1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Testing Infrastructure Guide</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
10 <div class="doc_title">
11 LLVM Testing Infrastructure Guide
15 <li><a href="#overview">Overview</a></li>
16 <li><a href="#requirements">Requirements</a></li>
17 <li><a href="#org">LLVM testing infrastructure organization</a>
19 <li><a href="#dejagnu">DejaGNU tests</a></li>
20 <li><a href="#testsuite">Test suite</a></li>
23 <li><a href="#quick">Quick start</a>
25 <li><a href="#quickdejagnu">DejaGNU tests</a></li>
26 <li><a href="#quicktestsuite">Test suite</a></li>
29 <li><a href="#dgstructure">DejaGNU structure</a>
31 <li><a href="#dgcustom">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="#testsuitestructure">Test suite structure</a></li>
37 <li><a href="#testsuiterun">Running the test suite</a>
39 <li><a href="#testsuiteexternal">Configuring External Tests</a></li>
40 <li><a href="#testsuitetests">Running different tests</a></li>
41 <li><a href="#testsuiteoutput">Generating test output</a></li>
42 <li><a href="#testsuitecustom">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 testing infrastructure. It documents
60 the structure of the LLVM testing infrastructure, the tools needed to use it,
61 and how to add and run tests.</p>
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 testing infrastructure, 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>
83 <li><tt>./configure --with-f2c=$DIR</tt><br>
84 This will specify a new <tt>$DIR</tt> for the above-described search
85 process. This will only work if the binary, header, and library are in their
86 respective subdirectories of <tt>$DIR</tt>.</li>
88 <li><tt>./configure --with-f2c-bin=/binary/path --with-f2c-inc=/include/path
89 --with-f2c-lib=/lib/path</tt><br>
90 This allows you to specify the F2C components separately. Note: if you choose
91 this route, you MUST specify all three components, and you need to only specify
92 <em>directories</em> where the files are located; do NOT include the
93 filenames themselves on the <tt>configure</tt> line.</li>
97 <p>Darwin (Mac OS X) developers can simplify the installation of Expect and tcl
98 by using fink. <tt>fink install expect</tt> will install both. Alternatively,
99 Darwinports users can use <tt>sudo port install expect</tt> to install Expect
104 <!--=========================================================================-->
105 <div class="doc_section"><a name="org">LLVM testing infrastructure organization</a></div>
106 <!--=========================================================================-->
108 <div class="doc_text">
110 <p>The LLVM testing infrastructure contains two major categories of tests: code
111 fragments and whole programs. Code fragments are referred to as the "DejaGNU
112 tests" and are in the <tt>llvm</tt> module in subversion under the
113 <tt>llvm/test</tt> directory. The whole programs tests are referred to as the
114 "Test suite" and are in the <tt>test-suite</tt> module in subversion.
119 <!-- _______________________________________________________________________ -->
120 <div class="doc_subsection"><a name="dejagnu">DejaGNU tests</a></div>
121 <!-- _______________________________________________________________________ -->
123 <div class="doc_text">
125 <p>Code fragments are small pieces of code that test a specific feature of LLVM
126 or trigger a specific bug in LLVM. They are usually written in LLVM assembly
127 language, but can be written in other languages if the test targets a particular
128 language front end. These tests are driven by the DejaGNU testing framework,
129 which is hidden behind a few simple makefiles.</p>
131 <p>These code fragments are not complete programs. The code generated from them is
132 never executed to determine correct behavior.</p>
134 <p>These code fragment tests are located in the <tt>llvm/test</tt>
137 <p>Typically when a bug is found in LLVM, a regression test containing
138 just enough code to reproduce the problem should be written and placed
139 somewhere underneath this directory. In most cases, this will be a small
140 piece of LLVM assembly language code, often distilled from an actual
141 application or benchmark.</p>
145 <!-- _______________________________________________________________________ -->
146 <div class="doc_subsection"><a name="testsuite">Test suite</a></div>
147 <!-- _______________________________________________________________________ -->
149 <div class="doc_text">
151 <p>The test suite contains whole programs, which are pieces of
152 code which can be compiled and linked into a stand-alone program that can be
153 executed. These programs are generally written in high level languages such as
154 C or C++, but sometimes they are written straight in LLVM assembly.</p>
156 <p>These programs are compiled and then executed using several different
157 methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation,
158 etc). The output of these programs is compared to ensure that LLVM is compiling
159 the program correctly.</p>
161 <p>In addition to compiling and executing programs, whole program tests serve as
162 a way of benchmarking LLVM performance, both in terms of the efficiency of the
163 programs generated as well as the speed with which LLVM compiles, optimizes, and
166 <p>The test-suite is located in the <tt>test-suite</tt> Subversion module.</p>
170 <!--=========================================================================-->
171 <div class="doc_section"><a name="quick">Quick start</a></div>
172 <!--=========================================================================-->
174 <div class="doc_text">
176 <p>The tests are located in two separate Subversion modules. The
177 DejaGNU tests are in the main "llvm" module under the directory
178 <tt>llvm/test</tt> (so you get these tests for free with the main llvm tree).
