1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>Extending LLVM: Adding instructions, intrinsics, types, etc.</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
11 <div class="doc_title">
12 Extending LLVM: Adding instructions, intrinsics, types, etc.
16 <li><a href="#introduction">Introduction and Warning</a></li>
17 <li><a href="#intrinsic">Adding a new intrinsic function</a></li>
18 <li><a href="#instruction">Adding a new instruction</a></li>
19 <li><a href="#sdnode">Adding a new SelectionDAG node</a></li>
20 <li><a href="#type">Adding a new type</a>
22 <li><a href="#fund_type">Adding a new fundamental type</a></li>
23 <li><a href="#derived_type">Adding a new derived type</a></li>
27 <div class="doc_author">
28 <p>Written by <a href="http://misha.brukman.net">Misha Brukman</a>,
29 Brad Jones, Nate Begeman,
30 and <a href="http://nondot.org/sabre">Chris Lattner</a></p>
33 <!-- *********************************************************************** -->
34 <div class="doc_section">
35 <a name="introduction">Introduction and Warning</a>
37 <!-- *********************************************************************** -->
39 <div class="doc_text">
41 <p>During the course of using LLVM, you may wish to customize it for your
42 research project or for experimentation. At this point, you may realize that
43 you need to add something to LLVM, whether it be a new fundamental type, a new
44 intrinsic function, or a whole new instruction.</p>
46 <p>When you come to this realization, stop and think. Do you really need to
47 extend LLVM? Is it a new fundamental capability that LLVM does not support at
48 its current incarnation or can it be synthesized from already pre-existing LLVM
49 elements? If you are not sure, ask on the <a
50 href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM-dev</a> list. The
51 reason is that extending LLVM will get involved as you need to update all the
52 different passes that you intend to use with your extension, and there are
53 <em>many</em> LLVM analyses and transformations, so it may be quite a bit of
56 <p>Adding an <a href="#intrinsic">intrinsic function</a> is easier than adding
57 an instruction, and is transparent to optimization passes which treat it as an
58 unanalyzable function. If your added functionality can be expressed as a
59 function call, an intrinsic function is the method of choice for LLVM
62 <p>Before you invest a significant amount of effort into a non-trivial
63 extension, <span class="doc_warning">ask on the list</span> if what you are
64 looking to do can be done with already-existing infrastructure, or if maybe
65 someone else is already working on it. You will save yourself a lot of time and
66 effort by doing so.</p>
70 <!-- *********************************************************************** -->
71 <div class="doc_section">
72 <a name="intrinsic">Adding a new intrinsic function</a>
74 <!-- *********************************************************************** -->
76 <div class="doc_text">
78 <p>Adding a new intrinsic function to LLVM is much easier than adding a new
79 instruction. Almost all extensions to LLVM should start as an intrinsic
80 function and then be turned into an instruction if warranted.</p>
83 <li><tt>llvm/docs/LangRef.html</tt>:
84 Document the intrinsic. Decide whether it is code generator specific and
85 what the restrictions are. Talk to other people about it so that you are
86 sure it's a good idea.</li>
88 <li><tt>llvm/include/llvm/Intrinsics.td</tt>:
89 Add an entry for your intrinsic.</li>
91 <li><tt>llvm/lib/Analysis/BasicAliasAnalysis.cpp</tt>: If the new intrinsic does
92 not access memory or does not write to memory, add it to the relevant list
95 <li><tt>llvm/lib/Analysis/ConstantFolding.cpp</tt>: If it is possible to
96 constant fold your intrinsic, add support to it in the
97 <tt>canConstantFoldCallTo</tt> and <tt>ConstantFoldCall</tt> functions.</li>
99 <li><tt>llvm/lib/Transforms/Utils/Local.cpp</tt>: If your intrinsic has no side-
100 effects, add it to the list of intrinsics in the
101 <tt>isInstructionTriviallyDead</tt> function.</li>
103 <li><tt>llvm/test/Regression/*</tt>: Add test cases for your test cases to the
107 <p>Once the intrinsic has been added to the system, you must add code generator
108 support for it. Generally you must do the following steps:</p>
111 <dt>Add support to the C backend in <tt>lib/Target/CBackend/</tt></dt>
113 <dd>Depending on the intrinsic, there are a few ways to implement this. For
114 most intrinsics, it makes sense to add code to lower your intrinsic in
115 <tt>LowerIntrinsicCall</tt> in <tt>lib/CodeGen/IntrinsicLowering.cpp</tt>.
