1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Test Suite Guide</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
10 <div class="doc_title">
15 <li><a href="#overview">Overview</a></li>
16 <li><a href="#Requirements">Requirements</a></li>
17 <li><a href="#org">LLVM Test Suite Organization</a>
19 <li><a href="#codefragments">Code Fragments</a></li>
20 <li><a href="#wholeprograms">Whole Programs</a></li>
23 <li><a href="#quick">Quick Start</a>
25 <li><a href="#quickdejagnu">DejaGNU tests</a></li>
26 <li><a href="#quickllvmtest">llvm-test tests</a></li>
29 <li><a href="#dgstructure">DejaGNU Structure</a>
31 <li><a href="#customdg">Writing new DejaGNU tests</a></li>
32 <li><a href="#dgvars">Variables and substitutions</a></li>
33 <li><a href="#dgfeatures">Other features</a></li>
36 <li><a href="#llvmteststructure"><tt>llvm-test</tt> Structure</a></li>
37 <li><a href="#runllvmtest">Running the LLVM Tests</a>
39 <li><a href="#externaltests">Configuring external tests</a></li>
40 <li><a href="#llvmtesttests">Running different tests</a></li>
41 <li><a href="#llvmtestoutput">Generating test output</a></li>
42 <li><a href="#customtest">Writing custom tests for llvm-test</a></li>
45 <li><a href="#nightly">Running the nightly tester</a></li>
48 <div class="doc_author">
49 <p>Written by John T. Criswell, <a
50 href="http://llvm.x10sys.com/rspencer">Reid Spencer</a>, and Tanya Lattner</p>
53 <!--=========================================================================-->
54 <div class="doc_section"><a name="overview">Overview</a></div>
55 <!--=========================================================================-->
57 <div class="doc_text">
59 <p>This document is the reference manual for the LLVM test suite. It documents
60 the structure of the LLVM test suite, the tools needed to use it, and how to add
65 <!--=========================================================================-->
66 <div class="doc_section"><a name="Requirements">Requirements</a></div>
67 <!--=========================================================================-->
69 <div class="doc_text">
71 <p>In order to use the LLVM test suite, you will need all of the software
72 required to build LLVM, plus the following:</p>
75 <dt><a href="http://www.gnu.org/software/dejagnu/">DejaGNU</a></dt>
76 <dd>The Feature and Regressions tests are organized and run by DejaGNU.</dd>
77 <dt><a href="http://expect.nist.gov/">Expect</a></dt>
78 <dd>Expect is required by DejaGNU.</dd>
79 <dt><a href="http://www.tcl.tk/software/tcltk/">tcl</a></dt>
80 <dd>Tcl is required by DejaGNU. </dd>
82 <dt><a href="http://www.netlib.org/f2c">F2C</a></dt>
83 <dd>For now, LLVM does not have a Fortran front-end, but using F2C, we can run
84 Fortran benchmarks. F2C support must be enabled via <tt>configure</tt> if not
85 installed in a standard place. F2C requires three items: the <tt>f2c</tt>
86 executable, <tt>f2c.h</tt> to compile the generated code, and <tt>libf2c.a</tt>
87 to link generated code. By default, given an F2C directory <tt>$DIR</tt>, the
88 configure script will search <tt>$DIR/bin</tt> for <tt>f2c</tt>,
89 <tt>$DIR/include</tt> for <tt>f2c.h</tt>, and <tt>$DIR/lib</tt> for
90 <tt>libf2c.a</tt>. The default <tt>$DIR</tt> values are: <tt>/usr</tt>,
91 <tt>/usr/local</tt>, <tt>/sw</tt>, and <tt>/opt</tt>. If you installed F2C in a
92 different location, you must tell <tt>configure</tt>:
95 <li><tt>./configure --with-f2c=$DIR</tt><br>
96 This will specify a new <tt>$DIR</tt> for the above-described search
97 process. This will only work if the binary, header, and library are in their
98 respective subdirectories of <tt>$DIR</tt>.</li>
100 <li><tt>./configure --with-f2c-bin=/binary/path --with-f2c-inc=/include/path
101 --with-f2c-lib=/lib/path</tt><br>
102 This allows you to specify the F2C components separately. Note: if you choose
103 this route, you MUST specify all three components, and you need to only specify
104 <em>directories</em> where the files are located; do NOT include the
105 filenames themselves on the <tt>configure</tt> line.</li>
109 <p>Darwin (Mac OS X) developers can simplify the installation of Expect and tcl
110 by using fink. <tt>fink install expect</tt> will install both. Alternatively,
111 Darwinports users can use <tt>sudo port install expect</tt> to install Expect
116 <!--=========================================================================-->
117 <div class="doc_section"><a name="org">LLVM Test Suite Organization</a></div>
118 <!--=========================================================================-->
120 <div class="doc_text">
122 <p>The LLVM test suite contains two major categories of tests: code
123 fragments and whole programs. Code fragments are in the <tt>llvm</tt> module
124 under the <tt>llvm/test</tt> directory. The whole programs
125 test suite is in the <tt>llvm-test</tt> module under the main directory.</p>
129 <!-- _______________________________________________________________________ -->
130 <div class="doc_subsection"><a name="codefragments">Code Fragments (a.k.a.
