1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <link rel="stylesheet" href="llvm.css" type="text/css">
7 <title>LLVM 2.6 Release Notes</title>
11 <div class="doc_title">LLVM 2.6 Release Notes</div>
14 <li><a href="#intro">Introduction</a></li>
15 <li><a href="#subproj">Sub-project Status Update</a></li>
16 <li><a href="#externalproj">External Projects Using LLVM 2.6</a></li>
17 <li><a href="#whatsnew">What's New in LLVM 2.6?</a></li>
18 <li><a href="GettingStarted.html">Installation Instructions</a></li>
19 <li><a href="#portability">Portability and Supported Platforms</a></li>
20 <li><a href="#knownproblems">Known Problems</a></li>
21 <li><a href="#additionalinfo">Additional Information</a></li>
24 <div class="doc_author">
25 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
28 <!-- *********************************************************************** -->
29 <div class="doc_section">
30 <a name="intro">Introduction</a>
32 <!-- *********************************************************************** -->
34 <div class="doc_text">
36 <p>This document contains the release notes for the LLVM Compiler
37 Infrastructure, release 2.6. Here we describe the status of LLVM, including
38 major improvements from the previous release and significant known problems.
39 All LLVM releases may be downloaded from the <a
40 href="http://llvm.org/releases/">LLVM releases web site</a>.</p>
42 <p>For more information about LLVM, including information about the latest
43 release, please check out the <a href="http://llvm.org/">main LLVM
44 web site</a>. If you have questions or comments, the <a
45 href="http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVM Developer's Mailing
46 List</a> is a good place to send them.</p>
48 <p>Note that if you are reading this file from a Subversion checkout or the
49 main LLVM web page, this document applies to the <i>next</i> release, not the
50 current one. To see the release notes for a specific release, please see the
51 <a href="http://llvm.org/releases/">releases page</a>.</p>
58 include/llvm/Analysis/LiveValues.h => Dan
59 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
60 llvm/Analysis/PointerTracking.h => Edwin wants this, consider for 2.8.
64 <!-- Unfinished features in 2.6:
67 variable debug info for optimized code
68 postalloc scheduler: anti dependence breaking, hazard recognizer?
70 loop dependence analysis
71 ELF Writer? How stable?
72 <li>PostRA scheduler improvements, ARM adoption (David Goodwin).</li>
73 2.7 supports the GDB 7.0 jit interfaces for debug info.
76 <!-- for announcement email:
80 KLEE web page at klee.llvm.org
81 Many new papers added to /pubs/
86 <!-- *********************************************************************** -->
87 <div class="doc_section">
88 <a name="subproj">Sub-project Status Update</a>
90 <!-- *********************************************************************** -->
92 <div class="doc_text">
94 The LLVM 2.6 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.
104 <!--=========================================================================-->
105 <div class="doc_subsection">
106 <a name="clang">Clang: C/C++/Objective-C Frontend Toolkit</a>
109 <div class="doc_text">
111 <p>The <a href="http://clang.llvm.org/">Clang project</a> is an effort to build
112 a set of new 'LLVM native' front-end technologies for the C family of languages.
113 LLVM 2.6 is the first release to officially include Clang, and it provides a
114 production quality C and Objective-C compiler. If you are interested in <a
115 href="http://clang.llvm.org/performance.html">fast compiles</a> and
116 <a href="http://clang.llvm.org/diagnostics.html">good diagnostics</a>, we
117 encourage you to try it out. Clang currently compiles typical Objective-C code
118 3x faster than GCC and compiles C code about 30% faster than GCC at -O0 -g
119 (which is when the most pressure is on the frontend).</p>
121 <p>In addition to supporting these languages, C++ support is also <a
122 href="http://clang.llvm.org/cxx_status.html">well under way</a>, and mainline
