| blob | b048b72485a0b08ab3bf3f54c51069dce85b7523 |
3 <html>
4 <head>
7 </head>
8 <body>
11 LLVM Testing Infrastructure Guide
12 </div>
14 <ol>
18 <ul>
22 </ul>
23 </li>
25 <ul>
29 </ul>
30 </li>
32 <ul>
37 </ul>
38 </li>
41 <ul>
46 </ul>
47 </li>
48 </ol>
52 </div>
54 <!--=========================================================================-->
56 <!--=========================================================================-->
60 <p>This document is the reference manual for the LLVM testing infrastructure. It
61 documents the structure of the LLVM testing infrastructure, the tools needed to
62 use it, and how to add and run tests.</p>
64 </div>
66 <!--=========================================================================-->
68 <!--=========================================================================-->
72 <p>In order to use the LLVM testing infrastructure, you will need all of the
73 software required to build LLVM, as well
76 </div>
78 <!--=========================================================================-->
80 <!--=========================================================================-->
84 <p>The LLVM testing infrastructure contains two major categories of tests:
85 regression tests and whole programs. The regression tests are contained inside
87 pass -- they should be run before every commit. The whole programs tests are
89 in subversion.
90 </p>
92 </div>
94 <!-- _______________________________________________________________________ -->
96 <!-- _______________________________________________________________________ -->
100 <p>The regression tests are small pieces of code that test a specific feature of
101 LLVM or trigger a specific bug in LLVM. They are usually written in LLVM
102 assembly language, but can be written in other languages if the test targets a
105 tests are driven by the 'lit' testing tool, which is part of LLVM.</p>
107 <p>These code fragments are not complete programs. The code generated
108 from them is never executed to determine correct behavior.</p>
111 directory.</p>
113 <p>Typically when a bug is found in LLVM, a regression test containing
114 just enough code to reproduce the problem should be written and placed
115 somewhere underneath this directory. In most cases, this will be a small
116 piece of LLVM assembly language code, often distilled from an actual
117 application or benchmark.</p>
119 </div>
121 <!-- _______________________________________________________________________ -->
123 <!-- _______________________________________________________________________ -->
127 <p>The test suite contains whole programs, which are pieces of
128 code which can be compiled and linked into a stand-alone program that can be
129 executed. These programs are generally written in high level languages such as
130 C or C++, but sometimes they are written straight in LLVM assembly.</p>
132 <p>These programs are compiled and then executed using several different
133 methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation,
134 etc). The output of these programs is compared to ensure that LLVM is compiling
135 the program correctly.</p>
137 <p>In addition to compiling and executing programs, whole program tests serve as
138 a way of benchmarking LLVM performance, both in terms of the efficiency of the
139 programs generated as well as the speed with which LLVM compiles, optimizes, and
140 generates code.</p>
144 </div>
146 <!-- _______________________________________________________________________ -->
148 tests</a></div>
149 <!-- _______________________________________________________________________ -->
153 <p>The test suite contains tests to check quality of debugging information.