131 DejaGNU tests)</a></div>
132 <!-- _______________________________________________________________________ -->
134 <div class="doc_text">
136 <p>Code fragments are small pieces of code that test a specific feature of LLVM
137 or trigger a specific bug in LLVM. They are usually written in LLVM assembly
138 language, but can be written in other languages if the test targets a particular
139 language front end. These tests are driven by the DejaGNU testing framework,
140 which is hidden behind a few simple makefiles.</p>
142 <p>These code fragments are not complete programs. The code generated from them is
143 never executed to determine correct behavior.</p>
145 <p>These code fragment tests are located in the <tt>llvm/test</tt>
148 <p>Typically when a bug is found in LLVM, a regression test containing
149 just enough code to reproduce the problem should be written and placed
150 somewhere underneath this directory. In most cases, this will be a small
151 piece of LLVM assembly language code, often distilled from an actual
152 application or benchmark.</p>
156 <!-- _______________________________________________________________________ -->
157 <div class="doc_subsection"><a name="wholeprograms">Whole Programs (a.k.a.
158 <tt>llvm-test</tt> tests)</a></div>
159 <!-- _______________________________________________________________________ -->
161 <div class="doc_text">
163 <p>The <tt>llvm-test</tt> suite contains whole programs, which are pieces of
164 code which can be compiled and linked into a stand-alone program that can be
165 executed. These programs are generally written in high level languages such as
166 C or C++, but sometimes they are written straight in LLVM assembly.</p>
168 <p>These programs are compiled and then executed using several different
169 methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation,
170 etc). The output of these programs is compared to ensure that LLVM is compiling
171 the program correctly.</p>
173 <p>In addition to compiling and executing programs, whole program tests serve as
174 a way of benchmarking LLVM performance, both in terms of the efficiency of the
175 programs generated as well as the speed with which LLVM compiles, optimizes, and
178 <p>All "whole program" tests are located in the <tt>test-suite</tt> Subversion
183 <!--=========================================================================-->
184 <div class="doc_section"><a name="quick">Quick Start</a></div>
185 <!--=========================================================================-->
187 <div class="doc_text">
189 <p>The tests are located in two separate Subversion modules. The code fragment
190 DejaGNU tests are in the main "llvm" module under the directory
191 <tt>llvm/test</tt> (so you get these tests for free with the main llvm tree).
192 The more comprehensive <tt>llvm-test</tt> suite that includes whole
193 programs in C and C++ is in the <tt>test-suite</tt> module. This module should
194 be checked out to the <tt>llvm/projects</tt> directory as llvm-test (don't use
195 another name, for then the test suite will be run every time you run
196 <tt>make</tt> in the main <tt>llvm</tt> directory).
197 When you <tt>configure</tt> the <tt>llvm</tt> module,
198 the <tt>llvm-test</tt> directory will be automatically configured.
199 Alternatively, you can configure the <tt>test-suite</tt> module manually.</p>
201 <!-- _______________________________________________________________________ -->
202 <div class="doc_subsection"><a name="quickdejagnu">DejaGNU tests</a></div>
203 <!-- _______________________________________________________________________ -->
204 <p>To run all of the simple tests in LLVM using DejaGNU, use the master Makefile
205 in the <tt>llvm/test</tt> directory:</p>
207 <div class="doc_code">
215 <div class="doc_code">
221 <p>To run only a subdirectory of tests in llvm/test using DejaGNU (ie.
222 Regression/Transforms), just set the TESTSUITE variable to the path of the
223 subdirectory (relative to <tt>llvm/test</tt>):</p>
225 <div class="doc_code">
227 % gmake TESTSUITE=Transforms check
231 <p><b>Note: If you are running the tests with <tt>objdir != subdir</tt>, you
232 must have run the complete testsuite before you can specify a
233 subdirectory.</b></p>
235 <p>To run only a single test, set TESTONE to its path (relative to
236 <tt>llvm/test</tt>) and make the check-one target:</p>
238 <div class="doc_code">
240 % gmake TESTONE=Feature/basictest.ll check-one
244 <!-- _______________________________________________________________________ -->
245 <div class="doc_subsection"><a name="quickllvmtest"><tt>llvm-test</tt> tests</a></div>
246 <!-- _______________________________________________________________________ -->
248 <p>To run the comprehensive test suite (tests that compile and execute whole
249 programs), first checkout and setup the <tt>test-suite</tt> module:</p>
251 <div class="doc_code">
254 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk llvm-test
256 % ./configure --with-llvmgccdir=$LLVM_GCC_DIR
258 <p>where <tt>$LLVM_GCC_DIR</tt> is the directory where you <em>installed</em>
259 llvm-gcc, not it's src or obj dir.</p>
262 <p>Then, run the entire test suite by running make in the <tt>llvm-test</tt>
265 <div class="doc_code">
267 % cd projects/llvm-test
272 <p>Usually, running the "nightly" set of tests is a good idea, and you can also
273 let it generate a report by running:</p>
275 <div class="doc_code">
277 % cd projects/llvm-test
278 % gmake TEST=nightly report report.html
282 <p>Any of the above commands can also be run in a subdirectory of
283 <tt>projects/llvm-test</tt> to run the specified test only on the programs in
284 that subdirectory.</p>
288 <!--=========================================================================-->
289 <div class="doc_section"><a name="dgstructure">DejaGNU Structure</a></div>
290 <!--=========================================================================-->
291 <div class="doc_text">
292 <p>The LLVM DejaGNU tests are driven by DejaGNU together with GNU Make and are
293 located in the <tt>llvm/test</tt> directory.