179 The more comprehensive test suite that includes whole
180 programs in C and C++ is in the <tt>test-suite</tt> module. This module should
181 be checked out to the <tt>llvm/projects</tt> directory (don't use another name
182 then the default "test-suite", for then the test suite will be run every time
183 you run <tt>make</tt> in the main <tt>llvm</tt> directory).
184 When you <tt>configure</tt> the <tt>llvm</tt> module,
185 the <tt>test-suite</tt> directory will be automatically configured.
186 Alternatively, you can configure the <tt>test-suite</tt> module manually.</p>
188 <!-- _______________________________________________________________________ -->
189 <div class="doc_subsection"><a name="quickdejagnu">DejaGNU tests</a></div>
190 <!-- _______________________________________________________________________ -->
191 <p>To run all of the simple tests in LLVM using DejaGNU, use the master Makefile
192 in the <tt>llvm/test</tt> directory:</p>
194 <div class="doc_code">
202 <div class="doc_code">
208 <p>To run only a subdirectory of tests in <tt>llvm/test</tt> using DejaGNU (ie.
209 Transforms), just set the TESTSUITE variable to the path of the
210 subdirectory (relative to <tt>llvm/test</tt>):</p>
212 <div class="doc_code">
214 % gmake TESTSUITE=Transforms check
218 <p><b>Note: If you are running the tests with <tt>objdir != subdir</tt>, you
219 must have run the complete testsuite before you can specify a
220 subdirectory.</b></p>
222 <p>To run only a single test, set <tt>TESTONE</tt> to its path (relative to
223 <tt>llvm/test</tt>) and make the <tt>check-one</tt> target:</p>
225 <div class="doc_code">
227 % gmake TESTONE=Feature/basictest.ll check-one
231 <p>To run the tests with Valgrind (Memcheck by default), just append
232 <tt>VG=1</tt> to the commands above, e.g.:</p>
234 <div class="doc_code">
240 <!-- _______________________________________________________________________ -->
241 <div class="doc_subsection"><a name="quicktestsuite">Test suite</a></div>
242 <!-- _______________________________________________________________________ -->
244 <p>To run the comprehensive test suite (tests that compile and execute whole
245 programs), first checkout and setup the <tt>test-suite</tt> module:</p>
247 <div class="doc_code">
250 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
252 % ./configure --with-llvmgccdir=$LLVM_GCC_DIR
254 <p>where <tt>$LLVM_GCC_DIR</tt> is the directory where you <em>installed</em>
255 llvm-gcc, not it's src or obj dir.</p>
258 <p>Then, run the entire test suite by running make in the <tt>test-suite</tt>
261 <div class="doc_code">
263 % cd projects/test-suite
268 <p>Usually, running the "nightly" set of tests is a good idea, and you can also
269 let it generate a report by running:</p>
271 <div class="doc_code">
273 % cd projects/test-suite
274 % gmake TEST=nightly report report.html
278 <p>Any of the above commands can also be run in a subdirectory of
279 <tt>projects/test-suite</tt> to run the specified test only on the programs in
280 that subdirectory.</p>
284 <!--=========================================================================-->
285 <div class="doc_section"><a name="dgstructure">DejaGNU structure</a></div>
286 <!--=========================================================================-->
287 <div class="doc_text">
288 <p>The LLVM DejaGNU tests are driven by DejaGNU together with GNU Make and are
289 located in the <tt>llvm/test</tt> directory.
