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="#quick">Quick Start</a></li>
18 <li><a href="#org">LLVM Test Suite Organization</a>
20 <li><a href="#codefragments">Code Fragments</a></li>
21 <li><a href="#wholeprograms">Whole Programs</a></li>
24 <li><a href="#tree">LLVM Test Suite Tree</a></li>
25 <li><a href="#dgstructure">DejaGNU Structure</a></li>
26 <li><a href="#progstructure"><tt>llvm-test</tt> Structure</a></li>
27 <li><a href="#run">Running the LLVM Tests</a>
29 <li><a href="#customtest">Writing custom tests for llvm-test</a></li>
32 <li><a href="#nightly">Running the nightly tester</a></li>
35 <div class="doc_author">
36 <p>Written by John T. Criswell, <a
37 href="http://llvm.x10sys.com/rspencer">Reid Spencer</a>, and Tanya Lattner</p>
40 <!--=========================================================================-->
41 <div class="doc_section"><a name="overview">Overview</a></div>
42 <!--=========================================================================-->
44 <div class="doc_text">
46 <p>This document is the reference manual for the LLVM test suite. It documents
47 the structure of the LLVM test suite, the tools needed to use it, and how to add
52 <!--=========================================================================-->
53 <div class="doc_section"><a name="Requirements">Requirements</a></div>
54 <!--=========================================================================-->
56 <div class="doc_text">
58 <p>In order to use the LLVM test suite, you will need all of the software
59 required to build LLVM, plus the following:</p>
62 <dt><a href="http://www.gnu.org/software/dejagnu/">DejaGNU</a></dt>
63 <dd>The Feature and Regressions tests are organized and run by DejaGNU.</dd>
64 <dt><a href="http://expect.nist.gov/">Expect</a></dt>
65 <dd>Expect is required by DejaGNU.</dd>
66 <dt><a href="http://www.tcl.tk/software/tcltk/">tcl</a></dt>
67 <dd>Tcl is required by DejaGNU. </dd>
69 <dt><a href="http://www.netlib.org/f2c">F2C</a></dt>
70 <dd>For now, LLVM does not have a Fortran front-end, but using F2C, we can run
71 Fortran benchmarks. F2C support must be enabled via <tt>configure</tt> if not
72 installed in a standard place. F2C requires three items: the <tt>f2c</tt>
73 executable, <tt>f2c.h</tt> to compile the generated code, and <tt>libf2c.a</tt>
74 to link generated code. By default, given an F2C directory <tt>$DIR</tt>, the
75 configure script will search <tt>$DIR/bin</tt> for <tt>f2c</tt>,
76 <tt>$DIR/include</tt> for <tt>f2c.h</tt>, and <tt>$DIR/lib</tt> for
77 <tt>libf2c.a</tt>. The default <tt>$DIR</tt> values are: <tt>/usr</tt>,
78 <tt>/usr/local</tt>, <tt>/sw</tt>, and <tt>/opt</tt>. If you installed F2C in a
79 different location, you must tell <tt>configure</tt>:
82 <li><tt>./configure --with-f2c=$DIR</tt><br>
83 This will specify a new <tt>$DIR</tt> for the above-described search
84 process. This will only work if the binary, header, and library are in their
85 respective subdirectories of <tt>$DIR</tt>.</li>
87 <li><tt>./configure --with-f2c-bin=/binary/path --with-f2c-inc=/include/path
88 --with-f2c-lib=/lib/path</tt><br>
89 This allows you to specify the F2C components separately. Note: if you choose
90 this route, you MUST specify all three components, and you need to only specify
91 <em>directories</em> where the files are located; do NOT include the
92 filenames themselves on the <tt>configure</tt> line.</li>
96 <p>Darwin (Mac OS X) developers can simplify the installation of Expect and tcl
97 by using fink. <tt>fink install expect</tt> will install both. Alternatively,
98 Darwinports users can use <tt>sudo port install expect</tt> to install Expect
103 <!--=========================================================================-->
104 <div class="doc_section"><a name="quick">Quick Start</a></div>
105 <!--=========================================================================-->
107 <div class="doc_text">
109 <p>The tests are located in two separate Subversion modules. The basic feature
110 and regression tests are in the main "llvm" module under the directory
111 <tt>llvm/test</tt>. A more comprehensive test suite that includes whole
112 programs in C and C++ is in the <tt>test-suite</tt> module. This module should
113 be checked out to the <tt>llvm/projects</tt> directory as llvm-test (for
114 historical purpose). When you <tt>configure</tt> the <tt>llvm</tt> module,
115 the <tt>llvm-test</tt> directory will be automatically configured.