295 <p>This directory contains a large array of small tests
296 that exercise various features of LLVM and to ensure that regressions do not
297 occur. The directory is broken into several sub-directories, each focused on
298 a particular area of LLVM. A few of the important ones are:</p>
301 <li><tt>Analysis</tt>: checks Analysis passes.</li>
302 <li><tt>Archive</tt>: checks the Archive library.</li>
303 <li><tt>Assembler</tt>: checks Assembly reader/writer functionality.</li>
304 <li><tt>Bitcode</tt>: checks Bitcode reader/writer functionality.</li>
305 <li><tt>CodeGen</tt>: checks code generation and each target.</li>
306 <li><tt>Features</tt>: checks various features of the LLVM language.</li>
307 <li><tt>Linker</tt>: tests bitcode linking.</li>
308 <li><tt>Transforms</tt>: tests each of the scalar, IPO, and utility
309 transforms to ensure they make the right transformations.</li>
310 <li><tt>Verifier</tt>: tests the IR verifier.</li>
315 <!-- _______________________________________________________________________ -->
316 <div class="doc_subsection"><a name="customdg">Writing new DejaGNU tests</a></div>
317 <!-- _______________________________________________________________________ -->
318 <div class="doc_text">
319 <p>The DejaGNU structure is very simple, but does require some information to
320 be set. This information is gathered via <tt>configure</tt> and is written
321 to a file, <tt>site.exp</tt> in <tt>llvm/test</tt>. The <tt>llvm/test</tt>
322 Makefile does this work for you.</p>
324 <p>In order for DejaGNU to work, each directory of tests must have a
325 <tt>dg.exp</tt> file. DejaGNU looks for this file to determine how to run the
326 tests. This file is just a Tcl script and it can do anything you want, but
327 we've standardized it for the LLVM regression tests. If you're adding a
328 directory of tests, just copy <tt>dg.exp</tt> from another directory to get
329 running. The standard <tt>dg.exp</tt> simply loads a Tcl
330 library (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt>
331 function defined in that library with a list of file names to run. The names
332 are obtained by using Tcl's glob command. Any directory that contains only
333 directories does not need the <tt>dg.exp</tt> file.</p>
335 <p>The <tt>llvm-runtests</tt> function lookas at each file that is passed to
336 it and gathers any lines together that match "RUN:". This are the "RUN" lines
337 that specify how the test is to be run. So, each test script must contain
338 RUN lines if it is to do anything. If there are no RUN lines, the
339 <tt>llvm-runtests</tt> function will issue an error and the test will
342 <p>RUN lines are specified in the comments of the test program using the
343 keyword <tt>RUN</tt> followed by a colon, and lastly the command (pipeline)
344 to execute. Together, these lines form the "script" that
345 <tt>llvm-runtests</tt> executes to run the test case. The syntax of the
346 RUN lines is similar to a shell's syntax for pipelines including I/O
347 redirection and variable substitution. However, even though these lines
348 may <i>look</i> like a shell script, they are not. RUN lines are interpreted
349 directly by the Tcl <tt>exec</tt> command. They are never executed by a
350 shell. Consequently the syntax differs from normal shell script syntax in a
351 few ways. You can specify as many RUN lines as needed.</p>
353 <p>Each RUN line is executed on its own, distinct from other lines unless
354 its last character is <tt>\</tt>. This continuation character causes the RUN
355 line to be concatenated with the next one. In this way you can build up long
356 pipelines of commands without making huge line lengths. The lines ending in
357 <tt>\</tt> are concatenated until a RUN line that doesn't end in <tt>\</tt> is
358 found. This concatenated set of RUN lines then constitutes one execution.
359 Tcl will substitute variables and arrange for the pipeline to be executed. If
360 any process in the pipeline fails, the entire line (and test case) fails too.
