1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2
"http://www.w3.org/TR/html4/strict.dtd">
5
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6
<title>LLVM test-suite Makefile Guide</title>
7
<link rel="stylesheet" href="llvm.css" type="text/css">
12
LLVM test-suite Makefile Guide
16
<li><a href="#overview">Overview</a></li>
17
<li><a href="#testsuitestructure">Test suite structure</a></li>
18
<li><a href="#testsuiterun">Running the test suite</a>
20
<li><a href="#testsuiteexternal">Configuring External Tests</a></li>
21
<li><a href="#testsuitetests">Running different tests</a></li>
22
<li><a href="#testsuiteoutput">Generating test output</a></li>
23
<li><a href="#testsuitecustom">Writing custom tests for test-suite</a></li>
28
<div class="doc_author">
29
<p>Written by John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner</p>
32
<!--=========================================================================-->
33
<h2><a name="overview">Overview</a></h2>
34
<!--=========================================================================-->
38
<p>This document describes the features of the Makefile-based LLVM
39
test-suite. This way of interacting with the test-suite is deprecated in favor
40
of running the test-suite using LNT, but may continue to prove useful for some
41
users. See the Testing
42
Guide's <a href="TestingGuide.html#testsuitequickstart">test-suite
43
Quickstart</a> section for more information.</p>
47
<!--=========================================================================-->
48
<h2><a name="testsuitestructure">Test suite Structure</a></h2>
49
<!--=========================================================================-->
53
<p>The <tt>test-suite</tt> module contains a number of programs that can be compiled
54
with LLVM and executed. These programs are compiled using the native compiler
55
and various LLVM backends. The output from the program compiled with the
56
native compiler is assumed correct; the results from the other programs are
57
compared to the native program output and pass if they match.</p>
59
<p>When executing tests, it is usually a good idea to start out with a subset of
60
the available tests or programs. This makes test run times smaller at first and
61
later on this is useful to investigate individual test failures. To run some
62
test only on a subset of programs, simply change directory to the programs you
63
want tested and run <tt>gmake</tt> there. Alternatively, you can run a different
64
test using the <tt>TEST</tt> variable to change what tests or run on the
65
selected programs (see below for more info).</p>
67
<p>In addition for testing correctness, the <tt>test-suite</tt> directory also
68
performs timing tests of various LLVM optimizations. It also records
69
compilation times for the compilers and the JIT. This information can be
70
used to compare the effectiveness of LLVM's optimizations and code
73
<p><tt>test-suite</tt> tests are divided into three types of tests: MultiSource,
74
SingleSource, and External.</p>
77
<li><tt>test-suite/SingleSource</tt>
78
<p>The SingleSource directory contains test programs that are only a single
79
source file in size. These are usually small benchmark programs or small
80
programs that calculate a particular value. Several such programs are grouped
81
together in each directory.</p></li>
83
<li><tt>test-suite/MultiSource</tt>
84
<p>The MultiSource directory contains subdirectories which contain entire
85
programs with multiple source files. Large benchmarks and whole applications
88
<li><tt>test-suite/External</tt>
89
<p>The External directory contains Makefiles for building code that is external
90
to (i.e., not distributed with) LLVM. The most prominent members of this
91
directory are the SPEC 95 and SPEC 2000 benchmark suites. The <tt>External</tt>
92
directory does not contain these actual tests, but only the Makefiles that know
93
how to properly compile these programs from somewhere else. The presence and
94
location of these external programs is configured by the test-suite
95
<tt>configure</tt> script.</p></li>
98
<p>Each tree is then subdivided into several categories, including applications,
99
benchmarks, regression tests, code that is strange grammatically, etc. These
100
organizations should be relatively self explanatory.</p>
102
<p>Some tests are known to fail. Some are bugs that we have not fixed yet;
103
others are features that we haven't added yet (or may never add). In the
104
regression tests, the result for such tests will be XFAIL (eXpected FAILure).
