1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 2 "http://www.w3.org/TR/html4/strict.dtd"> 3<html> 4<head> 5 <title>LLVM Testing Infrastructure Guide</title> 6 <link rel="stylesheet" href="llvm.css" type="text/css"> 7</head> 8<body> 9 10<h1> 11 LLVM Testing Infrastructure Guide 12</h1> 13 14<ol> 15 <li><a href="#overview">Overview</a></li> 16 <li><a href="#requirements">Requirements</a></li> 17 <li><a href="#org">LLVM testing infrastructure organization</a> 18 <ul> 19 <li><a href="#regressiontests">Regression tests</a></li> 20 <li><a href="#testsuite">Test suite</a></li> 21 <li><a href="#debuginfotests">Debugging Information tests</a></li> 22 </ul> 23 </li> 24 <li><a href="#quick">Quick start</a> 25 <ul> 26 <li><a href="#quickregressiontests">Regression tests</a></li> 27 <li><a href="#quicktestsuite">Test suite</a></li> 28 <li><a href="#quickdebuginfotests">Debugging Information tests</a></li> 29 </ul> 30 </li> 31 <li><a href="#rtstructure">Regression test structure</a> 32 <ul> 33 <li><a href="#rtcustom">Writing new regression tests</a></li> 34 <li><a href="#FileCheck">The FileCheck utility</a></li> 35 <li><a href="#rtvars">Variables and substitutions</a></li> 36 <li><a href="#rtfeatures">Other features</a></li> 37 </ul> 38 </li> 39 <li><a href="#testsuitestructure">Test suite structure</a></li> 40 <li><a href="#testsuiterun">Running the test suite</a> 41 <ul> 42 <li><a href="#testsuiteexternal">Configuring External Tests</a></li> 43 <li><a href="#testsuitetests">Running different tests</a></li> 44 <li><a href="#testsuiteoutput">Generating test output</a></li> 45 <li><a href="#testsuitecustom">Writing custom tests for test-suite</a></li> 46 </ul> 47 </li> 48</ol> 49 50<div class="doc_author"> 51 <p>Written by John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner</p> 52</div> 53 54<!--=========================================================================--> 55<h2><a name="overview">Overview</a></h2> 56<!--=========================================================================--> 57 58<div> 59 60<p>This document is the reference manual for the LLVM testing infrastructure. It 61documents the structure of the LLVM testing infrastructure, the tools needed to 62use it, and how to add and run tests.</p> 63 64</div> 65 66<!--=========================================================================--> 67<h2><a name="requirements">Requirements</a></h2> 68<!--=========================================================================--> 69 70<div> 71 72<p>In order to use the LLVM testing infrastructure, you will need all of the 73software required to build LLVM, as well 74as <a href="http://python.org">Python</a> 2.4 or later.</p> 75 76</div> 77 78<!--=========================================================================--> 79<h2><a name="org">LLVM testing infrastructure organization</a></h2> 80<!--=========================================================================--> 81 82<div> 83 84<p>The LLVM testing infrastructure contains two major categories of tests: 85regression tests and whole programs. The regression tests are contained inside 86the LLVM repository itself under <tt>llvm/test</tt> and are expected to always 87pass -- they should be run before every commit. The whole programs tests are 88referred to as the "LLVM test suite" and are in the <tt>test-suite</tt> module 89in subversion. 90</p> 91 92<!-- _______________________________________________________________________ --> 93<h3><a name="regressiontests">Regression tests</a></h3> 94<!-- _______________________________________________________________________ --> 95 96<div> 97 98<p>The regression tests are small pieces of code that test a specific feature of 99LLVM or trigger a specific bug in LLVM. They are usually written in LLVM 100assembly language, but can be written in other languages if the test targets a 101particular language front end (and the appropriate <tt>--with-llvmgcc</tt> 102options were used at <tt>configure</tt> time of the <tt>llvm</tt> module). These 103tests are driven by the 'lit' testing tool, which is part of LLVM.</p> 104 105<p>These code fragments are not complete programs. The code generated 106from them is never executed to determine correct behavior.</p> 107 108<p>These code fragment tests are located in the <tt>llvm/test</tt> 109directory.</p> 110 111<p>Typically when a bug is found in LLVM, a regression test containing 112just enough code to reproduce the problem should be written and placed 113somewhere underneath this directory. In most cases, this will be a small 114piece of LLVM assembly language code, often distilled from an actual 115application or benchmark.</p> 116 117</div> 118 119<!-- _______________________________________________________________________ --> 120<h3><a name="testsuite">Test suite</a></h3> 121<!-- _______________________________________________________________________ --> 122 123<div> 124 125<p>The test suite contains whole programs, which are pieces of 126code which can be compiled and linked into a stand-alone program that can be 127executed. These programs are generally written in high level languages such as 128C or C++, but sometimes they are written straight in LLVM assembly.</p> 129 130<p>These programs are compiled and then executed using several different 131methods (native compiler, LLVM C backend, LLVM JIT, LLVM native code generation, 132etc). The output of these programs is compared to ensure that LLVM is compiling 133the program correctly.</p> 134 135<p>In addition to compiling and executing programs, whole program tests serve as 136a way of benchmarking LLVM performance, both in terms of the efficiency of the 137programs generated as well as the speed with which LLVM compiles, optimizes, and 138generates code.</p> 139 140<p>The test-suite is located in the <tt>test-suite</tt> Subversion module.</p> 141 142</div> 143 144<!-- _______________________________________________________________________ --> 145<h3><a name="debuginfotests">Debugging Information tests</a></h3> 146<!-- _______________________________________________________________________ --> 147 148<div> 149 150<p>The test suite contains tests to check quality of debugging information. 151The test are written in C based languages or in LLVM assembly language. </p> 152 153<p>These tests are compiled and run under a debugger. The debugger output 154is checked to validate of debugging information. See README.txt in the 155test suite for more information . This test suite is located in the 156<tt>debuginfo-tests</tt> Subversion module. </p> 157 158</div> 159 160</div> 161 162<!