154 The test are written in C based languages or in LLVM assembly language. </p>
156 <p>These tests are compiled and run under a debugger. The debugger output
157 is checked to validate of debugging information. See README.txt in the
158 test suite for more information . This test suite is located in the
161 </div>
163 <!--=========================================================================-->
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
172 The more comprehensive test suite that includes whole
175 than the default "test-suite", for then the test suite will be run every time
181 <!-- _______________________________________________________________________ -->
183 <!-- _______________________________________________________________________ -->
184 <p>To run all of the LLVM regression tests, use master Makefile in
188 <pre>
189 % gmake -C llvm/test
190 </pre>
191 </div>
196 <pre>
197 % gmake check
198 </pre>
199 </div>
202 you can run the LLVM and Clang tests simultaneously using:</p>
207 <pre>
208 % gmake check-all
209 </pre>
210 </div>
212 <p>To run the tests with Valgrind (Memcheck by default), just append
216 <pre>
217 % gmake check VG=1
218 </pre>
219 </div>
221 <p>To run individual tests or subsets of tests, you can use the 'llvm-lit'
222 script which is built as part of LLVM. For example, to run the
223 'Integer/BitCast.ll' test by itself you can run:</p>
226 <pre>
227 % llvm-lit ~/llvm/test/Integer/BitCast.ll
228 </pre>
229 </div>
234 <pre>
235 % llvm-lit ~/llvm/test/CodeGen/ARM
236 </pre>
237 </div>
239 <p>For more information on using the 'lit' tool, see 'llvm-lit --help' or the
240 'lit' man page.</p>
242 <!-- _______________________________________________________________________ -->
244 <!-- _______________________________________________________________________ -->
246 <p>To run the comprehensive test suite (tests that compile and execute whole
250 <pre>
251 % cd llvm/projects
252 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
253 % cd ..
254 % ./configure --with-llvmgccdir=$LLVM_GCC_DIR
255 </pre>
256 </div>
264 respectively. If this is not the case,
266 executable's location.</p>
269 directory:</p>
272 <pre>
273 % cd projects/test-suite
274 % gmake
275 </pre>
276 </div>
279 let it generate a report by running:</p>
282 <pre>
283 % cd projects/test-suite
284 % gmake TEST=nightly report report.html
285 </pre>
286 </div>
288 <p>Any of the above commands can also be run in a subdirectory of
290 that subdirectory.</p>
292 </div>
294 <!-- _______________________________________________________________________ -->
296 tests</a></div>
297 <!-- _______________________________________________________________________ -->
299 <p> To run debugging information tests simply checkout the tests inside
300 clang/test directory. </p>
303 <pre>
304 %cd clang/test
305 % svn co http://llvm.org/svn/llvm-project/debuginfo-tests/trunk debuginfo-tests
306 </pre>
307 </div>
311 </div>
313 <!--=========================================================================-->
315 <!--=========================================================================-->
317 <p>The LLVM regression tests are driven by 'lit' and are located in
320 <p>This directory contains a large array of small tests
321 that exercise various features of LLVM and to ensure that regressions do not
322 occur. The directory is broken into several sub-directories, each focused on
323 a particular area of LLVM. A few of the important ones are:</p>
325 <ul>
334 transforms to ensure they make the right transformations.</li>
336 </ul>
338 </div>
340 <!-- _______________________________________________________________________ -->
342 <!-- _______________________________________________________________________ -->
344 <p>The regression test structure is very simple, but does require some
348 you.</p>
350 <p>In order for the regression tests to work, each directory of tests must
352 run the tests. This file is just a Tcl script and it can do anything you want,
353 but we've standardized it for the LLVM regression tests. If you're adding a
357 defined in that library with a list of file names to run. The names are
358 obtained by using Tcl's glob command. Any directory that contains only
363 that specify how the test is to be run. So, each test script must contain
364 RUN lines if it is to do anything. If there are no RUN lines, the
366 fail.</p>
368 <p>RUN lines are specified in the comments of the test program using the
370 to execute. Together, these lines form the "script" that
372 RUN lines is similar to a shell's syntax for pipelines including I/O
373 redirection and variable substitution. However, even though these lines
376 shell. Consequently the syntax differs from normal shell script syntax in a
377 few ways. You can specify as many RUN lines as needed.</p>
379 <p>lit performs substitution on each RUN line to replace LLVM tool
380 names with the full paths to the executable built for each tool (in
381 $(LLVM_OBJ_ROOT)/$(BuildMode)/bin). This ensures that lit does not
382 invoke any stray LLVM tools in the user's path during testing.</p>
384 <p>Each RUN line is executed on its own, distinct from other lines unless
386 line to be concatenated with the next one. In this way you can build up long
387 pipelines of commands without making huge line lengths. The lines ending in
389 found. This concatenated set of RUN lines then constitutes one execution.