363 <p> Below is an example of legal RUN lines in a <tt>.ll</tt> file:</p>
365 <div class="doc_code">
367 ; RUN: llvm-as < %s | llvm-dis > %t1
368 ; RUN: llvm-dis < %s.bc-13 > %t2
373 <p>As with a Unix shell, the RUN: lines permit pipelines and I/O redirection
374 to be used. However, the usage is slightly different than for Bash. To check
375 what's legal, see the documentation for the
376 <a href="http://www.tcl.tk/man/tcl8.5/TclCmd/exec.htm#M2">Tcl exec</a>
378 <a href="http://www.tcl.tk/man/tcl8.5/tutorial/Tcl26.html">tutorial</a>.
379 The major differences are:</p>
381 <li>You can't do <tt>2>&1</tt>. That will cause Tcl to write to a
382 file named <tt>&1</tt>. Usually this is done to get stderr to go through
383 a pipe. You can do that in tcl with <tt>|&</tt> so replace this idiom:
384 <tt>... 2>&1 | grep</tt> with <tt>... |& grep</tt></li>
385 <li>You can only redirect to a file, not to another descriptor and not from
386 a here document.</li>
387 <li>tcl supports redirecting to open files with the @ syntax but you
388 shouldn't use that here.</li>
391 <p>There are some quoting rules that you must pay attention to when writing
392 your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any
393 ' or " so they will get passed to the invoked program. For example:</p>
395 <div class="doc_code">
397 ... | grep 'find this string'
401 <p>This will fail because the ' characters are passed to grep. This would
402 instruction grep to look for <tt>'find</tt> in the files <tt>this</tt> and
403 <tt>string'</tt>. To avoid this use curly braces to tell Tcl that it should
404 treat everything enclosed as one value. So our example would become:</p>
406 <div class="doc_code">
408 ... | grep {find this string}
412 <p>Additionally, the characters <tt>[</tt> and <tt>]</tt> are treated
413 specially by Tcl. They tell Tcl to interpret the content as a command to
414 execute. Since these characters are often used in regular expressions this can
415 have disastrous results and cause the entire test run in a directory to fail.
416 For example, a common idiom is to look for some basicblock number:</p>
418 <div class="doc_code">
424 <p>This, however, will cause Tcl to fail because its going to try to execute
425 a program named "2-8". Instead, what you want is this:</p>
427 <div class="doc_code">
429 ... | grep {bb\[2-8\]}
433 <p>Finally, if you need to pass the <tt>\</tt> character down to a program,
434 then it must be doubled. This is another Tcl special character. So, suppose
437 <div class="doc_code">
443 <p>This will fail to match what you want (a pointer to i32). First, the
444 <tt>'</tt> do not get stripped off. Second, the <tt>\</tt> gets stripped off
445 by Tcl so what grep sees is: <tt>'i32*'</tt>. That's not likely to match
446 anything. To resolve this you must use <tt>\\</tt> and the <tt>{}</tt>, like
449 <div class="doc_code">
457 <!-- _______________________________________________________________________ -->
458 <div class="doc_subsection"><a name="dgvars">Variables and substitutions</a></div>
459 <!-- _______________________________________________________________________ -->
460 <div class="doc_text">
461 <p>With a RUN line there are a number of substitutions that are permitted. In
462 general, any Tcl variable that is available in the <tt>substitute</tt>
463 function (in <tt>test/lib/llvm.exp</tt>) can be substituted into a RUN line.
464 To make a substitution just write the variable's name preceded by a $.
465 Additionally, for compatibility reasons with previous versions of the test
466 library, certain names can be accessed with an alternate syntax: a % prefix.
467 These alternates are deprecated and may go away in a future version.
469 <p>Here are the available variable names. The alternate syntax is listed in
472 <dl style="margin-left: 25px">
473 <dt><b>$test</b> (%s)</dt>
474 <dd>The full path to the test case's source. This is suitable for passing
475 on the command line as the input to an llvm tool.</dd>
477 <dt><b>$srcdir</b></dt>
478 <dd>The source directory from where the "<tt>make check</tt>" was run.</dd>
480 <dt><b>objdir</b></dt>
481 <dd>The object directory that corresponds to the <tt>$srcdir</tt>.</dd>
483 <dt><b>subdir</b></dt>
484 <dd>A partial path from the <tt>test</tt> directory that contains the
485 sub-directory that contains the test source being executed.</dd>
487 <dt><b>srcroot</b></dt>
488 <dd>The root directory of the LLVM src tree.</dd>
490 <dt><b>objroot</b></dt>
491 <dd>The root directory of the LLVM object tree. This could be the same
495 <dd>The path to the directory that contains the test case source. This is
496 for locating any supporting files that are not generated by the test, but
497 used by the test.</dd>
500 <dd>The path to a temporary file name that could be used for this test case.