--=========================================================================--> 163<h2><a name="quick">Quick start</a></h2> 164<!--=========================================================================--> 165 166<div> 167 168 <p>The tests are located in two separate Subversion modules. The regressions 169 tests are in the main "llvm" module under the directory 170 <tt>llvm/test</tt> (so you get these tests for free with the main llvm tree). 171 The more comprehensive test suite that includes whole 172programs in C and C++ is in the <tt>test-suite</tt> module. This module should 173be checked out to the <tt>llvm/projects</tt> directory (don't use another name 174than the default "test-suite", for then the test suite will be run every time 175you run <tt>make</tt> in the main <tt>llvm</tt> directory). 176When you <tt>configure</tt> the <tt>llvm</tt> module, 177the <tt>test-suite</tt> directory will be automatically configured. 178Alternatively, you can configure the <tt>test-suite</tt> module manually.</p> 179 180<!-- _______________________________________________________________________ --> 181<h3><a name="quickregressiontests">Regression tests</a></h3> 182<div> 183<!-- _______________________________________________________________________ --> 184<p>To run all of the LLVM regression tests, use master Makefile in 185 the <tt>llvm/test</tt> directory:</p> 186 187<div class="doc_code"> 188<pre> 189% gmake -C llvm/test 190</pre> 191</div> 192 193<p>or</p> 194 195<div class="doc_code"> 196<pre> 197% gmake check 198</pre> 199</div> 200 201<p>If you have <a href="http://clang.llvm.org/">Clang</a> checked out and built, 202you can run the LLVM and Clang tests simultaneously using:</p> 203 204<p>or</p> 205 206<div class="doc_code"> 207<pre> 208% gmake check-all 209</pre> 210</div> 211 212<p>To run the tests with Valgrind (Memcheck by default), just append 213<tt>VG=1</tt> to the commands above, e.g.:</p> 214 215<div class="doc_code"> 216<pre> 217% gmake check VG=1 218</pre> 219</div> 220 221<p>To run individual tests or subsets of tests, you can use the 'llvm-lit' 222script which is built as part of LLVM. For example, to run the 223'Integer/BitCast.ll' test by itself you can run:</p> 224 225<div class="doc_code"> 226<pre> 227% llvm-lit ~/llvm/test/Integer/BitCast.ll 228</pre> 229</div> 230 231<p>or to run all of the ARM CodeGen tests:</p> 232 233<div class="doc_code"> 234<pre> 235% llvm-lit ~/llvm/test/CodeGen/ARM 236</pre> 237</div> 238 239<p>For more information on using the 'lit' tool, see 'llvm-lit --help' or the 240'lit' man page.</p> 241 242</div> 243 244<!-- _______________________________________________________________________ --> 245<h3><a name="quicktestsuite">Test suite</a></h3> 246<!-- _______________________________________________________________________ --> 247 248<div> 249 250<p>To run the comprehensive test suite (tests that compile and execute whole 251programs), first checkout and setup the <tt>test-suite</tt> module:</p> 252 253<div class="doc_code"> 254<pre> 255% cd llvm/projects 256% svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite 257% cd .. 258% ./configure --with-llvmgccdir=$LLVM_GCC_DIR 259</pre> 260</div> 261 262<p>where <tt>$LLVM_GCC_DIR</tt> is the directory where 263you <em>installed</em> llvm-gcc, not its src or obj 264dir. The <tt>--with-llvmgccdir</tt> option assumes that 265the <tt>llvm-gcc-4.2</tt> module was configured with 266<tt>--program-prefix=llvm-</tt>, and therefore that the C and C++ 267compiler drivers are called <tt>llvm-gcc</tt> and <tt>llvm-g++</tt> 268respectively. If this is not the case, 269use <tt>--with-llvmgcc</tt>/<tt>--with-llvmgxx</tt> to specify each 270executable's location.</p> 271 272<p>Then, run the entire test suite by running make in the <tt>test-suite</tt> 273directory:</p> 274 275<div class="doc_code"> 276<pre> 277% cd projects/test-suite 278% gmake 279</pre> 280</div> 281 282<p>Usually, running the "nightly" set of tests is a good idea, and you can also 283let it generate a report by running:</p> 284 285<div class="doc_code"> 286<pre> 287% cd projects/test-suite 288% gmake TEST=nightly report report.html 289</pre> 290</div> 291 292<p>Any of the above commands can also be run in a subdirectory of 293<tt>projects/test-suite</tt> to run the specified test only on the programs in 294that subdirectory.</p> 295 296</div> 297 298<!-- _______________________________________________________________________ --> 299<h3><a name="quickdebuginfotests">Debugging Information tests</a></h3> 300<div> 301<!-- _______________________________________________________________________ --> 302<div> 303 304<p> To run debugging information tests simply checkout the tests inside 305clang/test directory. </p> 306 307<div class="doc_code"> 308<pre> 309%cd clang/test 310% svn co http://llvm.org/svn/llvm-project/debuginfo-tests/trunk debuginfo-tests 311</pre> 312</div> 313 314<p> These tests are already set up to run as part of clang regression tests.</p> 315 316</div> 317 318</div> 319 320</div> 321 322<!--=========================================================================--> 323<h2><a name="rtstructure">Regression test structure</a></h2> 324<!--=========================================================================--> 325<div> 326 <p>The LLVM regression tests are driven by 'lit' and are located in 327 the <tt>llvm/test</tt> directory. 328 329 <p>This directory contains a large array of small tests 330 that exercise various features of LLVM and to ensure that regressions do not 331 occur. The directory is broken into several sub-directories, each focused on 332 a particular area of LLVM. A few of the important ones are:</p> 333 334 <ul> 335 <li><tt>Analysis</tt>: checks Analysis passes.</li> 336 <li><tt>Archive</tt>: checks the Archive library.</li> 337 <li><tt>Assembler</tt>: checks Assembly reader/writer functionality.</li> 338 <li><tt>Bitcode</tt>: checks Bitcode reader/writer functionality.</li> 339 <li><tt>CodeGen</tt>: checks code generation and each target.</li> 340 <li><tt>Features</tt>: checks various features of the LLVM language.</li> 341 <li><tt>Linker</tt>: tests bitcode linking.</li> 342 <li><tt>Transforms</tt>: tests each of the scalar, IPO, and utility 343 transforms to ensure they make the right transformations.