docs/ReleaseNotes.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
   2                       "http://www.w3.org/TR/html4/strict.dtd">
   3 <html>
   4 <head>
   5   <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
   6   <meta encoding="utf8">
   7   <link rel="stylesheet" href="llvm.css" type="text/css">
   8   <title>LLVM 2.8 Release Notes</title>
   9 </head>
  10 <body>
  11
  12 <div class="doc_title">LLVM 2.8 Release Notes</div>
  13
  14 <img align=right src="http://llvm.org/img/DragonSmall.png"
  15     width="136" height="136" alt="LLVM Dragon Logo">
  16
  17 <ol>
  18   <li><a href="#intro">Introduction</a></li>
  19   <li><a href="#subproj">Sub-project Status Update</a></li>
  20   <li><a href="#externalproj">External Projects Using LLVM 2.8</a></li>
  21   <li><a href="#whatsnew">What's New in LLVM 2.8?</a></li>
  22   <li><a href="GettingStarted.html">Installation Instructions</a></li>
  23   <li><a href="#knownproblems">Known Problems</a></li>
  24   <li><a href="#additionalinfo">Additional Information</a></li>
  25 </ol>
  26
  27 <div class="doc_author">
  28   <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
  29 </div>
  30
  31 <!--
  32 <h1 style="color:red">These are in-progress notes for the upcoming LLVM 2.8
  33 release.<br>
  34 You may prefer the
  35 <a href="http://llvm.org/releases/2.7/docs/ReleaseNotes.html">LLVM 2.7
  36 Release Notes</a>.</h1>
  37 -->
  38
  39 <!-- *********************************************************************** -->
  40 <div class="doc_section">
  41   <a name="intro">Introduction</a>
  42 </div>
  43 <!-- *********************************************************************** -->
  44
  45 <div class="doc_text">
  46
  47 <p>This document contains the release notes for the LLVM Compiler
  48 Infrastructure, release 2.8.  Here we describe the status of LLVM, including
  49 major improvements from the previous release and significant known problems.
  50 All LLVM releases may be downloaded from the <a
  51 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
  52
  53 <p>For more information about LLVM, including information about the latest
  54 release, please check out the <a href="http://llvm.org/">main LLVM
  55 web site</a>.  If you have questions or comments, the <a
  56 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's
  57 Mailing List</a> is a good place to send them.</p>
  58
  59 <p>Note that if you are reading this file from a Subversion checkout or the
  60 main LLVM web page, this document applies to the <i>next</i> release, not the
  61 current one.  To see the release notes for a specific release, please see the
  62 <a href="http://llvm.org/releases/">releases page</a>.</p>
  63
  64 </div>
  65
  66
  67 <!--
  68 Almost dead code.
  69   include/llvm/Analysis/LiveValues.h => Dan
  70   lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
  71   GEPSplitterPass
  72 -->
  73
  74
  75 <!-- Features that need text if they're finished for 2.9:
  76   combiner-aa?
  77   strong phi elim
  78   loop dependence analysis
  79   TBAA
  80   CorrelatedValuePropagation
  81  -->
  82
  83  <!-- Announcement, lldb, libc++ -->
  84
  85
  86 <!-- *********************************************************************** -->
  87 <div class="doc_section">
  88   <a name="subproj">Sub-project Status Update</a>
  89 </div>
  90 <!-- *********************************************************************** -->
  91
  92 <div class="doc_text">
  93 <p>
  94 The LLVM 2.8 distribution currently consists of code from the core LLVM
  95 repository (which roughly includes the LLVM optimizers, code generators
  96 and supporting tools), the Clang repository and the llvm-gcc repository.  In
  97 addition to this code, the LLVM Project includes other sub-projects that are in
  98 development.  Here we include updates on these subprojects.
  99 </p>
 100
 101 </div>
 102
 103
 104 <!--=========================================================================-->
 105 <div class="doc_subsection">
 106 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
 107 </div>
 108
 109 <div class="doc_text">
 110
 111 <p><a href="http://clang.llvm.org/">Clang</a> is an LLVM front end for the C,
 112 C++, and Objective-C languages. Clang aims to provide a better user experience
 113 through expressive diagnostics, a high level of conformance to language
 114 standards, fast compilation, and low memory use. Like LLVM, Clang provides a
 115 modular, library-based architecture that makes it suitable for creating or
 116 integrating with other development tools. Clang is considered a
 117 production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
 118 (32- and 64-bit), and for darwin-arm targets.</p>
 119
 120 <p>In the LLVM 2.8 time-frame, the Clang team has made many improvements:</p>
 121
 122   <ul>
 123     <li>Clang C++ is now feature-complete with respect to the ISO C++ 1998 and 2003 standards.</li>
 124     <li>Added support for Objective-C++.</li>
 125     <li>Clang now uses LLVM-MC to directly generate object code and to parse inline assembly (on Darwin).</li>
 126     <li>Introduced many new warnings, including <code>-Wmissing-field-initializers</code>, <code>-Wshadow</code>, <code>-Wno-protocol</code>, <code>-Wtautological-compare</code>, <code>-Wstrict-selector-match</code>, <code>-Wcast-align</code>, <code>-Wunused</code> improvements, and greatly improved format-string checking.</li>
 127     <li>Introduced the "libclang" library, a C interface to Clang intended to support IDE clients.</li>
 128     <li>Added support for <code>#pragma GCC visibility</code>, <code>#pragma align</code>, and others.</li>
 129     <li>Added support for SSE, AVX, ARM NEON, and AltiVec.</li>
 130     <li>Improved support for many Microsoft extensions.</li>
 131     <li>Implemented support for blocks in C++.</li>
 132     <li>Implemented precompiled headers for C++.</li>
 133     <li>Improved abstract syntax trees to retain more accurate source information.</li>
 134     <li>Added driver support for handling LLVM IR and bitcode files directly.</li>
 135     <li>Major improvements to compiler correctness for exception handling.</li>
 136     <li>Improved generated code quality in some areas:
 137       <ul>
 138         <li>Good code generation for X86-32 and X86-64 ABI handling.</li>
 139         <li>Improved code generation for bit-fields, although important work remains.</li>
 140       </ul>
 141     </li>
 142   </ul>
 143 </div>
 144
 145 <!--=========================================================================-->
 146 <div class="doc_subsection">
 147 <a name="clangsa">Clang Static Analyzer</a>
 148 </div>
 149
 150 <div class="doc_text">
 151
 152 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
 153    project is an effort to use static source code analysis techniques to
 154    automatically find bugs in C and Objective-C programs (and hopefully <a
 155    href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
 156    future</a>!).  The tool is very good at finding bugs that occur on specific
 157    paths through code, such as on error conditions.</p>
 158
 159 <p>The LLVM 2.8 release fixes a number of bugs and slightly improves precision
 160    over 2.7, but there are no major new features in the release.