390 Tcl will substitute variables and arrange for the pipeline to be executed. If
391 any process in the pipeline fails, the entire line (and test case) fails too.
392 </p>
397 <pre>
400 ; RUN: diff %t1 %t2
401 </pre>
402 </div>
404 <p>As with a Unix shell, the RUN: lines permit pipelines and I/O redirection
405 to be used. However, the usage is slightly different than for Bash. To check
406 what's legal, see the documentation for the
408 command and the
410 The major differences are:</p>
411 <ul>
416 <li>You can only redirect to a file, not to another descriptor and not from
417 a here document.</li>
418 <li>tcl supports redirecting to open files with the @ syntax but you
419 shouldn't use that here.</li>
420 </ul>
422 <p>There are some quoting rules that you must pay attention to when writing
423 your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any
424 quote characters so they will get passed to the invoked program. For
425 example:</p>
428 <pre>
429 ... | grep 'find this string'
430 </pre>
431 </div>
433 <p>This will fail because the ' characters are passed to grep. This would
436 treat everything enclosed as one value. So our example would become:</p>
439 <pre>
440 ... | grep {find this string}
441 </pre>
442 </div>
445 specially by Tcl. They tell Tcl to interpret the content as a command to
446 execute. Since these characters are often used in regular expressions this can
447 have disastrous results and cause the entire test run in a directory to fail.
448 For example, a common idiom is to look for some basicblock number:</p>
451 <pre>
453 </pre>
454 </div>
456 <p>This, however, will cause Tcl to fail because its going to try to execute
460 <pre>
462 </pre>
463 </div>
466 then it must be doubled. This is another Tcl special character. So, suppose
467 you had:
470 <pre>
471 ... | grep 'i32\*'
472 </pre>
473 </div>
475 <p>This will fail to match what you want (a pointer to i32). First, the
479 this:</p>
482 <pre>
483 ... | grep {i32\\*}
484 </pre>
485 </div>
489 you may get invalid results (both false positives and false
490 negatives).</p>
492 </div>
494 <!-- _______________________________________________________________________ -->
496 <!-- _______________________________________________________________________ -->
500 <p>A powerful feature of the RUN: lines is that it allows any arbitrary commands
501 to be executed as part of the test harness. While standard (portable) unix
502 tools like 'grep' work fine on run lines, as you see above, there are a lot
503 of caveats due to interaction with Tcl syntax, and we want to make sure the
504 run lines are portable to a wide range of systems. Another major problem is
505 that grep is not very good at checking to verify that the output of a tools
506 contains a series of different output in a specific order. The FileCheck
507 tool was designed to help with these problems.</p>
511 designed to read a file to check from standard input, and the set of things
512 to verify from a file specified as a command line argument. A simple example
513 of using FileCheck from a RUN line looks like this:</p>
516 <pre>
518 </pre>
519 </div>
522 llc, then pipe the output of llc into FileCheck. This means that FileCheck will
523 be verifying its standard input (the llc output) against the filename argument
524 specified (the original .ll file specified by "%s"). To see how this works,
525 lets look at the rest of the .ll file (after the RUN line):</p>
528 <pre>
529 define void @sub1(i32* %p, i32 %v) {
530 entry:
533 %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v)
534 ret void
535 }
537 define void @inc4(i64* %p) {
538 entry:
542 ret void
543 }
544 </pre>
545 </div>
548 how the file is piped into llvm-as, then llc, and the machine code output is
549 what we are verifying. FileCheck checks the machine code output to verify that
552 <p>The syntax of the CHECK: lines is very simple: they are fixed strings that
553 must occur in order. FileCheck defaults to ignoring horizontal whitespace
554 differences (e.g. a space is allowed to match a tab) but otherwise, the contents
555 of the CHECK: line is required to match some thing in the test file exactly.</p>
557 <p>One nice thing about FileCheck (compared to grep) is that it allows merging
558 test cases together into logical groups. For example, because the test above
560 is a "subl" in between those labels. If it existed somewhere else in the file,
561 that would not count: "grep subl" matches if subl exists anywhere in the