116 Second, if it makes sense to lower the intrinsic to an expanded sequence of C
117 code in all cases, just emit the expansion in <tt>visitCallInst</tt> in
118 <tt>Writer.cpp</tt>. If the intrinsic has some way to express it with GCC
119 (or any other compiler) extensions, it can be conditionally supported based on
120 the compiler compiling the CBE output (see <tt>llvm.prefetch</tt> for an
122 Third, if the intrinsic really has no way to be lowered, just have the code
123 generator emit code that prints an error message and calls abort if executed.
127 <dt>Add support to the SelectionDAG Instruction Selector in
128 <tt>lib/CodeGen/SelectionDAG/</tt></dt>
130 <dd>Since most targets in LLVM use the SelectionDAG framework for generating
131 code, you will likely need to add support for your intrinsic there as well.
132 This is usually accomplished by adding a new node, and then teaching the
133 SelectionDAG code how to handle that node. To do this, follow the steps in
134 the <a href="#sdnode">Adding a new SelectionDAG node</a> section.</dd>
137 <dt>Once you have added the new node, add code to
138 <tt>SelectionDAG/SelectionDAGISel.cpp</tt> to recognize the intrinsic. In most
139 cases, the intrinsic will just be turned into the node you just added. For an
140 example of this, see how <tt>visitIntrinsicCall</tt> handles
141 <tt>Intrinsic::ctpop_*</tt>.
146 <!-- *********************************************************************** -->
147 <div class="doc_section">
148 <a name="sdnode">Adding a new SelectionDAG node</a>
150 <!-- *********************************************************************** -->
152 <div class="doc_text">
154 <p>As with intrinsics, adding a new SelectionDAG node to LLVM is much easier
155 than adding a new instruction. New nodes are often added to help represent
156 instructions common to many targets. These nodes often map to an LLVM
157 instruction (add, sub) or intrinsic (byteswap, population count). In other
158 cases, new nodes have been added to allow many targets to perform a common task
159 (converting between floating point and integer representation) or capture more
160 complicated behavior in a single node (rotate).</p>
163 <li><tt>include/llvm/CodeGen/SelectionDAGNodes.h</tt>:
164 Add an enum value for the new SelectionDAG node.</li>
165 <li><tt>lib/CodeGen/SelectionDAG/SelectionDAG.cpp</tt>:
166 Add code to print the node to <tt>getOperationName</tt>. If your new node
167 can be evaluated at compile time when given constant arguments (such as an
168 add of a constant with another constant), find the <tt>getNode</tt> method
169 that takes the appropriate number of arguments, and add a case for your node
170 to the switch statement that performs constant folding for nodes that take
171 the same number of arguments as your new node.</li>
172 <li><tt>lib/CodeGen/SelectionDAG/LegalizeDAG.cpp</tt>:
173 Add code to <a href="CodeGenerator.html#selectiondag_legalize">legalize,
174 promote, and expand</a> the node as necessary. At a minimum, you will need
175 to add a case statement for your node in <tt>LegalizeOp</tt> which calls
176 LegalizeOp on the node's operands, and returns a new node if any of the
177 operands changed as a result of being legalized. It is likely that not all
178 targets supported by the SelectionDAG framework will natively support the
179 new node. In this case, you must also add code in your node's case
180 statement in <tt>LegalizeOp</tt> to Expand your node into simpler, legal
181 operations. The case for <tt>ISD::UREM</tt> for expanding a remainder into
182 a divide, multiply, and a subtract is a good example.</li>
183 <li><tt>lib/CodeGen/SelectionDAG/LegalizeDAG.cpp</tt>:
184 If targets may support the new node being added only at certain sizes, you
185 will also need to add code to your node's case statement in
186 <tt>LegalizeOp</tt> to Promote your node's operands to a larger size, and
187 perform the correct operation. You will also need to add code to
188 <tt>PromoteOp</tt> to do this as well. For a good example, see
190 which promotes its operand to a wider size, performs the byteswap, and then
191 shifts the correct bytes right to emulate the narrower byteswap in the
193 <li><tt>lib/CodeGen/SelectionDAG/LegalizeDAG.cpp</tt>:
194 Add a case for your node in <tt>ExpandOp</tt> to teach the legalizer how to
195 perform the action represented by the new node on a value that has been
196 split into high and low halves. This case will be used to support your
197 node with a 64 bit operand on a 32 bit target.</li>
198 <li><tt>lib/CodeGen/SelectionDAG/DAGCombiner.cpp</tt>:
199 If your node can be combined with itself, or other existing nodes in a
200 peephole-like fashion, add a visit function for it, and call that function
201 from <tt></tt>. There are several good examples for simple combines you
202 can do; <tt>visitFABS</tt> and <tt>visitSRL</tt> are good starting places.