 161 </p>
 162
 163 </div>
 164
 165 <!--=========================================================================-->
 166 <div class="doc_subsection">
 167 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
 168 </div>
 169
 170 <div class="doc_text">
 171 <p>
 172 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
 173 gcc-4.5.  Unlike llvm-gcc, dragonegg in theory does not require any gcc-4.5
 174 modifications whatsoever (currently one small patch is needed) thanks to the
 175 new <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>.
 176 DragonEgg is a gcc plugin that makes gcc-4.5 use the LLVM optimizers and code
 177 generators instead of gcc's, just like with llvm-gcc.
 178 </p>
 179
 180 <p>
 181 DragonEgg is still a work in progress, but it is able to compile a lot of code,
 182 for example all of gcc, LLVM and clang.  Currently Ada, C, C++ and Fortran work
 183 well, while all other languages either don't work at all or only work poorly.
 184 For the moment only the x86-32 and x86-64 targets are supported, and only on
 185 linux and darwin (darwin may need additional gcc patches).
 186 </p>
 187
 188 <p>
 189 The 2.8 release has the following notable changes:
 190 <ul>
 191 <li>The plugin loads faster due to exporting fewer symbols.</li>
 192 <li>Additional vector operations such as addps256 are now supported.</li>
 193 <li>Ada global variables with no initial value are no longer zero initialized,
 194 resulting in better optimization.</li>
 195 <li>The '-fplugin-arg-dragonegg-enable-gcc-optzns' flag now runs all gcc
 196 optimizers, rather than just a handful.</li>
 197 <li>Fortran programs using common variables now link correctly.</li>
 198 <li>GNU OMP constructs no longer crash the compiler.</li>
 199 </ul>
 200
 201 </div>
 202
 203 <!--=========================================================================-->
 204 <div class="doc_subsection">
 205 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
 206 </div>
 207
 208 <div class="doc_text">
 209 <p>
 210 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
 211 a Java Virtual Machine (Java VM or JVM) that uses LLVM for static and
 212 just-in-time compilation.  As of LLVM 2.8, VMKit now supports copying garbage
 213 collectors, and can be configured to use MMTk's copy mark-sweep garbage
 214 collector.  In LLVM 2.8, the VMKit .NET VM is no longer being maintained.
 215 </p>
 216 </div>
 217
 218 <!--=========================================================================-->
 219 <div class="doc_subsection">
 220 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
 221 </div>
 222
 223 <div class="doc_text">
 224 <p>
 225 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
 226 is a simple library that provides an implementation of the low-level
 227 target-specific hooks required by code generation and other runtime components.
 228 For example, when compiling for a 32-bit target, converting a double to a 64-bit
 229 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
 230 function. The compiler-rt library provides highly optimized implementations of
 231 this and other low-level routines (some are 3x faster than the equivalent
 232 libgcc routines).</p>
 233
 234 <p>
 235 All of the code in the compiler-rt project is available under the standard LLVM
 236 License, a "BSD-style" license.  New in LLVM 2.8, compiler_rt now supports
 237 soft floating point (for targets that don't have a real floating point unit),
 238 and includes an extensive testsuite for the "blocks" language feature and the
 239 blocks runtime included in compiler_rt.</p>
 240
 241 </div>
 242
 243 <!--=========================================================================-->
 244 <div class="doc_subsection">
 245 <a name="lldb">LLDB: Low Level Debugger</a>
 246 </div>
 247
 248 <div class="doc_text">
 249 <p>
 250 <a href="http://lldb.llvm.org/">LLDB</a> is a brand new member of the LLVM
 251 umbrella of projects. LLDB is a next generation, high-performance debugger. It
 252 is built as a set of reusable components which highly leverage existing
 253 libraries in the larger LLVM Project, such as the Clang expression parser, the
 254 LLVM disassembler and the LLVM JIT.</p>
 255
 256 <p>
 257 LLDB is in early development and not included as part of the LLVM 2.8 release,
 258 but is mature enough to support basic debugging scenarios on Mac OS X in C,
 259 Objective-C and C++.  We'd really like help extending and expanding LLDB to
 260 support new platforms, new languages, new architectures, and new features.
 261 </p>
 262
 263 </div>
 264
 265 <!--=========================================================================-->
 266 <div class="doc_subsection">
 267 <a name="libc++">libc++: C++ Standard Library</a>
 268 </div>
 269
 270 <div class="doc_text">
 271 <p>
 272 <a href="http://libcxx.llvm.org/">libc++</a> is another new member of the LLVM
 273 family.  It is an implementation of the C++ standard library, written from the
 274 ground up to specifically target the forthcoming C++'0X standard and focus on
 275 delivering great performance.</p>
 276
 277 <p>
 278 As of the LLVM 2.8 release, libc++ is virtually feature complete, but would
 279 benefit from more testing and better integration with Clang++.  It is also
 280 looking forward to the C++ committee finalizing the C++'0x standard.
 281 </p>
 282
 283 </div>
 284
 285
 286
 287 <!--=========================================================================-->
 288 <div class="doc_subsection">
 289 <a name="klee">KLEE: A Symbolic Execution Virtual Machine</a>
 290 </div>
 291
 292 <div class="doc_text">
 293 <p>
 294 <a href="http://klee.llvm.org/">KLEE</a> is a symbolic execution framework for
 295 programs in LLVM bitcode form. KLEE tries to symbolically evaluate "all" paths
 296 through the application and records state transitions that lead to fault
 297 states. This allows it to construct testcases that lead to faults and can even
 298 be used to verify some algorithms.
 299 </p>
 300
 301 <p>Although KLEE does not have any major new features as of 2.8, we have made
 302 various minor improvements, particular to ease development:</p>
 303 <ul>
 304   <li>Added support for LLVM 2.8. KLEE currently maintains compatibility with
 305     LLVM 2.6, 2.7, and 2.8.</li>
 306   <li>Added a buildbot for 2.6, 2.7, and trunk. A 2.8 buildbot will be coming
 307     soon following release.</li>
 308   <li>Fixed many C++ code issues to allow building with Clang++. Mostly
 309     complete, except for the version of MiniSAT which is inside the KLEE STP
 310     version.</li>
 311   <li>Improved support for building with separate source and build
 312     directories.</li>
 313   <li>Added support for "long double" on x86.</li>
 314   <li>Initial work on KLEE support for using 'lit' test runner instead of
 315     DejaGNU.</li>
 316   <li>Added <tt>configure</tt> support for using an external version of
 317     STP.</li>
 318 </ul>
 319
 320 </div>
 321
 322
 323 <!-- *********************************************************************** -->
 324 <div class="doc_section">
 325   <a name="externalproj">External Open Source Projects Using LLVM 2.8</a>
 326 </div>
 327 <!-- *********************************************************************** -->
 328
 329 <div class="doc_text">
 330
 331 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
 332    a lot of other language and tools projects.  This section lists some of the
 333    projects that have already been updated to work with LLVM 2.8.</p>
 334 </div>
 335
 336 <!--=========================================================================-->
 337 <div class="doc_subsection">
 338 <a name="tce">TTA-based Codesign Environment (TCE)</a>
 339 </div>
 340
 341 <div class="doc_text">
 342 <p>
 343 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
 344 application-specific processors (ASP) based on the Transport triggered
 345 architecture (TTA). The toolset provides a complete co-design flow from C/C++
 346 programs down to synthesizable VHDL and parallel program binaries. Processor
 347 customization points include the register files, function units, supported
 348 operations, and the interconnection network.</p>
 349
 350 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
 351 independent optimizations and also for parts of code generation. It generates
 352 new LLVM-based code generators "on the fly" for the designed TTA processors and
 353 loads them in to the compiler backend as runtime libraries to avoid per-target
 354 recompilation of larger parts of the compiler chain.</p>
 355
 356 </div>
 357
 358 <!--=========================================================================-->
 359 <div class="doc_subsection">
 360 <a name="Horizon">Horizon Bytecode Compiler</a>
 361 </div>
 362
 363 <div class="doc_text">
 364 <p>
 365 <a href="http://www.quokforge.org/projects/horizon">Horizon</a> is a bytecode
 366 language and compiler written on top of LLVM, intended for producing
 367 single-address-space managed code operating systems that
 368 run faster than the equivalent multiple-address-space C systems.
