blob: dbfd4b8ec703577dacafd0ac155f38e3e4622c18 [file] [log] [blame]
Misha Brukmana3ce4292004-04-06 03:53:49 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5 <title>Extending LLVM: Adding instructions, intrinsics, types, etc.</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
7</head>
8
9<body>
10
11<div class="doc_title">
12 Extending LLVM: Adding instructions, intrinsics, types, etc.
13</div>
14
15<ol>
16 <li><a href="#introduction">Introduction and Warning</a></li>
Misha Brukmana3ce4292004-04-06 03:53:49 +000017 <li><a href="#intrinsic">Adding a new intrinsic function</a></li>
Chris Lattner36365402004-04-09 19:24:20 +000018 <li><a href="#instruction">Adding a new instruction</a></li>
Misha Brukmana3ce4292004-04-06 03:53:49 +000019 <li><a href="#type">Adding a new type</a>
20 <ol>
21 <li><a href="#fund_type">Adding a new fundamental type</a></li>
22 <li><a href="#derived_type">Adding a new derived type</a></li>
23 </ol></li>
24</ol>
25
Chris Lattner7911ce22004-05-23 21:07:27 +000026<div class="doc_author">
Chris Lattner5eb9f0d2005-05-11 03:53:53 +000027 <p>Written by <a href="http://misha.brukman.net">Misha Brukman</a>,
28 Brad Jones, and <a href="http://nondot.org/sabre">Chris Lattner</a></p>
Misha Brukmana3ce4292004-04-06 03:53:49 +000029</div>
30
31<!-- *********************************************************************** -->
32<div class="doc_section">
33 <a name="introduction">Introduction and Warning</a>
34</div>
35<!-- *********************************************************************** -->
36
37<div class="doc_text">
38
39<p>During the course of using LLVM, you may wish to customize it for your
40research project or for experimentation. At this point, you may realize that
41you need to add something to LLVM, whether it be a new fundamental type, a new
42intrinsic function, or a whole new instruction.</p>
43
44<p>When you come to this realization, stop and think. Do you really need to
45extend LLVM? Is it a new fundamental capability that LLVM does not support at
46its current incarnation or can it be synthesized from already pre-existing LLVM
47elements? If you are not sure, ask on the <a
48href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM-dev</a> list. The
49reason is that extending LLVM will get involved as you need to update all the
50different passes that you intend to use with your extension, and there are
51<em>many</em> LLVM analyses and transformations, so it may be quite a bit of
52work.</p>
53
Misha Brukmanb3b28272004-04-06 04:17:51 +000054<p>Adding an <a href="#intrinsic">intrinsic function</a> is easier than adding
55an instruction, and is transparent to optimization passes which treat it as an
56unanalyzable function. If your added functionality can be expressed as a
57function call, an intrinsic function is the method of choice for LLVM
58extension.</p>
59
Misha Brukmana3ce4292004-04-06 03:53:49 +000060<p>Before you invest a significant amount of effort into a non-trivial
61extension, <span class="doc_warning">ask on the list</span> if what you are
62looking to do can be done with already-existing infrastructure, or if maybe
63someone else is already working on it. You will save yourself a lot of time and
64effort by doing so.</p>
65
Misha Brukmana3ce4292004-04-06 03:53:49 +000066</div>
67
68<!-- *********************************************************************** -->
69<div class="doc_section">
Chris Lattner36365402004-04-09 19:24:20 +000070 <a name="intrinsic">Adding a new intrinsic function</a>
71</div>
72<!-- *********************************************************************** -->
73
74<div class="doc_text">
75
76<p>Adding a new intrinsic function to LLVM is much easier than adding a new
77instruction. Almost all extensions to LLVM should start as an intrinsic
78function and then be turned into an instruction if warranted.</p>
79
80<ol>
81<li><tt>llvm/docs/LangRef.html</tt>:
82 Document the intrinsic. Decide whether it is code generator specific and
