blob: c597c5bf09718dd5689efd895920e18bcb8b48ed [file] [log] [blame]
<title>LLVM Test Suite Guide</title>
<link rel="stylesheet" href="llvm.css" type="text/css">
<div class="doc_title">
LLVM Test Suite Guide
<li><a href="#overview">Overview</a></li>
<li><a href="#Requirements">Requirements</a></li>
<li><a href="#quick">Quick Start</a></li>
<li><a href="#org">LLVM Test Suite Organization</a>
<li><a href="#codefragments">Code Fragments</a></li>
<li><a href="#wholeprograms">Whole Programs</a></li>
<li><a href="#tree">LLVM Test Suite Tree</a></li>
<li><a href="#dgstructure">DejaGNU Structure</a></li>
<li><a href="#progstructure"><tt>llvm-test</tt> Structure</a></li>
<li><a href="#run">Running the LLVM Tests</a>
<li><a href="#customtest">Writing custom tests for llvm-test</a></li>
<li><a href="#nightly">Running the nightly tester</a></li>
<div class="doc_author">
<p>Written by John T. Criswell, <a
href="">Reid Spencer</a>, and Tanya Lattner</p>
<div class="doc_section"><a name="overview">Overview</a></div>
<div class="doc_text">
<p>This document is the reference manual for the LLVM test suite. It documents
the structure of the LLVM test suite, the tools needed to use it, and how to add
and run tests.</p>
<div class="doc_section"><a name="Requirements">Requirements</a></div>
<div class="doc_text">
<p>In order to use the LLVM test suite, you will need all of the software
required to build LLVM, plus the following:</p>
<dt><a href="">DejaGNU</a></dt>
<dd>The Feature and Regressions tests are organized and run by DejaGNU.</dd>
<dt><a href="">Expect</a></dt>
<dd>Expect is required by DejaGNU.</dd>
<dt><a href="">tcl</a></dt>
<dd>Tcl is required by DejaGNU. </dd>
<dt><a href="">F2C</a></dt>
<dd>For now, LLVM does not have a Fortran front-end, but using F2C, we can run
Fortran benchmarks. F2C support must be enabled via <tt>configure</tt> if not
installed in a standard place. F2C requires three items: the <tt>f2c</tt>
executable, <tt>f2c.h</tt> to compile the generated code, and <tt>libf2c.a</tt>
to link generated code. By default, given an F2C directory <tt>$DIR</tt>, the
configure script will search <tt>$DIR/bin</tt> for <tt>f2c</tt>,
<tt>$DIR/include</tt> for <tt>f2c.h</tt>, and <tt>$DIR/lib</tt> for
<tt>libf2c.a</tt>. The default <tt>$DIR</tt> values are: <tt>/usr</tt>,
<tt>/usr/local</tt>, <tt>/sw</tt>, and <tt>/opt</tt>. If you installed F2C in a
different location, you must tell <tt>configure</tt>:
<li><tt>./configure --with-f2c=$DIR</tt><br>
This will specify a new <tt>$DIR</tt> for the above-described search
process. This will only work if the binary, header, and library are in their
respective subdirectories of <tt>$DIR</tt>.</li>
<li><tt>./configure --with-f2c-bin=/binary/path --with-f2c-inc=/include/path
This allows you to specify the F2C components separately. Note: if you choose
this route, you MUST specify all three components, and you need to only specify
<em>directories</em> where the files are located; do NOT include the
filenames themselves on the <tt>configure</tt> line.</li>
<p>Darwin (Mac OS X) developers can simplify the installation of Expect and tcl
by using fink. <tt>fink install expect</tt> will install both. Alternatively,
Darwinports users can use <tt>sudo port install expect</tt> to install Expect
and tcl.</p>
<div class="doc_section"><a name="quick">Quick Start</a></div>
<div class="doc_text">
<p>The tests are located in two separate CVS modules. The basic feature and
regression tests are in the main "llvm" module under the directory
<tt>llvm/test</tt>. A more comprehensive test suite that includes whole
programs in C and C++ is in the <tt>llvm-test</tt> module. This module should
be checked out to the <tt>llvm/projects</tt> directory. When you
<tt>configure</tt> the <tt>llvm</tt> module, the <tt>llvm-test</tt> module
will be automatically configured. Alternatively, you can configure the
<tt>llvm-test</tt> module manually.</p>
<p>To run all of the simple tests in LLVM using DejaGNU, use the master Makefile
in the <tt>llvm/test</tt> directory:</p>
% gmake -C llvm/test
% gmake check
<p>To run only a subdirectory of tests in llvm/test using DejaGNU (ie.
