1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <title>LLVM Testing Infrastructure Guide</title>
7 <link rel="stylesheet" href="llvm.css" type="text/css">
12 LLVM Testing Infrastructure Guide
16 <li><a href="#overview">Overview</a></li>
17 <li><a href="#requirements">Requirements</a></li>
18 <li><a href="#org">LLVM testing infrastructure organization</a>
20 <li><a href="#regressiontests">Regression tests</a></li>
21 <li><a href="#testsuite">Test suite</a></li>
22 <li><a href="#debuginfotests">Debugging Information tests</a></li>
25 <li><a href="#quick">Quick start</a>
27 <li><a href="#quickregressiontests">Regression tests</a></li>
28 <li><a href="#quicktestsuite">Test suite</a></li>
29 <li><a href="#quickdebuginfotests">Debugging Information tests</a></li>
32 <li><a href="#rtstructure">Regression test structure</a>
34 <li><a href="#rtcustom">Writing new regression tests</a></li>
35 <li><a href="#FileCheck">The FileCheck utility</a></li>
36 <li><a href="#rtvars">Variables and substitutions</a></li>
37 <li><a href="#rtfeatures">Other features</a></li>
40 <li><a href="#testsuitestructure">Test suite structure</a></li>
41 <li><a href="#testsuiterun">Running the test suite</a>
43 <li><a href="#testsuiteexternal">Configuring External Tests</a></li>
44 <li><a href="#testsuitetests">Running different tests</a></li>
45 <li><a href="#testsuiteoutput">Generating test output</a></li>
46 <li><a href="#testsuitecustom">Writing custom tests for test-suite</a></li>
51 <div class="doc_author">
52 <p>Written by John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner</p>
55 <!--=========================================================================-->
56 <h2><a name="overview">Overview</a></h2>
57 <!--=========================================================================-->
61 <p>This document is the reference manual for the LLVM testing infrastructure. It
62 documents the structure of the LLVM testing infrastructure, the tools needed to
63 use it, and how to add and run tests.</p>
67 <!--=========================================================================-->
68 <h2><a name="requirements">Requirements</a></h2>
69 <!--=========================================================================-->
73 <p>In order to use the LLVM testing infrastructure, you will need all of the
74 software required to build LLVM, as well
75 as <a href="http://python.org">Python</a> 2.4 or later.</p>
79 <!--=========================================================================-->
80 <h2><a name="org">LLVM testing infrastructure organization</a></h2>
81 <!--=========================================================================-->
85 <p>The LLVM testing infrastructure contains two major categories of tests:
86 regression tests and whole programs. The regression tests are contained inside
87 the LLVM repository itself under <tt>llvm/test</tt> and are expected to always
88 pass -- they should be run before every commit. The whole programs tests are
89 referred to as the "LLVM test suite" and are in the <tt>test-suite</tt> module
93 <!-- _______________________________________________________________________ -->
94 <h3><a name="regressiontests">Regression tests</a></h3>
95 <!-- _______________________________________________________________________ -->
99 <p>The regression tests are small pieces of code that test a specific feature of
100 LLVM or trigger a specific bug in LLVM. They are usually written in LLVM
101 assembly language, but can be written in other languages if the test targets a
102 particular language front end (and the appropriate <tt>--with-llvmgcc</tt>
103 options were used at <tt>configure</tt> time of the <tt>llvm</tt> module). These
104 tests are driven by the 'lit' testing tool, which is part of LLVM.</p>
106 <p>These code fragments are not complete programs. The code generated
107 from them is never executed to determine correct behavior.</p>
109 <p>These code fragment tests are located in the <tt>llvm/test</tt>
112 <p>Typically when a bug is found in LLVM, a regression test containing
113 just enough code to reproduce the problem should be written and placed
114 somewhere underneath this directory. In most cases, this will be a small
115 piece of LLVM assembly language code, often distilled from an actual
116 application or benchmark.</p>
120 <!-- _______________________________________________________________________ -->
121 <h3><a name="testsuite">Test suite</a></h3>
122 <!-- _______________________________________________________________________ -->
126 <p>The test suite contains whole programs, which are pieces of
127 code which can be compiled and linked into a stand-alone program that can be
128 executed. These programs are generally written in high level languages such as
129 C or C++, but sometimes they are written straight in LLVM assembly.</p>
131 <p>These programs are compiled and then executed using several different
132 methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation,
133 etc). The output of these programs is compared to ensure that LLVM is compiling
134 the program correctly.</p>
136 <p>In addition to compiling and executing programs, whole program tests serve as
137 a way of benchmarking LLVM performance, both in terms of the efficiency of the
138 programs generated as well as the speed with which LLVM compiles, optimizes, and
141 <p>The test-suite is located in the <tt>test-suite</tt> Subversion module.</p>
145 <!-- _______________________________________________________________________ -->
146 <h3><a name="debuginfotests">Debugging Information tests</a></h3>
147 <!-- _______________________________________________________________________ -->
151 <p>The test suite contains tests to check quality of debugging information.
152 The test are written in C based languages or in LLVM assembly language. </p>
154 <p>These tests are compiled and run under a debugger. The debugger output
155 is checked to validate of debugging information. See README.txt in the
156 test suite for more information . This test suite is located in the
157 <tt>debuginfo-tests</tt> Subversion module. </p>
163 <!--=========================================================================-->
164 <h2><a name="quick">Quick start</a></h2>
165 <!--=========================================================================-->
169 <p>The tests are located in two separate Subversion modules. The regressions
170 tests are in the main "llvm" module under the directory
171 <tt>llvm/test</tt> (so you get these tests for free with the main llvm tree).