83 what the restrictions are. Talk to other people about it so that you are
84 sure it's a good idea.</li>
85
86<li><tt>llvm/include/llvm/Intrinsics.h</tt>:
87 add an enum in the <tt>llvm::Intrinsic</tt> namespace</li>
88
Chris Lattner36365402004-04-09 19:24:20 +000089<li><tt>llvm/lib/VMCore/Verifier.cpp</tt>:
90 Add code to check the invariants of the intrinsic are respected.</li>
91
92<li><tt>llvm/lib/VMCore/Function.cpp (<tt>Function::getIntrinsicID()</tt>)</tt>:
93 Identify the new intrinsic function, returning the enum for the intrinsic
94 that you added.</li>
Chris Lattner0190fdb2004-04-10 06:56:53 +000095
96<li><tt>llvm/lib/Analysis/BasicAliasAnalysis.cpp</tt>: If the new intrinsic does
Chris Lattner81519d92004-06-20 07:53:22 +000097 not access memory or does not write to memory, add it to the relevant list
Chris Lattner0190fdb2004-04-10 06:56:53 +000098 of functions.</li>
99
Nate Begeman2f86c222006-01-14 01:27:10 +0000100<li><tt>llvm/lib/Analysis/ConstantFolding.cpp</tt>: If it is possible to
101 constant fold your intrinsic, add support to it in the
102 <tt>canConstantFoldCallTo</tt> and <tt>ConstantFoldCall</tt> functions.</li>
103
104<li><tt>llvm/lib/Transforms/Utils/Local.cpp</tt>: If your intrinsic has no side-
105 effects, add it to the list of intrinsics in the
106 <tt>isInstructionTriviallyDead</tt> function.</li>
Chris Lattnerd828bc62004-04-13 19:48:55 +0000107
Chris Lattner36365402004-04-09 19:24:20 +0000108<li>Test your intrinsic</li>
Misha Brukmana4242282004-12-01 20:58:54 +0000109
110<li><tt>llvm/test/Regression/*</tt>: add your test cases to the test suite</li>
Chris Lattner36365402004-04-09 19:24:20 +0000111</ol>
112
Chris Lattner5eb9f0d2005-05-11 03:53:53 +0000113<p>Once the intrinsic has been added to the system, you must add code generator
114support for it. Generally you must do the following steps:</p>
115
116<dl>
117<dt>Add support to the C backend in <tt>lib/Target/CBackend/</tt></dt>
118
119<dd>Depending on the intrinsic, there are a few ways to implement this. First,
120if it makes sense to lower the intrinsic to an expanded sequence of C code in
121all cases, just emit the expansion in <tt>visitCallInst</tt>. Second, if the
122intrinsic has some way to express it with GCC (or any other compiler)
123extensions, it can be conditionally supported based on the compiler compiling
124the CBE output (see llvm.prefetch for an example). Third, if the intrinsic
125really has no way to be lowered, just have the code generator emit code that
126prints an error message and calls abort if executed.
127</dd>
128
129<dt>Add a enum value for the SelectionDAG node in
130<tt>include/llvm/CodeGen/SelectionDAGNodes.h</tt></dt>
131
132<dd>Also, add code to <tt>lib/CodeGen/SelectionDAG/SelectionDAG.cpp</tt> (and
133<tt>SelectionDAGPrinter.cpp</tt>) to print the node.</dd>
134
135<dt>Add code to <tt>SelectionDAG/SelectionDAGISel.cpp</tt> to recognize the
136intrinsic.</dt>
137
138<dd>Presumably the intrinsic should be recognized and turned into the node you
139added above.</dd>
140
141<dt>Add code to <tt>SelectionDAG/LegalizeDAG.cpp</tt> to <a
142href="CodeGenerator.html#selectiondag_legalize">legalize, promote, and
143expand</a> the node as necessary.</dt>
144
145<dd>If the intrinsic can be expanded to primitive operations, legalize can break
146the node down into other elementary operations that are be supported.</dd>
147
148<dt>Add target-specific support to specific code generators.</dt>
149
150<dd>Extend the code generators you are interested in to recognize and support
151the node, emitting the code you want.</dd>
152</dl>
153
154<p>
155Unfortunately, the process of extending the code generator to support a new node
156is not extremely well documented. As such, it is often helpful to look at other
157intrinsics (e.g. <tt>llvm.ctpop</tt>) to see how they are recognized and turned
158into a node by <tt>SelectionDAGISel.cpp</tt>, legalized by
159<tt>LegalizeDAG.cpp</tt>, then finally emitted by the various code generators.