204 <li><tt>lib/Target/PowerPC/PPCISelLowering.cpp</tt>:
205 Each target has an implementation of the <tt>TargetLowering</tt> class,
206 usually in its own file (although some targets include it in the same
207 file as the DAGToDAGISel). The default behavior for a target is to
208 assume that your new node is legal for all types that are legal for
209 that target. If this target does not natively support your node, then
210 tell the target to either Promote it (if it is supported at a larger
211 type) or Expand it. This will cause the code you wrote in
212 <tt>LegalizeOp</tt> above to decompose your new node into other legal
213 nodes for this target.</li>
214 <li><tt>lib/Target/TargetSelectionDAG.td</tt>:
215 Most current targets supported by LLVM generate code using the DAGToDAG
216 method, where SelectionDAG nodes are pattern matched to target-specific
217 nodes, which represent individual instructions. In order for the targets
218 to match an instruction to your new node, you must add a def for that node
219 to the list in this file, with the appropriate type constraints. Look at
220 <tt>add</tt>, <tt>bswap</tt>, and <tt>fadd</tt> for examples.</li>
221 <li><tt>lib/Target/PowerPC/PPCInstrInfo.td</tt>:
222 Each target has a tablegen file that describes the target's instruction
223 set. For targets that use the DAGToDAG instruction selection framework,
224 add a pattern for your new node that uses one or more target nodes.
225 Documentation for this is a bit sparse right now, but there are several
226 decent examples. See the patterns for <tt>rotl</tt> in
227 <tt>PPCInstrInfo.td</tt>.</li>
228 <li>TODO: document complex patterns.</li>
229 <li><tt>llvm/test/Regression/CodeGen/*</tt>: Add test cases for your new node
230 to the test suite. <tt>llvm/test/Regression/CodeGen/X86/bswap.ll</tt> is
236 <!-- *********************************************************************** -->
237 <div class="doc_section">
238 <a name="instruction">Adding a new instruction</a>
240 <!-- *********************************************************************** -->
242 <div class="doc_text">
244 <p><span class="doc_warning">WARNING: adding instructions changes the bytecode
245 format, and it will take some effort to maintain compatibility with
246 the previous version.</span> Only add an instruction if it is absolutely
251 <li><tt>llvm/include/llvm/Instruction.def</tt>:
252 add a number for your instruction and an enum name</li>
254 <li><tt>llvm/include/llvm/Instructions.h</tt>:
255 add a definition for the class that will represent your instruction</li>
257 <li><tt>llvm/include/llvm/Support/InstVisitor.h</tt>:
258 add a prototype for a visitor to your new instruction type</li>
260 <li><tt>llvm/lib/AsmParser/Lexer.l</tt>:
261 add a new token to parse your instruction from assembly text file</li>
263 <li><tt>llvm/lib/AsmParser/llvmAsmParser.y</tt>:
264 add the grammar on how your instruction can be read and what it will
265 construct as a result</li>
267 <li><tt>llvm/lib/Bytecode/Reader/Reader.cpp</tt>:
268 add a case for your instruction and how it will be parsed from bytecode</li>
270 <li><tt>llvm/lib/VMCore/Instruction.cpp</tt>:
271 add a case for how your instruction will be printed out to assembly</li>
273 <li><tt>llvm/lib/VMCore/Instructions.cpp</tt>:
274 implement the class you defined in
275 <tt>llvm/include/llvm/Instructions.h</tt></li>