116 Alternatively, you can configure the <tt>test-suite</tt> module manually.</p>
117 <p>To run all of the simple tests in LLVM using DejaGNU, use the master Makefile
118 in the <tt>llvm/test</tt> directory:</p>
120 <div class="doc_code">
128 <div class="doc_code">
134 <p>To run only a subdirectory of tests in llvm/test using DejaGNU (ie.
135 Regression/Transforms), just set the TESTSUITE variable to the path of the
136 subdirectory (relative to <tt>llvm/test</tt>):</p>
138 <div class="doc_code">
140 % gmake -C llvm/test TESTSUITE=Regression/Transforms
144 <p><b>Note: If you are running the tests with <tt>objdir != subdir</tt>, you
145 must have run the complete testsuite before you can specify a
146 subdirectory.</b></p>
148 <p>To run the comprehensive test suite (tests that compile and execute whole
149 programs), run the <tt>llvm-test</tt> tests:</p>
151 <div class="doc_code">
154 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk llvm-test
156 % ./configure --with-llvmgccdir=$LLVM_GCC_DIR
157 % cd projects/llvm-test
160 <p>where <tt>$LLVM_GCC_DIR</tt> is the directory where you <em>installed</em>
161 llvm-gcc, not it's src or obj dir.</p>
166 <!--=========================================================================-->
167 <div class="doc_section"><a name="org">LLVM Test Suite Organization</a></div>
168 <!--=========================================================================-->
170 <div class="doc_text">
172 <p>The LLVM test suite contains two major categories of tests: code
173 fragments and whole programs. Code fragments are in the <tt>llvm</tt> module
174 under the <tt>llvm/test</tt> directory. The whole programs
175 test suite is in the <tt>llvm-test</tt> module under the main directory.</p>
179 <!-- _______________________________________________________________________ -->
180 <div class="doc_subsection"><a name="codefragments">Code Fragments</a></div>
181 <!-- _______________________________________________________________________ -->
183 <div class="doc_text">
185 <p>Code fragments are small pieces of code that test a specific feature of LLVM
186 or trigger a specific bug in LLVM. They are usually written in LLVM assembly
187 language, but can be written in other languages if the test targets a particular
188 language front end.</p>
190 <p>Code fragments are not complete programs, and they are never executed to
191 determine correct behavior.</p>
193 <p>These code fragment tests are located in the <tt>llvm/test/Features</tt> and
194 <tt>llvm/test/Regression</tt> directories.</p>
198 <!-- _______________________________________________________________________ -->
199 <div class="doc_subsection"><a name="wholeprograms">Whole Programs</a></div>
200 <!-- _______________________________________________________________________ -->
202 <div class="doc_text">
204 <p>Whole Programs are pieces of code which can be compiled and linked into a
205 stand-alone program that can be executed. These programs are generally written
206 in high level languages such as C or C++, but sometimes they are written
207 straight in LLVM assembly.</p>
209 <p>These programs are compiled and then executed using several different
210 methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation,
211 etc). The output of these programs is compared to ensure that LLVM is compiling
212 the program correctly.</p>
214 <p>In addition to compiling and executing programs, whole program tests serve as
215 a way of benchmarking LLVM performance, both in terms of the efficiency of the
216 programs generated as well as the speed with which LLVM compiles, optimizes, and
219 <p>All "whole program" tests are located in the <tt>test-suite</tt> Subversion
224 <!--=========================================================================-->
225 <div class="doc_section"><a name="tree">LLVM Test Suite Tree</a></div>
226 <!--=========================================================================-->
228 <div class="doc_text">
230 <p>Each type of test in the LLVM test suite has its own directory. The major
231 subtrees of the test suite directory tree are as follows:</p>
234 <li><tt>llvm/test</tt>
235 <p>This directory contains a large array of small tests
236 that exercise various features of LLVM and to ensure that regressions do not
237 occur. The directory is broken into several sub-directories, each focused on
238 a particular area of LLVM. A few of the important ones are:</p>
240 <li><tt>Analysis</tt>: checks Analysis passes.</li>
241 <li><tt>Archive</tt>: checks the Archive library.</li>
242 <li><tt>Assembler</tt>: checks Assembly reader/writer functionality.</li>
243 <li><tt>Bitcode</tt>: checks Bitcode reader/writer functionality.</li>
244 <li><tt>CodeGen</tt>: checks code generation and each target.</li>
245 <li><tt>Features</tt>: checks various features of the LLVM language.</li>
246 <li><tt>Linker</tt>: tests bitcode linking.</li>
247 <li><tt>Transforms</tt>: tests each of the scalar, IPO, and utility
248 transforms to ensure they make the right transformations.</li>