123 Clang is able to parse the libstdc++ 4.2 headers and even codegen simple apps.
124 If you are interested in Clang C++ support or any other Clang feature, we
125 strongly encourage you to get involved on the <a
126 href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev">Clang front-end mailing
129 <p>In the LLVM 2.6 time-frame, the Clang team has made many improvements:</p>
132 <li>C and Objective-C support are now considered production quality.</li>
133 <li>AuroraUX, FreeBSD, and OpenBSD are now supported.</li>
134 <li>Most of Objective-C 2.0 is now supported with the GNU runtime.</li>
135 <li>Many many bugs are fixed and many features have been added.</li>
139 <!--=========================================================================-->
140 <div class="doc_subsection">
141 <a name="clangsa">Clang Static Analyzer</a>
144 <div class="doc_text">
146 <p><b>UPDATE!</b> Previously announced in the 2.4 and 2.5 LLVM releases, the Clang project also
147 includes an early stage static source code analysis tool for <a
148 href="http://clang.llvm.org/StaticAnalysis.html">automatically finding bugs</a>
149 in C and Objective-C programs. The tool performs checks to find
150 bugs that occur on a specific path within a program.
151 The set of checks performed by the static analyzer continues to expand, and
152 future plans for the tool include full source-level inter-procedural analysis
153 and deeper checks such as buffer overrun detection. There are many opportunities
154 to extend and enhance the static analyzer, and anyone interested in working on
155 this project is encouraged to get involved!</p>
159 <!--=========================================================================-->
160 <div class="doc_subsection">
161 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
164 <div class="doc_text">
166 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
167 a JVM and a CLI Virtual Machine (Microsoft .NET is an
168 implementation of the CLI) using LLVM for static and just-in-time
172 VMKit version 0.26 builds with LLVM 2.6 and you can find it on its
173 <a href="http://vmkit.llvm.org/releases/">web page</a>. The release includes
174 bug fixes, cleanup and new features. The major changes are:</p>
178 <li>A new llcj tool to generate shared libraries or executables of Java
180 <li>Cooperative garbage collection. </li>
181 <li>Fast subtype checking (paper from Click et al [JGI'02]). </li>
182 <li>Implementation of a two-word header for Java objects instead of the original
183 three-word header. </li>
184 <li>Better Java specification-compliance: division by zero checks, stack
185 overflow checks, finalization and references support. </li>
191 <!--=========================================================================-->
192 <div class="doc_subsection">
193 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
196 <div class="doc_text">
198 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
199 is a simple library that provides an implementation of the low-level
200 target-specific hooks required by code generation and other runtime components.
201 For example, when compiling for a 32-bit target, converting a double to a 64-bit
202 unsigned integer is compiling into a runtime call to the "__fixunsdfdi"
203 function. The compiler-rt library provides highly optimized implementations of
204 this and other low-level routines (some are 3x faster than the equivalent
205 libgcc routines).</p>
208 All of the code in the compiler-rt project is available under the standard LLVM
209 License, a "BSD-style" license.</p>
213 <!--=========================================================================-->
214 <div class="doc_subsection">
215 <a name="klee">KLEE: Symbolic Execution and Automatic Test Case Generator</a>
218 <div class="doc_text">
220 The new LLVM <a href="http://klee.llvm.org/">KLEE project</a> is a symbolic
221 execution framework for programs in LLVM bitcode form. KLEE tries to
222 symbolically evaluate "all" paths through the application and records state
223 transitions that lead to fault states. This allows it to construct testcases
224 that lead to faults and can even be used to verify algorithms. For more
225 details, please see the <a
226 href="http://llvm.org/pubs/2008-12-OSDI-KLEE.html">OSDI 2008 paper</a> about
231 <!--=========================================================================-->
232 <div class="doc_subsection">
233 <a name="dragonegg">Dragon Egg: An LLVM backend plugin for GCC</a>
236 <div class="doc_text">
238 <b>Duncan needs to write me</b>.