291 <p>This directory contains a large array of small tests
292 that exercise various features of LLVM and to ensure that regressions do not
293 occur. The directory is broken into several sub-directories, each focused on
294 a particular area of LLVM. A few of the important ones are:</p>
297 <li><tt>Analysis</tt>: checks Analysis passes.</li>
298 <li><tt>Archive</tt>: checks the Archive library.</li>
299 <li><tt>Assembler</tt>: checks Assembly reader/writer functionality.</li>
300 <li><tt>Bitcode</tt>: checks Bitcode reader/writer functionality.</li>
301 <li><tt>CodeGen</tt>: checks code generation and each target.</li>
302 <li><tt>Features</tt>: checks various features of the LLVM language.</li>
303 <li><tt>Linker</tt>: tests bitcode linking.</li>
304 <li><tt>Transforms</tt>: tests each of the scalar, IPO, and utility
305 transforms to ensure they make the right transformations.</li>
306 <li><tt>Verifier</tt>: tests the IR verifier.</li>
311 <!-- _______________________________________________________________________ -->
312 <div class="doc_subsection"><a name="dgcustom">Writing new DejaGNU tests</a></div>
313 <!-- _______________________________________________________________________ -->
314 <div class="doc_text">
315 <p>The DejaGNU structure is very simple, but does require some information to
316 be set. This information is gathered via <tt>configure</tt> and is written
317 to a file, <tt>site.exp</tt> in <tt>llvm/test</tt>. The <tt>llvm/test</tt>
318 Makefile does this work for you.</p>
320 <p>In order for DejaGNU to work, each directory of tests must have a
321 <tt>dg.exp</tt> file. DejaGNU looks for this file to determine how to run the
322 tests. This file is just a Tcl script and it can do anything you want, but
323 we've standardized it for the LLVM regression tests. If you're adding a
324 directory of tests, just copy <tt>dg.exp</tt> from another directory to get
325 running. The standard <tt>dg.exp</tt> simply loads a Tcl
326 library (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt>
327 function defined in that library with a list of file names to run. The names
328 are obtained by using Tcl's glob command. Any directory that contains only
329 directories does not need the <tt>dg.exp</tt> file.</p>
331 <p>The <tt>llvm-runtests</tt> function lookas at each file that is passed to
332 it and gathers any lines together that match "RUN:". This are the "RUN" lines
333 that specify how the test is to be run. So, each test script must contain
334 RUN lines if it is to do anything. If there are no RUN lines, the
335 <tt>llvm-runtests</tt> function will issue an error and the test will
338 <p>RUN lines are specified in the comments of the test program using the
339 keyword <tt>RUN</tt> followed by a colon, and lastly the command (pipeline)
340 to execute. Together, these lines form the "script" that
341 <tt>llvm-runtests</tt> executes to run the test case. The syntax of the
342 RUN lines is similar to a shell's syntax for pipelines including I/O
343 redirection and variable substitution. However, even though these lines
344 may <i>look</i> like a shell script, they are not. RUN lines are interpreted
345 directly by the Tcl <tt>exec</tt> command. They are never executed by a
346 shell. Consequently the syntax differs from normal shell script syntax in a
347 few ways. You can specify as many RUN lines as needed.</p>
349 <p>Each RUN line is executed on its own, distinct from other lines unless
350 its last character is <tt>\</tt>. This continuation character causes the RUN
351 line to be concatenated with the next one. In this way you can build up long
352 pipelines of commands without making huge line lengths. The lines ending in
353 <tt>\</tt> are concatenated until a RUN line that doesn't end in <tt>\</tt> is
354 found. This concatenated set of RUN lines then constitutes one execution.
355 Tcl will substitute variables and arrange for the pipeline to be executed. If
356 any process in the pipeline fails, the entire line (and test case) fails too.
359 <p> Below is an example of legal RUN lines in a <tt>.ll</tt> file:</p>
361 <div class="doc_code">
363 ; RUN: llvm-as < %s | llvm-dis > %t1
364 ; RUN: llvm-dis < %s.bc-13 > %t2
369 <p>As with a Unix shell, the RUN: lines permit pipelines and I/O redirection
370 to be used. However, the usage is slightly different than for Bash. To check
371 what's legal, see the documentation for the
372 <a href="http://www.tcl.tk/man/tcl8.5/TclCmd/exec.htm#M2">Tcl exec</a>
374 <a href="http://www.tcl.tk/man/tcl8.5/tutorial/Tcl26.html">tutorial</a>.
375 The major differences are:</p>
377 <li>You can't do <tt>2>&1</tt>. That will cause Tcl to write to a
378 file named <tt>&1</tt>. Usually this is done to get stderr to go through
379 a pipe. You can do that in tcl with <tt>|&</tt> so replace this idiom:
380 <tt>... 2>&1 | grep</tt> with <tt>... |& grep</tt></li>
381 <li>You can only redirect to a file, not to another descriptor and not from
382 a here document.</li>
383 <li>tcl supports redirecting to open files with the @ syntax but you
384 shouldn't use that here.</li>
387 <p>There are some quoting rules that you must pay attention to when writing
388 your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any
389 ' or " so they will get passed to the invoked program. For example:</p>
391 <div class="doc_code">
393 ... | grep 'find this string'
397 <p>This will fail because the ' characters are passed to grep. This would
398 instruction grep to look for <tt>'find</tt> in the files <tt>this</tt> and
399 <tt>string'</tt>. To avoid this use curly braces to tell Tcl that it should
400 treat everything enclosed as one value. So our example would become:</p>
402 <div class="doc_code">
404 ... | grep {find this string}
408 <p>Additionally, the characters <tt>[</tt> and <tt>]</tt> are treated
409 specially by Tcl. They tell Tcl to interpret the content as a command to
410 execute. Since these characters are often used in regular expressions this can
411 have disastrous results and cause the entire test run in a directory to fail.