249 <li><tt>Verifier</tt>: tests the IR verifier.</li>
251 <p>Typically when a bug is found in LLVM, a regression test containing
252 just enough code to reproduce the problem should be written and placed
253 somewhere underneath this directory. In most cases, this will be a small
254 piece of LLVM assembly language code, often distilled from an actual
255 application or benchmark.</p></li>
257 <li><tt>test-suite</tt>
258 <p>The <tt>test-suite</tt> module contains programs that can be compiled
259 with LLVM and executed. These programs are compiled using the native compiler
260 and various LLVM backends. The output from the program compiled with the
261 native compiler is assumed correct; the results from the other programs are
262 compared to the native program output and pass if they match.</p>
264 <p>In addition for testing correctness, the <tt>llvm-test</tt> directory also
265 performs timing tests of various LLVM optimizations. It also records
266 compilation times for the compilers and the JIT. This information can be
267 used to compare the effectiveness of LLVM's optimizations and code
270 <li><tt>llvm-test/SingleSource</tt>
271 <p>The SingleSource directory contains test programs that are only a single
272 source file in size. These are usually small benchmark programs or small
273 programs that calculate a particular value. Several such programs are grouped
274 together in each directory.</p></li>
276 <li><tt>llvm-test/MultiSource</tt>
277 <p>The MultiSource directory contains subdirectories which contain entire
278 programs with multiple source files. Large benchmarks and whole applications
281 <li><tt>llvm-test/External</tt>
282 <p>The External directory contains Makefiles for building code that is external
283 to (i.e., not distributed with) LLVM. The most prominent members of this
284 directory are the SPEC 95 and SPEC 2000 benchmark suites. The presence and
285 location of these external programs is configured by the llvm-test
286 <tt>configure</tt> script.</p></li>
291 <!--=========================================================================-->
292 <div class="doc_section"><a name="dgstructure">DejaGNU Structure</a></div>
293 <!--=========================================================================-->
294 <div class="doc_text">
295 <p>The LLVM test suite is partially driven by DejaGNU and partially driven by
296 GNU Make. Specifically, the Features and Regression tests are all driven by
297 DejaGNU. The <tt>llvm-test</tt> module is currently driven by a set of
300 <p>The DejaGNU structure is very simple, but does require some information to
301 be set. This information is gathered via <tt>configure</tt> and is written
302 to a file, <tt>site.exp</tt> in <tt>llvm/test</tt>. The <tt>llvm/test</tt>
303 Makefile does this work for you.</p>
305 <p>In order for DejaGNU to work, each directory of tests must have a
306 <tt>dg.exp</tt> file. DejaGNU looks for this file to determine how to run the
307 tests. This file is just a Tcl script and it can do anything you want, but
308 we've standardized it for the LLVM regression tests. It simply loads a Tcl
309 library (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt>
310 function defined in that library with a list of file names to run. The names
311 are obtained by using Tcl's glob command. Any directory that contains only
312 directories does not need the <tt>dg.exp</tt> file.</p>
314 <p>The <tt>llvm-runtests</tt> function lookas at each file that is passed to
315 it and gathers any lines together that match "RUN:". This are the "RUN" lines
316 that specify how the test is to be run. So, each test script must contain
317 RUN lines if it is to do anything. If there are no RUN lines, the
318 <tt>llvm-runtests</tt> function will issue an error and the test will
321 <p>RUN lines are specified in the comments of the test program using the
322 keyword <tt>RUN</tt> followed by a colon, and lastly the command (pipeline)
323 to execute. Together, these lines form the "script" that
324 <tt>llvm-runtests</tt> executes to run the test case. The syntax of the
325 RUN lines is similar to a shell's syntax for pipelines including I/O
326 redirection and variable substitution. However, even though these lines
327 may <i>look</i> like a shell script, they are not. RUN lines are interpreted
328 directly by the Tcl <tt>exec</tt> command. They are never executed by a
329 shell. Consequently the syntax differs from normal shell script syntax in a
330 few ways. You can specify as many RUN lines as needed.</p>
332 <p>Each RUN line is executed on its own, distinct from other lines unless
333 its last character is <tt>\</tt>. This continuation character causes the RUN
334 line to be concatenated with the next one. In this way you can build up long
335 pipelines of commands without making huge line lengths. The lines ending in
336 <tt>\</tt> are concatenated until a RUN line that doesn't end in <tt>\</tt> is
337 found. This concatenated set or RUN lines then constitutes one execution.
