blob: dcea204e47cd02d7f322004a1e4aa9c693f1a1d7 [file] [log] [blame]
Misha Brukman8eb67192004-09-06 22:58:13 +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>Writing an LLVM backend</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
7</head>
8
9<body>
10
11<div class="doc_title">
12 Writing an LLVM backend
13</div>
14
15<ol>
16 <li><a href="#intro">Introduction</a>
17 <li><a href="#backends">Writing a backend</a>
18 <ol>
Chris Lattner7a2fd892004-09-18 06:28:07 +000019 <li><a href="#machine">Machine backends</a>
Misha Brukman8eb67192004-09-06 22:58:13 +000020 <ol>
21 <li><a href="#machineTOC">Outline</a></li>
22 <li><a href="#machineDetails">Implementation details</a></li>
23 </ol></li>
Misha Brukman8eb67192004-09-06 22:58:13 +000024 <li><a href="#lang">Language backends</a></li>
25 </ol></li>
26 <li><a href="#related">Related reading material</a>
27</ol>
28
29<div class="doc_author">
30 <p>Written by <a href="http://misha.brukman.net">Misha Brukman</a></p>
31</div>
32
33<!-- *********************************************************************** -->
34<div class="doc_section">
35 <a name="intro">Introduction</a>
36</div>
37<!-- *********************************************************************** -->
38
39<div class="doc_text">
40
41<p>This document describes techniques for writing backends for LLVM which
42convert the LLVM representation to machine assembly code or other languages.</p>
43
44</div>
45
46<!-- *********************************************************************** -->
47<div class="doc_section">
48 <a name="backends">Writing a backend</a>
49</div>
50<!-- *********************************************************************** -->
51
52<!-- ======================================================================= -->
53<div class="doc_subsection">
54 <a name="machine">Machine backends</a>
55</div>
56
57<!-- _______________________________________________________________________ -->
58<div class="doc_subsubsection">
59 <a name="machineTOC">Outline</a>
60</div>
61
62<div class="doc_text">
63
64<p>In general, you want to follow the format of X86 or PowerPC (in
65<tt>lib/Target</tt>).</p>
66
67<p>To create a static compiler (one that emits text assembly), you need to
68implement the following:</p>
69
70<ul>
71<li>Describe the register set
72 <ul>
73 <li>Create a <a href="TableGenFundamentals.html">TableGen</a> description of
74 the register set and register classes</li>
75 <li>Implement a subclass of <tt><a
76 href="CodeGenerator.html#mregisterinfo">MRegisterInfo</a></tt></li>
77 </ul></li>
78<li>Describe the instruction set
79 <ul>
80 <li>Create a <a href="TableGenFundamentals.html">TableGen</a> description of
81 the instruction set</li>
82 <li>Implement a subclass of <tt><a
83 href="CodeGenerator.html#targetinstrinfo">TargetInstrInfo</a></tt></li>
84 </ul></li>
85<li>Describe the target machine
86 <ul>
87 <li>Create a <a href="TableGenFundamentals.html">TableGen</a> description of
88 the target that describes the pointer size and references the instruction
89 set</li>
90 <li>Implement a subclass of <tt><a
91 href="CodeGenerator.html#targetmachine">TargetMachine</a></tt>, which
92 configures <tt><a href="CodeGenerator.html#targetdata">TargetData</a></tt>
93 correctly</li>
Misha Brukman93d416f2004-12-27 19:05:16 +000094 <li>Register your new target using the <tt>RegisterTarget</tt>
95 template:<br><br>
96<div class="doc_code"><pre>
97RegisterTarget&lt;<em>MyTargetMachine</em>&gt; M("short_name", " Target name");
98</pre></div>
99 <br>Here, <em>MyTargetMachine</em> is the name of your implemented
100 subclass of <tt><a
101 href="CodeGenerator.html#targetmachine">TargetMachine</a></tt>,
102 <em>short_name</em> is the option that will be active following
103 <tt>-march=</tt> to select a target in llc and lli, and the last string
104 is the description of your target to appear in <tt>-help</tt>
105 listing.</li>
Misha Brukman8eb67192004-09-06 22:58:13 +0000106 </ul></li>
107<li>Implement the assembly printer for the architecture. Usually, if you have
108described the instruction set with the assembly printer generator in mind, that
109step can be almost automated.</li>
110</ul>
111
112<p>Now, for static code generation you also need to write an instruction
113selector for your platform: see <tt>lib/Target/*/*ISelSimple.cpp</tt> which
114is no longer "simple" but it gives you the idea: you have to be able to create
115MachineInstrs for any given LLVM instruction using the <tt>InstVisitor</tt>
116pattern, and produce a <tt>MachineFunction</tt> with
117<tt>MachineBasicBlock</tt>s full of <tt><a
118href="CodeGenerator.html#machineinstr">MachineInstr</a></tt>s for a
119corresponding LLVM Function. Creating an instruction selector is perhaps the
120most time-consuming part of creating a back-end.</p>
121
122<p>To create a JIT for your platform:</p>
123
124<ul>
125<li>Create a subclass of <tt><a