412 For example, a common idiom is to look for some basicblock number:</p>
414 <div class="doc_code">
420 <p>This, however, will cause Tcl to fail because its going to try to execute
421 a program named "2-8". Instead, what you want is this:</p>
423 <div class="doc_code">
425 ... | grep {bb\[2-8\]}
429 <p>Finally, if you need to pass the <tt>\</tt> character down to a program,
430 then it must be doubled. This is another Tcl special character. So, suppose
433 <div class="doc_code">
439 <p>This will fail to match what you want (a pointer to i32). First, the
440 <tt>'</tt> do not get stripped off. Second, the <tt>\</tt> gets stripped off
441 by Tcl so what grep sees is: <tt>'i32*'</tt>. That's not likely to match
442 anything. To resolve this you must use <tt>\\</tt> and the <tt>{}</tt>, like
445 <div class="doc_code">
453 <!-- _______________________________________________________________________ -->
454 <div class="doc_subsection"><a name="dgvars">Variables and substitutions</a></div>
455 <!-- _______________________________________________________________________ -->
456 <div class="doc_text">
457 <p>With a RUN line there are a number of substitutions that are permitted. In
458 general, any Tcl variable that is available in the <tt>substitute</tt>
459 function (in <tt>test/lib/llvm.exp</tt>) can be substituted into a RUN line.
460 To make a substitution just write the variable's name preceded by a $.
461 Additionally, for compatibility reasons with previous versions of the test
462 library, certain names can be accessed with an alternate syntax: a % prefix.
463 These alternates are deprecated and may go away in a future version.
465 <p>Here are the available variable names. The alternate syntax is listed in
468 <dl style="margin-left: 25px">
469 <dt><b>$test</b> (%s)</dt>
470 <dd>The full path to the test case's source. This is suitable for passing
471 on the command line as the input to an llvm tool.</dd>
473 <dt><b>$srcdir</b></dt>
474 <dd>The source directory from where the "<tt>make check</tt>" was run.</dd>
476 <dt><b>objdir</b></dt>
477 <dd>The object directory that corresponds to the <tt>$srcdir</tt>.</dd>
479 <dt><b>subdir</b></dt>
480 <dd>A partial path from the <tt>test</tt> directory that contains the
481 sub-directory that contains the test source being executed.</dd>
483 <dt><b>srcroot</b></dt>
484 <dd>The root directory of the LLVM src tree.</dd>
486 <dt><b>objroot</b></dt>
487 <dd>The root directory of the LLVM object tree. This could be the same
491 <dd>The path to the directory that contains the test case source. This is
492 for locating any supporting files that are not generated by the test, but
493 used by the test.</dd>
496 <dd>The path to a temporary file name that could be used for this test case.
497 The file name won't conflict with other test cases. You can append to it if
498 you need multiple temporaries. This is useful as the destination of some
499 redirected output.</dd>
501 <dt><b>llvmlibsdir</b> (%llvmlibsdir)</dt>
502 <dd>The directory where the LLVM libraries are located.</dd>
504 <dt><b>target_triplet</b> (%target_triplet)</dt>
505 <dd>The target triplet that corresponds to the current host machine (the one
506 running the test cases). This should probably be called "host".<dd>
508 <dt><b>prcontext</b> (%prcontext)</dt>
509 <dd>Path to the prcontext tcl script that prints some context around a
510 line that matches a pattern. This isn't strictly necessary as the test suite
511 is run with its PATH altered to include the test/Scripts directory where
512 the prcontext script is located. Note that this script is similar to
513 <tt>grep -C</tt> but you should use the <tt>prcontext</tt> script because
514 not all platforms support <tt>grep -C</tt>.</dd>
516 <dt><b>llvmgcc</b> (%llvmgcc)</dt>
517 <dd>The full path to the <tt>llvm-gcc</tt> executable as specified in the
518 configured LLVM environment</dd>
520 <dt><b>llvmgxx</b> (%llvmgxx)</dt>
521 <dd>The full path to the <tt>llvm-gxx</tt> executable as specified in the
522 configured LLVM environment</dd>
524 <dt><b>llvmgcc_version</b> (%llvmgcc_version)</dt>
525 <dd>The full version number of the <tt>llvm-gcc</tt> executable.</dd>
527 <dt><b>llvmgccmajvers</b> (%llvmgccmajvers)</dt>
528 <dd>The major version number of the <tt>llvm-gcc</tt> executable.</dd>
530 <dt><b>gccpath</b></dt>
531 <dd>The full path to the C compiler used to <i>build </i> LLVM. Note that
532 this might not be gcc.</dd>
534 <dt><b>gxxpath</b></dt>
535 <dd>The full path to the C++ compiler used to <i>build </i> LLVM. Note that
536 this might not be g++.</dd>
538 <dt><b>compile_c</b> (%compile_c)</dt>
539 <dd>The full command line used to compile LLVM C source code. This has all
540 the configured -I, -D and optimization options.</dd>
542 <dt><b>compile_cxx</b> (%compile_cxx)</dt>
543 <dd>The full command used to compile LLVM C++ source code. This has
544 all the configured -I, -D and optimization options.</dd>
546 <dt><b>link</b> (%link)</dt>
547 <dd>This full link command used to link LLVM executables. This has all the
548 configured -I, -L and -l options.</dd>
550 <dt><b>shlibext</b> (%shlibext)</dt>
551 <dd>The suffix for the host platforms share library (dll) files. This
552 includes the period as the first character.</dd>