Regression/Transforms), just set the TESTSUITE variable to the path of the
subdirectory (relative to <tt>llvm/test</tt>):</p>
% gmake -C llvm/test TESTSUITE=Regression/Transforms
<p><b>Note: If you are running the tests with <tt>objdir != subdir</tt>, you
must have run the complete testsuite before you can specify a
<p>To run the comprehensive test suite (tests that compile and execute whole
programs), run the <tt>llvm-test</tt> tests:</p>
% cd llvm/projects
% cvs co llvm-test
% cd llvm-test
% ./configure --with-llvmsrc=$LLVM_SRC_ROOT --with-llvmobj=$LLVM_OBJ_ROOT
% gmake
<div class="doc_section"><a name="org">LLVM Test Suite Organization</a></div>
<div class="doc_text">
<p>The LLVM test suite contains two major categories of tests: code
fragments and whole programs. Code fragments are in the <tt>llvm</tt> module
under the <tt>llvm/test</tt> directory. The whole programs
test suite is in the <tt>llvm-test</tt> module under the main directory.</p>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="codefragments">Code Fragments</a></div>
<!-- _______________________________________________________________________ -->
<div class="doc_text">
<p>Code fragments are small pieces of code that test a specific feature of LLVM
or trigger a specific bug in LLVM. They are usually written in LLVM assembly
language, but can be written in other languages if the test targets a particular
language front end.</p>
<p>Code fragments are not complete programs, and they are never executed to
determine correct behavior.</p>
<p>These code fragment tests are located in the <tt>llvm/test/Features</tt> and
<tt>llvm/test/Regression</tt> directories.</p>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection"><a name="wholeprograms">Whole Programs</a></div>
<!-- _______________________________________________________________________ -->
<div class="doc_text">
<p>Whole Programs are pieces of code which can be compiled and linked into a
stand-alone program that can be executed. These programs are generally written
in high level languages such as C or C++, but sometimes they are written
straight in LLVM assembly.</p>
<p>These programs are compiled and then executed using several different
methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation,
etc). The output of these programs is compared to ensure that LLVM is compiling
the program correctly.</p>
<p>In addition to compiling and executing programs, whole program tests serve as
a way of benchmarking LLVM performance, both in terms of the efficiency of the
programs generated as well as the speed with which LLVM compiles, optimizes, and
generates code.</p>
<p>All "whole program" tests are located in the <tt>llvm-test</tt> CVS
<div class="doc_section"><a name="tree">LLVM Test Suite Tree</a></div>
<div class="doc_text">
<p>Each type of test in the LLVM test suite has its own directory. The major
subtrees of the test suite directory tree are as follows:</p>
<p>This directory contains sample codes that test various features of the
LLVM language. These pieces of sample code are run through various
assembler, disassembler, and optimizer passes.</p>
<p>This directory contains regression tests for LLVM. When a bug is found
in LLVM, a regression test containing just enough code to reproduce the
problem should be written and placed somewhere underneath this directory.