160</p>
Chris Lattner36365402004-04-09 19:24:20 +0000161
162</div>
163
164<!-- *********************************************************************** -->
165<div class="doc_section">
Misha Brukmana3ce4292004-04-06 03:53:49 +0000166 <a name="instruction">Adding a new instruction</a>
167</div>
168<!-- *********************************************************************** -->
169
170<div class="doc_text">
171
172<p><span class="doc_warning">WARNING: adding instructions changes the bytecode
Misha Brukmanb3b28272004-04-06 04:17:51 +0000173format, and it will take some effort to maintain compatibility with
174the previous version.</span> Only add an instruction if it is absolutely
Misha Brukmana3ce4292004-04-06 03:53:49 +0000175necessary.</p>
176
177<ol>
Misha Brukmanb3b28272004-04-06 04:17:51 +0000178
Misha Brukmana3ce4292004-04-06 03:53:49 +0000179<li><tt>llvm/include/llvm/Instruction.def</tt>:
180 add a number for your instruction and an enum name</li>
181
Misha Brukman47b14a42004-07-29 17:30:56 +0000182<li><tt>llvm/include/llvm/Instructions.h</tt>:
Misha Brukmana3ce4292004-04-06 03:53:49 +0000183 add a definition for the class that will represent your instruction</li>
184
185<li><tt>llvm/include/llvm/Support/InstVisitor.h</tt>:
186 add a prototype for a visitor to your new instruction type</li>
187
188<li><tt>llvm/lib/AsmParser/Lexer.l</tt>:
189 add a new token to parse your instruction from assembly text file</li>
190
191<li><tt>llvm/lib/AsmParser/llvmAsmParser.y</tt>:
192 add the grammar on how your instruction can be read and what it will
193 construct as a result</li>
194
Misha Brukmane39cd632004-09-28 16:58:12 +0000195<li><tt>llvm/lib/Bytecode/Reader/Reader.cpp</tt>:
Misha Brukmana3ce4292004-04-06 03:53:49 +0000196 add a case for your instruction and how it will be parsed from bytecode</li>
197
198<li><tt>llvm/lib/VMCore/Instruction.cpp</tt>:
199 add a case for how your instruction will be printed out to assembly</li>
200
Chris Lattner8f363212004-07-29 17:31:57 +0000201<li><tt>llvm/lib/VMCore/Instructions.cpp</tt>:
Misha Brukmane39cd632004-09-28 16:58:12 +0000202 implement the class you defined in
203 <tt>llvm/include/llvm/Instructions.h</tt></li>
Misha Brukmana3ce4292004-04-06 03:53:49 +0000204
Misha Brukmana4242282004-12-01 20:58:54 +0000205<li>Test your instruction</li>
206
207<li><tt>llvm/lib/Target/*</tt>:
208 Add support for your instruction to code generators, or add a lowering
209 pass.</li>
210
211<li><tt>llvm/test/Regression/*</tt>: add your test cases to the test suite.</li>
212
Misha Brukmana3ce4292004-04-06 03:53:49 +0000213</ol>
214
215<p>Also, you need to implement (or modify) any analyses or passes that you want
216to understand this new instruction.</p>
217
218</div>
219
Misha Brukmana3ce4292004-04-06 03:53:49 +0000220
221<!-- *********************************************************************** -->
222<div class="doc_section">
223 <a name="type">Adding a new type</a>
224</div>
225<!-- *********************************************************************** -->
226
227<div class="doc_text">
228
229<p><span class="doc_warning">WARNING: adding new types changes the bytecode
230format, and will break compatibility with currently-existing LLVM
231installations.</span> Only add new types if it is absolutely necessary.</p>
232
233</div>
234
235<!-- ======================================================================= -->
236<div class="doc_subsection">
237 <a name="fund_type">Adding a fundamental type</a>
238</div>
239
240<div class="doc_text">
241
242<ol>
243