172 The more comprehensive test suite that includes whole
173 programs in C and C++ is in the <tt>test-suite</tt> module. This module should
174 be checked out to the <tt>llvm/projects</tt> directory (don't use another name
175 than the default "test-suite", for then the test suite will be run every time
176 you run <tt>make</tt> in the main <tt>llvm</tt> directory).
177 When you <tt>configure</tt> the <tt>llvm</tt> module,
178 the <tt>test-suite</tt> directory will be automatically configured.
179 Alternatively, you can configure the <tt>test-suite</tt> module manually.</p>
181 <!-- _______________________________________________________________________ -->
182 <h3><a name="quickregressiontests">Regression tests</a></h3>
184 <!-- _______________________________________________________________________ -->
185 <p>To run all of the LLVM regression tests, use master Makefile in
186 the <tt>llvm/test</tt> directory:</p>
188 <div class="doc_code">
196 <div class="doc_code">
202 <p>If you have <a href="http://clang.llvm.org/">Clang</a> checked out and built,
203 you can run the LLVM and Clang tests simultaneously using:</p>
207 <div class="doc_code">
213 <p>To run the tests with Valgrind (Memcheck by default), just append
214 <tt>VG=1</tt> to the commands above, e.g.:</p>
216 <div class="doc_code">
222 <p>To run individual tests or subsets of tests, you can use the 'llvm-lit'
223 script which is built as part of LLVM. For example, to run the
224 'Integer/BitCast.ll' test by itself you can run:</p>
226 <div class="doc_code">
228 % llvm-lit ~/llvm/test/Integer/BitCast.ll
232 <p>or to run all of the ARM CodeGen tests:</p>
234 <div class="doc_code">
236 % llvm-lit ~/llvm/test/CodeGen/ARM
240 <p>For more information on using the 'lit' tool, see 'llvm-lit --help' or the
245 <!-- _______________________________________________________________________ -->
246 <h3><a name="quicktestsuite">Test suite</a></h3>
247 <!-- _______________________________________________________________________ -->
251 <p>To run the comprehensive test suite (tests that compile and execute whole
252 programs), first checkout and setup the <tt>test-suite</tt> module:</p>
254 <div class="doc_code">
257 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
262 <p>and then configure and build normally as you would from the
263 <a href="http://llvm.org/docs/GettingStarted.html#quickstart">Getting Started
264 Guide</a>. This will autodetect first the built clang if you are building
265 clang, then <tt>clang</tt> in your path and finally look for <tt>llvm-gcc</tt>
268 <p>Then, run the entire test suite by running make in the <tt>test-suite</tt>
269 subdirectory of your build directory:</p>
271 <div class="doc_code">
273 % cd <i>where-you-built-llvm</i>/projects/test-suite
278 <p>Usually, running the "simple" set of tests is a good idea, and you can also
279 let it generate a report by running:</p>
281 <div class="doc_code">
283 % cd <i>where-you-built-llvm</i>/projects/test-suite
284 % gmake TEST=simple report report.html
288 <p>Any of the above commands can also be run in a subdirectory of
289 <tt>projects/test-suite</tt> to run the specified test only on the programs in
290 that subdirectory.</p>
294 <!-- _______________________________________________________________________ -->
295 <h3><a name="quickdebuginfotests">Debugging Information tests</a></h3>
297 <!-- _______________________________________________________________________ -->
300 <p> To run debugging information tests simply checkout the tests inside
301 clang/test directory. </p>
303 <div class="doc_code">
306 % svn co http://llvm.org/svn/llvm-project/debuginfo-tests/trunk debuginfo-tests
310 <p> These tests are already set up to run as part of clang regression tests.</p>
318 <!--=========================================================================-->
319 <h2><a name="rtstructure">Regression test structure</a></h2>
320 <!--=========================================================================-->
322 <p>The LLVM regression tests are driven by 'lit' and are located in
323 the <tt>llvm/test</tt> directory.