 369 More in-depth blurb is available on the <a
 370 href="http://www.quokforge.org/projects/horizon/wiki/Wiki">wiki</a>.</p>
 371
 372 </div>
 373
 374 <!--=========================================================================-->
 375 <div class="doc_subsection">
 376 <a name="clamav">Clam AntiVirus</a>
 377 </div>
 378
 379 <div class="doc_text">
 380 <p>
 381 <a href="http://www.clamav.net">Clam AntiVirus</a> is an open source (GPL)
 382 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
 383 gateways.  Since version 0.96 it has <a
 384 href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
 385 signatures</a> that allow writing detections for complex malware. It
 386 uses LLVM's JIT to speed up the execution of bytecode on
 387 X86, X86-64, PPC32/64, falling back to its own interpreter otherwise.
 388 The git version was updated to work with LLVM 2.8.
 389 </p>
 390
 391 <p>The <a
 392 href="http://git.clamav.net/gitweb?p=clamav-bytecode-compiler.git;a=blob_plain;f=docs/user/clambc-user.pdf">
 393 ClamAV bytecode compiler</a> uses Clang and LLVM to compile a C-like
 394 language, insert runtime checks, and generate ClamAV bytecode.</p>
 395
 396 </div>
 397
 398 <!--=========================================================================-->
 399 <div class="doc_subsection">
 400 <a name="pure">Pure</a>
 401 </div>
 402
 403 <div class="doc_text">
 404 <p>
 405 <a href="http://pure-lang.googlecode.com/">Pure</a>
 406 is an algebraic/functional
 407 programming language based on term rewriting. Programs are collections
 408 of equations which are used to evaluate expressions in a symbolic
 409 fashion. Pure offers dynamic typing, eager and lazy evaluation, lexical
 410 closures, a hygienic macro system (also based on term rewriting),
 411 built-in list and matrix support (including list and matrix
 412 comprehensions) and an easy-to-use C interface. The interpreter uses
 413 LLVM as a backend to JIT-compile Pure programs to fast native code.</p>
 414
 415 <p>Pure versions 0.44 and later have been tested and are known to work with
 416 LLVM 2.8 (and continue to work with older LLVM releases >= 2.5).</p>
 417
 418 </div>
 419
 420 <!--=========================================================================-->
 421 <div class="doc_subsection">
 422 <a name="GHC">Glasgow Haskell Compiler (GHC)</a>
 423 </div>
 424
 425 <div class="doc_text">
 426 <p>
 427 <a href="http://www.haskell.org/ghc/">GHC</a> is an open source,
 428 state-of-the-art programming suite for
 429 Haskell, a standard lazy functional programming language. It includes
 430 an optimizing static compiler generating good code for a variety of
 431 platforms, together with an interactive system for convenient, quick
 432 development.</p>
 433
 434 <p>In addition to the existing C and native code generators, GHC 7.0 now
 435 supports an <a
 436 href="http://hackage.haskell.org/trac/ghc/wiki/Commentary/Compiler/Backends/LLVM">LLVM
 437 code generator</a>. GHC supports LLVM 2.7 and later.</p>
 438
 439 </div>
 440
 441 <!--=========================================================================-->
 442 <div class="doc_subsection">
 443 <a name="Clay">Clay Programming Language</a>
 444 </div>
 445
 446 <div class="doc_text">
 447 <p>
 448 <a href="http://tachyon.in/clay/">Clay</a> is a new systems programming
 449 language that is specifically designed for generic programming. It makes
 450 generic programming very concise thanks to whole program type propagation. It
 451 uses LLVM as its backend.</p>
 452
 453 </div>
 454
 455 <!--=========================================================================-->
 456 <div class="doc_subsection">
 457 <a name="llvm-py">llvm-py Python Bindings for LLVM</a>
 458 </div>
 459
 460 <div class="doc_text">
 461 <p>
 462 <a href="http://www.mdevan.org/llvm-py/">llvm-py</a> has been updated to work
 463 with LLVM 2.8.  llvm-py provides Python bindings for LLVM, allowing you to write a
 464 compiler backend or a VM in Python.</p>
 465
 466 </div>
 467
 468
 469 <!--=========================================================================-->
 470 <div class="doc_subsection">
 471 <a name="FAUST">FAUST Real-Time Audio Signal Processing Language</a>
 472 </div>
 473
 474 <div class="doc_text">
 475 <p>
 476 <a href="http://faust.grame.fr">FAUST</a> is a compiled language for real-time
 477 audio signal processing. The name FAUST stands for Functional AUdio STream. Its
 478 programming model combines two approaches: functional programming and block
 479 diagram composition. In addition with the C, C++, JAVA output formats, the
 480 Faust compiler can now generate LLVM bitcode, and works with LLVM 2.7 and
 481 2.8.</p>
 482
 483 </div>
 484
 485 <!--=========================================================================-->
 486 <div class="doc_subsection">
 487 <a name="jade">Jade Just-in-time Adaptive Decoder Engine</a>
 488 </div>
 489
 490 <div class="doc_text">
 491 <p><a
 492 href="http://sourceforge.net/apps/trac/orcc/wiki/JadeDocumentation">Jade</a>
 493 (Just-in-time Adaptive Decoder Engine) is a generic video decoder engine using
 494 LLVM for just-in-time compilation of video decoder configurations. Those
 495 configurations are designed by MPEG Reconfigurable Video Coding (RVC) committee.