338 Tcl will substitute variables and arrange for the pipeline to be executed. If
339 any process in the pipeline fails, the entire line (and test case) fails too.
342 <p> Below is an example of legal RUN lines in a <tt>.ll</tt> file:</p>
344 <div class="doc_code">
346 ; RUN: llvm-as < %s | llvm-dis > %t1
347 ; RUN: llvm-dis < %s.bc-13 > %t2
352 <p>As with a Unix shell, the RUN: lines permit pipelines and I/O redirection
353 to be used. However, the usage is slightly different than for Bash. To check
354 what's legal, see the documentation for the
355 <a href="http://www.tcl.tk/man/tcl8.5/TclCmd/exec.htm#M2">Tcl exec</a>
357 <a href="http://www.tcl.tk/man/tcl8.5/tutorial/Tcl26.html">tutorial</a>.
358 The major differences are:</p>
360 <li>You can't do <tt>2>&1</tt>. That will cause Tcl to write to a
361 file named <tt>&1</tt>. Usually this is done to get stderr to go through
362 a pipe. You can do that in tcl with <tt>|&</tt> so replace this idiom:
363 <tt>... 2>&1 | grep</tt> with <tt>... |& grep</tt></li>
364 <li>You can only redirect to a file, not to another descriptor and not from
365 a here document.</li>
366 <li>tcl supports redirecting to open files with the @ syntax but you
367 shouldn't use that here.</li>
370 <p>There are some quoting rules that you must pay attention to when writing
371 your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any
372 ' or " so they will get passed to the invoked program. For example:</p>
374 <div class="doc_code">
376 ... | grep 'find this string'
380 <p>This will fail because the ' characters are passed to grep. This would
381 instruction grep to look for <tt>'find</tt> in the files <tt>this</tt> and
382 <tt>string'</tt>. To avoid this use curly braces to tell Tcl that it should
383 treat everything enclosed as one value. So our example would become:</p>
385 <div class="doc_code">
387 ... | grep {find this string}
391 <p>Additionally, the characters <tt>[</tt> and <tt>]</tt> are treated
392 specially by Tcl. They tell Tcl to interpret the content as a command to
393 execute. Since these characters are often used in regular expressions this can
394 have disastrous results and cause the entire test run in a directory to fail.
395 For example, a common idiom is to look for some basicblock number:</p>
397 <div class="doc_code">
403 <p>This, however, will cause Tcl to fail because its going to try to execute
404 a program named "2-8". Instead, what you want is this:</p>
406 <div class="doc_code">
408 ... | grep {bb\[2-8\]}
412 <p>Finally, if you need to pass the <tt>\</tt> character down to a program,
413 then it must be doubled. This is another Tcl special character. So, suppose
416 <div class="doc_code">
422 <p>This will fail to match what you want (a pointer to i32). First, the
423 <tt>'</tt> do not get stripped off. Second, the <tt>\</tt> gets stripped off
424 by Tcl so what grep sees is: <tt>'i32*'</tt>. That's not likely to match
425 anything. To resolve this you must use <tt>\\</tt> and the <tt>{}</tt>, like
428 <div class="doc_code">
436 <!-- _______________________________________________________________________ -->
437 <div class="doc_subsection"><a name="dgvars">Vars And Substitutions</a></div>
438 <div class="doc_text">
439 <p>With a RUN line there are a number of substitutions that are permitted. In
440 general, any Tcl variable that is available in the <tt>substitute</tt>
441 function (in <tt>test/lib/llvm.exp</tt>) can be substituted into a RUN line.
442 To make a substitution just write the variable's name preceded by a $.
443 Additionally, for compatibility reasons with previous versions of the test
444 library, certain names can be accessed with an alternate syntax: a % prefix.
445 These alternates are deprecated and may go away in a future version.
447 <p>Here are the available variable names. The alternate syntax is listed in
450 <dl style="margin-left: 25px">
451 <dt><b>$test</b> (%s)</dt>
452 <dd>The full path to the test case's source. This is suitable for passing
453 on the command line as the input to an llvm tool.</dd>
455 <dt><b>$srcdir</b></dt>
456 <dd>The source directory from where the "<tt>make check</tt>" was run.</dd>
458 <dt><b>objdir</b></dt>
459 <dd>The object directory that corresponds to the <tt>$srcdir</tt>.</dd>
461 <dt><b>subdir</b></dt>
462 <dd>A partial path from the <tt>test</tt> directory that contains the
463 sub-directory that contains the test source being executed.</dd>
465 <dt><b>srcroot</b></dt>
466 <dd>The root directory of the LLVM src tree.</dd>
468 <dt><b>objroot</b></dt>
469 <dd>The root directory of the LLVM object tree. This could be the same
473 <dd>The path to the directory that contains the test case source. This is
474 for locating any supporting files that are not generated by the test, but
475 used by the test.</dd>
478 <dd>The path to a temporary file name that could be used for this test case.