325 <p>This directory contains a large array of small tests
326 that exercise various features of LLVM and to ensure that regressions do not
327 occur. The directory is broken into several sub-directories, each focused on
328 a particular area of LLVM. A few of the important ones are:</p>
331 <li><tt>Analysis</tt>: checks Analysis passes.</li>
332 <li><tt>Archive</tt>: checks the Archive library.</li>
333 <li><tt>Assembler</tt>: checks Assembly reader/writer functionality.</li>
334 <li><tt>Bitcode</tt>: checks Bitcode reader/writer functionality.</li>
335 <li><tt>CodeGen</tt>: checks code generation and each target.</li>
336 <li><tt>Features</tt>: checks various features of the LLVM language.</li>
337 <li><tt>Linker</tt>: tests bitcode linking.</li>
338 <li><tt>Transforms</tt>: tests each of the scalar, IPO, and utility
339 transforms to ensure they make the right transformations.</li>
340 <li><tt>Verifier</tt>: tests the IR verifier.</li>
343 <!-- _______________________________________________________________________ -->
344 <h3><a name="rtcustom">Writing new regression tests</a></h3>
345 <!-- _______________________________________________________________________ -->
347 <p>The regression test structure is very simple, but does require some
348 information to be set. This information is gathered via <tt>configure</tt> and
349 is written to a file, <tt>lit.site.cfg</tt>
350 in <tt>llvm/test</tt>. The <tt>llvm/test</tt> Makefile does this work for
353 <p>In order for the regression tests to work, each directory of tests must
354 have a <tt>dg.exp</tt> file. Lit looks for this file to determine how to
355 run the tests. This file is just a Tcl script and it can do anything you want,
356 but we've standardized it for the LLVM regression tests. If you're adding a
357 directory of tests, just copy <tt>dg.exp</tt> from another directory to get
358 running. The standard <tt>dg.exp</tt> simply loads a Tcl library
359 (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt> function
360 defined in that library with a list of file names to run. The names are
361 obtained by using Tcl's glob command. Any directory that contains only
362 directories does not need the <tt>dg.exp</tt> file.</p>
364 <p>The <tt>llvm-runtests</tt> function looks at each file that is passed to
365 it and gathers any lines together that match "RUN:". These are the "RUN" lines
366 that specify how the test is to be run. So, each test script must contain
367 RUN lines if it is to do anything. If there are no RUN lines, the
368 <tt>llvm-runtests</tt> function will issue an error and the test will
371 <p>RUN lines are specified in the comments of the test program using the
372 keyword <tt>RUN</tt> followed by a colon, and lastly the command (pipeline)
373 to execute. Together, these lines form the "script" that
374 <tt>llvm-runtests</tt> executes to run the test case. The syntax of the
375 RUN lines is similar to a shell's syntax for pipelines including I/O
376 redirection and variable substitution. However, even though these lines
377 may <i>look</i> like a shell script, they are not. RUN lines are interpreted
378 directly by the Tcl <tt>exec</tt> command. They are never executed by a
379 shell. Consequently the syntax differs from normal shell script syntax in a
380 few ways. You can specify as many RUN lines as needed.</p>
382 <p>lit performs substitution on each RUN line to replace LLVM tool
383 names with the full paths to the executable built for each tool (in
384 $(LLVM_OBJ_ROOT)/$(BuildMode)/bin). This ensures that lit does not
385 invoke any stray LLVM tools in the user's path during testing.</p>
387 <p>Each RUN line is executed on its own, distinct from other lines unless
388 its last character is <tt>\</tt>. This continuation character causes the RUN
389 line to be concatenated with the next one. In this way you can build up long
390 pipelines of commands without making huge line lengths. The lines ending in
391 <tt>\</tt> are concatenated until a RUN line that doesn't end in <tt>\</tt> is
392 found. This concatenated set of RUN lines then constitutes one execution.
393 Tcl will substitute variables and arrange for the pipeline to be executed. If
394 any process in the pipeline fails, the entire line (and test case) fails too.
397 <p> Below is an example of legal RUN lines in a <tt>.ll</tt> file:</p>
399 <div class="doc_code">
401 ; RUN: llvm-as < %s | llvm-dis > %t1
402 ; RUN: llvm-dis < %s.bc-13 > %t2
407 <p>As with a Unix shell, the RUN: lines permit pipelines and I/O redirection
408 to be used. However, the usage is slightly different than for Bash. To check
409 what's legal, see the documentation for the
410 <a href="http://www.tcl.tk/man/tcl8.5/TclCmd/exec.htm#M2">Tcl exec</a>
412 <a href="http://www.tcl.tk/man/tcl8.5/tutorial/Tcl26.html">tutorial</a>.
413 The major differences are:</p>
415 <li>You can't do <tt>2>&1</tt>. That will cause Tcl to write to a
416 file named <tt>&1</tt>. Usually this is done to get stderr to go through
417 a pipe. You can do that in tcl with <tt>|&</tt> so replace this idiom:
418 <tt>... 2>&1 | grep</tt> with <tt>... |& grep</tt></li>
419 <li>You can only redirect to a file, not to another descriptor and not from
420 a here document.</li>
421 <li>tcl supports redirecting to open files with the @ syntax but you
422 shouldn't use that here.</li>
425 <p>There are some quoting rules that you must pay attention to when writing
426 your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any
427 quote characters so they will get passed to the invoked program. For
430 <div class="doc_code">
432 ... | grep 'find this string'
436 <p>This will fail because the ' characters are passed to grep. This would
437 instruction grep to look for <tt>'find</tt> in the files <tt>this</tt> and
438 <tt>string'</tt>. To avoid this use curly braces to tell Tcl that it should
439 treat everything enclosed as one value. So our example would become:</p>
441 <div class="doc_code">
443 ... | grep {find this string}
447 <p>Additionally, the characters <tt>[</tt> and <tt>]</tt> are treated
448 specially by Tcl. They tell Tcl to interpret the content as a command to
449 execute. Since these characters are often used in regular expressions this can
450 have disastrous results and cause the entire test run in a directory to fail.