 496 MPEG RVC standard is built on a stream-based dataflow representation of
 497 decoders. It is composed of a standard library of coding tools written in
 498 RVC-CAL language and a dataflow configuration &#8212; block diagram &#8212;
 499 of a decoder.</p>
 500
 501 <p>Jade project is hosted as part of the <a href="http://orcc.sf.net">Open
 502 RVC-CAL Compiler</a> and requires it to translate the RVC-CAL standard library
 503 of video coding tools into an LLVM assembly code.</p>
 504
 505 </div>
 506
 507 <!--=========================================================================-->
 508 <div class="doc_subsection">
 509 <a name="neko_llvm_jit">LLVM JIT for Neko VM</a>
 510 </div>
 511
 512 <div class="doc_text">
 513 <p><a href="http://github.com/vava/neko_llvm_jit">Neko LLVM JIT</a>
 514 replaces the standard Neko JIT with an LLVM-based implementation.  While not
 515 fully complete, it is already providing a 1.5x speedup on 64-bit systems.
 516 Neko LLVM JIT requires LLVM 2.8 or later.</p>
 517
 518 </div>
 519
 520 <!--=========================================================================-->
 521 <div class="doc_subsection">
 522 <a name="crack">Crack Scripting Language</a>
 523 </div>
 524
 525 <div class="doc_text">
 526 <p>
 527 <a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide
 528 the ease of development of a scripting language with the performance of a
 529 compiled language. The language derives concepts from C++, Java and Python,
 530 incorporating object-oriented programming, operator overloading and strong
 531 typing.  Crack 0.2 works with LLVM 2.7, and the forthcoming Crack 0.2.1 release
 532 builds on LLVM 2.8.</p>
 533
 534 </div>
 535
 536 <!--=========================================================================-->
 537 <div class="doc_subsection">
 538 <a name="DresdenTM">Dresden TM Compiler (DTMC)</a>
 539 </div>
 540
 541 <div class="doc_text">
 542 <p>
 543 <a href="http://tm.inf.tu-dresden.de">DTMC</a> provides support for
 544 Transactional Memory, which is an easy-to-use and efficient way to synchronize
 545 accesses to shared memory. Transactions can contain normal C/C++ code (e.g.,
 546 <code>__transaction { list.remove(x); x.refCount--; }</code>) and will be executed
 547 virtually atomically and isolated from other transactions.</p>
 548
 549 </div>
 550
 551 <!--=========================================================================-->
 552 <div class="doc_subsection">
 553 <a name="Kai">Kai Programming Language</a>
 554 </div>
 555
 556 <div class="doc_text">
 557 <p>
 558 <a href="http://www.oriontransfer.co.nz/research/kai">Kai</a> (Japanese 会 for
 559 meeting/gathering) is an experimental interpreter that provides a highly
 560 extensible runtime environment and explicit control over the compilation
 561 process. Programs are defined using nested symbolic expressions, which are all
 562 parsed into first-class values with minimal intrinsic semantics. Kai can
 563 generate optimised code at run-time (using LLVM) in order to exploit the nature
 564 of the underlying hardware and to integrate with external software libraries.
 565 It is a unique exploration into world of dynamic code compilation, and the
 566 interaction between high level and low level semantics.</p>
 567
 568 </div>
 569
 570 <!--=========================================================================-->
 571 <div class="doc_subsection">
 572 <a name="OSL">OSL: Open Shading Language</a>
 573 </div>
 574
 575 <div class="doc_text">
 576 <p>
 577 <a href="http://code.google.com/p/openshadinglanguage/">OSL</a> is a shading
 578 language designed for use in physically based renderers and in particular
 579 production rendering. By using LLVM instead of the interpreter, it was able to
 580 meet its performance goals (&gt;= C-code) while retaining the benefits of
 581 runtime specialization and a portable high-level language.
 582 </p>
 583
 584 </div>
 585
 586
 587
 588 <!-- *********************************************************************** -->
 589 <div class="doc_section">
 590   <a name="whatsnew">What's New in LLVM 2.8?</a>
 591 </div>
 592 <!-- *********************************************************************** -->
 593
 594 <div class="doc_text">
 595
 596 <p>This release includes a huge number of bug fixes, performance tweaks and
 597 minor improvements.  Some of the major improvements and new features are listed
 598 in this section.
 599 </p>
 600
 601 </div>
 602
 603 <!--=========================================================================-->
 604 <div class="doc_subsection">
 605 <a name="majorfeatures">Major New Features</a>
 606 </div>
 607
 608 <div class="doc_text">
 609
 610 <p>LLVM 2.8 includes several major new capabilities:</p>
 611
 612 <ul>
 613 <li>As mentioned above, <a href="#libc++">libc++</a> and <a
 614    href="#lldb">LLDB</a> are major new additions to the LLVM collective.</li>
 615 <li>LLVM 2.8 now has pretty decent support for debugging optimized code.  You
 616     should be able to reliably get debug info for function arguments, assuming
 617     that the value is actually available where you have stopped.</li>
 618 <li>A new 'llvm-diff' tool is available that does a semantic diff of .ll
 619     files.</li>
 620 <li>The <a href="#mc">MC subproject</a> has made major progress in this release.
 621     Direct .o file writing support for darwin/x86[-64] is now reliable and
 622     support for other targets and object file formats are in progress.</li>
 623 </ul>
 624
 625 </div>
 626
 627 <!--=========================================================================-->
 628 <div class="doc_subsection">
 629 <a name="coreimprovements">LLVM IR and Core Improvements</a>
 630 </div>
 631
 632 <div class="doc_text">
 633 <p>LLVM IR has several new features for better support of new targets and that
 634 expose new optimization opportunities:</p>
 635
 636 <ul>
 637 <li>The <a href="LangRef.html#int_libc">memcpy, memmove, and memset</a>
 638   intrinsics now take address space qualified pointers and a bit to indicate
 639   whether the transfer is "<a href="LangRef.html#volatile">volatile</a>" or not.
 640 </li>
 641 <li>Per-instruction debug info metadata is much faster and uses less memory by
 642     using the new DebugLoc class.</li>
 643 <li>LLVM IR now has a more formalized concept of "<a
 644     href="LangRef.html#trapvalues">trap values</a>", which allow the optimizer
 645     to optimize more aggressively in the presence of undefined behavior, while
 646     still producing predictable results.</li>
 647 <li>LLVM IR now supports two new <a href="LangRef.html#linkage">linkage
 648     types</a> (linker_private_weak and linker_private_weak_def_auto) which map
 649     onto some obscure MachO concepts.</li>
 650 </ul>
 651
 652 </div>
 653
 654 <!--=========================================================================-->
 655 <div class="doc_subsection">
 656 <a name="optimizer">Optimizer Improvements</a>
 657 </div>
 658
 659 <div class="doc_text">
 660
 661 <p>In addition to a large array of minor performance tweaks and bug fixes, this
 662 release includes a few major enhancements and additions to the optimizers:</p>
 663
 664 <ul>
 665 <li>As mentioned above, the optimizer now has support for updating debug
 666    information as it goes.  A key aspect of this is the new <a
 667    href="SourceLevelDebugging.html#format_common_value">llvm.dbg.value</a>
 668    intrinsic.  This intrinsic represents debug info for variables that are
 669    promoted to SSA values (typically by mem2reg or the -scalarrepl passes).</li>
 670
 671 <li>The JumpThreading pass is now much more aggressive about implied value
 672     relations, allowing it to thread conditions like "a == 4" when a is known to
 673     be 13 in one of the predecessors of a block.  It does this in conjunction
 674     with the new LazyValueInfo analysis pass.</li>
 675 <li>The new RegionInfo analysis pass identifies single-entry single-exit regions
 676     in the CFG.  You can play with it with the "opt -regions -analyze" or
 677     "opt -view-regions" commands.</li>
 678 <li>The loop optimizer has significantly improved strength reduction and analysis
 679   capabilities.  Notably it is able to build on the trap value and signed
 680   integer overflow information to optimize &lt;= and &gt;= loops.</li>
 681 <li>The CallGraphSCCPassManager now has some basic support for iterating within
 682     an SCC when a optimizer devirtualizes a function call.  This allows inlining
 683     through indirect call sites that are devirtualized by store-load forwarding
 684     and other optimizations.</li>
 685 <li>The new <A href="Passes.html#loweratomic">-loweratomic</a> pass is available
 686     to lower atomic instructions into their non-atomic form.  This can be useful