In most cases, this will be a small piece of LLVM assembly language code,
often distilled from an actual application or benchmark.</p>
<p>The <tt>llvm-test</tt> CVS module contains programs that can be compiled
with LLVM and executed. These programs are compiled using the native compiler
and various LLVM backends. The output from the program compiled with the
native compiler is assumed correct; the results from the other programs are
compared to the native program output and pass if they match.</p>
<p>In addition for testing correctness, the <tt>llvm-test</tt> directory also
performs timing tests of various LLVM optimizations. It also records
compilation times for the compilers and the JIT. This information can be
used to compare the effectiveness of LLVM's optimizations and code
<p>The SingleSource directory contains test programs that are only a single
source file in size. These are usually small benchmark programs or small
programs that calculate a particular value. Several such programs are grouped
together in each directory.</p></li>
<p>The MultiSource directory contains subdirectories which contain entire
programs with multiple source files. Large benchmarks and whole applications
go here.</p></li>
<p>The External directory contains Makefiles for building code that is external
to (i.e., not distributed with) LLVM. The most prominent members of this
directory are the SPEC 95 and SPEC 2000 benchmark suites. The presence and
location of these external programs is configured by the llvm-test
<tt>configure</tt> script.</p></li>
<div class="doc_section"><a name="dgstructure">DejaGNU Structure</a></div>
<div class="doc_text">
<p>The LLVM test suite is partially driven by DejaGNU and partially
driven by GNU Make. Specifically, the Features and Regression tests
are all driven by DejaGNU. The <tt>llvm-test</tt>
module is currently driven by a set of Makefiles.</p>
<p>The DejaGNU structure is very simple, but does require some
information to be set. This information is gathered via <tt>configure</tt> and
is written to a file, <tt>site.exp</tt> in <tt>llvm/test</tt>. The
Makefile does this work for you.</p>
<p>In order for DejaGNU to work, each directory of tests must have a
<tt>dg.exp</tt> file. This file is a program written in tcl that calls
the <tt>llvm-runtests</tt> procedure on each test file. The
llvm-runtests procedure is defined in
<tt>llvm/test/lib/llvm-dg.exp</tt>. Any directory that contains only
directories does not need the <tt>dg.exp</tt> file.</p>
<p>In order for a test to be run, it must contain information within
the test file on how to run the test. These are called <tt>RUN</tt>
lines. Run lines are specified in the comments of the test program
using the keyword <tt>RUN</tt> followed by a colon, and lastly the
commands to execute. These commands will be executed in a bash script,
so any bash syntax is acceptable. You can specify as many RUN lines as
necessary. Each RUN line translates to one line in the resulting bash
script. Below is an example of legal RUN lines in a <tt>.ll</tt>
; RUN: llvm-as < %s | llvm-dis > %t1
; RUN: llvm-dis < %s.bc-13 > %t2
; RUN: diff %t1 %t2
<p>There are a couple patterns within a <tt>RUN</tt> line that the
llvm-runtest procedure looks for and replaces with the appropriate
<dl style="margin-left: 25px">
<dd>The path to the source directory. This is for locating
any supporting files that are not generated by the test, but used by
the test.</dd>
<dd>The test file.</dd>
<dd>Temporary filename: testscript.test_filename.tmp, where
test_filename is the name of the test file. All temporary files are
placed in the Output directory within the directory the test is
<dd>Path to a script that performs grep -C. Use this since not all
platforms support grep -C.</dd>
<dt>%llvmgcc</dt> <dd>Full path to the llvm-gcc executable.</dd>
<dt>%llvmgxx</dt> <dd>Full path to the llvm-g++ executable.</dd>
<p>There are also several scripts in the llvm/test/Scripts directory
that you might find useful when writing <tt>RUN</tt> lines.</p>
<p>Lastly, you can easily mark a test that is expected to fail on a
specific platform or with a specific version of llvmgcc by using the
<tt>XFAIL</tt> keyword. Xfail lines are
specified in the comments of the test program using <tt>XFAIL</tt>,
followed by a colon, and one or more regular expressions (separated by
a comma) that will match against the target triplet or llvmgcc version for the
machine. You can use * to match all targets. You can specify the major or full
version (i.e. 3.4) for llvmgcc. Here is an example of an
<tt>XFAIL</tt> line:</p>
; XFAIL: darwin,sun,llvmgcc4
<div class="doc_section"><a name="progstructure"><tt>llvm-test</tt>
<div class="doc_text">
<p>As mentioned previously, the <tt>llvm-test</tt> module provides three types
of tests: MultiSource, SingleSource, and External. Each tree is then subdivided
into several categories, including applications, benchmarks, regression tests,
code that is strange grammatically, etc. These organizations should be
relatively self explanatory.</p>
<p>In addition to the regular "whole program" tests, the <tt>llvm-test</tt>
module also provides a mechanism for compiling the programs in different ways.
If the variable TEST is defined on the gmake command line, the test system will
include a Makefile named <tt>TEST.&lt;value of TEST variable&gt;.Makefile</tt>.