451 For example, a common idiom is to look for some basicblock number:</p>
453 <div class="doc_code">
459 <p>This, however, will cause Tcl to fail because its going to try to execute
460 a program named "2-8". Instead, what you want is this:</p>
462 <div class="doc_code">
464 ... | grep {bb\[2-8\]}
468 <p>Finally, if you need to pass the <tt>\</tt> character down to a program,
469 then it must be doubled. This is another Tcl special character. So, suppose
472 <div class="doc_code">
478 <p>This will fail to match what you want (a pointer to i32). First, the
479 <tt>'</tt> do not get stripped off. Second, the <tt>\</tt> gets stripped off
480 by Tcl so what grep sees is: <tt>'i32*'</tt>. That's not likely to match
481 anything. To resolve this you must use <tt>\\</tt> and the <tt>{}</tt>, like
484 <div class="doc_code">
490 <p>If your system includes GNU <tt>grep</tt>, make sure
491 that <tt>GREP_OPTIONS</tt> is not set in your environment. Otherwise,
492 you may get invalid results (both false positives and false
497 <!-- _______________________________________________________________________ -->
498 <h3><a name="FileCheck">The FileCheck utility</a></h3>
499 <!-- _______________________________________________________________________ -->
503 <p>A powerful feature of the RUN: lines is that it allows any arbitrary commands
504 to be executed as part of the test harness. While standard (portable) unix
505 tools like 'grep' work fine on run lines, as you see above, there are a lot
506 of caveats due to interaction with Tcl syntax, and we want to make sure the
507 run lines are portable to a wide range of systems. Another major problem is
508 that grep is not very good at checking to verify that the output of a tools
509 contains a series of different output in a specific order. The FileCheck
510 tool was designed to help with these problems.</p>
512 <p>FileCheck (whose basic command line arguments are described in <a
513 href="http://llvm.org/cmds/FileCheck.html">the FileCheck man page</a> is
514 designed to read a file to check from standard input, and the set of things
515 to verify from a file specified as a command line argument. A simple example
516 of using FileCheck from a RUN line looks like this:</p>
518 <div class="doc_code">
520 ; RUN: llvm-as < %s | llc -march=x86-64 | <b>FileCheck %s</b>
524 <p>This syntax says to pipe the current file ("%s") into llvm-as, pipe that into
525 llc, then pipe the output of llc into FileCheck. This means that FileCheck will
526 be verifying its standard input (the llc output) against the filename argument
527 specified (the original .ll file specified by "%s"). To see how this works,
528 lets look at the rest of the .ll file (after the RUN line):</p>
530 <div class="doc_code">
532 define void @sub1(i32* %p, i32 %v) {
534 ; <b>CHECK: sub1:</b>
536 %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v)
540 define void @inc4(i64* %p) {
542 ; <b>CHECK: inc4:</b>
544 %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1)
550 <p>Here you can see some "CHECK:" lines specified in comments. Now you can see
551 how the file is piped into llvm-as, then llc, and the machine code output is
552 what we are verifying. FileCheck checks the machine code output to verify that
553 it matches what the "CHECK:" lines specify.</p>
555 <p>The syntax of the CHECK: lines is very simple: they are fixed strings that
556 must occur in order. FileCheck defaults to ignoring horizontal whitespace
557 differences (e.g. a space is allowed to match a tab) but otherwise, the contents
558 of the CHECK: line is required to match some thing in the test file exactly.</p>
560 <p>One nice thing about FileCheck (compared to grep) is that it allows merging
561 test cases together into logical groups. For example, because the test above
562 is checking for the "sub1:" and "inc4:" labels, it will not match unless there
563 is a "subl" in between those labels. If it existed somewhere else in the file,
564 that would not count: "grep subl" matches if subl exists anywhere in the
567 <!-- _______________________________________________________________________ -->
569 <a name="FileCheck-check-prefix">The FileCheck -check-prefix option</a>
574 <p>The FileCheck -check-prefix option allows multiple test configurations to be
575 driven from one .ll file. This is useful in many circumstances, for example,
576 testing different architectural variants with llc. Here's a simple example:</p>
578 <div class="doc_code">
580 ; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \
581 ; RUN: | <b>FileCheck %s -check-prefix=X32</b>
582 ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \
583 ; RUN: | <b>FileCheck %s -check-prefix=X64</b>
585 define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind {
586 %tmp1 = insertelement <4 x i32> %tmp, i32 %s, i32 1
587 ret <4 x i32> %tmp1
588 ; <b>X32:</b> pinsrd_1:
589 ; <b>X32:</b> pinsrd $1, 4(%esp), %xmm0
591 ; <b>X64:</b> pinsrd_1:
592 ; <b>X64:</b> pinsrd $1, %edi, %xmm0
597 <p>In this case, we're testing that we get the expected code generation with
598 both 32-bit and 64-bit code generation.</p>
602 <!-- _______________________________________________________________________ -->
604 <a name="FileCheck-CHECK-NEXT">The "CHECK-NEXT:" directive</a>
609 <p>Sometimes you want to match lines and would like to verify that matches
610 happen on exactly consecutive lines with no other lines in between them. In