 687     to optimize generic code that expects to run in a single-threaded
 688     environment.</li>
 689 </ul>
 690
 691 <!--
 692 <p>In addition to these features that are done in 2.8, there is preliminary
 693    support in the release for Type Based Alias Analysis
 694   Preliminary work on TBAA but not usable in 2.8.
 695   New CorrelatedValuePropagation pass, not on by default in 2.8 yet.
 696 -->
 697
 698 </div>
 699
 700 <!--=========================================================================-->
 701 <div class="doc_subsection">
 702 <a name="mc">MC Level Improvements</a>
 703 </div>
 704
 705 <div class="doc_text">
 706 <p>
 707 The LLVM Machine Code (aka MC) subsystem was created to solve a number
 708 of problems in the realm of assembly, disassembly, object file format handling,
 709 and a number of other related areas that CPU instruction-set level tools work
 710 in.</p>
 711
 712 <p>The MC subproject has made great leaps in LLVM 2.8.  For example, support for
 713    directly writing .o files from LLC (and clang) now works reliably for
 714    darwin/x86[-64] (including inline assembly support) and the integrated
 715    assembler is turned on by default in Clang for these targets.  This provides
 716    improved compile times among other things.</p>
 717
 718 <ul>
 719 <li>The entire compiler has converted over to using the MCStreamer assembler API
 720     instead of writing out a .s file textually.</li>
 721 <li>The "assembler parser" is far more mature than in 2.7, supporting a full
 722     complement of directives, now supports assembler macros, etc.</li>
 723 <li>The "assembler backend" has been completed, including support for relaxation
 724     relocation processing and all the other things that an assembler does.</li>
 725 <li>The MachO file format support is now fully functional and works.</li>
 726 <li>The MC disassembler now fully supports ARM and Thumb.  ARM assembler support
 727     is still in early development though.</li>
 728 <li>The X86 MC assembler now supports the X86 AES and AVX instruction set.</li>
 729 <li>Work on ELF and COFF object files and ARM target support is well underway,
 730     but isn't useful yet in LLVM 2.8.  Please contact the llvmdev mailing list
 731     if you're interested in this.</li>
 732 </ul>
 733
 734 <p>For more information, please see the <a
 735 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
 736 LLVM MC Project Blog Post</a>.
 737 </p>
 738
 739 </div>
 740
 741
 742 <!--=========================================================================-->
 743 <div class="doc_subsection">
 744 <a name="codegen">Target Independent Code Generator Improvements</a>
 745 </div>
 746
 747 <div class="doc_text">
 748
 749 <p>We have put a significant amount of work into the code generator
 750 infrastructure, which allows us to implement more aggressive algorithms and make
 751 it run faster:</p>
 752
 753 <ul>
 754 <li>The clang/gcc -momit-leaf-frame-pointer argument is now supported.</li>
 755 <li>The clang/gcc -ffunction-sections and -fdata-sections arguments are now
 756     supported on ELF targets (like GCC).</li>
 757 <li>The MachineCSE pass is now tuned and on by default.  It eliminates common
 758     subexpressions that are exposed when lowering to machine instructions.</li>
 759 <li>The "local" register allocator was replaced by a new "fast" register
 760     allocator.  This new allocator (which is often used at -O0) is substantially
 761     faster and produces better code than the old local register allocator.</li>
 762 <li>A new LLC "-regalloc=default" option is available, which automatically
 763     chooses a register allocator based on the -O optimization level.</li>
 764 <li>The common code generator code was modified to promote illegal argument and
 765     return value vectors to wider ones when possible instead of scalarizing
 766     them.  For example, &lt;3 x float&gt; will now pass in one SSE register
 767     instead of 3 on X86.  This generates substantially better code since the
 768     rest of the code generator was already expecting this.</li>
 769 <li>The code generator uses a new "COPY" machine instruction.  This speeds up
 770     the code generator and eliminates the need for targets to implement the
 771     isMoveInstr hook.  Also, the copyRegToReg hook was renamed to copyPhysReg
 772     and simplified.</li>
 773 <li>The code generator now has a "LocalStackSlotPass", which optimizes stack
 774     slot access for targets (like ARM) that have limited stack displacement
 775     addressing.</li>
 776 <li>A new "PeepholeOptimizer" is available, which eliminates sign and zero
 777     extends, and optimizes away compare instructions when the condition result
 778     is available from a previous instruction.</li>
 779 <li>Atomic operations now get legalized into simpler atomic operations if not
 780     natively supported, easing the implementation burden on targets.</li>
 781 <li>We have added two new bottom-up pre-allocation register pressure aware schedulers:
 782 <ol>
 783 <li>The hybrid scheduler schedules aggressively to minimize schedule length when registers are available and avoid overscheduling in high pressure situations.</li>
 784 <li>The instruction-level-parallelism scheduler schedules for maximum ILP when registers are available and avoid overscheduling in high pressure situations.</li>
 785 </ol></li>
 786 <li>The tblgen type inference algorithm was rewritten to be more consistent and
 787      diagnose more target bugs.  If you have an out-of-tree backend, you may
 788      find that it finds bugs in your target description.  This support also
 789      allows limited support for writing patterns for instructions that return
 790      multiple results (e.g. a virtual register and a flag result).  The
 791      'parallel' modifier in tblgen was removed, you should use the new support
 792      for multiple results instead.</li>
 793 <li>A new (experimental) "-rendermf" pass is available which renders a
 794     MachineFunction into HTML, showing live ranges and other useful
 795     details.</li>
 796 <li>The new SubRegIndex tablegen class allows subregisters to be indexed
 797     symbolically instead of numerically.  If your target uses subregisters you
 798     will need to adapt to use SubRegIndex when you upgrade to 2.8.</li>
