2, FP3, FP4, FP5, FP6] in {
513 <b>def</b> CALLpcrel32 : X86Inst<"call", 0xE8, RawFrm, NoArg>;
514 <b>def</b> CALLr32 : X86Inst<"call", 0xFF, MRMS2r, Arg32>;
515 <b>def</b> CALLm32 : X86Inst<"call", 0xFF, MRMS2m, Arg32>;
520 File-scope "let" expressions are often useful when a couple of definitions need
521 to be added to several records, and the records do not otherwise need to be
522 opened, as in the case with the CALL* instructions above.
527 <!-- *********************************************************************** -->
528 <div class="doc_section"><a name="backends">TableGen backends</a></div>
529 <!-- *********************************************************************** -->
531 <div class="doc_text">
534 How they work, how to write one. This section should not contain details about
535 any particular backend, except maybe -print-enums as an example. This should
536 highlight the APIs in TableGen/Record.h.
542 <!-- *********************************************************************** -->
543 <div class="doc_section"><a name="codegenerator">The LLVM code generator</a></div>
544 <!-- *********************************************************************** -->
546 <div class="doc_text">
549 This is just a temporary, convenient, place to put stuff about the code
550 generator before it gets its own document. This should describe all of the
551 tablegen backends used by the code generator and the classes/definitions they
559 <!-- *********************************************************************** -->
561 <div class="doc_footer">
562 <address><a href="mailto:sabre@nondot.org">Chris Lattner</a></address>
563 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
565 Last modified: $Date$