This Makefile can modify build rules to yield different results.</p>
<p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to
create the nightly test reports. To run the nightly tests, run <tt>gmake
<p>There are several TEST Makefiles available in the tree. Some of them are
designed for internal LLVM research and will not work outside of the LLVM
research group. They may still be valuable, however, as a guide to writing your
own TEST Makefile for any optimization or analysis passes that you develop with
<p>Note, when configuring the <tt>llvm-test</tt> module, you might want to
specify the following configuration options:</p>
Enable the use of SPEC2000 when testing LLVM. This is disabled by default
(unless <tt>configure</tt> finds SPEC2000 installed). By specifying
<tt>directory</tt>, you can tell configure where to find the SPEC2000
benchmarks. If <tt>directory</tt> is left unspecified, <tt>configure</tt>
uses the default value
Enable the use of SPEC95 when testing LLVM. It is similar to the
<i>--enable-spec2000</i> option.
Enable the use of Povray as an external test. Versions of Povray written
in C should work. This option is similar to the <i>--enable-spec2000</i>
<div class="doc_section"><a name="run">Running the LLVM Tests</a></div>
<div class="doc_text">
<p>First, all tests are executed within the LLVM object directory tree. They
<i>are not</i> executed inside of the LLVM source tree. This is because the
test suite creates temporary files during execution.</p>
<p>The master Makefile in llvm/test is capable of running only the DejaGNU
driven tests. By default, it will run all of these tests.</p>
<p>To run only the DejaGNU driven tests, run <tt>gmake</tt> at the
command line in <tt>llvm/test</tt>. To run a specific directory of tests, use
the TESTSUITE variable.
<p>For example, to run the Regression tests, type
<tt>gmake TESTSUITE=Regression</tt> in <tt>llvm/tests</tt>.</p>
<p>Note that there are no Makefiles in <tt>llvm/test/Features</tt> and
<tt>llvm/test/Regression</tt>. You must use DejaGNU from the <tt>llvm/test</tt>
directory to run them.</p>
<p>To run the <tt>llvm-test</tt> suite, you need to use the following steps:
<li>cd into the llvm/projects directory</li>
<li>check out the <tt>llvm-test</tt> module with:<br/>
<tt>cvs -d co -PR llvm-test</tt><br>
This will get the test suite into <tt>llvm/projects/llvm-test</tt></li>
<li>configure the test suite. You can do this one of two ways:
<li>Use the regular llvm configure:<br/>
<tt>cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure</tt><br/>
This will ensure that the <tt>projects/llvm-test</tt> directory is also
properly configured.</li>
<li>Use the <tt>configure</tt> script found in the <tt>llvm-test</tt> source
--with-llvmsrc=$LLVM_SRC_ROOT --with-llvmobj=$LLVM_OBJ_ROOT</tt>
<p>Note that the second and third steps only need to be done once. After you
have the suite checked out and configured, you don't need to do it again (unless
the test code or configure script changes).</p>
<p>To make a specialized test (use one of the
<tt>llvm-test/TEST.&lt;type&gt;.Makefile</tt>s), just run:<br/>
<tt>gmake TEST=&lt;type&gt; test</tt><br/>For example, you could run the
nightly tester tests using the following commands:</p>
% cd llvm/projects/llvm-test
% gmake TEST=nightly test
<p>Regardless of which test you're running, the results are printed on standard
output and standard error. You can redirect these results to a file if you
<p>Some tests are known to fail. Some are bugs that we have not fixed yet;
others are features that we haven't added yet (or may never add). In DejaGNU,
the result for such tests will be XFAIL (eXpected FAILure). In this way, you
can tell the difference between an expected and unexpected failure.</p>
<p>The tests in <tt>llvm-test</tt> have no such feature at this time. If the
test passes, only warnings and other miscellaneous output will be generated. If
a test fails, a large &lt;program&gt; FAILED message will be displayed. This
will help you separate benign warnings from actual test failures.</p>
<!-- _______________________________________________________________________ -->
<div class="doc_subsection">
<a name="customtest">Writing custom tests for llvm-test</a></div>
<!-- _______________________________________________________________________ -->
<div class="doc_text">
<p>Assuming you can run llvm-test, (e.g. "<tt>gmake TEST=nightly report</tt>"
should work), it is really easy to run optimizations or code generator