479 The file name won't conflict with other test cases. You can append to it if
480 you need multiple temporaries. This is useful as the destination of some
481 redirected output.</dd>
483 <dt><b>llvmlibsdir</b> (%llvmlibsdir)</dt>
484 <dd>The directory where the LLVM libraries are located.</dd>
486 <dt><b>target_triplet</b> (%target_triplet)</dt>
487 <dd>The target triplet that corresponds to the current host machine (the one
488 running the test cases). This should probably be called "host".<dd>
490 <dt><b>prcontext</b> (%prcontext)</dt>
491 <dd>Path to the prcontext tcl script that prints some context around a
492 line that matches a pattern. This isn't strictly necessary as the test suite
493 is run with its PATH altered to include the test/Scripts directory where
494 the prcontext script is located. Note that this script is similar to
495 <tt>grep -C</tt> but you should use the <tt>prcontext</tt> script because
496 not all platforms support <tt>grep -C</tt>.</dd>
498 <dt><b>llvmgcc</b> (%llvmgcc)</dt>
499 <dd>The full path to the <tt>llvm-gcc</tt> executable as specified in the
500 configured LLVM environment</dd>
502 <dt><b>llvmgxx</b> (%llvmgxx)</dt>
503 <dd>The full path to the <tt>llvm-gxx</tt> executable as specified in the
504 configured LLVM environment</dd>
506 <dt><b>llvmgcc_version</b> (%llvmgcc_version)</dt>
507 <dd>The full version number of the <tt>llvm-gcc</tt> executable.</dd>
509 <dt><b>llvmgccmajvers</b> (%llvmgccmajvers)</dt>
510 <dd>The major version number of the <tt>llvm-gcc</tt> executable.</dd>
512 <dt><b>gccpath</b></dt>
513 <dd>The full path to the C compiler used to <i>build </i> LLVM. Note that
514 this might not be gcc.</dd>
516 <dt><b>gxxpath</b></dt>
517 <dd>The full path to the C++ compiler used to <i>build </i> LLVM. Note that
518 this might not be g++.</dd>
520 <dt><b>compile_c</b> (%compile_c)</dt>
521 <dd>The full command line used to compile LLVM C source code. This has all
522 the configured -I, -D and optimization options.</dd>
524 <dt><b>compile_cxx</b> (%compile_cxx)</dt>
525 <dd>The full command used to compile LLVM C++ source code. This has
526 all the configured -I, -D and optimization options.</dd>
528 <dt><b>link</b> (%link)</dt>
529 <dd>This full link command used to link LLVM executables. This has all the
530 configured -I, -L and -l options.</dd>
532 <dt><b>shlibext</b> (%shlibext)</dt>
533 <dd>The suffix for the host platforms share library (dll) files. This
534 includes the period as the first character.</dd>
536 <p>To add more variables, two things need to be changed. First, add a line in
537 the <tt>test/Makefile</tt> that creates the <tt>site.exp</tt> file. This will
538 "set" the variable as a global in the site.exp file. Second, in the
539 <tt>test/lib/llvm.exp</tt> file, in the substitute proc, add the variable name
540 to the list of "global" declarations at the beginning of the proc. That's it,
541 the variable can then be used in test scripts.</p>
544 <!-- _______________________________________________________________________ -->
545 <div class="doc_subsection"><a name="dgfeatures">Other Features</a></div>
546 <div class="doc_text">
547 <p>To make RUN line writing easier, there are several shell scripts located
548 in the <tt>llvm/test/Scripts</tt> directory. For example:</p>
550 <dt><b>ignore</b></dt>
551 <dd>This script runs its arguments and then always returns 0. This is useful
552 in cases where the test needs to cause a tool to generate an error (e.g. to
553 check the error output). However, any program in a pipeline that returns a
554 non-zero result will cause the test to fail. This script overcomes that
555 issue and nicely documents that the test case is purposefully ignoring the
556 result code of the tool</dd>
559 <dd>This script runs its arguments and then inverts the result code from
560 it. Zero result codes become 1. Non-zero result codes become 0. This is
561 useful to invert the result of a grep. For example "not grep X" means
562 succeed only if you don't find X in the input.</dd>
565 <p>Sometimes it is necessary to mark a test case as "expected fail" or XFAIL.