562 file.</p>
564 </div>
566 <!-- _______________________________________________________________________ -->
572 <p>The FileCheck -check-prefix option allows multiple test configurations to be
573 driven from one .ll file. This is useful in many circumstances, for example,
574 testing different architectural variants with llc. Here's a simple example:</p>
577 <pre>
578 ; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \
580 ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \
591 }
592 </pre>
593 </div>
595 <p>In this case, we're testing that we get the expected code generation with
598 </div>
600 <!-- _______________________________________________________________________ -->
606 <p>Sometimes you want to match lines and would like to verify that matches
607 happen on exactly consecutive lines with no other lines in between them. In
608 this case, you can use CHECK: and CHECK-NEXT: directives to specify this. If
609 you specified a custom check prefix, just use "<PREFIX>-NEXT:". For
610 example, something like this works as you'd expect:</p>
613 <pre>
621 ret void
630 }
631 </pre>
632 </div>
634 <p>CHECK-NEXT: directives reject the input unless there is exactly one newline
635 between it an the previous directive. A CHECK-NEXT cannot be the first
636 directive in a file.</p>
638 </div>
640 <!-- _______________________________________________________________________ -->
646 <p>The CHECK-NOT: directive is used to verify that a string doesn't occur
647 between two matches (or the first match and the beginning of the file). For
648 example, to verify that a load is removed by a transformation, a test like this
649 can be used:</p>
652 <pre>
653 define i8 @coerce_offset0(i32 %V, i32* %P) {
654 store i32 %V, i32* %P
656 %P2 = bitcast i32* %P to i8*
657 %P3 = getelementptr i8* %P2, i32 2
659 %A = load i8* %P3
660 ret i8 %A
664 }
665 </pre>
666 </div>
668 </div>
670 <!-- _______________________________________________________________________ -->
676 <p>The CHECK: and CHECK-NOT: directives both take a pattern to match. For most
677 uses of FileCheck, fixed string matching is perfectly sufficient. For some
678 things, a more flexible form of matching is desired. To support this, FileCheck
679 allows you to specify regular expressions in matching strings, surrounded by
681 matching for a majority of what we do, FileCheck has been designed to support
682 mixing and matching fixed string matching with regular expressions. This allows
683 you to write things like this:</p>
686 <pre>
688 </pre>
689 </div>
691 <p>In this case, any offset from the ESP register will be allowed, and any xmm
692 register will be allowed.</p>
694 <p>Because regular expressions are enclosed with double braces, they are
695 visually distinct, and you don't need to use escape characters within the double
696 braces like you would in C. In the rare case that you want to match double
697 braces explicitly from the input, you can use something ugly like
700 </div>
702 <!-- _______________________________________________________________________ -->
708 <p>It is often useful to match a pattern and then verify that it occurs again
709 later in the file. For codegen tests, this can be useful to allow any register,
710 but verify that that register is used consistently later. To do this, FileCheck
711 allows named variables to be defined and substituted into patterns. Here is a
712 simple example:</p>
715 <pre>
716 ; CHECK: test5:
719 </pre>
720 </div>
723 the variables "REGISTER". The second line verifies that whatever is in REGISTER
724 occurs later in the file after an "andw". FileCheck variable references are
726 formed with the regex "<tt>[a-zA-Z][a-zA-Z0-9]*</tt>". If a colon follows the
727 name, then it is a definition of the variable, if not, it is a use.</p>
729 <p>FileCheck variables can be defined multiple times, and uses always get the
730 latest value. Note that variables are all read at the start of a "CHECK" line
731 and are all defined at the end. This means that if you have something like
732 "<tt>CHECK: [[XYZ:.*]]x[[XYZ]]</tt>" that the check line will read the previous
733 value of the XYZ variable and define a new one after the match is performed. If
734 you need to do something like this you can probably take advantage of the fact
735 that FileCheck is not actually line-oriented when it matches, this allows you to
736 define two separate CHECK lines that match on the same line.