</li> 344 <li><tt>Verifier</tt>: tests the IR verifier.</li> 345 </ul> 346 347<!-- _______________________________________________________________________ --> 348<h3><a name="rtcustom">Writing new regression tests</a></h3> 349<!-- _______________________________________________________________________ --> 350<div> 351 <p>The regression test structure is very simple, but does require some 352 information to be set. This information is gathered via <tt>configure</tt> and 353 is written to a file, <tt>lit.site.cfg</tt> 354 in <tt>llvm/test</tt>. The <tt>llvm/test</tt> Makefile does this work for 355 you.</p> 356 357 <p>In order for the regression tests to work, each directory of tests must 358 have a <tt>dg.exp</tt> file. Lit looks for this file to determine how to 359 run the tests. This file is just a Tcl script and it can do anything you want, 360 but we've standardized it for the LLVM regression tests. If you're adding a 361 directory of tests, just copy <tt>dg.exp</tt> from another directory to get 362 running. The standard <tt>dg.exp</tt> simply loads a Tcl library 363 (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt> function 364 defined in that library with a list of file names to run. The names are 365 obtained by using Tcl's glob command. Any directory that contains only 366 directories does not need the <tt>dg.exp</tt> file.</p> 367 368 <p>The <tt>llvm-runtests</tt> function looks at each file that is passed to 369 it and gathers any lines together that match "RUN:". These are the "RUN" lines 370 that specify how the test is to be run. So, each test script must contain 371 RUN lines if it is to do anything. If there are no RUN lines, the 372 <tt>llvm-runtests</tt> function will issue an error and the test will 373 fail.</p> 374 375 <p>RUN lines are specified in the comments of the test program using the 376 keyword <tt>RUN</tt> followed by a colon, and lastly the command (pipeline) 377 to execute. Together, these lines form the "script" that 378 <tt>llvm-runtests</tt> executes to run the test case. The syntax of the 379 RUN lines is similar to a shell's syntax for pipelines including I/O 380 redirection and variable substitution. However, even though these lines 381 may <i>look</i> like a shell script, they are not. RUN lines are interpreted 382 directly by the Tcl <tt>exec</tt> command. They are never executed by a 383 shell. Consequently the syntax differs from normal shell script syntax in a 384 few ways. You can specify as many RUN lines as needed.</p> 385 386 <p>lit performs substitution on each RUN line to replace LLVM tool 387 names with the full paths to the executable built for each tool (in 388 $(LLVM_OBJ_ROOT)/$(BuildMode)/bin). This ensures that lit does not 389 invoke any stray LLVM tools in the user's path during testing.</p> 390 391 <p>Each RUN line is executed on its own, distinct from other lines unless 392 its last character is <tt>\</tt>. This continuation character causes the RUN 393 line to be concatenated with the next one. In this way you can build up long 394 pipelines of commands without making huge line lengths. The lines ending in 395 <tt>\</tt> are concatenated until a RUN line that doesn't end in <tt>\</tt> is 396 found. This concatenated set of RUN lines then constitutes one execution. 397 Tcl will substitute variables and arrange for the pipeline to be executed. If 398 any process in the pipeline fails, the entire line (and test case) fails too. 399 </p> 400 401 <p> Below is an example of legal RUN lines in a <tt>.ll</tt> file:</p> 402 403<div class="doc_code"> 404<pre> 405; RUN: llvm-as < %s | llvm-dis > %t1 406; RUN: llvm-dis < %s.bc-13 > %t2 407; RUN: diff %t1 %t2 408</pre> 409</div> 410 411 <p>As with a Unix shell, the RUN: lines permit pipelines and I/O redirection 412 to be used. However, the usage is slightly different than for Bash. To check 413 what's legal, see the documentation for the 414 <a href="http://www.tcl.tk/man/tcl8.5/TclCmd/exec.htm#M2">Tcl exec</a> 415 command and the 416 <a href="http://www.tcl.tk/man/tcl8.5/tutorial/Tcl26.html">tutorial</a>. 417 The major differences are:</p> 418 <ul> 419 <li>You can't do <tt>2>&1</tt>. That will cause Tcl to write to a 420 file named <tt>&1</tt>. Usually this is done to get stderr to go through 421 a pipe. You can do that in tcl with <tt>|&</tt> so replace this idiom: 422 <tt>... 2>&1 | grep</tt> with <tt>... |& grep</tt></li> 423 <li>You can only redirect to a file, not to another descriptor and not from 424 a here document.</li> 425 <li>tcl supports redirecting to open files with the @ syntax but you 426 shouldn't use that here.</li> 427 </ul> 428 429 <p>There are some quoting rules that you must pay attention to when writing 430 your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any 431 quote characters so they will get passed to the invoked program. For 432 example:</p> 433 434<div class="doc_code"> 435<pre> 436... | grep 'find this string' 437</pre> 438</div> 439 440 <p>This will fail because the ' characters are passed to grep. This would 441 instruction grep to look for <tt>'find</tt> in the files <tt>this</tt> and 442 <tt>string'</tt>. To avoid this use curly braces to tell Tcl that it should 443 treat everything enclosed as one value. So our example would become:</p> 444 445<div class="doc_code"> 446<pre> 447... | grep {find this string} 448</pre> 449</div> 450 451 <p>Additionally, the characters <tt>[</tt> and <tt>]</tt> are treated 452 specially by Tcl. They tell Tcl to interpret the content as a command to 453 execute. Since these characters are often used in regular expressions this can 454 have disastrous results and cause the entire test run in a directory to fail. 455 For example, a common idiom is to look for some basicblock number:</p> 456 457<div class="doc_code"> 458<pre> 459... | grep bb[2-8] 460</pre> 461</div> 462 463 <p>This, however, will cause Tcl to fail because its going to try to execute 464 a program named "2-8". Instead, what you want is this:</p> 465 466<div class="doc_code"> 467<pre> 468... | grep {bb\[2-8\]} 469</pre> 470</div> 471 472 <p>Finally, if you need to pass the <tt>\</tt> character down to a program, 473 then it must be doubled. This is another Tcl special character. So, suppose 474 you had: 475 476<div class="doc_code"> 477<pre> 478... | grep 'i32\*' 479</pre> 480</div> 481 482 <p>This will fail to match what you want (a pointer to i32). First, the 483 <tt>'</tt> do not get stripped off. Second, the <tt>\</tt> gets stripped off 484 by Tcl so what grep sees is: <tt>'i32*'</tt>. That's not likely to match 485 anything. To resolve this you must use <tt>\\</tt> and the <tt>{}</tt>, like 486 this:</p> 487 488<div class="doc_code"> 489<pre> 490... | grep {i32\\*} 491</pre> 492</div> 493 494<p>If your system includes GNU <tt>grep</tt>, make sure 495that <tt>GREP_OPTIONS</tt> is not set in your environment. Otherwise, 496you may get invalid results (both false positives and false 497negatives).