501 The file name won't conflict with other test cases. You can append to it if
502 you need multiple temporaries. This is useful as the destination of some
503 redirected output.</dd>
505 <dt><b>llvmlibsdir</b> (%llvmlibsdir)</dt>
506 <dd>The directory where the LLVM libraries are located.</dd>
508 <dt><b>target_triplet</b> (%target_triplet)</dt>
509 <dd>The target triplet that corresponds to the current host machine (the one
510 running the test cases). This should probably be called "host".<dd>
512 <dt><b>prcontext</b> (%prcontext)</dt>
513 <dd>Path to the prcontext tcl script that prints some context around a
514 line that matches a pattern. This isn't strictly necessary as the test suite
515 is run with its PATH altered to include the test/Scripts directory where
516 the prcontext script is located. Note that this script is similar to
517 <tt>grep -C</tt> but you should use the <tt>prcontext</tt> script because
518 not all platforms support <tt>grep -C</tt>.</dd>
520 <dt><b>llvmgcc</b> (%llvmgcc)</dt>
521 <dd>The full path to the <tt>llvm-gcc</tt> executable as specified in the
522 configured LLVM environment</dd>
524 <dt><b>llvmgxx</b> (%llvmgxx)</dt>
525 <dd>The full path to the <tt>llvm-gxx</tt> executable as specified in the
526 configured LLVM environment</dd>
528 <dt><b>llvmgcc_version</b> (%llvmgcc_version)</dt>
529 <dd>The full version number of the <tt>llvm-gcc</tt> executable.</dd>
531 <dt><b>llvmgccmajvers</b> (%llvmgccmajvers)</dt>
532 <dd>The major version number of the <tt>llvm-gcc</tt> executable.</dd>
534 <dt><b>gccpath</b></dt>
535 <dd>The full path to the C compiler used to <i>build </i> LLVM. Note that
536 this might not be gcc.</dd>
538 <dt><b>gxxpath</b></dt>
539 <dd>The full path to the C++ compiler used to <i>build </i> LLVM. Note that
540 this might not be g++.</dd>
542 <dt><b>compile_c</b> (%compile_c)</dt>
543 <dd>The full command line used to compile LLVM C source code. This has all
544 the configured -I, -D and optimization options.</dd>
546 <dt><b>compile_cxx</b> (%compile_cxx)</dt>
547 <dd>The full command used to compile LLVM C++ source code. This has
548 all the configured -I, -D and optimization options.</dd>
550 <dt><b>link</b> (%link)</dt>
551 <dd>This full link command used to link LLVM executables. This has all the
552 configured -I, -L and -l options.</dd>
554 <dt><b>shlibext</b> (%shlibext)</dt>
555 <dd>The suffix for the host platforms share library (dll) files. This
556 includes the period as the first character.</dd>
558 <p>To add more variables, two things need to be changed. First, add a line in
559 the <tt>test/Makefile</tt> that creates the <tt>site.exp</tt> file. This will
560 "set" the variable as a global in the site.exp file. Second, in the
561 <tt>test/lib/llvm.exp</tt> file, in the substitute proc, add the variable name
562 to the list of "global" declarations at the beginning of the proc. That's it,
563 the variable can then be used in test scripts.</p>
566 <!-- _______________________________________________________________________ -->
567 <div class="doc_subsection"><a name="dgfeatures">Other Features</a></div>
568 <!-- _______________________________________________________________________ -->
569 <div class="doc_text">
570 <p>To make RUN line writing easier, there are several shell scripts located
571 in the <tt>llvm/test/Scripts</tt> directory. This directory is in the PATH
572 when running tests, so you can just call these scripts using their name. For
575 <dt><b>ignore</b></dt>
576 <dd>This script runs its arguments and then always returns 0. This is useful
577 in cases where the test needs to cause a tool to generate an error (e.g. to
578 check the error output). However, any program in a pipeline that returns a
579 non-zero result will cause the test to fail. This script overcomes that
580 issue and nicely documents that the test case is purposefully ignoring the
581 result code of the tool</dd>
584 <dd>This script runs its arguments and then inverts the result code from
585 it. Zero result codes become 1. Non-zero result codes become 0. This is
586 useful to invert the result of a grep. For example "not grep X" means
587 succeed only if you don't find X in the input.</dd>
590 <p>Sometimes it is necessary to mark a test case as "expected fail" or XFAIL.