554 <p>To add more variables, two things need to be changed. First, add a line in
555 the <tt>test/Makefile</tt> that creates the <tt>site.exp</tt> file. This will
556 "set" the variable as a global in the site.exp file. Second, in the
557 <tt>test/lib/llvm.exp</tt> file, in the substitute proc, add the variable name
558 to the list of "global" declarations at the beginning of the proc. That's it,
559 the variable can then be used in test scripts.</p>
562 <!-- _______________________________________________________________________ -->
563 <div class="doc_subsection"><a name="dgfeatures">Other Features</a></div>
564 <!-- _______________________________________________________________________ -->
565 <div class="doc_text">
566 <p>To make RUN line writing easier, there are several shell scripts located
567 in the <tt>llvm/test/Scripts</tt> directory. This directory is in the PATH
568 when running tests, so you can just call these scripts using their name. For
571 <dt><b>ignore</b></dt>
572 <dd>This script runs its arguments and then always returns 0. This is useful
573 in cases where the test needs to cause a tool to generate an error (e.g. to
574 check the error output). However, any program in a pipeline that returns a
575 non-zero result will cause the test to fail. This script overcomes that
576 issue and nicely documents that the test case is purposefully ignoring the
577 result code of the tool</dd>
580 <dd>This script runs its arguments and then inverts the result code from
581 it. Zero result codes become 1. Non-zero result codes become 0. This is
582 useful to invert the result of a grep. For example "not grep X" means
583 succeed only if you don't find X in the input.</dd>
586 <p>Sometimes it is necessary to mark a test case as "expected fail" or XFAIL.
587 You can easily mark a test as XFAIL just by including <tt>XFAIL: </tt> on a
588 line near the top of the file. This signals that the test case should succeed
589 if the test fails. Such test cases are counted separately by DejaGnu. To
590 specify an expected fail, use the XFAIL keyword in the comments of the test
591 program followed by a colon and one or more regular expressions (separated by
592 a comma). The regular expressions allow you to XFAIL the test conditionally
593 by host platform. The regular expressions following the : are matched against
594 the target triplet or llvmgcc version number for the host machine. If there is
595 a match, the test is expected to fail. If not, the test is expected to
596 succeed. To XFAIL everywhere just specify <tt>XFAIL: *</tt>. When matching
597 the llvm-gcc version, you can specify the major (e.g. 3) or full version
598 (i.e. 3.4) number. Here is an example of an <tt>XFAIL</tt> line:</p>
600 <div class="doc_code">
602 ; XFAIL: darwin,sun,llvmgcc4
606 <p>To make the output more useful, the <tt>llvm_runtest</tt> function wil
607 scan the lines of the test case for ones that contain a pattern that matches
608 PR[0-9]+. This is the syntax for specifying a PR (Problem Report) number that
609 is related to the test case. The number after "PR" specifies the LLVM bugzilla
610 number. When a PR number is specified, it will be used in the pass/fail
611 reporting. This is useful to quickly get some context when a test fails.</p>
613 <p>Finally, any line that contains "END." will cause the special
614 interpretation of lines to terminate. This is generally done right after the
615 last RUN: line. This has two side effects: (a) it prevents special
616 interpretation of lines that are part of the test program, not the
617 instructions to the test case, and (b) it speeds things up for really big test
618 cases by avoiding interpretation of the remainder of the file.</p>
622 <!--=========================================================================-->
623 <div class="doc_section"><a name="testsuitestructure">Test suite
625 <!--=========================================================================-->
627 <div class="doc_text">
629 <p>The <tt>test-suite</tt> module contains a number of programs that can be compiled
630 with LLVM and executed. These programs are compiled using the native compiler
631 and various LLVM backends. The output from the program compiled with the
632 native compiler is assumed correct; the results from the other programs are
633 compared to the native program output and pass if they match.</p>
635 <p>When executing tests, it is usually a good idea to start out with a subset of
636 the available tests or programs. This makes test run times smaller at first and
637 later on this is useful to investigate individual test failures. To run some
638 test only on a subset of programs, simply change directory to the programs you
639 want tested and run <tt>gmake</tt> there. Alternatively, you can run a different
640 test using the <tt>TEST</tt> variable to change what tests or run on the
641 selected programs (see below for more info).</p>
643 <p>In addition for testing correctness, the <tt>llvm-test</tt> directory also
644 performs timing tests of various LLVM optimizations. It also records
645 compilation times for the compilers and the JIT. This information can be
646 used to compare the effectiveness of LLVM's optimizations and code