244 <!--=========================================================================-->
245 <div class="doc_subsection">
246 <a name="mc">llvm-mc: Machine Code Toolkit</a>
249 <div class="doc_text">
251 The LLVM Machine Code (MC) Toolkit project is a (very early) effort to build
252 better tools for dealing with machine code, object file formats, etc. The idea
253 is to be able to generate most of the target specific details of assemblers and
254 disassemblers from existing LLVM target .td files (with suitable enhancements),
255 and to build infrastructure for reading and writing common object file formats.
256 One of the first deliverables is to build a full assembler and integrate it into
257 the compiler, which is predicted to substantially reduce compile time in some
261 <p>In the LLVM 2.6 timeframe, the MC framework has grown to the point where it
262 can reliably parse and pretty print (with some encoding information) a
263 darwin/x86 .s file successfully, and has the very early phases of a Mach-O
264 assembler in progress. Beyond the MC framework itself, major refactoring of the
265 LLVM code generator has started. The idea is to make the code generator reason
266 about the code it is producing in a much more semantic way, rather than a
267 textual way. For example, the code generator now uses MCSection objects to
268 represent section assignments, instead of text strings that print to .section
271 <p>MC is an early and ongoing project that will hopefully continue to lead to
272 many improvements in the code generator and build infrastructure useful for many
279 <!-- *********************************************************************** -->
280 <div class="doc_section">
281 <a name="externalproj">External Projects Using LLVM 2.6</a>
283 <!-- *********************************************************************** -->
285 <div class="doc_text">
287 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
288 a lot of other language and tools projects. This section lists some of the
289 projects that have already been updated to work with LLVM 2.6.</p>
293 <!--=========================================================================-->
294 <div class="doc_subsection">
295 <a name="Rubinius">Rubinius</a>
298 <div class="doc_text">
299 <p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
300 for running Ruby code which strives to write as much of the core class
301 implementation in Ruby as possible. Combined with a bytecode interpreting VM, it
302 uses LLVM to optimize and compile ruby code down to machine code. Techniques
303 such as type feedback, method inlining, and uncommon traps are all used to
304 remove dynamism from ruby execution and increase performance.</p>
306 <p>Since LLVM 2.5, Rubinius has made several major leaps forward, implementing
307 a counter based JIT, type feedback, and speculative method inlining.
312 <!--=========================================================================-->
313 <div class="doc_subsection">
314 <a name="macruby">MacRuby</a>
317 <div class="doc_text">
320 <a href="http://macruby.org">MacRuby</a> is an implementation of Ruby on top of
321 core Mac OS X technologies, such as the Objective-C common runtime and garbage
322 collector, and the CoreFoundation framework. It is principally developed by
323 Apple and aims at enabling the creation of full-fledged Mac OS X applications.
327 MacRuby uses LLVM for optimization passes, JIT and AOT compilation of Ruby
328 expressions. It also uses zero-cost DWARF exceptions to implement Ruby exception
334 <!--=========================================================================-->
335 <div class="doc_subsection">
336 <a name="pure">Pure</a>
339 <div class="doc_text">
341 <a href="http://pure-lang.googlecode.com/">Pure</a>
342 is an algebraic/functional programming language based on term rewriting.
343 Programs are collections of equations which are used to evaluate expressions in
344 a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation,
345 lexical closures, a hygienic macro system (also based on term rewriting),
346 built-in list and matrix support (including list and matrix comprehensions) and
347 an easy-to-use C interface. The interpreter uses LLVM as a backend to
348 JIT-compile Pure programs to fast native code.</p>
350 <p>Pure versions 0.31 and later have been tested and are known to work with
351 LLVM 2.6 (and continue to work with older LLVM releases >= 2.3 as well).
356 <!--=========================================================================-->
357 <div class="doc_subsection">
358 <a name="ldc">LLVM D Compiler</a>
361 <div class="doc_text">
363 <a href="http://www.dsource.org/projects/ldc">LDC</a> is an implementation of
364 the D Programming Language using the LLVM optimizer and code generator.