591 You can easily mark a test as XFAIL just by including <tt>XFAIL: </tt> on a
592 line near the top of the file. This signals that the test case should succeed
593 if the test fails. Such test cases are counted separately by DejaGnu. To
594 specify an expected fail, use the XFAIL keyword in the comments of the test
595 program followed by a colon and one or more regular expressions (separated by
596 a comma). The regular expressions allow you to XFAIL the test conditionally
597 by host platform. The regular expressions following the : are matched against
598 the target triplet or llvmgcc version number for the host machine. If there is
599 a match, the test is expected to fail. If not, the test is expected to
600 succeed. To XFAIL everywhere just specify <tt>XFAIL: *</tt>. When matching
601 the llvm-gcc version, you can specify the major (e.g. 3) or full version
602 (i.e. 3.4) number. Here is an example of an <tt>XFAIL</tt> line:</p>
604 <div class="doc_code">
606 ; XFAIL: darwin,sun,llvmgcc4
610 <p>To make the output more useful, the <tt>llvm_runtest</tt> function wil
611 scan the lines of the test case for ones that contain a pattern that matches
612 PR[0-9]+. This is the syntax for specifying a PR (Problem Report) number that
613 is related to the test case. The number after "PR" specifies the LLVM bugzilla
614 number. When a PR number is specified, it will be used in the pass/fail
615 reporting. This is useful to quickly get some context when a test fails.</p>
617 <p>Finally, any line that contains "END." will cause the special
618 interpretation of lines to terminate. This is generally done right after the
619 last RUN: line. This has two side effects: (a) it prevents special
620 interpretation of lines that are part of the test program, not the
621 instructions to the test case, and (b) it speeds things up for really big test
622 cases by avoiding interpretation of the remainder of the file.</p>
626 <!--=========================================================================-->
627 <div class="doc_section"><a name="llvmteststructure"><tt>llvm-test</tt>
629 <!--=========================================================================-->
631 <div class="doc_text">
633 <p>The <tt>test-suite</tt> module contains a number of programs that can be compiled
634 with LLVM and executed. These programs are compiled using the native compiler
635 and various LLVM backends. The output from the program compiled with the
636 native compiler is assumed correct; the results from the other programs are
637 compared to the native program output and pass if they match.</p>
639 <p>When executing tests, it is usually a good idea to start out with a subset of
640 the available tests or programs. This makes test run times smaller at first and
641 later on this is useful to investigate individual test failures. To run some
642 test only on a subset of programs, simply change directory to the programs you
643 want tested and run <tt>gmake</tt> there. Alternatively, you can run a different
644 test using the <tt>TEST</tt> variable to change what tests or run on the
645 selected programs (see below for more info).</p>
647 <p>In addition for testing correctness, the <tt>llvm-test</tt> directory also
648 performs timing tests of various LLVM optimizations. It also records
649 compilation times for the compilers and the JIT. This information can be
650 used to compare the effectiveness of LLVM's optimizations and code
653 <p><tt>llvm-test</tt> tests are divided into three types of tests: MultiSource,
654 SingleSource, and External.</p>
657 <li><tt>llvm-test/SingleSource</tt>
658 <p>The SingleSource directory contains test programs that are only a single
659 source file in size. These are usually small benchmark programs or small
660 programs that calculate a particular value. Several such programs are grouped
661 together in each directory.</p></li>
663 <li><tt>llvm-test/MultiSource</tt>
664 <p>The MultiSource directory contains subdirectories which contain entire
665 programs with multiple source files. Large benchmarks and whole applications
668 <li><tt>llvm-test/External</tt>
669 <p>The External directory contains Makefiles for building code that is external
670 to (i.e., not distributed with) LLVM. The most prominent members of this
671 directory are the SPEC 95 and SPEC 2000 benchmark suites. The <tt>External</tt>
672 directory does not contain these actual tests,but only the Makefiles that know
673 how to properly compile these programs from somewhere else. The presence and
674 location of these external programs is configured by the llvm-test
675 <tt>configure</tt> script.</p></li>
678 <p>Each tree is then subdivided into several categories, including applications,
679 benchmarks, regression tests, code that is strange grammatically, etc. These
680 organizations should be relatively self explanatory.</p>
682 <p>Some tests are known to fail. Some are bugs that we have not fixed yet;
683 others are features that we haven't added yet (or may never add). In DejaGNU,
684 the result for such tests will be XFAIL (eXpected FAILure). In this way, you
685 can tell the difference between an expected and unexpected failure.</p>
687 <p>The tests in <tt>llvm-test</tt> have no such feature at this time. If the
688 test passes, only warnings and other miscellaneous output will be generated. If
689 a test fails, a large <program> FAILED message will be displayed. This
690 will help you separate benign warnings from actual test failures.</p>
694 <!--=========================================================================-->
695 <div class="doc_section"><a name="runllvmtest">Running the LLVM Tests</a></div>
696 <!--=========================================================================-->
698 <div class="doc_text">
700 <p>First, all tests are executed within the LLVM object directory tree. They
701 <i>are not</i> executed inside of the LLVM source tree. This is because the
702 test suite creates temporary files during execution.</p>
704 <p>To run the <tt>llvm-test</tt> suite, you need to use the following steps:</p>
707 <li><tt>cd</tt> into the <tt>llvm/projects</tt> directory</li>
709 <li><p>Check out the <tt>test-suite</tt> module with:</p>
711 <div class="doc_code">
713 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk llvm-test
717 <p>This will get the test suite into <tt>llvm/projects/llvm-test</tt></p>
719 <li><p>Configure the test suite using llvm configure. This will automatically configure llvm-test.