Misha Brukmana3ce4292004-04-06 03:53:49 +0000244<li><tt>llvm/include/llvm/Type.h</tt>:
Chris Lattner55f95012005-04-23 21:59:11 +0000245 add enum for the new type; add static <tt>Type*</tt> for this type</li>
Misha Brukmana3ce4292004-04-06 03:53:49 +0000246
247<li><tt>llvm/lib/VMCore/Type.cpp</tt>:
248 add mapping from <tt>TypeID</tt> =&gt; <tt>Type*</tt>;
249 initialize the static <tt>Type*</tt></li>
250
251<li><tt>llvm/lib/AsmReader/Lexer.l</tt>:
252 add ability to parse in the type from text assembly</li>
253
254<li><tt>llvm/lib/AsmReader/llvmAsmParser.y</tt>:
255 add a token for that type</li>
256
257</ol>
258
259</div>
260
261<!-- ======================================================================= -->
262<div class="doc_subsection">
263 <a name="derived_type">Adding a derived type</a>
264</div>
265
266<div class="doc_text">
267
Chris Lattner8dad40c2004-08-12 19:06:24 +0000268<ol>
Chris Lattner8dad40c2004-08-12 19:06:24 +0000269<li><tt>llvm/include/llvm/Type.h</tt>:
Chris Lattner55f95012005-04-23 21:59:11 +0000270 add enum for the new type; add a forward declaration of the type
Misha Brukman7cc8a892004-08-12 19:58:43 +0000271 also</li>
Chris Lattner8dad40c2004-08-12 19:06:24 +0000272
Chris Lattner47746aa2005-11-13 02:09:55 +0000273<li><tt>llvm/include/llvm/DerivedTypes.h</tt>:
Chris Lattner8dad40c2004-08-12 19:06:24 +0000274 add new class to represent new class in the hierarchy; add forward
275 declaration to the TypeMap value type</li>
276
277<li><tt>llvm/lib/VMCore/Type.cpp</tt>:
278 add support for derived type to:
Misha Brukman7cc8a892004-08-12 19:58:43 +0000279<div class="doc_code">
280<pre>
281std::string getTypeDescription(const Type &amp;Ty,
282 std::vector&lt;const Type*&gt; &amp;TypeStack)
283bool TypesEqual(const Type *Ty, const Type *Ty2,
284 std::map&lt;const Type*, const Type*&gt; &amp; EqTypes)
285</pre>
286</div>
287 add necessary member functions for type, and factory methods</li>
Chris Lattner8dad40c2004-08-12 19:06:24 +0000288
289<li><tt>llvm/lib/AsmReader/Lexer.l</tt>:
290 add ability to parse in the type from text assembly</li>
291
292<li><tt>llvm/lib/ByteCode/Writer/Writer.cpp</tt>:
Misha Brukman7cc8a892004-08-12 19:58:43 +0000293 modify <tt>void BytecodeWriter::outputType(const Type *T)</tt> to serialize
294 your type</li>
Chris Lattner8dad40c2004-08-12 19:06:24 +0000295
296<li><tt>llvm/lib/ByteCode/Reader/Reader.cpp</tt>:
Misha Brukman7cc8a892004-08-12 19:58:43 +0000297 modify <tt>const Type *BytecodeReader::ParseType()</tt> to read your data
298 type</li>
Chris Lattner8dad40c2004-08-12 19:06:24 +0000299
300<li><tt>llvm/lib/VMCore/AsmWriter.cpp</tt>:
Misha Brukman7cc8a892004-08-12 19:58:43 +0000301 modify
302<div class="doc_code">
303<pre>
304void calcTypeName(const Type *Ty,
305 std::vector&lt;const Type*&gt; &amp;TypeStack,
306 std::map&lt;const Type*,std::string&gt; &amp;TypeNames,
307 std::string &amp; Result)
308</pre>
309</div>
Chris Lattner8dad40c2004-08-12 19:06:24 +0000310 to output the new derived type
311</li>
312
313
314</ol>
Misha Brukmana3ce4292004-04-06 03:53:49 +0000315
316</div>
317
318<!-- *********************************************************************** -->
319
320<hr>
321<address>
322 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
323 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
324 <a href="http://validator.w3.org/check/referer"><img
325 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
326
Misha Brukmana3ce4292004-04-06 03:53:49 +0000327 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
328 <br>
329 Last modified: $Date$
330</address>
331
332</body>
333</html>