365 The LDC project works great with the LLVM 2.6 release. General improvements in
367 cycle have included new inline asm constraint handling, better debug info
368 support, general bug fixes, and better x86-64 support. This has allowed
369 some major improvements in LDC, getting us much closer to being as
370 fully featured as the original DMD compiler from DigitalMars.
374 <!--=========================================================================-->
375 <div class="doc_subsection">
376 <a name="RoadsendPHP">Roadsend PHP</a>
379 <div class="doc_text">
381 <a href="http://code.roadsend.com/rphp">Roadsend PHP</a> (rphp) is an open
382 source implementation of the PHP programming
383 language that uses LLVM for its optimizer, JIT, and static compiler. This is a
384 reimplementation of an earlier project that is now based on LLVM.</p>
387 <!--=========================================================================-->
388 <div class="doc_subsection">
389 <a name="UnladenSwallow">Unladen Swallow</a>
392 <div class="doc_text">
394 <a href="http://code.google.com/p/unladen-swallow/">Unladen Swallow</a> is a
395 branch of <a href="http://python.org/">Python</a> intended to be fully
396 compatible and significantly faster. It uses LLVM's optimization passes and JIT
400 <!--=========================================================================-->
401 <div class="doc_subsection">
402 <a name="llvm-lua">llvm-lua</a>
405 <div class="doc_text">
407 <a href="http://code.google.com/p/llvm-lua/">LLVM-Lua</a> uses LLVM to add JIT
408 & static compiling support to the Lua VM. Lua bytecode is analyzed to
409 remove type checks, then LLVM is used to compile those bytecodes down to machine
415 <!-- *********************************************************************** -->
416 <div class="doc_section">
417 <a name="whatsnew">What's New in LLVM 2.6?</a>
419 <!-- *********************************************************************** -->
421 <div class="doc_text">
423 <p>This release includes a huge number of bug fixes, performance tweaks, and
424 minor improvements. Some of the major improvements and new features are listed
430 <!--=========================================================================-->
431 <div class="doc_subsection">
432 <a name="majorfeatures">Major New Features</a>
435 <div class="doc_text">
437 <p>LLVM 2.6 includes several major new capabilities:</p>
440 <li>New <a href="#compiler-rt">compiler-rt</a>, <A href="#klee">KLEE</a>,
441 and <a href="#mc">machine code toolkit</a> sub-projects.</li>
442 <li>Debug information now includes line numbers when optimizations are enabled.
443 This allows statistical sampling tools like oprofile and Shark to map
444 samples back to source lines.</li>
445 <li>LLVM now includes new experimental backends to support the MSP430, SystemZ,
446 and BlackFin architectures.</li>
447 <li>LLVM supports a new <a href="GoldPlugin.html">Gold Linker Plugin</a> which
448 enables support for <a href="LinkTimeOptimization.html">transparent
449 link-time optimization</a> on ELF targets when used with the Gold binutils
451 <li>LLVM now supports doing optimization and code generation on multiple
452 threads. Please see the <a href="ProgrammersManual.html#threading">LLVM
453 Programmer's Manual</a> for more information.</li>
454 <li>LLVM now has experimental support for <a
455 href="http://nondot.org/~sabre/LLVMNotes/EmbeddedMetadata.txt">embedded
456 metadata</a> in LLVM IR, though the implementation is not guaranteed to be
457 final and the .bc file format may change in future releases. Debug info
458 does not yet use this format in LLVM 2.6.</li>
463 <!--=========================================================================-->
464 <div class="doc_subsection">
465 <a name="coreimprovements">LLVM IR and Core Improvements</a>
468 <div class="doc_text">
469 <p>LLVM IR has several new features for better support of new targets and that
470 expose new optimization opportunities:</p>
473 <li>The <a href="LangRef.html#i_add">add</a>, <a
474 href="LangRef.html#i_sub">sub</a>, and <a href="LangRef.html#i_mul">mul</a>
475 instructions have been split into integer and floating point version (like
476 divide and remainder), introducing new <a
477 href="LangRef.html#i_fadd">fadd</a>, <a href="LangRef.html#i_fsub">fsub</a>,
478 and <a href="LangRef.html#i_fmul">fmul</a> instructions.</li>