649 <p><tt>llvm-test</tt> tests are divided into three types of tests: MultiSource,
650 SingleSource, and External.</p>
653 <li><tt>llvm-test/SingleSource</tt>
654 <p>The SingleSource directory contains test programs that are only a single
655 source file in size. These are usually small benchmark programs or small
656 programs that calculate a particular value. Several such programs are grouped
657 together in each directory.</p></li>
659 <li><tt>llvm-test/MultiSource</tt>
660 <p>The MultiSource directory contains subdirectories which contain entire
661 programs with multiple source files. Large benchmarks and whole applications
664 <li><tt>llvm-test/External</tt>
665 <p>The External directory contains Makefiles for building code that is external
666 to (i.e., not distributed with) LLVM. The most prominent members of this
667 directory are the SPEC 95 and SPEC 2000 benchmark suites. The <tt>External</tt>
668 directory does not contain these actual tests, but only the Makefiles that know
669 how to properly compile these programs from somewhere else. The presence and
670 location of these external programs is configured by the llvm-test
671 <tt>configure</tt> script.</p></li>
674 <p>Each tree is then subdivided into several categories, including applications,
675 benchmarks, regression tests, code that is strange grammatically, etc. These
676 organizations should be relatively self explanatory.</p>
678 <p>Some tests are known to fail. Some are bugs that we have not fixed yet;
679 others are features that we haven't added yet (or may never add). In DejaGNU,
680 the result for such tests will be XFAIL (eXpected FAILure). In this way, you
681 can tell the difference between an expected and unexpected failure.</p>
683 <p>The tests in the test suite have no such feature at this time. If the
684 test passes, only warnings and other miscellaneous output will be generated. If
685 a test fails, a large <program> FAILED message will be displayed. This
686 will help you separate benign warnings from actual test failures.</p>
690 <!--=========================================================================-->
691 <div class="doc_section"><a name="testsuiterun">Running the test suite</a></div>
692 <!--=========================================================================-->
694 <div class="doc_text">
696 <p>First, all tests are executed within the LLVM object directory tree. They
697 <i>are not</i> executed inside of the LLVM source tree. This is because the
698 test suite creates temporary files during execution.</p>
700 <p>To run the test suite, you need to use the following steps:</p>
703 <li><tt>cd</tt> into the <tt>llvm/projects</tt> directory in your source tree.
706 <li><p>Check out the <tt>test-suite</tt> module with:</p>
708 <div class="doc_code">
710 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
713 <p>This will get the test suite into <tt>llvm/projects/test-suite</tt>.</p>
715 <li><p>Configure and build <tt>llvm</tt>.</p></li>
716 <li><p>Configure and build <tt>llvm-gcc</tt>.</p></li>
717 <li><p>Install <tt>llvm-gcc</tt> somewhere.</p></li>
718 <li><p><em>Re-configure</em> <tt>llvm</tt> from the top level of
719 each build tree (LLVM object directory tree) in which you want
720 to run the test suite, just as you do before building LLVM.</p>
721 <p>During the <em>re-configuration</em>, you must either: (1)
722 have <tt>llvm-gcc</tt> you just built in your path, or (2)
723 specify the directory where your just-built <tt>llvm-gcc</tt> is
724 installed using <tt>--with-llvmgccdir=$LLVM_GCC_DIR</tt>.</p>
725 <p>You must also tell the configure machinery that the test suite
726 is available so it can be configured for your build tree:</p>
727 <div class="doc_code">
729 % cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure [--with-llvmgccdir=$LLVM_GCC_DIR]
732 <p>[Remember that <tt>$LLVM_GCC_DIR</tt> is the directory where you
733 <em>installed</em> llvm-gcc, not its src or obj directory.]</p>
736 <li><p>You can now run the test suite from your build tree as follows:</p>
737 <div class="doc_code">
739 % cd $LLVM_OBJ_ROOT/projects/test-suite
745 <p>Note that the second and third steps only need to be done once. After you
746 have the suite checked out and configured, you don't need to do it again (unless
747 the test code or configure script changes).</p>
749 <!-- _______________________________________________________________________ -->
750 <div class="doc_subsection">
751 <a name="testsuiteexternal">Configuring External Tests</a></div>
752 <!-- _______________________________________________________________________ -->
754 <div class="doc_text">
755 <p>In order to run the External tests in the <tt>test-suite</tt>
756 module, you must specify <i>--with-externals</i>. This
757 must be done during the <em>re-configuration</em> step (see above),
758 and the <tt>llvm</tt> re-configuration must recognize the
759 previously-built <tt>llvm-gcc</tt>. If any of these is missing or
760 neglected, the External tests won't work.</p>
762 <dt><i>--with-externals</i></dt>
763 <dt><i>--with-externals=<<tt>directory</tt>></i></dt>
765 This tells LLVM where to find any external tests. They are expected to be
766 in specifically named subdirectories of <<tt>directory</tt>>.