737 </p>
739 </div>
741 <!-- _______________________________________________________________________ -->
743 substitutions</a></div>
744 <!-- _______________________________________________________________________ -->
746 <p>With a RUN line there are a number of substitutions that are permitted. In
749 To make a substitution just write the variable's name preceded by a $.
750 Additionally, for compatibility reasons with previous versions of the test
751 library, certain names can be accessed with an alternate syntax: a % prefix.
752 These alternates are deprecated and may go away in a future version.
753 </p>
754 <p>Here are the available variable names. The alternate syntax is listed in
755 parentheses.</p>
759 <dd>The full path to the test case's source. This is suitable for passing
760 on the command line as the input to an llvm tool.</dd>
770 sub-directory that contains the test source being executed.</dd>
776 <dd>The root directory of the LLVM object tree. This could be the same
777 as the srcroot.</dd>
780 <dd>The path to the directory that contains the test case source. This is
781 for locating any supporting files that are not generated by the test, but
782 used by the test.</dd>
785 <dd>The path to a temporary file name that could be used for this test case.
786 The file name won't conflict with other test cases. You can append to it if
787 you need multiple temporaries. This is useful as the destination of some
788 redirected output.</dd>
794 <dd>The target triplet that corresponds to the current host machine (the one
799 configured LLVM environment</dd>
803 configured LLVM environment</dd>
807 this might not be gcc.</dd>
811 this might not be g++.</dd>
814 <dd>The full command line used to compile LLVM C source code. This has all
815 the configured -I, -D and optimization options.</dd>
818 <dd>The full command used to compile LLVM C++ source code. This has
819 all the configured -I, -D and optimization options.</dd>
822 <dd>This full link command used to link LLVM executables. This has all the
823 configured -I, -L and -l options.</dd>
826 <dd>The suffix for the host platforms share library (dll) files. This
827 includes the period as the first character.</dd>
828 </dl>
829 <p>To add more variables, two things need to be changed. First, add a line in
831 "set" the variable as a global in the site.exp file. Second, in the
833 to the list of "global" declarations at the beginning of the proc. That's it,
834 the variable can then be used in test scripts.</p>
835 </div>
837 <!-- _______________________________________________________________________ -->
839 <!-- _______________________________________________________________________ -->
841 <p>To make RUN line writing easier, there are several shell scripts located
843 when running tests, so you can just call these scripts using their name. For
844 example:</p>
845 <dl>
848 in cases where the test needs to cause a tool to generate an error (e.g. to
849 check the error output). However, any program in a pipeline that returns a
850 non-zero result will cause the test to fail. This script overcomes that
851 issue and nicely documents that the test case is purposefully ignoring the
852 result code of the tool</dd>
855 <dd>This script runs its arguments and then inverts the result code from
857 useful to invert the result of a grep. For example "not grep X" means
858 succeed only if you don't find X in the input.</dd>
859 </dl>
863 line near the top of the file. This signals that the test case should succeed
864 if the test fails. Such test cases are counted separately by the testing tool. To
865 specify an expected fail, use the XFAIL keyword in the comments of the test
866 program followed by a colon and one or more regular expressions (separated by
867 a comma). The regular expressions allow you to XFAIL the test conditionally by
868 host platform. The regular expressions following the : are matched against the
869 target triplet for the host machine. If there is a match, the test is expected
870 to fail. If not, the test is expected to succeed. To XFAIL everywhere just
874 <pre>
875 ; XFAIL: darwin,sun
876 </pre>
877 </div>
880 scan the lines of the test case for ones that contain a pattern that matches
882 is related to the test case. The number after "PR" specifies the LLVM bugzilla
883 number. When a PR number is specified, it will be used in the pass/fail
884 reporting. This is useful to quickly get some context when a test fails.</p>
887 interpretation of lines to terminate. This is generally done right after the
888 last RUN: line. This has two side effects: (a) it prevents special
889 interpretation of lines that are part of the test program, not the
890 instructions to the test case, and (b) it speeds things up for really big test
891 cases by avoiding interpretation of the remainder of the file.</p>