720 You must do it from the top level otherwise llvm-gcc will not be set which is required to
722 <div class="doc_code">
724 % cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure --with-llvmgccdir=$LLVM_GCC_DIR
727 <p>Note that that <tt>$LLVM_GCC_DIR</tt> is the directory where you
728 <em>installed</em> llvm-gcc, not its src or obj directory.</p>
731 <li><p>Change back to the <tt>llvm/projects/llvm-test</tt> directory you created before
732 and run <tt>gmake</tt> (or just "<tt>make</tt>" on systems where GNU make is
733 the default, such as linux.</p></li>
735 <p>Note that the second and third steps only need to be done once. After you
736 have the suite checked out and configured, you don't need to do it again (unless
737 the test code or configure script changes).</p>
739 <!-- _______________________________________________________________________ -->
740 <div class="doc_subsection">
741 <a name="externaltests">Configuring external tests</a></div>
742 <!-- _______________________________________________________________________ -->
744 <div class="doc_text">
745 <p>Note, when configuring the <tt>llvm-test</tt> module, you might want to
746 specify the following configuration options:</p>
748 <dt><i>--enable-spec2000</i>
749 <dt><i>--enable-spec2000=<<tt>directory</tt>></i>
751 Enable the use of SPEC2000 when testing LLVM. This is disabled by default
752 (unless <tt>configure</tt> finds SPEC2000 installed). By specifying
753 <tt>directory</tt>, you can tell configure where to find the SPEC2000
754 benchmarks. If <tt>directory</tt> is left unspecified, <tt>configure</tt>
755 uses the default value
756 <tt>/home/vadve/shared/benchmarks/speccpu2000/benchspec</tt>.
759 <dt><i>--enable-spec95</i>
760 <dt><i>--enable-spec95=<<tt>directory</tt>></i>
762 Enable the use of SPEC95 when testing LLVM. It is similar to the
763 <i>--enable-spec2000</i> option.
766 <dt><i>--enable-povray</i>
767 <dt><i>--enable-povray=<<tt>directory</tt>></i>
769 Enable the use of Povray as an external test. Versions of Povray written
770 in C should work. This option is similar to the <i>--enable-spec2000</i>
775 <!-- _______________________________________________________________________ -->
776 <div class="doc_subsection">
777 <a name="llvmtesttests">Running different tests</a></div>
778 <!-- _______________________________________________________________________ -->
779 <div class="doc_text">
780 <p>In addition to the regular "whole program" tests, the <tt>llvm-test</tt>
781 module also provides a mechanism for compiling the programs in different ways.
782 If the variable TEST is defined on the gmake command line, the test system will
783 include a Makefile named <tt>TEST.<value of TEST variable>.Makefile</tt>.
784 This Makefile can modify build rules to yield different results.</p>
786 <p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to
787 create the nightly test reports. To run the nightly tests, run <tt>gmake
788 TEST=nightly</tt>.</p>
790 <p>There are several TEST Makefiles available in the tree. Some of them are
791 designed for internal LLVM research and will not work outside of the LLVM
792 research group. They may still be valuable, however, as a guide to writing your
793 own TEST Makefile for any optimization or analysis passes that you develop with
798 <!-- _______________________________________________________________________ -->
799 <div class="doc_subsection">
800 <a name="llvmtestoutput">Generating test output</a></div>
801 <!-- _______________________________________________________________________ -->
802 <div class="doc_text">
803 <p>There are a number of ways to run the tests and generate output. The most
804 simple one is simply running <tt>gmake</tt> with no arguments. This will
805 compile and run all programs in the tree using a number of different methods
806 and compare results. Any failures are reported in the output, but are likely
807 drowned in the other output. Passes are not reported explicitely.</p>
809 <p>Somewhat better is running <tt>gmake TEST=sometest test</tt>, which runs
810 the specified test and usually adds per-program summaries to the output
811 (depending on which sometest you use). For example, the <tt>nightly</tt> test
812 explicitely outputs TEST-PASS or TEST-FAIL for every test after each program.
813 Though these lines are still drowned in the output, it's easy to grep the
814 output logs in the Output directories.</p>
816 <p>Even better are the <tt>report</tt> and <tt>report.format</tt> targets
817 (where <tt>format</tt> is one of <tt>html</tt>, <tt>csv</tt>, <tt>text</tt> or
818 <tt>graphs</tt>). The exact contents of the report are dependent on which
819 <tt>TEST</tt> you are running, but the text results are always shown at the
820 end of the run and the results are always stored in the
821 <tt>report.<type>.format</tt> file (when running with
822 <tt>TEST=<type></tt>).