</p> 498 499</div> 500 501<!-- _______________________________________________________________________ --> 502<h3><a name="FileCheck">The FileCheck utility</a></h3> 503<!-- _______________________________________________________________________ --> 504 505<div> 506 507<p>A powerful feature of the RUN: lines is that it allows any arbitrary commands 508 to be executed as part of the test harness. While standard (portable) unix 509 tools like 'grep' work fine on run lines, as you see above, there are a lot 510 of caveats due to interaction with Tcl syntax, and we want to make sure the 511 run lines are portable to a wide range of systems. Another major problem is 512 that grep is not very good at checking to verify that the output of a tools 513 contains a series of different output in a specific order. The FileCheck 514 tool was designed to help with these problems.</p> 515 516<p>FileCheck (whose basic command line arguments are described in <a 517 href="http://llvm.org/cmds/FileCheck.html">the FileCheck man page</a> is 518 designed to read a file to check from standard input, and the set of things 519 to verify from a file specified as a command line argument. A simple example 520 of using FileCheck from a RUN line looks like this:</p> 521 522<div class="doc_code"> 523<pre> 524; RUN: llvm-as < %s | llc -march=x86-64 | <b>FileCheck %s</b> 525</pre> 526</div> 527 528<p>This syntax says to pipe the current file ("%s") into llvm-as, pipe that into 529llc, then pipe the output of llc into FileCheck. This means that FileCheck will 530be verifying its standard input (the llc output) against the filename argument 531specified (the original .ll file specified by "%s"). To see how this works, 532lets look at the rest of the .ll file (after the RUN line):</p> 533 534<div class="doc_code"> 535<pre> 536define void @sub1(i32* %p, i32 %v) { 537entry: 538; <b>CHECK: sub1:</b> 539; <b>CHECK: subl</b> 540 %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v) 541 ret void 542} 543 544define void @inc4(i64* %p) { 545entry: 546; <b>CHECK: inc4:</b> 547; <b>CHECK: incq</b> 548 %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1) 549 ret void 550} 551</pre> 552</div> 553 554<p>Here you can see some "CHECK:" lines specified in comments. Now you can see 555how the file is piped into llvm-as, then llc, and the machine code output is 556what we are verifying. FileCheck checks the machine code output to verify that 557it matches what the "CHECK:" lines specify.</p> 558 559<p>The syntax of the CHECK: lines is very simple: they are fixed strings that 560must occur in order. FileCheck defaults to ignoring horizontal whitespace 561differences (e.g. a space is allowed to match a tab) but otherwise, the contents 562of the CHECK: line is required to match some thing in the test file exactly.</p> 563 564<p>One nice thing about FileCheck (compared to grep) is that it allows merging 565test cases together into logical groups. For example, because the test above 566is checking for the "sub1:" and "inc4:" labels, it will not match unless there 567is a "subl" in between those labels. If it existed somewhere else in the file, 568that would not count: "grep subl" matches if subl exists anywhere in the 569file.</p> 570 571<!-- _______________________________________________________________________ --> 572<h4> 573 <a name="FileCheck-check-prefix">The FileCheck -check-prefix option</a> 574</h4> 575 576<div> 577 578<p>The FileCheck -check-prefix option allows multiple test configurations to be 579driven from one .ll file. This is useful in many circumstances, for example, 580testing different architectural variants with llc. Here's a simple example:</p> 581 582<div class="doc_code"> 583<pre> 584; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \ 585; RUN: | <b>FileCheck %s -check-prefix=X32</b> 586; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \ 587; RUN: | <b>FileCheck %s -check-prefix=X64</b> 588 589define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind { 590 %tmp1 = insertelement <4 x i32> %tmp, i32 %s, i32 1 591 ret <4 x i32> %tmp1 592; <b>X32:</b> pinsrd_1: 593; <b>X32:</b> pinsrd $1, 4(%esp), %xmm0 594 595; <b>X64:</b> pinsrd_1: 596; <b>X64:</b> pinsrd $1, %edi, %xmm0 597} 598</pre> 599</div> 600 601<p>In this case, we're testing that we get the expected code generation with 602both 32-bit and 64-bit code generation.</p> 603 604</div> 605 606<!-- _______________________________________________________________________ --> 607<h4> 608 <a name="FileCheck-CHECK-NEXT">The "CHECK-NEXT:" directive</a> 609</h4> 610 611<div> 612 613<p>Sometimes you want to match lines and would like to verify that matches 614happen on exactly consecutive lines with no other lines in between them. In 615this case, you can use CHECK: and CHECK-NEXT: directives to specify this. If 616you specified a custom check prefix, just use "<PREFIX>-NEXT:". For 617example, something like this works as you'd expect:</p> 618 619<div class="doc_code"> 620<pre> 621define void @t2(<2 x double>* %r, <2 x double>* %A, double %B) { 622 %tmp3 = load <2 x double>* %A, align 16 623 %tmp7 = insertelement <2 x double> undef, double %B, i32 0 624 %tmp9 = shufflevector <2 x double> %tmp3, 625 <2 x double> %tmp7, 626 <2 x i32> < i32 0, i32 2 > 627 store <2 x double> %tmp9, <2 x double>* %r, align 16 628 ret void 629 630; <b>CHECK:</b> t2: 631; <b>CHECK:</b> movl 8(%esp), %eax 632; <b>CHECK-NEXT:</b> movapd (%eax), %xmm0 633; <b>CHECK-NEXT:</b> movhpd 12(%esp), %xmm0 634; <b>CHECK-NEXT:</b> movl 4(%esp), %eax 635; <b>CHECK-NEXT:</b> movapd %xmm0, (%eax) 636; <b>CHECK-NEXT:</b> ret 637} 638</pre> 639</div> 640 641<p>CHECK-NEXT: directives reject the input unless there is exactly one newline 642between it an the previous directive. A CHECK-NEXT cannot be the first 643directive in a file.</p> 644 645</div> 646 647<!