893 </div>
895 <!--=========================================================================-->
897 Structure</a></div>
898 <!--=========================================================================-->
903 with LLVM and executed. These programs are compiled using the native compiler
904 and various LLVM backends. The output from the program compiled with the
905 native compiler is assumed correct; the results from the other programs are
906 compared to the native program output and pass if they match.</p>
908 <p>When executing tests, it is usually a good idea to start out with a subset of
909 the available tests or programs. This makes test run times smaller at first and
910 later on this is useful to investigate individual test failures. To run some
911 test only on a subset of programs, simply change directory to the programs you
914 selected programs (see below for more info).</p>
917 performs timing tests of various LLVM optimizations. It also records
918 compilation times for the compilers and the JIT. This information can be
919 used to compare the effectiveness of LLVM's optimizations and code
920 generation.</p>
923 SingleSource, and External.</p>
925 <ul>
927 <p>The SingleSource directory contains test programs that are only a single
928 source file in size. These are usually small benchmark programs or small
929 programs that calculate a particular value. Several such programs are grouped
930 together in each directory.</p></li>
933 <p>The MultiSource directory contains subdirectories which contain entire
934 programs with multiple source files. Large benchmarks and whole applications
935 go here.</p></li>
938 <p>The External directory contains Makefiles for building code that is external
939 to (i.e., not distributed with) LLVM. The most prominent members of this
941 directory does not contain these actual tests, but only the Makefiles that know
942 how to properly compile these programs from somewhere else. The presence and
943 location of these external programs is configured by the test-suite
945 </ul>
947 <p>Each tree is then subdivided into several categories, including applications,
948 benchmarks, regression tests, code that is strange grammatically, etc. These
949 organizations should be relatively self explanatory.</p>
951 <p>Some tests are known to fail. Some are bugs that we have not fixed yet;
952 others are features that we haven't added yet (or may never add). In the
953 regression tests, the result for such tests will be XFAIL (eXpected FAILure).
954 In this way, you can tell the difference between an expected and unexpected
955 failure.</p>
957 <p>The tests in the test suite have no such feature at this time. If the
958 test passes, only warnings and other miscellaneous output will be generated. If
960 will help you separate benign warnings from actual test failures.</p>
962 </div>
964 <!--=========================================================================-->
966 <!--=========================================================================-->
970 <p>First, all tests are executed within the LLVM object directory tree. They
972 test suite creates temporary files during execution.</p>
976 <ol>
978 </li>
983 <pre>
984 % svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite
985 </pre>
986 </div>
988 </li>
993 each build tree (LLVM object directory tree) in which you want
994 to run the test suite, just as you do before building LLVM.</p>
999 <p>You must also tell the configure machinery that the test suite
1000 is available so it can be configured for your build tree:</p>
1002 <pre>
1003 % cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure [--with-llvmgccdir=$LLVM_GCC_DIR]
1004 </pre>
1005 </div>
1008 </li>
1012 <pre>
1013 % cd $LLVM_OBJ_ROOT/projects/test-suite
1014 % make
1015 </pre>
1016 </div>
1017 </li>
1018 </ol>
1019 <p>Note that the second and third steps only need to be done once. After you
1020 have the suite checked out and configured, you don't need to do it again (unless
1021 the test code or configure script changes).</p>
1023 </div>
1025 <!-- _______________________________________________________________________ -->
1028 <!-- _______________________________________________________________________ -->
1036 neglected, the External tests won't work.</p>
1037 <dl>
1040 </dl>
1041 This tells LLVM where to find any external tests. They are expected to be
1046 Subdirectory names known to LLVM include:
1047 <dl>
1052 </dl>
1053 Others are added from time to time, and can be determined from
1055 </div>
1057 <!-- _______________________________________________________________________ -->
1060 <!-- _______________________________________________________________________ -->
1063 module also provides a mechanism for compiling the programs in different ways.