 799 <!-- SplitKit -->
 800
 801 <li>The -fast-isel instruction selection path (used at -O0 on X86) was rewritten
 802     to work bottom-up on basic blocks instead of top down.  This makes it
 803     slightly faster (because the MachineDCE pass is not needed any longer) and
 804     allows it to generate better code in some cases.</li>
 805
 806 </ul>
 807 </div>
 808
 809 <!--=========================================================================-->
 810 <div class="doc_subsection">
 811 <a name="x86">X86-32 and X86-64 Target Improvements</a>
 812 </div>
 813
 814 <div class="doc_text">
 815 <p>New features and major changes in the X86 target include:
 816 </p>
 817
 818 <ul>
 819 <li>The X86 backend now supports holding X87 floating point stack values
 820     in registers across basic blocks, dramatically improving performance of code
 821     that uses long double, and when targeting CPUs that don't support SSE.</li>
 822
 823 <li>The X86 backend now uses a SSEDomainFix pass to optimize SSE operations.  On
 824     Nehalem ("Core i7") and newer CPUs there is a 2 cycle latency penalty on
 825     using a register in a different domain than where it was defined. This pass
 826     optimizes away these stalls.</li>
 827
 828 <li>The X86 backend now promotes 16-bit integer operations to 32-bits when
 829     possible. This avoids 0x66 prefixes, which are slow on some
 830     microarchitectures and bloat the code on all of them.</li>
 831
 832 <li>The X86 backend now supports the Microsoft "thiscall" calling convention,
 833     and a <a href="LangRef.html#callingconv">calling convention</a> to support
 834     <a href="#GHC">ghc</a>.</li>
 835
 836 <li>The X86 backend supports a new "llvm.x86.int" intrinsic, which maps onto
 837     the X86 "int $42" and "int3" instructions.</li>
 838
 839 <li>At the IR level, the &lt;2 x float&gt; datatype is now promoted and passed
 840     around as a &lt;4 x float&gt; instead of being passed and returned as an MMX
 841     vector.  If you have a frontend that uses this, please pass and return a
 842     &lt;2 x i32&gt; instead (using bitcasts).</li>
 843
 844 <li>When printing .s files in verbose assembly mode (the default for clang -S),
 845     the X86 backend now decodes X86 shuffle instructions and prints human
 846     readable comments after the most inscrutable of them, e.g.:
 847
 848 <pre>
 849   insertps $113, %xmm3, %xmm0 <i># xmm0 = zero,xmm0[1,2],xmm3[1]</i>
 850   unpcklps %xmm1, %xmm0       <i># xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]</i>
 851   pshufd   $1, %xmm1, %xmm1   <i># xmm1 = xmm1[1,0,0,0]</i>
 852 </pre>
 853 </li>
 854
 855 </ul>
 856
 857 </div>
 858
 859 <!--=========================================================================-->
 860 <div class="doc_subsection">
 861 <a name="ARM">ARM Target Improvements</a>
 862 </div>
 863
 864 <div class="doc_text">
 865 <p>New features of the ARM target include:
 866 </p>
 867
 868 <ul>
 869 <li>The ARM backend now optimizes tail calls into jumps.</li>
 870 <li>Scheduling is improved through the new list-hybrid scheduler as well
 871     as through better modeling of structural hazards.</li>
 872 <li><a href="LangRef.html#int_fp16">Half float</a> instructions are now
 873     supported.</li>
 874 <li>NEON support has been improved to model instructions which operate onto
 875     multiple consecutive registers more aggressively.  This avoids lots of
 876     extraneous register copies.</li>
 877 <li>The ARM backend now uses a new "ARMGlobalMerge" pass, which merges several
 878     global variables into one, saving extra address computation (all the global
 879     variables can be accessed via same base address) and potentially reducing
 880     register pressure.</li>
 881
 882 <li>The ARM backend has received many minor improvements and tweaks which lead
 883     to substantially better performance in a wide range of different scenarios.
 884 </li>
 885
 886 <li>The ARM NEON intrinsics have been substantially reworked to reduce
 887     redundancy and improve code generation.  Some of the major changes are:
 888   <ol>
 889   <li>
 890     All of the NEON load and store intrinsics (llvm.arm.neon.vld* and
 891     llvm.arm.neon.vst*) take an extra parameter to specify the alignment in bytes
 892     of the memory being accessed.
 893   </li>
 894   <li>
 895     The llvm.arm.neon.vaba intrinsic (vector absolute difference and
 896     accumulate) has been removed.  This operation is now represented using
 897     the llvm.arm.neon.vabd intrinsic (vector absolute difference) followed by a
 898     vector add.
 899   </li>
 900   <li>
 901     The llvm.arm.neon.vabdl and llvm.arm.neon.vabal intrinsics (lengthening
 902     vector absolute difference with and without accumulation) have been removed.
 903     They are represented using the llvm.arm.neon.vabd intrinsic (vector absolute
 904     difference) followed by a vector zero-extend operation, and for vabal,
 905     a vector add.
 906   </li>
 907   <li>
 908     The llvm.arm.neon.vmovn intrinsic has been removed.  Calls of this intrinsic
 909     are now replaced by vector truncate operations.
 910   </li>
 911   <li>
 912     The llvm.arm.neon.vmovls and llvm.arm.neon.vmovlu intrinsics have been
 913     removed.  They are now represented as vector sign-extend (vmovls) and
 914     zero-extend (vmovlu) operations.
 915   </li>
 916   <li>
 917     The llvm.arm.neon.vaddl*, llvm.arm.neon.vaddw*, llvm.arm.neon.vsubl*, and
 918     llvm.arm.neon.vsubw* intrinsics (lengthening vector add and subtract) have
 919     been removed.  They are replaced by vector add and vector subtract operations
 920     where one (vaddw, vsubw) or both (vaddl, vsubl) of the operands are either
 921     sign-extended or zero-extended.
 922   </li>
 923   <li>
 924     The llvm.arm.neon.vmulls, llvm.arm.neon.vmullu, llvm.arm.neon.vmlal*, and
 925     llvm.arm.neon.vmlsl* intrinsics (lengthening vector multiply with and without
 926     accumulation and subtraction) have been removed.  These operations are now
 927     represented as vector multiplications where the operands are either
 928     sign-extended or zero-extended, followed by a vector add for vmlal or a
 929     vector subtract for vmlsl.  Note that the polynomial vector multiply
 930     intrinsic, llvm.arm.neon.vmullp, remains unchanged.