105
In this way, you can tell the difference between an expected and unexpected
108
<p>The tests in the test suite have no such feature at this time. If the
109
test passes, only warnings and other miscellaneous output will be generated. If
110
a test fails, a large <program> FAILED message will be displayed. This
111
will help you separate benign warnings from actual test failures.</p>
115
<!--=========================================================================-->
116
<h2><a name="testsuiterun">Running the test suite</a></h2>
117
<!--=========================================================================-->
121
<p>First, all tests are executed within the LLVM object directory tree. They
122
<i>are not</i> executed inside of the LLVM source tree. This is because the
123
test suite creates temporary files during execution.</p>
125
<p>To run the test suite, you need to use the following steps:</p>
128
<li><tt>cd</tt> into the <tt>llvm/projects</tt> directory in your source tree.
131
<li><p>Check out the <tt>test-suite</tt> module with:</p>
133
<div class="doc_code">
135
% svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
138
<p>This will get the test suite into <tt>llvm/projects/test-suite</tt>.</p>
140
<li><p>Configure and build <tt>llvm</tt>.</p></li>
141
<li><p>Configure and build <tt>llvm-gcc</tt>.</p></li>
142
<li><p>Install <tt>llvm-gcc</tt> somewhere.</p></li>
143
<li><p><em>Re-configure</em> <tt>llvm</tt> from the top level of
144
each build tree (LLVM object directory tree) in which you want
145
to run the test suite, just as you do before building LLVM.</p>
146
<p>During the <em>re-configuration</em>, you must either: (1)
147
have <tt>llvm-gcc</tt> you just built in your path, or (2)
148
specify the directory where your just-built <tt>llvm-gcc</tt> is
149
installed using <tt>--with-llvmgccdir=$LLVM_GCC_DIR</tt>.</p>
150
<p>You must also tell the configure machinery that the test suite
151
is available so it can be configured for your build tree:</p>
152
<div class="doc_code">
154
% cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure [--with-llvmgccdir=$LLVM_GCC_DIR]
157
<p>[Remember that <tt>$LLVM_GCC_DIR</tt> is the directory where you
158
<em>installed</em> llvm-gcc, not its src or obj directory.]</p>
161
<li><p>You can now run the test suite from your build tree as follows:</p>
162
<div class="doc_code">
164
% cd $LLVM_OBJ_ROOT/projects/test-suite
170
<p>Note that the second and third steps only need to be done once. After you
171
have the suite checked out and configured, you don't need to do it again (unless
172
the test code or configure script changes).</p>
174
<!-- _______________________________________________________________________ -->
176
<a name="testsuiteexternal">Configuring External Tests</a>
178
<!-- _______________________________________________________________________ -->
181
<p>In order to run the External tests in the <tt>test-suite</tt>
182
module, you must specify <i>--with-externals</i>. This
183
must be done during the <em>re-configuration</em> step (see above),
184
and the <tt>llvm</tt> re-configuration must recognize the
185
previously-built <tt>llvm-gcc</tt>. If any of these is missing or
186
neglected, the External tests won't work.</p>
188
<dt><i>--with-externals</i></dt>
189
<dt><i>--with-externals=<<tt>directory</tt>></i></dt>
191
This tells LLVM where to find any external tests. They are expected to be
192
in specifically named subdirectories of <<tt>directory</tt>>.
193
If <tt>directory</tt> is left unspecified,
194
<tt>configure</tt> uses the default value
195
<tt>/home/vadve/shared/benchmarks/speccpu2000/benchspec</tt>.
196
Subdirectory names known to LLVM include:
203
Others are added from time to time, and can be determined from
207
<!-- _______________________________________________________________________ -->
209
<a name="testsuitetests">Running different tests</a>
211
<!-- _______________________________________________________________________ -->
213
<p>In addition to the regular "whole program" tests, the <tt>test-suite</tt>
214
module also provides a mechanism for compiling the programs in different ways.
215
If the variable TEST is defined on the <tt>gmake</tt> command line, the test system will
216
include a Makefile named <tt>TEST.<value of TEST variable>.Makefile</tt>.