1066 This Makefile can modify build rules to yield different results.</p>
1069 create the nightly test reports. To run the nightly tests, run <tt>gmake
1072 <p>There are several TEST Makefiles available in the tree. Some of them are
1073 designed for internal LLVM research and will not work outside of the LLVM
1074 research group. They may still be valuable, however, as a guide to writing your
1075 own TEST Makefile for any optimization or analysis passes that you develop with
1076 LLVM.</p>
1078 </div>
1080 <!-- _______________________________________________________________________ -->
1083 <!-- _______________________________________________________________________ -->
1085 <p>There are a number of ways to run the tests and generate output. The most
1087 compile and run all programs in the tree using a number of different methods
1088 and compare results. Any failures are reported in the output, but are likely
1089 drowned in the other output. Passes are not reported explicitely.</p>
1092 the specified test and usually adds per-program summaries to the output
1094 explicitely outputs TEST-PASS or TEST-FAIL for every test after each program.
1095 Though these lines are still drowned in the output, it's easy to grep the
1096 output logs in the Output directories.</p>
1102 end of the run and the results are always stored in the
1108 run.
1109 </div>
1111 <!-- _______________________________________________________________________ -->
1114 <!-- _______________________________________________________________________ -->
1119 should work), it is really easy to run optimizations or code generator
1120 components against every program in the tree, collecting statistics or running
1121 custom checks for correctness. At base, this is how the nightly tester works,
1122 it's just one example of a general framework.</p>
1124 <p>Lets say that you have an LLVM optimization pass, and you want to see how
1125 many times it triggers. First thing you should do is add an LLVM
1127 will tally counts of things you care about.</p>
1129 <p>Following this, you can set up a test and a report that collects these and
1130 formats them for easy viewing. This consists of two files, a
1131 "<tt>test-suite/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your
1132 test) and a "<tt>test-suite/TEST.XXX.report</tt>" file that indicates how to
1133 format the output into a table. There are many example reports of various
1134 levels of sophistication included with the test suite, and the framework is very
1135 general.</p>
1137 <p>If you are interested in testing an optimization pass, check out the
1141 <pre>
1142 % cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level
1143 % make TEST=libcalls report
1144 </pre>
1145 </div>
1150 <pre>
1151 Name | total | #exit |
1152 ...
1157 MallocBench/cfrac/cfrac | 1 | * |
1159 MallocBench/gs/gs | 4 | * |
1160 Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * |
1162 Prolangs-C/allroots/allroots | * | * |
1163 Prolangs-C/assembler/assembler | 47 | * |
1164 Prolangs-C/bison/mybison | 74 | * |
1165 ...
1166 </pre>
1167 </div>
1169 <p>This basically is grepping the -stats output and displaying it in a table.
1170 You can also use the "TEST=libcalls report.html" target to get the table in HTML
1171 form, similarly for report.csv and report.tex.</p>
1173 <p>The source for this is in test-suite/TEST.libcalls.*. The format is pretty
1174 simple: the Makefile indicates how to run the test (in this case,
1175 "<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for
1176 each column of the output. The first value is the header for the column and the
1177 second is the regex to grep the output of the command for. There are lots of
1178 example reports that can do fancy stuff.</p>
1180 </div>
1182 <!-- *********************************************************************** -->
1184 <hr>
1185 <address>
1191 John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner<br>
1193 Last modified: $Date$
1194 </address>
1195 </body>
1196 </html>