126 href="CodeGenerator.html#targetjitinfo">TargetJITInfo</a></tt></li>
127<li>Create a machine code emitter that will be used to emit binary code
128 directly into memory, given <tt>MachineInstr</tt>s</li>
129</ul>
130
131<p>Note that <tt>lib/target/Skeleton</tt> is a clean skeleton for a new target,
132so you might want to start with that and adapt it for your target, and if you
133are wondering how things are done, peek in the X86 or PowerPC target.</p>
134
135<p>The Skeleton target is non-functional but provides the basic building blocks
136you will need for your endeavor.</p>
137
138</div>
139
140<!-- _______________________________________________________________________ -->
141<div class="doc_subsubsection">
142 <a name="machineDetails">Implementation details</a>
143</div>
144
145<div class="doc_text">
146
147<ul>
148
149<li><p><b>TableGen register info description</b> - describe a class which
150will store the register's number in the binary encoding of the instruction
151(e.g., for JIT purposes).</p>
152
153<p>You also need to define register classes to contain these registers, such as
154the integer register class and floating-point register class, so that you can
155allocate virtual registers to instructions from these sets, and let the
156target-independent register allocator automatically choose the actual
157architected registers.</p>
158
159<div class="doc_code">
160<pre>
161// class Register is defined in Target.td
Chris Lattner7a2fd892004-09-18 06:28:07 +0000162<b>class</b> <em>Target</em>Reg&lt;string name&gt; : Register&lt;name&gt; {
Misha Brukman8eb67192004-09-06 22:58:13 +0000163 <b>let</b> Namespace = "<em>Target</em>";
164}
165
Chris Lattner7a2fd892004-09-18 06:28:07 +0000166<b>class</b> IntReg&lt;<b>bits</b>&lt;5&gt; num, string name&gt; : <em>Target</em>Reg&lt;name&gt; {
Misha Brukman8eb67192004-09-06 22:58:13 +0000167 <b>field</b> <b>bits</b>&lt;5&gt; Num = num;
168}
169
Chris Lattner7a2fd892004-09-18 06:28:07 +0000170<b>def</b> R0 : IntReg&lt;0, "%R0"&gt;;
Misha Brukman8eb67192004-09-06 22:58:13 +0000171...
172
173// class RegisterClass is defined in Target.td
174<b>def</b> IReg : RegisterClass&lt;i64, 64, [R0, ... ]&gt;;
175</pre>
176</div>
177</li>
178
179<li><p><b>TableGen instruction info description</b> - break up instructions into
180classes, usually that's already done by the manufacturer (see instruction
181manual). Define a class for each instruction category. Define each opcode as a
182subclass of the category, with appropriate parameters such as the fixed binary
183encoding of opcodes and extended opcodes, and map the register bits to the bits
184of the instruction which they are encoded in (for the JIT). Also specify how
185the instruction should be printed so it can use the automatic assembly printer,
186e.g.:</p>
187
188<div class="doc_code">
189<pre>
190// class Instruction is defined in Target.td
191<b>class</b> Form&lt;<b>bits</b>&lt;6&gt; opcode, <b>dag</b> OL, <b>string</b> asmstr&gt; : Instruction {
192 <b>field</b> <b>bits</b>&lt;42&gt; Inst;
193
194 <b>let</b> Namespace = "<em>Target</em>";
195 <b>let</b> Inst{0-6} = opcode;
196 <b>let</b> OperandList = OL;
197 <b>let</b> AsmString = asmstr;
198}
199
200<b>def</b> ADD : Form&lt;42, (ops IReg:$rD, IReg:$rA, IReg:$rB), "add $rD, $rA, $rB"&gt;;
201</pre>
202</div>
203</li>
204
205</ul>
206
207</div>
208
209<!-- ======================================================================= -->
210<div class="doc_subsection">
211 <a name="lang">Language backends</a>
212</div>
213
214<div class="doc_text">
215
216<p>For now, just take a look at <tt>lib/Target/CBackend</tt> for an example of
217how the C backend is written.</p>
218
219</div>
220
221<!-- *********************************************************************** -->
222<div class="doc_section">
223 <a name="related">Related reading material</a>
224</div>
225<!-- *********************************************************************** -->
226
227<div class="doc_text">
228
229<ul>
230<li><a href="CodeGenerator.html">Code generator</a> -
231 describes some of the classes in code generation at a high level, but
232 it is not (yet) complete.</li>
233<li><a href="TableGenFundamentals.html">TableGen fundamentals</a> -
234 describes how to use TableGen to describe your target information succinctly
235</li>
236</ul>
237
238</div>
239
240<!-- *********************************************************************** -->
241
242<hr>
243<address>
244 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
245 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
246 <a href="http://validator.w3.org/check/referer"><img
247 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
248
249 <a href="http://misha.brukman.net">Misha Brukman</a><br>
250 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a>
251 <br>
252 Last modified: $Date$
253</address>
254
255</body>
256</html>