-- _______________________________________________________________________ --> 648<h4> 649 <a name="FileCheck-CHECK-NOT">The "CHECK-NOT:" directive</a> 650</h4> 651 652<div> 653 654<p>The CHECK-NOT: directive is used to verify that a string doesn't occur 655between two matches (or the first match and the beginning of the file). For 656example, to verify that a load is removed by a transformation, a test like this 657can be used:</p> 658 659<div class="doc_code"> 660<pre> 661define i8 @coerce_offset0(i32 %V, i32* %P) { 662 store i32 %V, i32* %P 663 664 %P2 = bitcast i32* %P to i8* 665 %P3 = getelementptr i8* %P2, i32 2 666 667 %A = load i8* %P3 668 ret i8 %A 669; <b>CHECK:</b> @coerce_offset0 670; <b>CHECK-NOT:</b> load 671; <b>CHECK:</b> ret i8 672} 673</pre> 674</div> 675 676</div> 677 678<!-- _______________________________________________________________________ --> 679<h4> 680 <a name="FileCheck-Matching">FileCheck Pattern Matching Syntax</a> 681</h4> 682 683<div> 684 685<p>The CHECK: and CHECK-NOT: directives both take a pattern to match. For most 686uses of FileCheck, fixed string matching is perfectly sufficient. For some 687things, a more flexible form of matching is desired. To support this, FileCheck 688allows you to specify regular expressions in matching strings, surrounded by 689double braces: <b>{{yourregex}}</b>. Because we want to use fixed string 690matching for a majority of what we do, FileCheck has been designed to support 691mixing and matching fixed string matching with regular expressions. This allows 692you to write things like this:</p> 693 694<div class="doc_code"> 695<pre> 696; CHECK: movhpd <b>{{[0-9]+}}</b>(%esp), <b>{{%xmm[0-7]}}</b> 697</pre> 698</div> 699 700<p>In this case, any offset from the ESP register will be allowed, and any xmm 701register will be allowed.</p> 702 703<p>Because regular expressions are enclosed with double braces, they are 704visually distinct, and you don't need to use escape characters within the double 705braces like you would in C. In the rare case that you want to match double 706braces explicitly from the input, you can use something ugly like 707<b>{{[{][{]}}</b> as your pattern.</p> 708 709</div> 710 711<!-- _______________________________________________________________________ --> 712<h4> 713 <a name="FileCheck-Variables">FileCheck Variables</a> 714</h4> 715 716<div> 717 718<p>It is often useful to match a pattern and then verify that it occurs again 719later in the file. For codegen tests, this can be useful to allow any register, 720but verify that that register is used consistently later. To do this, FileCheck 721allows named variables to be defined and substituted into patterns. Here is a 722simple example:</p> 723 724<div class="doc_code"> 725<pre> 726; CHECK: test5: 727; CHECK: notw <b>[[REGISTER:%[a-z]+]]</b> 728; CHECK: andw {{.*}}<b>[[REGISTER]]</b> 729</pre> 730</div> 731 732<p>The first check line matches a regex (<tt>%[a-z]+</tt>) and captures it into 733the variables "REGISTER". The second line verifies that whatever is in REGISTER 734occurs later in the file after an "andw". FileCheck variable references are 735always contained in <tt>[[ ]]</tt> pairs, are named, and their names can be 736formed with the regex "<tt>[a-zA-Z][a-zA-Z0-9]*</tt>". If a colon follows the 737name, then it is a definition of the variable, if not, it is a use.</p> 738 739<p>FileCheck variables can be defined multiple times, and uses always get the 740latest value. Note that variables are all read at the start of a "CHECK" line 741and are all defined at the end. This means that if you have something like 742"<tt>CHECK: [[XYZ:.*]]x[[XYZ]]</tt>" that the check line will read the previous 743value of the XYZ variable and define a new one after the match is performed. If 744you need to do something like this you can probably take advantage of the fact 745that FileCheck is not actually line-oriented when it matches, this allows you to 746define two separate CHECK lines that match on the same line. 747</p> 748 749</div> 750 751</div> 752 753<!-- _______________________________________________________________________ --> 754<h3><a name="rtvars">Variables and substitutions</a></h3> 755<!-- _______________________________________________________________________ --> 756<div> 757 <p>With a RUN line there are a number of substitutions that are permitted. In 758 general, any Tcl variable that is available in the <tt>substitute</tt> 759 function (in <tt>test/lib/llvm.exp</tt>) can be substituted into a RUN line. 760 To make a substitution just write the variable's name preceded by a $. 761 Additionally, for compatibility reasons with previous versions of the test 762 library, certain names can be accessed with an alternate syntax: a % prefix. 763 These alternates are deprecated and may go away in a future version. 764 </p> 765 <p>Here are the available variable names. The alternate syntax is listed in 766 parentheses.</p> 767 768 <dl style="margin-left: 25px"> 769 <dt><b>$test</b> (%s)</dt> 770 <dd>The full path to the test case's source. This is suitable for passing 771 on the command line as the input to an llvm tool.</dd> 772 773 <dt><b>$srcdir</b></dt> 774 <dd>The source directory from where the "<tt>make check</tt>" was run.</dd> 775 776 <dt><b>objdir</b></dt> 777 <dd>The object directory that corresponds to the <tt>$srcdir</tt>.</dd> 778 779 <dt><b>subdir</b></dt> 780 <dd>A partial path from the <tt>test</tt> directory that contains the 781 sub-directory that contains the test source being executed.</dd> 782 783 <dt><b>srcroot</b></dt> 784 <dd>The root directory of the LLVM src tree.</dd> 785 786 <dt><b>objroot</b></dt> 787 <dd>The root directory of the LLVM object tree. This could be the same 788 as the srcroot.</dd> 789 790 <dt><b>path</b><dt> 791 <dd>The path to the directory that contains the test case source. This is 792 for locating any supporting files that are not generated by the test, but 793 used by the test.</dd> 794 795 <dt><b>tmp</b></dt> 796 <dd>The path to a temporary file name that could be used for this test case. 797 The file name won't conflict with other test cases. You can append to it if 798 you need multiple temporaries. This is useful as the destination of some 799 redirected output.</dd> 800 801 <dt><b>llvmlibsdir</b> (%llvmlibsdir)</dt> 802 <dd>The directory where the LLVM libraries are located.</dd> 803 804 <dt><b>target_triplet</b> (%target_triplet)</dt> 805 <dd>The target triplet that corresponds to the current host machine (the one 806 running the test cases). This should probably be called "host".