566 You can easily mark a test as XFAIL just by including <tt>XFAIL: </tt> on a
567 line near the top of the file. This signals that the test case should succeed
568 if the test fails. Such test cases are counted separately by DejaGnu. To
569 specify an expected fail, use the XFAIL keyword in the comments of the test
570 program followed by a colon and one or more regular expressions (separated by
571 a comma). The regular expressions allow you to XFAIL the test conditionally
572 by host platform. The regular expressions following the : are matched against
573 the target triplet or llvmgcc version number for the host machine. If there is
574 a match, the test is expected to fail. If not, the test is expected to
575 succeed. To XFAIL everywhere just specify <tt>XFAIL: *</tt>. When matching
576 the llvm-gcc version, you can specify the major (e.g. 3) or full version
577 (i.e. 3.4) number. Here is an example of an <tt>XFAIL</tt> line:</p>
579 <div class="doc_code">
581 ; XFAIL: darwin,sun,llvmgcc4
585 <p>To make the output more useful, the <tt>llvm_runtest</tt> function wil
586 scan the lines of the test case for ones that contain a pattern that matches
587 PR[0-9]+. This is the syntax for specifying a PR (Problem Report) number that
588 is related to the test case. The numer after "PR" specifies the LLVM bugzilla
589 number. When a PR number is specified, it will be used in the pass/fail
590 reporting. This is useful to quickly get some context when a test fails.</p>
592 <p>Finally, any line that contains "END." will cause the special
593 interpretation of lines to terminate. This is generally done right after the
594 last RUN: line. This has two side effects: (a) it prevents special
595 interpretation of lines that are part of the test program, not the
596 instructions to the test case, and (b) it speeds things up for really big test
597 cases by avoiding interpretation of the remainder of the file.</p>
601 <!--=========================================================================-->
602 <div class="doc_section"><a name="progstructure"><tt>llvm-test</tt>
604 <!--=========================================================================-->
606 <div class="doc_text">
608 <p>As mentioned previously, the <tt>llvm-test</tt> module provides three types
609 of tests: MultiSource, SingleSource, and External. Each tree is then subdivided
610 into several categories, including applications, benchmarks, regression tests,
611 code that is strange grammatically, etc. These organizations should be
612 relatively self explanatory.</p>
614 <p>In addition to the regular "whole program" tests, the <tt>llvm-test</tt>
615 module also provides a mechanism for compiling the programs in different ways.
616 If the variable TEST is defined on the gmake command line, the test system will
617 include a Makefile named <tt>TEST.<value of TEST variable>.Makefile</tt>.
618 This Makefile can modify build rules to yield different results.</p>
620 <p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to
621 create the nightly test reports. To run the nightly tests, run <tt>gmake
622 TEST=nightly</tt>.</p>
624 <p>There are several TEST Makefiles available in the tree. Some of them are
625 designed for internal LLVM research and will not work outside of the LLVM
626 research group. They may still be valuable, however, as a guide to writing your
627 own TEST Makefile for any optimization or analysis passes that you develop with
630 <p>Note, when configuring the <tt>llvm-test</tt> module, you might want to
631 specify the following configuration options:</p>
633 <dt><i>--enable-spec2000</i>
634 <dt><i>--enable-spec2000=<<tt>directory</tt>></i>
636 Enable the use of SPEC2000 when testing LLVM. This is disabled by default
637 (unless <tt>configure</tt> finds SPEC2000 installed). By specifying
638 <tt>directory</tt>, you can tell configure where to find the SPEC2000
639 benchmarks. If <tt>directory</tt> is left unspecified, <tt>configure</tt>
640 uses the default value
641 <tt>/home/vadve/shared/benchmarks/speccpu2000/benchspec</tt>.
644 <dt><i>--enable-spec95</i>
645 <dt><i>--enable-spec95=<<tt>directory</tt>></i>
647 Enable the use of SPEC95 when testing LLVM. It is similar to the
648 <i>--enable-spec2000</i> option.