 931   </li>
 932   </ol>
 933 </li>
 934
 935 </ul>
 936 </div>
 937
 938
 939 <!--=========================================================================-->
 940 <div class="doc_subsection">
 941 <a name="changes">Major Changes and Removed Features</a>
 942 </div>
 943
 944 <div class="doc_text">
 945
 946 <p>If you're already an LLVM user or developer with out-of-tree changes based
 947 on LLVM 2.7, this section lists some "gotchas" that you may run into upgrading
 948 from the previous release.</p>
 949
 950 <ul>
 951 <li>The build configuration machinery changed the output directory names.  It
 952     wasn't clear to many people that a "Release-Asserts" build was a release build
 953     without asserts.  To make this more clear, "Release" does not include
 954     assertions and "Release+Asserts" does (likewise, "Debug" and
 955     "Debug+Asserts").</li>
 956 <li>The MSIL Backend was removed, it was unsupported and broken.</li>
 957 <li>The ABCD, SSI, and SCCVN passes were removed.  These were not fully
 958     functional and their behavior has been or will be subsumed by the
 959     LazyValueInfo  pass.</li>
 960 <li>The LLVM IR 'Union' feature was removed.  While this is a desirable feature
 961     for LLVM IR to support, the existing implementation was half baked and
 962     barely useful.  We'd really like anyone interested to resurrect the work and
 963     finish it for a future release.</li>
 964 <li>If you're used to reading .ll files, you'll probably notice that .ll file
 965     dumps don't produce #uses comments anymore.  To get them, run a .bc file
 966     through "llvm-dis --show-annotations".</li>
 967 <li>Target triples are now stored in a normalized form, and all inputs from
 968     humans are expected to be normalized by Triple::normalize before being
 969     stored in a module triple or passed to another library.</li>
 970 </ul>
 971
 972
 973
 974 <p>In addition, many APIs have changed in this release.  Some of the major LLVM
 975 API changes are:</p>
 976 <ul>
 977 <li>LLVM 2.8 changes the internal order of operands in <a
 978   href="http://llvm.org/doxygen/classllvm_1_1InvokeInst.html"><tt>InvokeInst</tt></a>
 979   and <a href="http://llvm.org/doxygen/classllvm_1_1CallInst.html"><tt>CallInst</tt></a>.
 980   To be portable across releases, please use the <tt>CallSite</tt> class and the
 981   high-level accessors, such as <tt>getCalledValue</tt> and
 982   <tt>setUnwindDest</tt>.
 983 </li>
 984 <li>
 985   You can no longer pass use_iterators directly to cast&lt;&gt; (and similar),
 986   because these routines tend to perform costly dereference operations more
 987   than once. You have to dereference the iterators yourself and pass them in.
 988 </li>
 989 <li>
 990   llvm.memcpy.*, llvm.memset.*, llvm.memmove.* intrinsics take an extra
 991   parameter now ("i1 isVolatile"), totaling 5 parameters, and the pointer
 992   operands are now address-space qualified.
 993   If you were creating these intrinsic calls and prototypes yourself (as opposed
 994   to using Intrinsic::getDeclaration), you can use
 995   UpgradeIntrinsicFunction/UpgradeIntrinsicCall to be portable across releases.
 996 </li>
 997 <li>
 998   SetCurrentDebugLocation takes a DebugLoc now instead of a MDNode.
 999   Change your code to use
1000   SetCurrentDebugLocation(DebugLoc::getFromDILocation(...)).
1001 </li>
1002 <li>
1003   The <tt>RegisterPass</tt> and <tt>RegisterAnalysisGroup</tt> templates are
1004   considered deprecated, but continue to function in LLVM 2.8.  Clients are
1005   strongly advised to use the upcoming <tt>INITIALIZE_PASS()</tt> and
1006   <tt>INITIALIZE_AG_PASS()</tt> macros instead.
1007 </li>
1008 <li>
1009   The constructor for the Triple class no longer tries to understand odd triple
1010   specifications.  Frontends should ensure that they only pass valid triples to
1011   LLVM.  The Triple::normalize utility method has been added to help front-ends
1012   deal with funky triples.
1013 </li>
1014 <li>
1015   The signature of the <tt>GCMetadataPrinter::finishAssembly</tt> virtual
1016   function changed: the <tt>raw_ostream</tt> and <tt>MCAsmInfo</tt> arguments
1017   were dropped.  GC plugins which compute stack maps must be updated to avoid
1018   having the old definition overload the new signature.
1019 </li>
1020 <li>
1021   The signature of <tt>MemoryBuffer::getMemBuffer</tt> changed.  Unfortunately
1022   calls intended for the old version still compile, but will not work correctly,
1023   leading to a confusing error about an invalid header in the bitcode.
1024 </li>
1025
1026 <li>
1027   Some APIs were renamed:
1028   <ul>
1029   <li>llvm_report_error -&gt; report_fatal_error</li>
1030   <li>llvm_install_error_handler -&gt; install_fatal_error_handler</li>
1031   <li>llvm::DwarfExceptionHandling -&gt; llvm::JITExceptionHandling</li>
1032   <li>VISIBILITY_HIDDEN -&gt; LLVM_LIBRARY_VISIBILITY</li>
1033   </ul>
1034 </li>
1035
1036 <li>
1037   Some public headers were renamed:
1038   <ul>
1039     <li><tt>llvm/Assembly/AsmAnnotationWriter.h</tt> was renamed
1040     to <tt>llvm/Assembly/AssemblyAnnotationWriter.h</tt>
1041     </li>
1042   </ul>
1043 </ul>
1044
1045 </div>
1046
1047 <!--=========================================================================-->
1048 <div class="doc_subsection">
1049 <a name="devtree_changes">Development Infrastructure Changes</a>
1050 </div>
1051
1052 <div class="doc_text">
1053
1054 <p>This section lists changes to the LLVM development infrastructure. This
1055 mostly impacts users who actively work on LLVM or follow development on
1056 mainline, but may also impact users who leverage the LLVM build infrastructure
1057 or are interested in LLVM qualification.</p>
1058
1059 <ul>
1060   <li>The default for <tt>make check</tt> is now to use
1061   the <a href="http://llvm.org/cmds/lit.html">lit</a> testing tool, which is
1062   part of LLVM itself. You can use <tt>lit</tt> directly as well, or use
1063   the <tt>llvm-lit</tt> tool which is created as part of a Makefile or CMake
1064   build (and knows how to find the appropriate tools). See the <tt>lit</tt>
1065   documentation and the <a href="http://blog.llvm.org/2009/12/lit-it.html">blog
1066   post</a>, and <a href="http://llvm.org/bugs/show_bug.cgi?id=5217">PR5217</a>
1067   for more information.</li>
1068
1069   <li>The LLVM <tt>test-suite</tt> infrastructure has a new "simple" test format
1070   (<tt>make TEST=simple</tt>). The new format is intended to require only a
1071   compiler and not a full set of LLVM tools. This makes it useful for testing
1072   released compilers, for running the test suite with other compilers (for
1073   performance comparisons), and makes sure that we are testing the compiler as
1074   users would see it. The new format is also designed to work using reference
1075   outputs instead of comparison to a baseline compiler, which makes it run much
1076   faster and makes it less system dependent.</li>
1077
1078   <li>Significant progress has been made on a new interface to running the
1079   LLVM <tt>test-suite</tt> (aka the LLVM "nightly tests") using
1080   the <a href="http://llvm.org/docs/lnt">LNT</a> infrastructure. The LNT
1081   interface to the <tt>test-suite</tt> brings significantly improved reporting
1082   capabilities for monitoring the correctness and generated code quality
1083   produced by LLVM over time.</li>
1084 </ul>
1085 </div>
1086
1087 <!-- *********************************************************************** -->
1088 <div class="doc_section">
1089   <a name="knownproblems">Known Problems</a>
1090 </div>
1091 <!-- *********************************************************************** -->
1092
1093 <div class="doc_text">
1094
1095 <p>This section contains significant known problems with the LLVM system,
1096 listed by component.  If you run into a problem, please check the <a