<dd> 807 808 <dt><b>llvmgcc</b> (%llvmgcc)</dt> 809 <dd>The full path to the <tt>llvm-gcc</tt> executable as specified in the 810 configured LLVM environment</dd> 811 812 <dt><b>llvmgxx</b> (%llvmgxx)</dt> 813 <dd>The full path to the <tt>llvm-gxx</tt> executable as specified in the 814 configured LLVM environment</dd> 815 816 <dt><b>gccpath</b></dt> 817 <dd>The full path to the C compiler used to <i>build </i> LLVM. Note that 818 this might not be gcc.</dd> 819 820 <dt><b>gxxpath</b></dt> 821 <dd>The full path to the C++ compiler used to <i>build </i> LLVM. Note that 822 this might not be g++.</dd> 823 824 <dt><b>compile_c</b> (%compile_c)</dt> 825 <dd>The full command line used to compile LLVM C source code. This has all 826 the configured -I, -D and optimization options.</dd> 827 828 <dt><b>compile_cxx</b> (%compile_cxx)</dt> 829 <dd>The full command used to compile LLVM C++ source code. This has 830 all the configured -I, -D and optimization options.</dd> 831 832 <dt><b>link</b> (%link)</dt> 833 <dd>This full link command used to link LLVM executables. This has all the 834 configured -I, -L and -l options.</dd> 835 836 <dt><b>shlibext</b> (%shlibext)</dt> 837 <dd>The suffix for the host platforms share library (dll) files. This 838 includes the period as the first character.</dd> 839 </dl> 840 <p>To add more variables, two things need to be changed. First, add a line in 841 the <tt>test/Makefile</tt> that creates the <tt>site.exp</tt> file. This will 842 "set" the variable as a global in the site.exp file. Second, in the 843 <tt>test/lib/llvm.exp</tt> file, in the substitute proc, add the variable name 844 to the list of "global" declarations at the beginning of the proc. That's it, 845 the variable can then be used in test scripts.</p> 846</div> 847 848<!-- _______________________________________________________________________ --> 849<h3><a name="rtfeatures">Other Features</a></h3> 850<!-- _______________________________________________________________________ --> 851<div> 852 <p>To make RUN line writing easier, there are several shell scripts located 853 in the <tt>llvm/test/Scripts</tt> directory. This directory is in the PATH 854 when running tests, so you can just call these scripts using their name. For 855 example:</p> 856 <dl> 857 <dt><b>ignore</b></dt> 858 <dd>This script runs its arguments and then always returns 0. This is useful 859 in cases where the test needs to cause a tool to generate an error (e.g. to 860 check the error output). However, any program in a pipeline that returns a 861 non-zero result will cause the test to fail. This script overcomes that 862 issue and nicely documents that the test case is purposefully ignoring the 863 result code of the tool</dd> 864 865 <dt><b>not</b></dt> 866 <dd>This script runs its arguments and then inverts the result code from 867 it. Zero result codes become 1. Non-zero result codes become 0. This is 868 useful to invert the result of a grep. For example "not grep X" means 869 succeed only if you don't find X in the input.</dd> 870 </dl> 871 872 <p>Sometimes it is necessary to mark a test case as "expected fail" or XFAIL. 873 You can easily mark a test as XFAIL just by including <tt>XFAIL: </tt> on a 874 line near the top of the file. This signals that the test case should succeed 875 if the test fails. Such test cases are counted separately by the testing tool. To 876 specify an expected fail, use the XFAIL keyword in the comments of the test 877 program followed by a colon and one or more regular expressions (separated by 878 a comma). The regular expressions allow you to XFAIL the test conditionally by 879 host platform. The regular expressions following the : are matched against the 880 target triplet for the host machine. If there is a match, the test is expected 881 to fail. If not, the test is expected to succeed. To XFAIL everywhere just 882 specify <tt>XFAIL: *</tt>. Here is an example of an <tt>XFAIL</tt> line:</p> 883 884<div class="doc_code"> 885<pre> 886; XFAIL: darwin,sun 887</pre> 888</div> 889 890 <p>To make the output more useful, the <tt>llvm_runtest</tt> function wil 891 scan the lines of the test case for ones that contain a pattern that matches 892 PR[0-9]+. This is the syntax for specifying a PR (Problem Report) number that 893 is related to the test case. The number after "PR" specifies the LLVM bugzilla 894 number. When a PR number is specified, it will be used in the pass/fail 895 reporting. This is useful to quickly get some context when a test fails.</p> 896 897 <p>Finally, any line that contains "END." will cause the special 898 interpretation of lines to terminate. This is generally done right after the 899 last RUN: line. This has two side effects: (a) it prevents special 900 interpretation of lines that are part of the test program, not the 901 instructions to the test case, and (b) it speeds things up for really big test 902 cases by avoiding interpretation of the remainder of the file.</p> 903 904</div> 905 906</div> 907 908<!--=========================================================================--> 909<h2><a name="testsuitestructure">Test suite Structure</a></h2> 910<!--=========================================================================--> 911 912<div> 913 914<p>The <tt>test-suite</tt> module contains a number of programs that can be compiled 915with LLVM and executed. These programs are compiled using the native compiler 916and various LLVM backends. The output from the program compiled with the 917native compiler is assumed correct; the results from the other programs are 918compared to the native program output and pass if they match.</p> 919 920<p>When executing tests, it is usually a good idea to start out with a subset of 921the available tests or programs. This makes test run times smaller at first and 922later on this is useful to investigate individual test failures. To run some 923test only on a subset of programs, simply change directory to the programs you 924want tested and run <tt>gmake</tt> there. Alternatively, you can run a different 925test using the <tt>TEST</tt> variable to change what tests or run on the 926selected programs (see below for more info).