611 this case, you can use CHECK: and CHECK-NEXT: directives to specify this. If
612 you specified a custom check prefix, just use "<PREFIX>-NEXT:". For
613 example, something like this works as you'd expect:</p>
615 <div class="doc_code">
617 define void @t2(<2 x double>* %r, <2 x double>* %A, double %B) {
618 %tmp3 = load <2 x double>* %A, align 16
619 %tmp7 = insertelement <2 x double> undef, double %B, i32 0
620 %tmp9 = shufflevector <2 x double> %tmp3,
621 <2 x double> %tmp7,
622 <2 x i32> < i32 0, i32 2 >
623 store <2 x double> %tmp9, <2 x double>* %r, align 16
627 ; <b>CHECK:</b> movl 8(%esp), %eax
628 ; <b>CHECK-NEXT:</b> movapd (%eax), %xmm0
629 ; <b>CHECK-NEXT:</b> movhpd 12(%esp), %xmm0
630 ; <b>CHECK-NEXT:</b> movl 4(%esp), %eax
631 ; <b>CHECK-NEXT:</b> movapd %xmm0, (%eax)
632 ; <b>CHECK-NEXT:</b> ret
637 <p>CHECK-NEXT: directives reject the input unless there is exactly one newline
638 between it an the previous directive. A CHECK-NEXT cannot be the first
639 directive in a file.</p>
643 <!-- _______________________________________________________________________ -->
645 <a name="FileCheck-CHECK-NOT">The "CHECK-NOT:" directive</a>
650 <p>The CHECK-NOT: directive is used to verify that a string doesn't occur
651 between two matches (or the first match and the beginning of the file). For
652 example, to verify that a load is removed by a transformation, a test like this
655 <div class="doc_code">
657 define i8 @coerce_offset0(i32 %V, i32* %P) {
658 store i32 %V, i32* %P
660 %P2 = bitcast i32* %P to i8*
661 %P3 = getelementptr i8* %P2, i32 2
665 ; <b>CHECK:</b> @coerce_offset0
666 ; <b>CHECK-NOT:</b> load
667 ; <b>CHECK:</b> ret i8
674 <!-- _______________________________________________________________________ -->
676 <a name="FileCheck-Matching">FileCheck Pattern Matching Syntax</a>
681 <p>The CHECK: and CHECK-NOT: directives both take a pattern to match. For most
682 uses of FileCheck, fixed string matching is perfectly sufficient. For some
683 things, a more flexible form of matching is desired. To support this, FileCheck
684 allows you to specify regular expressions in matching strings, surrounded by
685 double braces: <b>{{yourregex}}</b>. Because we want to use fixed string
686 matching for a majority of what we do, FileCheck has been designed to support
687 mixing and matching fixed string matching with regular expressions. This allows
688 you to write things like this:</p>
690 <div class="doc_code">
692 ; CHECK: movhpd <b>{{[0-9]+}}</b>(%esp), <b>{{%xmm[0-7]}}</b>
696 <p>In this case, any offset from the ESP register will be allowed, and any xmm
697 register will be allowed.</p>
699 <p>Because regular expressions are enclosed with double braces, they are
700 visually distinct, and you don't need to use escape characters within the double
701 braces like you would in C. In the rare case that you want to match double
702 braces explicitly from the input, you can use something ugly like
703 <b>{{[{][{]}}</b> as your pattern.</p>
707 <!-- _______________________________________________________________________ -->
709 <a name="FileCheck-Variables">FileCheck Variables</a>
714 <p>It is often useful to match a pattern and then verify that it occurs again
715 later in the file. For codegen tests, this can be useful to allow any register,
716 but verify that that register is used consistently later. To do this, FileCheck
717 allows named variables to be defined and substituted into patterns. Here is a
720 <div class="doc_code">
723 ; CHECK: notw <b>[[REGISTER:%[a-z]+]]</b>
724 ; CHECK: andw {{.*}}<b>[[REGISTER]]</b>
728 <p>The first check line matches a regex (<tt>%[a-z]+</tt>) and captures it into
729 the variables "REGISTER". The second line verifies that whatever is in REGISTER
730 occurs later in the file after an "andw". FileCheck variable references are
731 always contained in <tt>[[ ]]</tt> pairs, are named, and their names can be
732 formed with the regex "<tt>[a-zA-Z][a-zA-Z0-9]*</tt>". If a colon follows the
733 name, then it is a definition of the variable, if not, it is a use.</p>
735 <p>FileCheck variables can be defined multiple times, and uses always get the
736 latest value. Note that variables are all read at the start of a "CHECK" line
737 and are all defined at the end. This means that if you have something like
738 "<tt>CHECK: [[XYZ:.*]]x[[XYZ]]</tt>" that the check line will read the previous
739 value of the XYZ variable and define a new one after the match is performed. If
740 you need to do something like this you can probably take advantage of the fact
741 that FileCheck is not actually line-oriented when it matches, this allows you to
742 define two separate CHECK lines that match on the same line.
749 <!-- _______________________________________________________________________ -->
750 <h3><a name="rtvars">Variables and substitutions</a></h3>
751 <!-- _______________________________________________________________________ -->
753 <p>With a RUN line there are a number of substitutions that are permitted. In
754 general, any Tcl variable that is available in the <tt>substitute</tt>
755 function (in <tt>test/lib/llvm.exp</tt>) can be substituted into a RUN line.
756 To make a substitution just write the variable's name preceded by a $.
757 Additionally, for compatibility reasons with previous versions of the test
758 library, certain names can be accessed with an alternate syntax: a % prefix.