651 <dt><i>--enable-povray</i>
652 <dt><i>--enable-povray=<<tt>directory</tt>></i>
654 Enable the use of Povray as an external test. Versions of Povray written
655 in C should work. This option is similar to the <i>--enable-spec2000</i>
660 <!--=========================================================================-->
661 <div class="doc_section"><a name="run">Running the LLVM Tests</a></div>
662 <!--=========================================================================-->
664 <div class="doc_text">
666 <p>First, all tests are executed within the LLVM object directory tree. They
667 <i>are not</i> executed inside of the LLVM source tree. This is because the
668 test suite creates temporary files during execution.</p>
670 <p>The master Makefile in <tt>llvm/test</tt> is capable of running only the
671 DejaGNU driven tests. By default, it will run all of these tests.</p>
673 <p>To run only the DejaGNU driven tests, run <tt>gmake</tt> at the
674 command line in <tt>llvm/test</tt>. To run a specific directory of tests, use
675 the <tt>TESTSUITE</tt> variable.
678 <p>For example, to run the Regression tests, type
679 <tt>gmake TESTSUITE=Regression</tt> in <tt>llvm/tests</tt>.</p>
681 <p>Note that there are no Makefiles in <tt>llvm/test/Features</tt> and
682 <tt>llvm/test/Regression</tt>. You must use DejaGNU from the <tt>llvm/test</tt>
683 directory to run them.</p>
685 <p>To run the <tt>llvm-test</tt> suite, you need to use the following steps:</p>
688 <li><tt>cd</tt> into the <tt>llvm/projects</tt> directory</li>
690 <li><p>Check out the <tt>test-suite</tt> module with:</p>
692 <div class="doc_code">
694 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk llvm-test
698 <p>This will get the test suite into <tt>llvm/projects/llvm-test</tt></p>
700 <li><p>Configure the test suite using llvm configure. This will automatically configure llvm-test.
701 You must do it from the top level otherwise llvm-gcc will not be set which is required to
703 <div class="doc_code">
705 % cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure --with-llvmgccdir=$LLVM_GCC_DIR
708 <p>Note that that <tt>$LLVM_GCC_DIR</tt> is the directory where you
709 <em>installed</em> llvm-gcc, not its src or obj directory.</p>
712 <li><p>Change back to the <tt>llvm/projects/llvm-test</tt> directory you created before
713 and run <tt>gmake</tt> (or just "<tt>make</tt>" on systems where GNU make is
714 the default, such as linux.</p></li>
716 <p>Note that the second and third steps only need to be done once. After you
717 have the suite checked out and configured, you don't need to do it again (unless
718 the test code or configure script changes).</p>
720 <p>To make a specialized test (use one of the
721 <tt>llvm-test/TEST.<type>.Makefile</tt>s), just run:</p>
723 <div class="doc_code">
725 % cd llvm/projects/llvm-test
726 % gmake TEST=<type> test
730 <p>For example, you could run the nightly tester tests using the following
733 <div class="doc_code">
735 % cd llvm/projects/llvm-test
736 % gmake TEST=nightly test
740 <p>Regardless of which test you're running, the results are printed on standard
741 output and standard error. You can redirect these results to a file if you
744 <p>Some tests are known to fail. Some are bugs that we have not fixed yet;
745 others are features that we haven't added yet (or may never add). In DejaGNU,
746 the result for such tests will be XFAIL (eXpected FAILure). In this way, you
747 can tell the difference between an expected and unexpected failure.</p>
749 <p>The tests in <tt>llvm-test</tt> have no such feature at this time. If the
750 test passes, only warnings and other miscellaneous output will be generated. If
751 a test fails, a large <program> FAILED message will be displayed. This
752 will help you separate benign warnings from actual test failures.</p>
756 <!-- _______________________________________________________________________ -->
757 <div class="doc_subsection">
758 <a name="customtest">Writing custom tests for llvm-test</a></div>
759 <!-- _______________________________________________________________________ -->
761 <div class="doc_text">
763 <p>Assuming you can run llvm-test, (e.g. "<tt>gmake TEST=nightly report</tt>"
764 should work), it is really easy to run optimizations or code generator
765 components against every program in the tree, collecting statistics or running
766 custom checks for correctness. At base, this is how the nightly tester works,
767 it's just one example of a general framework.</p>
769 <p>Lets say that you have an LLVM optimization pass, and you want to see how
770 many times it triggers. First thing you should do is add an LLVM
771 <a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which
772 will tally counts of things you care about.</p>
774 <p>Following this, you can set up a test and a report that collects these and
775 formats them for easy viewing. This consists of two files, an
776 "<tt>llvm-test/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
777 test) and an "<tt>llvm-test/TEST.XXX.report</tt>" file that indicates how to
778 format the output into a table. There are many example reports of various
779 levels of sophistication included with llvm-test, and the framework is very
782 <p>If you are interested in testing an optimization pass, check out the
783 "libcalls" test as an example. It can be run like this:<p>
785 <div class="doc_code">
787 % cd llvm/projects/llvm-test/MultiSource/Benchmarks # or some other level
788 % make TEST=libcalls report
792 <p>This will do a bunch of stuff, then eventually print a table like this:</p>
794 <div class="doc_code">
796 Name | total | #exit |
798 FreeBench/analyzer/analyzer | 51 | 6 |
799 FreeBench/fourinarow/fourinarow | 1 | 1 |
800 FreeBench/neural/neural | 19 | 9 |
801 FreeBench/pifft/pifft | 5 | 3 |
802 MallocBench/cfrac/cfrac | 1 | * |
803 MallocBench/espresso/espresso | 52 | 12 |
804 MallocBench/gs/gs | 4 | * |
805 Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
806 Prolangs-C/agrep/agrep | 33 | 12 |
807 Prolangs-C/allroots/allroots | * | * |
808 Prolangs-C/assembler/assembler | 47 | * |
809 Prolangs-C/bison/mybison | 74 | * |
814 <p>This basically is grepping the -stats output and displaying it in a table.