</p> 927 928<p>In addition for testing correctness, the <tt>test-suite</tt> directory also 929performs timing tests of various LLVM optimizations. It also records 930compilation times for the compilers and the JIT. This information can be 931used to compare the effectiveness of LLVM's optimizations and code 932generation.</p> 933 934<p><tt>test-suite</tt> tests are divided into three types of tests: MultiSource, 935SingleSource, and External.</p> 936 937<ul> 938<li><tt>test-suite/SingleSource</tt> 939<p>The SingleSource directory contains test programs that are only a single 940source file in size. These are usually small benchmark programs or small 941programs that calculate a particular value. Several such programs are grouped 942together in each directory.</p></li> 943 944<li><tt>test-suite/MultiSource</tt> 945<p>The MultiSource directory contains subdirectories which contain entire 946programs with multiple source files. Large benchmarks and whole applications 947go here.</p></li> 948 949<li><tt>test-suite/External</tt> 950<p>The External directory contains Makefiles for building code that is external 951to (i.e., not distributed with) LLVM. The most prominent members of this 952directory are the SPEC 95 and SPEC 2000 benchmark suites. The <tt>External</tt> 953directory does not contain these actual tests, but only the Makefiles that know 954how to properly compile these programs from somewhere else. The presence and 955location of these external programs is configured by the test-suite 956<tt>configure</tt> script.</p></li> 957</ul> 958 959<p>Each tree is then subdivided into several categories, including applications, 960benchmarks, regression tests, code that is strange grammatically, etc. These 961organizations should be relatively self explanatory.</p> 962 963<p>Some tests are known to fail. Some are bugs that we have not fixed yet; 964others are features that we haven't added yet (or may never add). In the 965regression tests, the result for such tests will be XFAIL (eXpected FAILure). 966In this way, you can tell the difference between an expected and unexpected 967failure.</p> 968 969<p>The tests in the test suite have no such feature at this time. If the 970test passes, only warnings and other miscellaneous output will be generated. If 971a test fails, a large <program> FAILED message will be displayed. This 972will help you separate benign warnings from actual test failures.</p> 973 974</div> 975 976<!--=========================================================================--> 977<h2><a name="testsuiterun">Running the test suite</a></h2> 978<!--=========================================================================--> 979 980<div> 981 982<p>First, all tests are executed within the LLVM object directory tree. They 983<i>are not</i> executed inside of the LLVM source tree. This is because the 984test suite creates temporary files during execution.</p> 985 986<p>To run the test suite, you need to use the following steps:</p> 987 988<ol> 989 <li><tt>cd</tt> into the <tt>llvm/projects</tt> directory in your source tree. 990 </li> 991 992 <li><p>Check out the <tt>test-suite</tt> module with:</p> 993 994<div class="doc_code"> 995<pre> 996% svn co http://llvm.org/svn/llvm-project/test-suite/trunk test-suite 997</pre> 998</div> 999 <p>This will get the test suite into <tt>llvm/projects/test-suite</tt>.</p> 1000 </li> 1001 <li><p>Configure and build <tt>llvm</tt>.</p></li> 1002 <li><p>Configure and build <tt>llvm-gcc</tt>.</p></li> 1003 <li><p>Install <tt>llvm-gcc</tt> somewhere.</p></li> 1004 <li><p><em>Re-configure</em> <tt>llvm</tt> from the top level of 1005 each build tree (LLVM object directory tree) in which you want 1006 to run the test suite, just as you do before building LLVM.</p> 1007 <p>During the <em>re-configuration</em>, you must either: (1) 1008 have <tt>llvm-gcc</tt> you just built in your path, or (2) 1009 specify the directory where your just-built <tt>llvm-gcc</tt> is 1010 installed using <tt>--with-llvmgccdir=$LLVM_GCC_DIR</tt>.</p> 1011 <p>You must also tell the configure machinery that the test suite 1012 is available so it can be configured for your build tree:</p> 1013<div class="doc_code"> 1014<pre> 1015% cd $LLVM_OBJ_ROOT ; $LLVM_SRC_ROOT/configure [--with-llvmgccdir=$LLVM_GCC_DIR] 1016</pre> 1017</div> 1018 <p>[Remember that <tt>$LLVM_GCC_DIR</tt> is the directory where you 1019 <em>installed</em> llvm-gcc, not its src or obj directory.]</p> 1020 </li> 1021 1022 <li><p>You can now run the test suite from your build tree as follows:</p> 1023<div class="doc_code"> 1024<pre> 1025% cd $LLVM_OBJ_ROOT/projects/test-suite 1026% make 1027</pre> 1028</div> 1029 </li> 1030</ol> 1031<p>Note that the second and third steps only need to be done once. After you 1032have the suite checked out and configured, you don't need to do it again (unless 1033the test code or configure script changes).</p> 1034 1035<!-- _______________________________________________________________________ --> 1036<h3> 1037 <a name="testsuiteexternal">Configuring External Tests</a> 1038</h3> 1039<!-- _______________________________________________________________________ --> 1040 1041<div> 1042<p>In order to run the External tests in the <tt>test-suite</tt> 1043 module, you must specify <i>--with-externals</i>. This 1044 must be done during the <em>re-configuration</em> step (see above), 1045 and the <tt>llvm</tt> re-configuration must recognize the 1046 previously-built <tt>llvm-gcc</tt>. If any of these is missing or 1047 neglected, the External tests won't work.</p> 1048<dl> 1049<dt><i>--with-externals</i></dt> 1050<dt><i>--with-externals=<<tt>directory</tt>></i></dt> 1051</dl> 1052 This tells LLVM where to find any external tests. They are expected to be 1053 in specifically named subdirectories of <<tt>directory</tt>>. 1054 If <tt>directory</tt> is left unspecified, 1055 <tt>configure</tt> uses the default value 1056 <tt>/home/vadve/shared/benchmarks/speccpu2000/benchspec</tt>. 1057 Subdirectory names known to LLVM include: 1058 <dl> 1059 <dt>spec95</dt> 1060 <dt>speccpu2000</dt> 1061 <dt>speccpu2006</dt> 1062 <dt>povray31</dt> 1063 </dl> 1064 Others are added from time to time, and can be determined from 1065 <tt>configure</tt>. 1066</div> 1067 1068<!-- _______________________________________________________________________ --> 1069<h3> 1070 <a name="testsuitetests">Running different tests</a> 1071</h3> 1072<!