759 These alternates are deprecated and may go away in a future version.
761 <p>Here are the available variable names. The alternate syntax is listed in
764 <dl style="margin-left: 25px">
765 <dt><b>$test</b> (%s)</dt>
766 <dd>The full path to the test case's source. This is suitable for passing
767 on the command line as the input to an llvm tool.</dd>
769 <dt><b>$srcdir</b></dt>
770 <dd>The source directory from where the "<tt>make check</tt>" was run.</dd>
772 <dt><b>objdir</b></dt>
773 <dd>The object directory that corresponds to the <tt>$srcdir</tt>.</dd>
775 <dt><b>subdir</b></dt>
776 <dd>A partial path from the <tt>test</tt> directory that contains the
777 sub-directory that contains the test source being executed.</dd>
779 <dt><b>srcroot</b></dt>
780 <dd>The root directory of the LLVM src tree.</dd>
782 <dt><b>objroot</b></dt>
783 <dd>The root directory of the LLVM object tree. This could be the same
787 <dd>The path to the directory that contains the test case source. This is
788 for locating any supporting files that are not generated by the test, but
789 used by the test.</dd>
792 <dd>The path to a temporary file name that could be used for this test case.
793 The file name won't conflict with other test cases. You can append to it if
794 you need multiple temporaries. This is useful as the destination of some
795 redirected output.</dd>
797 <dt><b>target_triplet</b> (%target_triplet)</dt>
798 <dd>The target triplet that corresponds to the current host machine (the one
799 running the test cases). This should probably be called "host".<dd>
801 <dt><b>link</b> (%link)</dt>
802 <dd>This full link command used to link LLVM executables. This has all the
803 configured -I, -L and -l options.</dd>
805 <dt><b>shlibext</b> (%shlibext)</dt>
806 <dd>The suffix for the host platforms share library (dll) files. This
807 includes the period as the first character.</dd>
809 <p>To add more variables, two things need to be changed. First, add a line in
810 the <tt>test/Makefile</tt> that creates the <tt>site.exp</tt> file. This will
811 "set" the variable as a global in the site.exp file. Second, in the
812 <tt>test/lib/llvm.exp</tt> file, in the substitute proc, add the variable name
813 to the list of "global" declarations at the beginning of the proc. That's it,
814 the variable can then be used in test scripts.</p>
817 <!-- _______________________________________________________________________ -->
818 <h3><a name="rtfeatures">Other Features</a></h3>
819 <!-- _______________________________________________________________________ -->
821 <p>To make RUN line writing easier, there are several shell scripts located
822 in the <tt>llvm/test/Scripts</tt> directory. This directory is in the PATH
823 when running tests, so you can just call these scripts using their name. For
826 <dt><b>ignore</b></dt>
827 <dd>This script runs its arguments and then always returns 0. This is useful
828 in cases where the test needs to cause a tool to generate an error (e.g. to
829 check the error output). However, any program in a pipeline that returns a
830 non-zero result will cause the test to fail. This script overcomes that
831 issue and nicely documents that the test case is purposefully ignoring the
832 result code of the tool</dd>
835 <dd>This script runs its arguments and then inverts the result code from
836 it. Zero result codes become 1. Non-zero result codes become 0. This is
837 useful to invert the result of a grep. For example "not grep X" means
838 succeed only if you don't find X in the input.</dd>
841 <p>Sometimes it is necessary to mark a test case as "expected fail" or XFAIL.
842 You can easily mark a test as XFAIL just by including <tt>XFAIL: </tt> on a
843 line near the top of the file. This signals that the test case should succeed
844 if the test fails. Such test cases are counted separately by the testing tool. To
845 specify an expected fail, use the XFAIL keyword in the comments of the test
846 program followed by a colon and one or more regular expressions (separated by
847 a comma). The regular expressions allow you to XFAIL the test conditionally by
848 host platform. The regular expressions following the : are matched against the
849 target triplet for the host machine. If there is a match, the test is expected
850 to fail. If not, the test is expected to succeed. To XFAIL everywhere just
851 specify <tt>XFAIL: *</tt>. Here is an example of an <tt>XFAIL</tt> line:</p>
853 <div class="doc_code">
859 <p>To make the output more useful, the <tt>llvm_runtest</tt> function wil
860 scan the lines of the test case for ones that contain a pattern that matches
861 PR[0-9]+. This is the syntax for specifying a PR (Problem Report) number that
862 is related to the test case. The number after "PR" specifies the LLVM bugzilla
863 number. When a PR number is specified, it will be used in the pass/fail
864 reporting. This is useful to quickly get some context when a test fails.</p>
866 <p>Finally, any line that contains "END." will cause the special
867 interpretation of lines to terminate. This is generally done right after the
868 last RUN: line. This has two side effects: (a) it prevents special
869 interpretation of lines that are part of the test program, not the
870 instructions to the test case, and (b) it speeds things up for really big test
871 cases by avoiding interpretation of the remainder of the file.</p>
877 <!--=========================================================================-->
878 <h2><a name="testsuitestructure">Test suite Structure</a></h2>
879 <!--=========================================================================-->
883 <p>The <tt>test-suite</tt> module contains a number of programs that can be compiled
884 with LLVM and executed. These programs are compiled using the native compiler
885 and various LLVM backends. The output from the program compiled with the
886 native compiler is assumed correct; the results from the other programs are
887 compared to the native program output and pass if they match.</p>
889 <p>When executing tests, it is usually a good idea to start out with a subset of
890 the available tests or programs. This makes test run times smaller at first and
891 later on this is useful to investigate individual test failures. To run some
892 test only on a subset of programs, simply change directory to the programs you
893 want tested and run <tt>gmake</tt> there. Alternatively, you can run a different
894 test using the <tt>TEST</tt> variable to change what tests or run on the
895 selected programs (see below for more info).</p>
897 <p>In addition for testing correctness, the <tt>test-suite</tt> directory also
898 performs timing tests of various LLVM optimizations. It also records
899 compilation times for the compilers and the JIT. This information can be
900 used to compare the effectiveness of LLVM's optimizations and code
903 <p><tt>test-suite</tt> tests are divided into three types of tests: MultiSource,
904 SingleSource, and External.</p>
907 <li><tt>test-suite/SingleSource</tt>
908 <p>The SingleSource directory contains test programs that are only a single
909 source file in size. These are usually small benchmark programs or small
910 programs that calculate a particular value. Several such programs are grouped
911 together in each directory.</p></li>
913 <li><tt>test-suite/MultiSource</tt>
914 <p>The MultiSource directory contains subdirectories which contain entire
915 programs with multiple source files. Large benchmarks and whole applications
918 <li><tt>test-suite/External</tt>
919 <p>The External directory contains Makefiles for building code that is external
920 to (i.e., not distributed with) LLVM. The most prominent members of this
921 directory are the SPEC 95 and SPEC 2000 benchmark suites. The <tt>External</tt>
922 directory does not contain these actual tests, but only the Makefiles that know
923 how to properly compile these programs from somewhere else. The presence and
924 location of these external programs is configured by the test-suite
925 <tt>configure</tt> script.</p></li>
928 <p>Each tree is then subdivided into several categories, including applications,
929 benchmarks, regression tests, code that is strange grammatically, etc. These
930 organizations should be relatively self explanatory.</p>
932 <p>Some tests are known to fail. Some are bugs that we have not fixed yet;
933 others are features that we haven't added yet (or may never add). In the
934 regression tests, the result for such tests will be XFAIL (eXpected FAILure).