815 You can also use the "TEST=libcalls report.html" target to get the table in HTML
816 form, similarly for report.csv and report.tex.</p>
818 <p>The source for this is in llvm-test/TEST.libcalls.*. The format is pretty
819 simple: the Makefile indicates how to run the test (in this case,
820 "<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
821 each column of the output. The first value is the header for the column and the
822 second is the regex to grep the output of the command for. There are lots of
823 example reports that can do fancy stuff.</p>
828 <!--=========================================================================-->
829 <div class="doc_section"><a name="nightly">Running the nightly tester</a></div>
830 <!--=========================================================================-->
832 <div class="doc_text">
835 The <a href="http://llvm.org/nightlytest/">LLVM Nightly Testers</a>
836 automatically check out an LLVM tree, build it, run the "nightly"
837 program test (described above), run all of the feature and regression tests,
838 delete the checked out tree, and then submit the results to
839 <a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>.
840 After test results are submitted to
841 <a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>,
842 they are processed and displayed on the tests page. An email to
843 <a href="http://lists.cs.uiuc.edu/pipermail/llvm-testresults/">
844 llvm-testresults@cs.uiuc.edu</a> summarizing the results is also generated.
845 This testing scheme is designed to ensure that programs don't break as well
846 as keep track of LLVM's progress over time.</p>
848 <p>If you'd like to set up an instance of the nightly tester to run on your
849 machine, take a look at the comments at the top of the
850 <tt>utils/NewNightlyTest.pl</tt> file. If you decide to set up a nightly tester
851 please choose a unique nickname and invoke <tt>utils/NewNightlyTest.pl</tt>
852 with the "-nickname [yournickname]" command line option.
854 <p>You can create a shell script to encapsulate the running of the script.
855 The optimized x86 Linux nightly test is run from just such a script:</p>
857 <div class="doc_code">
860 BASE=/proj/work/llvm/nightlytest
861 export BUILDDIR=$BASE/build
862 export WEBDIR=$BASE/testresults
863 export LLVMGCCDIR=/proj/work/llvm/cfrontend/install
864 export PATH=/proj/install/bin:$LLVMGCCDIR/bin:$PATH
865 export LD_LIBRARY_PATH=/proj/install/lib
867 cp /proj/work/llvm/llvm/utils/NewNightlyTest.pl .
868 nice ./NewNightlyTest.pl -nice -release -verbose -parallel -enable-linscan \
869 -nickname NightlyTester -noexternals > output.log 2>&1
873 <p>It is also possible to specify the the location your nightly test results
874 are submitted. You can do this by passing the command line option
875 "-submit-server [server_address]" and "-submit-script [script_on_server]" to
876 <tt>utils/NewNightlyTest.pl</tt>. For example, to submit to the llvm.org
877 nightly test results page, you would invoke the nightly test script with
878 "-submit-server llvm.org -submit-script /nightlytest/NightlyTestAccept.cgi".
879 If these options are not specified, the nightly test script sends the results
880 to the llvm.org nightly test results page.</p>
882 <p>Take a look at the <tt>NewNightlyTest.pl</tt> file to see what all of the
883 flags and strings do. If you start running the nightly tests, please let us
888 <!-- *********************************************************************** -->
892 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
893 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
894 <a href="http://validator.w3.org/check/referer"><img
895 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
897 John T. Criswell, Reid Spencer, and Tanya Lattner<br>
898 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
899 Last modified: $Date$