824 The <tt>report</tt> also generate a file called
825 <tt>report.<type>.raw.out</tt> containing the output of the entire test
829 <!-- _______________________________________________________________________ -->
830 <div class="doc_subsection">
831 <a name="customtest">Writing custom tests for llvm-test</a></div>
832 <!-- _______________________________________________________________________ -->
834 <div class="doc_text">
836 <p>Assuming you can run llvm-test, (e.g. "<tt>gmake TEST=nightly report</tt>"
837 should work), it is really easy to run optimizations or code generator
838 components against every program in the tree, collecting statistics or running
839 custom checks for correctness. At base, this is how the nightly tester works,
840 it's just one example of a general framework.</p>
842 <p>Lets say that you have an LLVM optimization pass, and you want to see how
843 many times it triggers. First thing you should do is add an LLVM
844 <a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which
845 will tally counts of things you care about.</p>
847 <p>Following this, you can set up a test and a report that collects these and
848 formats them for easy viewing. This consists of two files, an
849 "<tt>llvm-test/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
850 test) and an "<tt>llvm-test/TEST.XXX.report</tt>" file that indicates how to
851 format the output into a table. There are many example reports of various
852 levels of sophistication included with llvm-test, and the framework is very
855 <p>If you are interested in testing an optimization pass, check out the
856 "libcalls" test as an example. It can be run like this:<p>
858 <div class="doc_code">
860 % cd llvm/projects/llvm-test/MultiSource/Benchmarks # or some other level
861 % make TEST=libcalls report
865 <p>This will do a bunch of stuff, then eventually print a table like this:</p>
867 <div class="doc_code">
869 Name | total | #exit |
871 FreeBench/analyzer/analyzer | 51 | 6 |
872 FreeBench/fourinarow/fourinarow | 1 | 1 |
873 FreeBench/neural/neural | 19 | 9 |
874 FreeBench/pifft/pifft | 5 | 3 |
875 MallocBench/cfrac/cfrac | 1 | * |
876 MallocBench/espresso/espresso | 52 | 12 |
877 MallocBench/gs/gs | 4 | * |
878 Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
879 Prolangs-C/agrep/agrep | 33 | 12 |
880 Prolangs-C/allroots/allroots | * | * |
881 Prolangs-C/assembler/assembler | 47 | * |
882 Prolangs-C/bison/mybison | 74 | * |
887 <p>This basically is grepping the -stats output and displaying it in a table.
888 You can also use the "TEST=libcalls report.html" target to get the table in HTML
889 form, similarly for report.csv and report.tex.</p>
891 <p>The source for this is in llvm-test/TEST.libcalls.*. The format is pretty
892 simple: the Makefile indicates how to run the test (in this case,
893 "<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
894 each column of the output. The first value is the header for the column and the
895 second is the regex to grep the output of the command for. There are lots of
896 example reports that can do fancy stuff.</p>
901 <!--=========================================================================-->
902 <div class="doc_section"><a name="nightly">Running the nightly tester</a></div>
903 <!--=========================================================================-->
905 <div class="doc_text">
908 The <a href="http://llvm.org/nightlytest/">LLVM Nightly Testers</a>
909 automatically check out an LLVM tree, build it, run the "nightly"
910 program test (described above), run all of the DejaGNU tests,
911 delete the checked out tree, and then submit the results to
912 <a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>.
913 After test results are submitted to
914 <a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>,
915 they are processed and displayed on the tests page. An email to
916 <a href="http://lists.cs.uiuc.edu/pipermail/llvm-testresults/">
917 llvm-testresults@cs.uiuc.edu</a> summarizing the results is also generated.
918 This testing scheme is designed to ensure that programs don't break as well
919 as keep track of LLVM's progress over time.</p>
921 <p>If you'd like to set up an instance of the nightly tester to run on your
922 machine, take a look at the comments at the top of the
923 <tt>utils/NewNightlyTest.pl</tt> file. If you decide to set up a nightly tester
924 please choose a unique nickname and invoke <tt>utils/NewNightlyTest.pl</tt>
925 with the "-nickname [yournickname]" command line option.
927 <p>You can create a shell script to encapsulate the running of the script.
928 The optimized x86 Linux nightly test is run from just such a script:</p>
930 <div class="doc_code">
933 BASE=/proj/work/llvm/nightlytest
934 export BUILDDIR=$BASE/build
935 export WEBDIR=$BASE/testresults
936 export LLVMGCCDIR=/proj/work/llvm/cfrontend/install
937 export PATH=/proj/install/bin:$LLVMGCCDIR/bin:$PATH
938 export LD_LIBRARY_PATH=/proj/install/lib
940 cp /proj/work/llvm/llvm/utils/NewNightlyTest.pl .
941 nice ./NewNightlyTest.pl -nice -release -verbose -parallel -enable-linscan \
942 -nickname NightlyTester -noexternals > output.log 2>&1
946 <p>It is also possible to specify the the location your nightly test results
947 are submitted. You can do this by passing the command line option
948 "-submit-server [server_address]" and "-submit-script [script_on_server]" to
949 <tt>utils/NewNightlyTest.pl</tt>. For example, to submit to the llvm.org
950 nightly test results page, you would invoke the nightly test script with
951 "-submit-server llvm.org -submit-script /nightlytest/NightlyTestAccept.cgi".
952 If these options are not specified, the nightly test script sends the results
953 to the llvm.org nightly test results page.</p>
955 <p>Take a look at the <tt>NewNightlyTest.pl</tt> file to see what all of the
956 flags and strings do. If you start running the nightly tests, please let us
961 <!-- *********************************************************************** -->
965 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
966 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
967 <a href="http://validator.w3.org/check/referer"><img
968 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
970 John T. Criswell, Reid Spencer, and Tanya Lattner<br>
971 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
972 Last modified: $Date$