935 In this way, you can tell the difference between an expected and unexpected
938 <p>The tests in the test suite have no such feature at this time. If the
939 test passes, only warnings and other miscellaneous output will be generated. If
940 a test fails, a large <program> FAILED message will be displayed. This
941 will help you separate benign warnings from actual test failures.</p>
945 <!--=========================================================================-->
946 <h2><a name="testsuiterun">Running the test suite</a></h2>
947 <!--=========================================================================-->
951 <p>First, all tests are executed within the LLVM object directory tree. They
952 <i>are not</i> executed inside of the LLVM source tree. This is because the
953 test suite creates temporary files during execution.</p>
955 <p>To run the test suite, you need to use the following steps:</p>
958 <li><tt>cd</tt> into the <tt>llvm/projects</tt> directory in your source tree.
961 <li><p>Check out the <tt>test-suite</tt> module with:</p>
963 <div class="doc_code">
965 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
968 <p>This will get the test suite into <tt>llvm/projects/test-suite</tt>.</p>
970 <li><p>Configure and build <tt>llvm</tt>.</p></li>
971 <li><p>Configure and build <tt>llvm-gcc</tt>.</p></li>
972 <li><p>Install <tt>llvm-gcc</tt> somewhere.</p></li>
973 <li><p><em>Re-configure</em> <tt>llvm</tt> from the top level of
974 each build tree (LLVM object directory tree) in which you want
975 to run the test suite, just as you do before building LLVM.</p>
976 <p>During the <em>re-configuration</em>, you must either: (1)
977 have <tt>llvm-gcc</tt> you just built in your path, or (2)
978 specify the directory where your just-built <tt>llvm-gcc</tt> is
979 installed using <tt>--with-llvmgccdir=$LLVM_GCC_DIR</tt>.</p>
980 <p>You must also tell the configure machinery that the test suite
981 is available so it can be configured for your build tree:</p>
982 <div class="doc_code">
984 % cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure [--with-llvmgccdir=$LLVM_GCC_DIR]
987 <p>[Remember that <tt>$LLVM_GCC_DIR</tt> is the directory where you
988 <em>installed</em> llvm-gcc, not its src or obj directory.]</p>
991 <li><p>You can now run the test suite from your build tree as follows:</p>
992 <div class="doc_code">
994 % cd $LLVM_OBJ_ROOT/projects/test-suite
1000 <p>Note that the second and third steps only need to be done once. After you
1001 have the suite checked out and configured, you don't need to do it again (unless
1002 the test code or configure script changes).</p>
1004 <!-- _______________________________________________________________________ -->
1006 <a name="testsuiteexternal">Configuring External Tests</a>
1008 <!-- _______________________________________________________________________ -->
1011 <p>In order to run the External tests in the <tt>test-suite</tt>
1012 module, you must specify <i>--with-externals</i>. This
1013 must be done during the <em>re-configuration</em> step (see above),
1014 and the <tt>llvm</tt> re-configuration must recognize the
1015 previously-built <tt>llvm-gcc</tt>. If any of these is missing or
1016 neglected, the External tests won't work.</p>
1018 <dt><i>--with-externals</i></dt>
1019 <dt><i>--with-externals=<<tt>directory</tt>></i></dt>
1021 This tells LLVM where to find any external tests. They are expected to be
1022 in specifically named subdirectories of <<tt>directory</tt>>.
1023 If <tt>directory</tt> is left unspecified,
1024 <tt>configure</tt> uses the default value
1025 <tt>/home/vadve/shared/benchmarks/speccpu2000/benchspec</tt>.