479 <li>The <a href="LangRef.html#i_add">add</a>, <a
480 href="LangRef.html#i_sub">sub</a>, and <a href="LangRef.html#i_mul">mul</a>
481 instructions now support optional "nsw" and "nuw" bits which indicate that
482 the operation is guaranteed to not overflow (in the signed or
483 unsigned case, respectively). This gives the optimizer more information and
484 can be used for things C signed integer values, which are undefined on
486 <li>The <a href="LangRef.html#i_sdiv">sdiv</a> instruction now supports an
487 optional "exact" flag which indicates that the result of the division is
488 guaranteed to have a remainder of zero. This is useful to optimize pointer
489 subtraction in C.</li>
490 <li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
491 supports arbitrary integer index values for array/pointer indices. This
492 allows for better better code generation on 16-bit targets like PIC16.</li>
493 <li>The <a href="LangRef.html#i_getelementptr">getelementptr</a> instruction now
494 supports an "inbounds" optimization hint that tells the optimizer that the
495 pointer is guaranteed to be within its allocated object.</li>
496 <li>LLVM now support a series of new linkage types for global values which allow
497 for better optimization and new capabilities:
499 <li><a href="LangRef.html#linkage_linkonce">linkonce_odr</a> and
500 <a href="LangRef.html#linkage_weak">weak_odr</a> have the same linkage
501 semantics as the non-"odr" linkage types. The difference is that these
502 linkage types indicate that all definitions of the specified function
503 are guaranteed to have the same semantics. This allows inlining
504 templates functions in C++ but not inlining weak functions in C,
505 which previously both got the same linkage type.</li>
506 <li><a href="LangRef.html#linkage_available_externally">available_externally
507 </a> is a new linkage type that gives the optimizer visibility into the
508 definition of a function (allowing inlining and side effect analysis)
509 but that does not cause code to be generated. This allows better
510 optimization of "GNU inline" functions, extern templates, etc.</li>
511 <li><a href="LangRef.html#linkage_linker_private">linker_private</a> is a
512 new linkage type (which is only useful on Mac OS X) that is used for
513 some metadata generation and other obscure things.</li>
515 <li>Finally, target-specific intrinsics can now return multiple values, which
516 is useful for modeling target operations with multiple results.</li>
521 <!--=========================================================================-->
522 <div class="doc_subsection">
523 <a name="optimizer">Optimizer Improvements</a>
526 <div class="doc_text">
528 <p>In addition to a large array of minor performance tweaks and bug fixes, this
529 release includes a few major enhancements and additions to the optimizers:</p>
533 <li>The <a href="Passes.html#scalarrepl">Scalar Replacement of Aggregates</a>
534 pass has many improvements that allow it to better promote vector unions,
535 variables which are memset, and much more strange code that can happen do
536 to bitfield accesses to register operations. An interesting change is that
537 it now produces "unusual" integer sizes (like i1704) in some cases and lets
538 other optimizers clean things up.</li>
539 <li>The <a href="Passes.html#loop-reduce">Loop Strength Reduction</a> pass now
540 promotes small integer induction variables to 64-bit on 64-bit targets,
541 which provides a major performance boost many for numerical code. It also
542 promotes shorts to int on 32-bit hosts, etc. LSR now also analyzes pointer
543 expressions (e.g. getelementptrs), as well as integers.</li>
544 <li>The <a href="Passes.html#gvn">GVN</a> pass now eliminates partial
545 redundancies of loads in simple cases.</li>
546 <li>The <a href="Passes.html#inline">Inliner</a> now reuses stack space when
547 inlining similar arrays from multiple callees into one caller.</li>
548 <li>LLVM includes a new experimental Static Single Information (SSI)
549 construction pass.</li>
556 <!--=========================================================================-->
557 <div class="doc_subsection">
558 <a name="codegen">Target Independent Code Generator Improvements</a>
561 <div class="doc_text">
563 <p>We have put a significant amount of work into the code generator