767 If <tt>directory</tt> is left unspecified,
768 <tt>configure</tt> uses the default value
769 <tt>/home/vadve/shared/benchmarks/speccpu2000/benchspec</tt>.
770 Subdirectory names known to LLVM include:
777 Others are added from time to time, and can be determined from
781 <!-- _______________________________________________________________________ -->
782 <div class="doc_subsection">
783 <a name="testsuitetests">Running different tests</a></div>
784 <!-- _______________________________________________________________________ -->
785 <div class="doc_text">
786 <p>In addition to the regular "whole program" tests, the <tt>test-suite</tt>
787 module also provides a mechanism for compiling the programs in different ways.
788 If the variable TEST is defined on the <tt>gmake</tt> command line, the test system will
789 include a Makefile named <tt>TEST.<value of TEST variable>.Makefile</tt>.
790 This Makefile can modify build rules to yield different results.</p>
792 <p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to
793 create the nightly test reports. To run the nightly tests, run <tt>gmake
794 TEST=nightly</tt>.</p>
796 <p>There are several TEST Makefiles available in the tree. Some of them are
797 designed for internal LLVM research and will not work outside of the LLVM
798 research group. They may still be valuable, however, as a guide to writing your
799 own TEST Makefile for any optimization or analysis passes that you develop with
804 <!-- _______________________________________________________________________ -->
805 <div class="doc_subsection">
806 <a name="testsuiteoutput">Generating test output</a></div>
807 <!-- _______________________________________________________________________ -->
808 <div class="doc_text">
809 <p>There are a number of ways to run the tests and generate output. The most
810 simple one is simply running <tt>gmake</tt> with no arguments. This will
811 compile and run all programs in the tree using a number of different methods
812 and compare results. Any failures are reported in the output, but are likely
813 drowned in the other output. Passes are not reported explicitely.</p>
815 <p>Somewhat better is running <tt>gmake TEST=sometest test</tt>, which runs
816 the specified test and usually adds per-program summaries to the output
817 (depending on which sometest you use). For example, the <tt>nightly</tt> test
818 explicitely outputs TEST-PASS or TEST-FAIL for every test after each program.
819 Though these lines are still drowned in the output, it's easy to grep the
820 output logs in the Output directories.</p>
822 <p>Even better are the <tt>report</tt> and <tt>report.format</tt> targets
823 (where <tt>format</tt> is one of <tt>html</tt>, <tt>csv</tt>, <tt>text</tt> or
824 <tt>graphs</tt>). The exact contents of the report are dependent on which
825 <tt>TEST</tt> you are running, but the text results are always shown at the
826 end of the run and the results are always stored in the
827 <tt>report.<type>.format</tt> file (when running with
828 <tt>TEST=<type></tt>).