217
This Makefile can modify build rules to yield different results.</p>
219
<p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to
220
create the nightly test reports. To run the nightly tests, run <tt>gmake
221
TEST=nightly</tt>.</p>
223
<p>There are several TEST Makefiles available in the tree. Some of them are
224
designed for internal LLVM research and will not work outside of the LLVM
225
research group. They may still be valuable, however, as a guide to writing your
226
own TEST Makefile for any optimization or analysis passes that you develop with
231
<!-- _______________________________________________________________________ -->
233
<a name="testsuiteoutput">Generating test output</a>
235
<!-- _______________________________________________________________________ -->
237
<p>There are a number of ways to run the tests and generate output. The most
238
simple one is simply running <tt>gmake</tt> with no arguments. This will
239
compile and run all programs in the tree using a number of different methods
240
and compare results. Any failures are reported in the output, but are likely
241
drowned in the other output. Passes are not reported explicitely.</p>
243
<p>Somewhat better is running <tt>gmake TEST=sometest test</tt>, which runs
244
the specified test and usually adds per-program summaries to the output
245
(depending on which sometest you use). For example, the <tt>nightly</tt> test
246
explicitely outputs TEST-PASS or TEST-FAIL for every test after each program.
247
Though these lines are still drowned in the output, it's easy to grep the
248
output logs in the Output directories.</p>
250
<p>Even better are the <tt>report</tt> and <tt>report.format</tt> targets
251
(where <tt>format</tt> is one of <tt>html</tt>, <tt>csv</tt>, <tt>text</tt> or
252
<tt>graphs</tt>). The exact contents of the report are dependent on which
253
<tt>TEST</tt> you are running, but the text results are always shown at the
254
end of the run and the results are always stored in the
255
<tt>report.<type>.format</tt> file (when running with
256
<tt>TEST=<type></tt>).
258
The <tt>report</tt> also generate a file called
259
<tt>report.<type>.raw.out</tt> containing the output of the entire test
263
<!-- _______________________________________________________________________ -->
265
<a name="testsuitecustom">Writing custom tests for the test suite</a>
267
<!-- _______________________________________________________________________ -->
271
<p>Assuming you can run the test suite, (e.g. "<tt>gmake TEST=nightly report</tt>"
272
should work), it is really easy to run optimizations or code generator
273
components against every program in the tree, collecting statistics or running
274
custom checks for correctness. At base, this is how the nightly tester works,
275
it's just one example of a general framework.</p>
277
<p>Lets say that you have an LLVM optimization pass, and you want to see how
278
many times it triggers. First thing you should do is add an LLVM
279
<a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which
280
will tally counts of things you care about.</p>
282
<p>Following this, you can set up a test and a report that collects these and
283
formats them for easy viewing. This consists of two files, a
284
"<tt>test-suite/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
285
test) and a "<tt>test-suite/TEST.XXX.report</tt>" file that indicates how to
286
format the output into a table. There are many example reports of various
287
levels of sophistication included with the test suite, and the framework is very
290
<p>If you are interested in testing an optimization pass, check out the
291
"libcalls" test as an example. It can be run like this:<p>
293
<div class="doc_code">
295
% cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
296
% make TEST=libcalls report
300
<p>This will do a bunch of stuff, then eventually print a table like this:</p>
302
<div class="doc_code">
304
Name | total | #exit |
306
FreeBench/analyzer/analyzer | 51 | 6 |
307
FreeBench/fourinarow/fourinarow | 1 | 1 |
308
FreeBench/neural/neural | 19 | 9 |
309
FreeBench/pifft/pifft | 5 | 3 |
310
MallocBench/cfrac/cfrac | 1 | * |
311
MallocBench/espresso/espresso | 52 | 12 |
312
MallocBench/gs/gs | 4 | * |
313
Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
314
Prolangs-C/agrep/agrep | 33 | 12 |
315
Prolangs-C/allroots/allroots | * | * |
316
Prolangs-C/assembler/assembler | 47 | * |
317
Prolangs-C/bison/mybison | 74 | * |
322
<p>This basically is grepping the -stats output and displaying it in a table.
323
You can also use the "TEST=libcalls report.html" target to get the table in HTML
324
form, similarly for report.csv and report.tex.</p>
326
<p>The source for this is in test-suite/TEST.libcalls.*. The format is pretty
327
simple: the Makefile indicates how to run the test (in this case,
328
"<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
329
each column of the output. The first value is the header for the column and the
330
second is the regex to grep the output of the command for. There are lots of
331
example reports that can do fancy stuff.</p>
337
<!-- *********************************************************************** -->
341
<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
342
src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
343
<a href="http://validator.w3.org/check/referer"><img
344
src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
346
John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner<br>
347
<a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
348
Last modified: $Date$