-- _______________________________________________________________________ --> 1073<div> 1074<p>In addition to the regular "whole program" tests, the <tt>test-suite</tt> 1075module also provides a mechanism for compiling the programs in different ways. 1076If the variable TEST is defined on the <tt>gmake</tt> command line, the test system will 1077include a Makefile named <tt>TEST.<value of TEST variable>.Makefile</tt>. 1078This Makefile can modify build rules to yield different results.</p> 1079 1080<p>For example, the LLVM nightly tester uses <tt>TEST.nightly.Makefile</tt> to 1081create the nightly test reports. To run the nightly tests, run <tt>gmake 1082TEST=nightly</tt>.</p> 1083 1084<p>There are several TEST Makefiles available in the tree. Some of them are 1085designed for internal LLVM research and will not work outside of the LLVM 1086research group. They may still be valuable, however, as a guide to writing your 1087own TEST Makefile for any optimization or analysis passes that you develop with 1088LLVM.</p> 1089 1090</div> 1091 1092<!-- _______________________________________________________________________ --> 1093<h3> 1094 <a name="testsuiteoutput">Generating test output</a> 1095</h3> 1096<!-- _______________________________________________________________________ --> 1097<div> 1098 <p>There are a number of ways to run the tests and generate output. The most 1099 simple one is simply running <tt>gmake</tt> with no arguments. This will 1100 compile and run all programs in the tree using a number of different methods 1101 and compare results. Any failures are reported in the output, but are likely 1102 drowned in the other output. Passes are not reported explicitely.</p> 1103 1104 <p>Somewhat better is running <tt>gmake TEST=sometest test</tt>, which runs 1105 the specified test and usually adds per-program summaries to the output 1106 (depending on which sometest you use). For example, the <tt>nightly</tt> test 1107 explicitely outputs TEST-PASS or TEST-FAIL for every test after each program. 1108 Though these lines are still drowned in the output, it's easy to grep the 1109 output logs in the Output directories.</p> 1110 1111 <p>Even better are the <tt>report</tt> and <tt>report.format</tt> targets 1112 (where <tt>format</tt> is one of <tt>html</tt>, <tt>csv</tt>, <tt>text</tt> or 1113 <tt>graphs</tt>). The exact contents of the report are dependent on which 1114 <tt>TEST</tt> you are running, but the text results are always shown at the 1115 end of the run and the results are always stored in the 1116 <tt>report.<type>.format</tt> file (when running with 1117 <tt>TEST=<type></tt>). 1118 1119 The <tt>report</tt> also generate a file called 1120 <tt>report.<type>.raw.out</tt> containing the output of the entire test 1121 run. 1122</div> 1123 1124<!-- _______________________________________________________________________ --> 1125<h3> 1126 <a name="testsuitecustom">Writing custom tests for the test suite</a> 1127</h3> 1128<!-- _______________________________________________________________________ --> 1129 1130<div> 1131 1132<p>Assuming you can run the test suite, (e.g. "<tt>gmake TEST=nightly report</tt>" 1133should work), it is really easy to run optimizations or code generator 1134components against every program in the tree, collecting statistics or running 1135custom checks for correctness. At base, this is how the nightly tester works, 1136it's just one example of a general framework.</p> 1137 1138<p>Lets say that you have an LLVM optimization pass, and you want to see how 1139many times it triggers. First thing you should do is add an LLVM 1140<a href="ProgrammersManual.html#Statistic">statistic</a> to your pass, which 1141will tally counts of things you care about.</p> 1142 1143<p>Following this, you can set up a test and a report that collects these and 1144formats them for easy viewing. This consists of two files, a 1145"<tt>test-suite/TEST.XXX.Makefile</tt>" fragment (where XXX is the name of your 1146test) and a "<tt>test-suite/TEST.XXX.report</tt>" file that indicates how to 1147format the output into a table. There are many example reports of various 1148levels of sophistication included with the test suite, and the framework is very 1149general.</p> 1150 1151<p>If you are interested in testing an optimization pass, check out the 1152"libcalls" test as an example. It can be run like this:<p> 1153 1154<div class="doc_code"> 1155<pre> 1156% cd llvm/projects/test-suite/MultiSource/Benchmarks # or some other level 1157% make TEST=libcalls report 1158</pre> 1159</div> 1160 1161<p>This will do a bunch of stuff, then eventually print a table like this:</p> 1162 1163<div class="doc_code"> 1164<pre> 1165Name | total | #exit | 1166... 1167FreeBench/analyzer/analyzer | 51 | 6 | 1168FreeBench/fourinarow/fourinarow | 1 | 1 | 1169FreeBench/neural/neural | 19 | 9 | 1170FreeBench/pifft/pifft | 5 | 3 | 1171MallocBench/cfrac/cfrac | 1 | * | 1172MallocBench/espresso/espresso | 52 | 12 | 1173MallocBench/gs/gs | 4 | * | 1174Prolangs-C/TimberWolfMC/timberwolfmc | 302 | * | 1175Prolangs-C/agrep/agrep | 33 | 12 | 1176Prolangs-C/allroots/allroots | * | * | 1177Prolangs-C/assembler/assembler | 47 | * | 1178Prolangs-C/bison/mybison | 74 | * | 1179... 1180</pre> 1181</div> 1182 1183<p>This basically is grepping the -stats output and displaying it in a table. 1184You can also use the "TEST=libcalls report.html" target to get the table in HTML 1185form, similarly for report.csv and report.tex.</p> 1186 1187<p>The source for this is in test-suite/TEST.libcalls.*. The format is pretty 1188simple: the Makefile indicates how to run the test (in this case, 1189"<tt>opt -simplify-libcalls -stats</tt>"), and the report contains one line for 1190each column of the output. The first value is the header for the column and the 1191second is the regex to grep the output of the command for. There are lots of 1192example reports that can do fancy stuff.</p> 1193 1194</div> 1195 1196</div> 1197 1198<!-- *********************************************************************** --> 1199 1200<hr> 1201<address> 1202 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img 1203 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a> 1204 <a href="http://validator.w3.org/check/referer"><img 1205 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a> 1206 1207 John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner<br> 1208 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br> 1209 Last modified: $Date: 2011-05-18 14:07:16 -0400 (Wed, 18 May 2011) $ 1210</address> 1211</body> 1212</html> 1213