830 The <tt>report</tt> also generate a file called
831 <tt>report.<type>.raw.out</tt> containing the output of the entire test
835 <!-- _______________________________________________________________________ -->
836 <div class="doc_subsection">
837 <a name="testsuitecustom">Writing custom tests for the test suite</a></div>
838 <!-- _______________________________________________________________________ -->
840 <div class="doc_text">
842 <p>Assuming you can run the test suite, (e.g. "<tt>gmake TEST=nightly report</tt>"
843 should work), it is really easy to run optimizations or code generator
844 components against every program in the tree, collecting statistics or running
845 custom checks for correctness. At base, this is how the nightly tester works,
846 it's just one example of a general framework.</p>
848 <p>Lets say that you have an LLVM optimization pass, and you want to see how
849 many times it triggers. First thing you should do is add an LLVM
850 <a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which
851 will tally counts of things you care about.</p>
853 <p>Following this, you can set up a test and a report that collects these and
854 formats them for easy viewing. This consists of two files, an
855 "<tt>test-suite/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
856 test) and an "<tt>llvm-test/TEST.XXX.report</tt>" file that indicates how to
857 format the output into a table. There are many example reports of various
858 levels of sophistication included with the test suite, and the framework is very
861 <p>If you are interested in testing an optimization pass, check out the
862 "libcalls" test as an example. It can be run like this:<p>
864 <div class="doc_code">
866 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
867 % make TEST=libcalls report
871 <p>This will do a bunch of stuff, then eventually print a table like this:</p>
873 <div class="doc_code">
875 Name | total | #exit |
877 FreeBench/analyzer/analyzer | 51 | 6 |
878 FreeBench/fourinarow/fourinarow | 1 | 1 |
879 FreeBench/neural/neural | 19 | 9 |
880 FreeBench/pifft/pifft | 5 | 3 |
881 MallocBench/cfrac/cfrac | 1 | * |
882 MallocBench/espresso/espresso | 52 | 12 |
883 MallocBench/gs/gs | 4 | * |
884 Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
885 Prolangs-C/agrep/agrep | 33 | 12 |
886 Prolangs-C/allroots/allroots | * | * |
887 Prolangs-C/assembler/assembler | 47 | * |
888 Prolangs-C/bison/mybison | 74 | * |
893 <p>This basically is grepping the -stats output and displaying it in a table.
894 You can also use the "TEST=libcalls report.html" target to get the table in HTML
895 form, similarly for report.csv and report.tex.</p>
897 <p>The source for this is in test-suite/TEST.libcalls.*. The format is pretty
898 simple: the Makefile indicates how to run the test (in this case,
899 "<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
900 each column of the output. The first value is the header for the column and the
901 second is the regex to grep the output of the command for. There are lots of
902 example reports that can do fancy stuff.</p>
907 <!--=========================================================================-->
908 <div class="doc_section"><a name="nightly">Running the nightly tester</a></div>
909 <!--=========================================================================-->
911 <div class="doc_text">
914 The <a href="http://llvm.org/nightlytest/">LLVM Nightly Testers</a>
915 automatically check out an LLVM tree, build it, run the "nightly"
916 program test (described above), run all of the DejaGNU tests,
917 delete the checked out tree, and then submit the results to
918 <a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>.
919 After test results are submitted to
920 <a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>,
921 they are processed and displayed on the tests page. An email to
922 <a href="http://lists.cs.uiuc.edu/pipermail/llvm-testresults/">
923 llvm-testresults@cs.uiuc.edu</a> summarizing the results is also generated.
924 This testing scheme is designed to ensure that programs don't break as well
925 as keep track of LLVM's progress over time.</p>
927 <p>If you'd like to set up an instance of the nightly tester to run on your
928 machine, take a look at the comments at the top of the
929 <tt>utils/NewNightlyTest.pl</tt> file. If you decide to set up a nightly tester
930 please choose a unique nickname and invoke <tt>utils/NewNightlyTest.pl</tt>
931 with the "-nickname [yournickname]" command line option.
933 <p>You can create a shell script to encapsulate the running of the script.
934 The optimized x86 Linux nightly test is run from just such a script:</p>
936 <div class="doc_code">
939 BASE=/proj/work/llvm/nightlytest
940 export BUILDDIR=$BASE/build
941 export WEBDIR=$BASE/testresults
942 export LLVMGCCDIR=/proj/work/llvm/cfrontend/install
943 export PATH=/proj/install/bin:$LLVMGCCDIR/bin:$PATH
944 export LD_LIBRARY_PATH=/proj/install/lib
946 cp /proj/work/llvm/llvm/utils/NewNightlyTest.pl .
947 nice ./NewNightlyTest.pl -nice -release -verbose -parallel -enable-linscan \
948 -nickname NightlyTester -noexternals > output.log 2>&1
952 <p>It is also possible to specify the the location your nightly test results
953 are submitted. You can do this by passing the command line option
954 "-submit-server [server_address]" and "-submit-script [script_on_server]" to
955 <tt>utils/NewNightlyTest.pl</tt>. For example, to submit to the llvm.org
956 nightly test results page, you would invoke the nightly test script with
957 "-submit-server llvm.org -submit-script /nightlytest/NightlyTestAccept.cgi".
958 If these options are not specified, the nightly test script sends the results
959 to the llvm.org nightly test results page.</p>
961 <p>Take a look at the <tt>NewNightlyTest.pl</tt> file to see what all of the
962 flags and strings do. If you start running the nightly tests, please let us
967 <!-- *********************************************************************** -->
971 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
972 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
973 <a href="http://validator.w3.org/check/referer"><img
974 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
976 John T. Criswell, Reid Spencer, and Tanya Lattner<br>
977 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
978 Last modified: $Date$