564 infrastructure, which allows us to implement more aggressive algorithms and make
569 <li>The <tt>llc -asm-verbose</tt> option (exposed from llvm-gcc and clang as
570 <tt>-fverbose-asm</tt>) now adds a lot of useful information in comments to
571 the generated .s file. This information includes location information (if
572 built with <tt>-g</tt>) and loop nest information.</li>
573 <li>The code generator now supports a new MachineVerifier pass which is useful
574 for finding bugs in targets and codegen passes.</li>
575 <li>The Machine LICM is now enabled by default. It hoists instructions out of
576 loops (such as constant pool loads, loads from read-only stubs, vector
577 constant synthesization code, etc) and is currently configured to only do so
578 when the hoisted operation can be rematerialized.</li>
579 <li>The Machine Sinking pass is now enabled by default. This pass moves
580 side-effect free operations down the CFG so that they are executed on fewer
581 paths through a function.</li>
582 <li>The code generator now performs "Stack slot coloring" of register spills,
583 which allows spill slots to be reused. This leads to smaller stack frames
584 in cases where there are lots of register spills.</li>
585 <li>The register allocator has many improvements to take better advantage of
586 commutable operations, various spiller peephole optimizations, and can now
587 coalesce cross-register-class copies.</li>
588 <li>Tblgen now supports multiclass inheritance and a number of new string and
589 list operations like !(subst), !(foreach), !car, !cdr, !null, !if, !cast.
590 These make the .td files more expressive and allow more aggressive factoring
591 of duplication across instruction patterns.</li>
592 <li>Target-specific intrinsics can now be added without having to hack VMCore to
593 add them. This makes it easier to maintain out-of-tree targets.</li>
594 <li>The instruction selector is better at propagating information about values
595 (such as whether they are sign/zero extended etc) across basic block
597 <li>SelectionDAGS: New BuildVectorSDNode (r65296), and ISD::VECTOR_SHUFFLE (r69952 / PR2957)</li>
598 <li>The Prolog/Epilog Insertion Pass now has experimental support for performing
599 the "shrink wrapping" optimization, which moves spills and reloads around in
600 the CFG to avoid doing saves on paths that don't need them.</li>
601 <li>LLVM includes new experimental support for writing ELF .o files directly
602 from the compiler. It works well for many simple C testcases, but doesn't
603 support exception handling, debug info, inline assembly, etc.</li>
604 <li>Targets can now specify register allocation hints through
605 MachineRegisterInfo:: setRegAllocationHint. A regalloc hint consists of hint
606 type and physical register number. A hint type of zero specifies a register
607 allocation preference. Other hint type values are target specific which are
608 resolved by TargetRegisterInfo::ResolveRegAllocHint. An example of which is
609 the ARM target can uses register hint to request that the register allocator
610 provide an even / odd register pair to two virtual registers.</li>
614 <!--=========================================================================-->
615 <div class="doc_subsection">
616 <a name="x86">X86-32 and X86-64 Target Improvements</a>
619 <div class="doc_text">
620 <p>New features of the X86 target include:
625 <li>Preliminary support for addrspace 256 -> GS, 257 -> FS, known problems: CodeGenerator.html#x86_memory</li>
626 <li>Support for softfloat modes, typically used by OS kernels.</li>
628 <li>X86-64: better modeling of implicit zero extensions, eliminates a lot of redundant zexts</li>
629 <li>X86-64 TLS support for local exec and initial exec.</li>
630 <li>Better modeling of H registerts as subregs.</li>
631 <li>Vector icmp/fcmp now work with SSE codegen.</li>
632 <li>SSE 4.2 support.</li>
633 <li>all global variable reference logic is now in ClassifyGlobalReference.</li>
639 <!--=========================================================================-->
640 <div class="doc_subsection">
641 <a name="pic16">PIC16 Target Improvements</a>
644 <div class="doc_text">
645 <p>New features of the PIC16 target include:
649 <li>Support for floating-point, indirect function calls, and
650 passing/returning aggregate types to functions.