1026 Subdirectory names known to LLVM include:
1029 <dt>speccpu2000</dt>
1030 <dt>speccpu2006</dt>
1033 Others are added from time to time, and can be determined from
1037 <!-- _______________________________________________________________________ -->
1039 <a name="testsuitetests">Running different tests</a>
1041 <!-- _______________________________________________________________________ -->
1043 <p>In addition to the regular "whole program" tests, the <tt>test-suite</tt>
1044 module also provides a mechanism for compiling the programs in different ways.
1045 If the variable TEST is defined on the <tt>gmake</tt> command line, the test system will
1046 include a Makefile named <tt>TEST.<value of TEST variable>.Makefile</tt>.
1047 This Makefile can modify build rules to yield different results.</p>
1049 <p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to
1050 create the nightly test reports. To run the nightly tests, run <tt>gmake
1051 TEST=nightly</tt>.</p>
1053 <p>There are several TEST Makefiles available in the tree. Some of them are
1054 designed for internal LLVM research and will not work outside of the LLVM
1055 research group. They may still be valuable, however, as a guide to writing your
1056 own TEST Makefile for any optimization or analysis passes that you develop with
1061 <!-- _______________________________________________________________________ -->
1063 <a name="testsuiteoutput">Generating test output</a>
1065 <!-- _______________________________________________________________________ -->
1067 <p>There are a number of ways to run the tests and generate output. The most
1068 simple one is simply running <tt>gmake</tt> with no arguments. This will
1069 compile and run all programs in the tree using a number of different methods
1070 and compare results. Any failures are reported in the output, but are likely
1071 drowned in the other output. Passes are not reported explicitely.</p>
1073 <p>Somewhat better is running <tt>gmake TEST=sometest test</tt>, which runs
1074 the specified test and usually adds per-program summaries to the output
1075 (depending on which sometest you use). For example, the <tt>nightly</tt> test
1076 explicitely outputs TEST-PASS or TEST-FAIL for every test after each program.
1077 Though these lines are still drowned in the output, it's easy to grep the
1078 output logs in the Output directories.</p>
1080 <p>Even better are the <tt>report</tt> and <tt>report.format</tt> targets
1081 (where <tt>format</tt> is one of <tt>html</tt>, <tt>csv</tt>, <tt>text</tt> or
1082 <tt>graphs</tt>). The exact contents of the report are dependent on which
1083 <tt>TEST</tt> you are running, but the text results are always shown at the
1084 end of the run and the results are always stored in the
1085 <tt>report.<type>.format</tt> file (when running with
1086 <tt>TEST=<type></tt>).
1088 The <tt>report</tt> also generate a file called
1089 <tt>report.<type>.raw.out</tt> containing the output of the entire test
1093 <!-- _______________________________________________________________________ -->
1095 <a name="testsuitecustom">Writing custom tests for the test suite</a>
1097 <!-- _______________________________________________________________________ -->
1101 <p>Assuming you can run the test suite, (e.g. "<tt>gmake TEST=nightly report</tt>"
1102 should work), it is really easy to run optimizations or code generator
1103 components against every program in the tree, collecting statistics or running
1104 custom checks for correctness. At base, this is how the nightly tester works,
1105 it's just one example of a general framework.</p>
1107 <p>Lets say that you have an LLVM optimization pass, and you want to see how
1108 many times it triggers. First thing you should do is add an LLVM
1109 <a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which
1110 will tally counts of things you care about.</p>
1112 <p>Following this, you can set up a test and a report that collects these and
1113 formats them for easy viewing. This consists of two files, a
1114 "<tt>test-suite/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
1115 test) and a "<tt>test-suite/TEST.XXX.report</tt>" file that indicates how to
1116 format the output into a table. There are many example reports of various
1117 levels of sophistication included with the test suite, and the framework is very
1120 <p>If you are interested in testing an optimization pass, check out the
1121 "libcalls" test as an example. It can be run like this:<p>
1123 <div class="doc_code">
1125 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
1126 % make TEST=libcalls report
1130 <p>This will do a bunch of stuff, then eventually print a table like this:</p>
1132 <div class="doc_code">
1134 Name | total | #exit |
1136 FreeBench/analyzer/analyzer | 51 | 6 |
1137 FreeBench/fourinarow/fourinarow | 1 | 1 |
1138 FreeBench/neural/neural | 19 | 9 |
1139 FreeBench/pifft/pifft | 5 | 3 |
1140 MallocBench/cfrac/cfrac | 1 | * |
1141 MallocBench/espresso/espresso | 52 | 12 |
1142 MallocBench/gs/gs | 4 | * |
1143 Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
1144 Prolangs-C/agrep/agrep | 33 | 12 |
1145 Prolangs-C/allroots/allroots | * | * |
1146 Prolangs-C/assembler/assembler | 47 | * |
1147 Prolangs-C/bison/mybison | 74 | * |
1152 <p>This basically is grepping the -stats output and displaying it in a table.
1153 You can also use the "TEST=libcalls report.html" target to get the table in HTML
1154 form, similarly for report.csv and report.tex.</p>
1156 <p>The source for this is in test-suite/TEST.libcalls.*. The format is pretty
1157 simple: the Makefile indicates how to run the test (in this case,
1158 "<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
1159 each column of the output. The first value is the header for the column and the
1160 second is the regex to grep the output of the command for. There are lots of
1161 example reports that can do fancy stuff.</p>
1167 <!-- *********************************************************************** -->
1171 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1172 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1173 <a href="http://validator.w3.org/check/referer"><img
1174 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1176 John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner<br>
1177 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
1178 Last modified: $Date$