277 <li>Test your instruction</li>
279 <li><tt>llvm/lib/Target/*</tt>:
280 Add support for your instruction to code generators, or add a lowering
283 <li><tt>llvm/test/Regression/*</tt>: add your test cases to the test suite.</li>
287 <p>Also, you need to implement (or modify) any analyses or passes that you want
288 to understand this new instruction.</p>
293 <!-- *********************************************************************** -->
294 <div class="doc_section">
295 <a name="type">Adding a new type</a>
297 <!-- *********************************************************************** -->
299 <div class="doc_text">
301 <p><span class="doc_warning">WARNING: adding new types changes the bytecode
302 format, and will break compatibility with currently-existing LLVM
303 installations.</span> Only add new types if it is absolutely necessary.</p>
307 <!-- ======================================================================= -->
308 <div class="doc_subsection">
309 <a name="fund_type">Adding a fundamental type</a>
312 <div class="doc_text">
316 <li><tt>llvm/include/llvm/Type.h</tt>:
317 add enum for the new type; add static <tt>Type*</tt> for this type</li>
319 <li><tt>llvm/lib/VMCore/Type.cpp</tt>:
320 add mapping from <tt>TypeID</tt> => <tt>Type*</tt>;
321 initialize the static <tt>Type*</tt></li>
323 <li><tt>llvm/lib/AsmReader/Lexer.l</tt>:
324 add ability to parse in the type from text assembly</li>
326 <li><tt>llvm/lib/AsmReader/llvmAsmParser.y</tt>:
327 add a token for that type</li>
333 <!-- ======================================================================= -->
334 <div class="doc_subsection">
335 <a name="derived_type">Adding a derived type</a>
338 <div class="doc_text">
341 <li><tt>llvm/include/llvm/Type.h</tt>:
342 add enum for the new type; add a forward declaration of the type
345 <li><tt>llvm/include/llvm/DerivedTypes.h</tt>:
346 add new class to represent new class in the hierarchy; add forward
347 declaration to the TypeMap value type</li>
349 <li><tt>llvm/lib/VMCore/Type.cpp</tt>:
350 add support for derived type to:
351 <div class="doc_code">
353 std::string getTypeDescription(const Type &Ty,
354 std::vector<const Type*> &TypeStack)
355 bool TypesEqual(const Type *Ty, const Type *Ty2,
356 std::map<const Type*, const Type*> & EqTypes)
359 add necessary member functions for type, and factory methods</li>
361 <li><tt>llvm/lib/AsmReader/Lexer.l</tt>:
362 add ability to parse in the type from text assembly</li>
364 <li><tt>llvm/lib/ByteCode/Writer/Writer.cpp</tt>:
365 modify <tt>void BytecodeWriter::outputType(const Type *T)</tt> to serialize
368 <li><tt>llvm/lib/ByteCode/Reader/Reader.cpp</tt>:
369 modify <tt>const Type *BytecodeReader::ParseType()</tt> to read your data
372 <li><tt>llvm/lib/VMCore/AsmWriter.cpp</tt>:
374 <div class="doc_code">
376 void calcTypeName(const Type *Ty,
377 std::vector<const Type*> &TypeStack,
378 std::map<const Type*,std::string> &TypeNames,
379 std::string & Result)
382 to output the new derived type
390 <!-- *********************************************************************** -->
394 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
395 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
396 <a href="http://validator.w3.org/check/referer"><img
397 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
399 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
401 Last modified: $Date$