651 <li>The code generator is able to generate debug info into output COFF files.
652 <li>Support for placing an object into a specific section or at a specific
653 address in memory.</li>
656 <p>Things not yet supported:</p>
659 <li>Variable arguments.</li>
660 <li>Interrupts/programs.</li>
665 <!--=========================================================================-->
666 <div class="doc_subsection">
667 <a name="ARM">ARM Target Improvements</a>
670 <div class="doc_text">
671 <p>New features of the ARM target include:
676 <li>Preliminary support for processors, such as the Cortex-A8 and Cortex-A9,
677 that implement version v7-A of the ARM architecture. The ARM backend now
678 supports both the Thumb2 and Advanced SIMD (Neon) instruction sets. The
679 AAPCS-VFP "hard float" calling conventions are also supported with the
680 <tt>-float-abi=hard</tt> flag. These features are still somewhat experimental
681 and subject to change. The Neon intrinsics, in particular, may change in future
685 ARM AAPCS-VFP hard float ABI is supported.
686 ARM calling convention code is now tblgen generated instead of manual.
687 ARM: NEON support. neonfp for doing single precision fp with neon instead of VFP.
693 <!--=========================================================================-->
694 <div class="doc_subsection">
695 <a name="OtherTarget">Other Target Specific Improvements</a>
698 <div class="doc_text">
699 <p>New features of other targets include:
703 <li>Mips now supports O32 Calling Convention.</li>
704 <li>Many improvements to the 32-bit PowerPC SVR4 ABI (used on powerpc-linux)
705 support, lots of bugs fixed.</li>
706 <li>Added support for the 64-bit PowerPC SVR4 ABI (used on powerpc64-linux).
707 Needs more testing.</li>
712 <!--=========================================================================-->
713 <div class="doc_subsection">
714 <a name="executionengine">Interpreter and JIT Improvements</a>
717 <div class="doc_text">
720 <li>The JIT now supports generating more than 16M of code.</li>
721 <li>When configured with --with-oprofile, the JIT can now inform oprofile about
722 JIT'd code, allowing oprofile to get line number and function name
723 information for JIT'd functions.</li>
724 <li>When "libffi" is available, the LLVM interpreter now uses it, which supports
725 calling almost arbitrary external (natively compiled) functions.</li>
726 <li>Clients of the JIT can now register a 'JITEventListener' object to receive
727 callbacks when the JIT emits or frees machine code. The OProfile support
728 uses this mechanism.</li>
734 <!--=========================================================================-->
735 <div class="doc_subsection">
736 <a name="newapis">New Useful APIs</a>
739 <div class="doc_text">
742 <li>New EngineBuilder class for creating JITs: r76276</li>
743 New PrettyStackTrace, crashes of llvm tools should give some indication of what the compiler was doing at the time of the crash (e.g. running a pass), and print out command line arguments.