1097 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
1098 there isn't already one.</p>
1099
1100 </div>
1101
1102 <!-- ======================================================================= -->
1103 <div class="doc_subsection">
1104   <a name="experimental">Experimental features included with this release</a>
1105 </div>
1106
1107 <div class="doc_text">
1108
1109 <p>The following components of this LLVM release are either untested, known to
1110 be broken or unreliable, or are in early development.  These components should
1111 not be relied on, and bugs should not be filed against them, but they may be
1112 useful to some people.  In particular, if you would like to work on one of these
1113 components, please contact us on the <a
1114 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
1115
1116 <ul>
1117 <li>The Alpha, Blackfin, CellSPU, MicroBlaze, MSP430, MIPS, SystemZ
1118     and XCore backends are experimental.</li>
1119 <li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets
1120     other than darwin-i386 and darwin-x86_64.</li>
1121 </ul>
1122
1123 </div>
1124
1125 <!-- ======================================================================= -->
1126 <div class="doc_subsection">
1127   <a name="x86-be">Known problems with the X86 back-end</a>
1128 </div>
1129
1130 <div class="doc_text">
1131
1132 <ul>
1133   <li>The X86 backend does not yet support
1134     all <a href="http://llvm.org/PR879">inline assembly that uses the X86
1135     floating point stack</a>.  It supports the 'f' and 't' constraints, but not
1136     'u'.</li>
1137   <li>Win64 code generation wasn't widely tested. Everything should work, but we
1138     expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1139     runtime currently due to lack of support for the 'u' inline assembly
1140     constraint and for X87 floating point inline assembly.</li>
1141   <li>The X86-64 backend does not yet support the LLVM IR instruction
1142       <tt>va_arg</tt>. Currently, front-ends support variadic
1143       argument constructs on X86-64 by lowering them manually.</li>
1144 </ul>
1145
1146 </div>
1147
1148 <!-- ======================================================================= -->
1149 <div class="doc_subsection">
1150   <a name="ppc-be">Known problems with the PowerPC back-end</a>
1151 </div>
1152
1153 <div class="doc_text">
1154
1155 <ul>
1156 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1157 compilation, and lacks support for debug information.</li>
1158 </ul>
1159
1160 </div>
1161
1162 <!-- ======================================================================= -->
1163 <div class="doc_subsection">
1164   <a name="arm-be">Known problems with the ARM back-end</a>
1165 </div>
1166
1167 <div class="doc_text">
1168
1169 <ul>
1170 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1171 processors, thumb programs can crash or produce wrong
1172 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1173 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1174 </li>
1175 </ul>
1176
1177 </div>
1178
1179 <!-- ======================================================================= -->
1180 <div class="doc_subsection">
1181   <a name="sparc-be">Known problems with the SPARC back-end</a>
1182 </div>
1183
1184 <div class="doc_text">
1185
1186 <ul>
1187 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1188     support the 64-bit SPARC ABI (-m64).</li>
1189 </ul>
1190
1191 </div>
1192
1193 <!-- ======================================================================= -->
1194 <div class="doc_subsection">
1195   <a name="mips-be">Known problems with the MIPS back-end</a>
1196 </div>
1197
1198 <div class="doc_text">
1199
1200 <ul>
1201 <li>64-bit MIPS targets are not supported yet.</li>
1202 </ul>
1203
1204 </div>
1205
1206 <!-- ======================================================================= -->
1207 <div class="doc_subsection">
1208   <a name="alpha-be">Known problems with the Alpha back-end</a>
1209 </div>
1210
1211 <div class="doc_text">
1212
1213 <ul>
1214
1215 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1216 appropriate nops inserted to ensure restartability.</li>
1217
1218 </ul>
1219 </div>
1220
1221 <!-- ======================================================================= -->
1222 <div class="doc_subsection">
1223   <a name="c-be">Known problems with the C back-end</a>
1224 </div>
1225
1226 <div class="doc_text">
1227
1228 <p>The C backend has numerous problems and is not being actively maintained.
1229 Depending on it for anything serious is not advised.</p>
1230
1231 <ul>
1232 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1233     inline assembly code</a>.</li>
1234 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1235     C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1236     C++ code compiled with <tt>llc</tt> or native compilers.</li>
1237 <li>The C backend does not support all exception handling constructs.</li>
1238 <li>The C backend does not support arbitrary precision integers.</li>
1239 </ul>
1240
1241 </div>
1242
1243
1244 <!-- ======================================================================= -->
1245 <div class="doc_subsection">
1246   <a name="llvm-gcc">Known problems with the llvm-gcc front-end</a>
1247 </div>
1248
1249 <div class="doc_text">
1250
1251 <p>llvm-gcc is generally very stable for the C family of languages.  The only
1252    major language feature of GCC not supported by llvm-gcc is the
1253    <tt>__builtin_apply</tt> family of builtins.   However, some extensions
1254    are only supported on some targets.  For example, trampolines are only
1255    supported on some targets (these are used when you take the address of a
1256    nested function).</p>
1257
1258 <p>Fortran support generally works, but there are still several unresolved bugs
1259    in <a href="http://llvm.org/bugs/">Bugzilla</a>.  Please see the
1260    tools/gfortran component for details.  Note that llvm-gcc is missing major
1261    Fortran performance work in the frontend and library that went into GCC after
1262    4.2.  If you are interested in Fortran, we recommend that you consider using
1263    <a href="#dragonegg">dragonegg</a> instead.</p>
1264
1265 <p>The llvm-gcc 4.2 Ada compiler has basic functionality, but is no longer being
1266 actively maintained.  If you are interested in Ada, we recommend that you
1267 consider using <a href="#dragonegg">dragonegg</a> instead.</p>
1268 </div>
1269
1270 <!-- *********************************************************************** -->
1271 <div class="doc_section">
1272   <a name="additionalinfo">Additional Information</a>
1273 </div>
1274 <!-- *********************************************************************** -->
1275
1276 <div class="doc_text">
1277
1278 <p>A wide variety of additional information is available on the <a
1279 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1280 href="http://llvm.org/docs/">documentation</a> section.  The web page also
1281 contains versions of the API documentation which is up-to-date with the
1282 Subversion version of the source code.
1283 You can access versions of these documents specific to this release by going
1284 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1285
1286 <p>If you have any questions or comments about LLVM, please feel free to contact
1287 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1288 lists</a>.</p>
1289
1290 </div>
1291
1292 <!-- *********************************************************************** -->
1293
1294 <hr>
1295 <address>
1296   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1297   src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1298   <a href="http://validator.w3.org/check/referer"><img
1299   src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1300
1301   <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1302   Last modified: $Date$
1303 </address>
1304
1305 </body>
1306 </html>