components against every program in the tree, collecting statistics or running
custom checks for correctness. At base, this is how the nightly tester works,
it's just one example of a general framework.</p>
<p>Lets say that you have an LLVM optimization pass, and you want to see how
many times it triggers. First thing you should do is add an LLVM
<a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which
will tally counts of things you care about.</p>
<p>Following this, you can set up a test and a report that collects these and
formats them for easy viewing. This consists of two files, an
"<tt>llvm-test/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
test) and an "<tt>llvm-test/</tt>" file that indicates how to
format the output into a table. There are many example reports of various
levels of sophistication included with llvm-test, and the framework is very
<p>If you are interested in testing an optimization pass, check out the
"libcalls" test as an example. It can be run like this:<p>
<div class="doc_code">
% cd llvm/projects/llvm-test/MultiSource/Benchmarks # or some other level
% make TEST=libcalls report
<p>This will do a bunch of stuff, then eventually print a table like this:</p>
<div class="doc_code">
Name | total | #exit |
FreeBench/analyzer/analyzer | 51 | 6 |
FreeBench/fourinarow/fourinarow | 1 | 1 |
FreeBench/neural/neural | 19 | 9 |
FreeBench/pifft/pifft | 5 | 3 |
MallocBench/cfrac/cfrac | 1 | * |
MallocBench/espresso/espresso | 52 | 12 |
MallocBench/gs/gs | 4 | * |
Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
Prolangs-C/agrep/agrep | 33 | 12 |
Prolangs-C/allroots/allroots | * | * |
Prolangs-C/assembler/assembler | 47 | * |
Prolangs-C/bison/mybison | 74 | * |
<p>This basically is grepping the -stats output and displaying it in a table.
You can also use the "TEST=libcalls report.html" target to get the table in HTML
form, similarly for report.csv and report.tex.</p>
<p>The source for this is in llvm-test/TEST.libcalls.*. The format is pretty
simple: the Makefile indicates how to run the test (in this case,
"<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
each column of the output. The first value is the header for the column and the
second is the regex to grep the output of the command for. There are lots of
example reports that can do fancy stuff.</p>
<div class="doc_section"><a name="nightly">Running the nightly tester</a></div>
<div class="doc_text">
The <a href="">LLVM Nightly Testers</a>
automatically check out an LLVM tree, build it, run the "nightly"
program test (described above), run all of the feature and regression tests,
delete the checked out tree, and finally submit the test results to the
llvm nightlytest results server. This tester is designed to ensure that
programs don't break as well as keep track of LLVM's progress over time.</p>
<p>If you'd like to set up an instance of the nightly tester to run on your
machine, take a look at the comments at the top of the
<tt>utils/</tt> file. We usually run it from a crontab entry
that looks like this:</p>
<div class="doc_code">
5 3 * * * $HOME/llvm/utils/ -parallel $CVSROOT \
$HOME/buildtest $HOME/cvs/testresults
<p>Or, you can create a shell script to encapsulate the running of the script.
The optimized x86 Linux nightly test is run from just such a script:</p>
<div class="doc_code">
export BUILDDIR=$BASE/build
export WEBDIR=$BASE/testresults
export LLVMGCCDIR=/proj/work/llvm/cfrontend/install
export PATH=/proj/install/bin:$LLVMGCCDIR/bin:$PATH
export LD_LIBRARY_PATH=/proj/install/lib
cd $BASE
cp /proj/work/llvm/llvm/utils/ .
nice ./ -nice -release -verbose -parallel -enable-linscan \
-noexternals 2&gt;&amp;1 &gt; output.log
mail -s 'X86 nightly tester results' <a href="\
listinfo/llvm-testresults"></a> &lt; output.log
<p>Please examine the LLVM Nightly Tester Page and choose a nickname for your
machine that has that has not already been used. You can specify the
nickname that will appear on the Nightly Tester Page for your machine by
using the flag <tt>-nickname YourNickname</tt> when you invoke
<p>Take a look at the <tt></tt> file to see what all of the flags
and strings do. If you start running the nightly tests, please let us know.
<!-- *********************************************************************** -->
<a href=""><img
src="" alt="Valid CSS!"></a>
<a href=""><img
src="" alt="Valid HTML 4.01!" /></a>
John T. Criswell, Reid Spencer, and Tanya Lattner<br>
<a href="">The LLVM Compiler Infrastructure</a><br/>
Last modified: $Date$