744 StringRef class, Twine class.
745 New WeakVH and AssertingVH and CallbackVH classes.
746 New llvm/ADT/Triple class.
747 llvm_report_error() error handling API (llvm/Support/ErrorHandling.h)
748 New llvm/System/Atomic.h, llvm/System/RWMutex.h for portable atomic ops, rw locks.
749 New SourceMgr, SMLoc classes for simple parsers with caret diagnostics and #include support, (used by
750 tablegen, llvm-mc, the .ll parser, FileCheck, etc)
757 <!--=========================================================================-->
758 <div class="doc_subsection">
759 <a name="otherimprovements">Other Improvements and New Features</a>
762 <div class="doc_text">
763 <p>Other miscellaneous features include:</p>
766 <li>LLVM now includes a new internal '<a
767 href="http://llvm.org/cmds/FileCheck.html">FileCheck</a>' tool which allows
768 writing much more accurate regression tests that run faster. Please see the
769 <a href="TestingGuide.html#FileCheck">FileCheck section of the Testing
770 Guide</a> for more information.</li>
771 <li>LLVM profile information support has been significantly improved to produce
772 correct use counts, and has support for edge profiling with reduced runtime
773 overhead. Combined, the generated profile information is both more correct and
774 imposes about half as much overhead (2.6. from 12% to 6% overhead on SPEC
776 <li>Many extensions to the C APIs.</li>
777 <li>LLVM 2.6 includes a brand new experimental LLVM bindings to the Ada2005
778 programming language.</li>
782 * Dynamic plugins now work on Windows.
783 * New option property: init. Makes possible to provide default values for
784 options defined in plugins (interface to cl::init).
785 * New example: Skeleton, shows how to create a standalone LLVMC-based driver.
786 * New example: mcc16, a driver for the PIC16 toolchain.</li>
793 <!--=========================================================================-->
794 <div class="doc_subsection">
795 <a name="changes">Major Changes and Removed Features</a>
798 <div class="doc_text">
800 <p>If you're already an LLVM user or developer with out-of-tree changes based
801 on LLVM 2.5, this section lists some "gotchas" that you may run into upgrading
802 from the previous release.</p>
806 <li>The Itanium (IA64) backend has been removed. It was not supported and
808 <li>The BigBlock register allocator has been removed, it also bitrotted.</li>
809 <li>The C Backend (-march=c) is no longer considered part of the LLVM release
810 criteria. We still want it to work, but no one is maintaining it and it lacks
811 support for arbitrary precision integers and other important IR features.</li>
814 LLVM build now builds all libraries as .a files instead of some
815 libraries as relinked .o files. This requires some APIs like
816 InitializeAllTargets.h. TargetRegistry!
821 <p>In addition, many APIs have changed in this release. Some of the major LLVM
826 no use of hash_set/hash_map, no more llvm::OStream
827 Use raw_ostream for everything, killed off llvm/Streams.h and DOUT
830 <li>LLVM's global uniquing tables for <tt>Type</tt>s and <tt>Constant</tt>s have
831 been privatized into members of an <tt>LLVMContext</tt>. A number of APIs
832 now take an <tt>LLVMContext</tt> as a parameter. To smooth the transition
833 for clients that will only ever use a single context, the new
834 <tt>getGlobalContext()</tt> API can be used to access a default global
835 context which can be passed in any and all cases where a context is
837 <li>The <tt>getABITypeSize</tt> methods are now called <tt>getAllocSize</tt>.</li>
838 <li>The <tt>Add</tt>, <tt>Sub</tt>, and <tt>Mul</tt> operators are no longer
839 overloaded for floating-point types. Floating-point addition, subtraction,
840 and multiplication are now represented with new operators <tt>FAdd</tt>,
841 <tt>FSub</tt>, and <tt>FMul</tt>. In the <tt>IRBuilder</tt> API,
842 <tt>CreateAdd</tt>, <tt>CreateSub</tt>, <tt>CreateMul</tt>, and
843 <tt>CreateNeg</tt> should only be used for integer arithmetic now;
844 <tt>CreateFAdd</tt>, <tt>CreateFSub</tt>, <tt>CreateFMul</tt>, and
845 <tt>CreateFNeg</tt> should now be used for floating-point arithmetic.</li>
846 <li>The DynamicLibrary class can no longer be constructed, its functionality has
847 moved to static member functions.</li>
848 <li><tt>raw_fd_ostream</tt>'s constructor for opening a given filename now
849 takes an extra <tt>Force</tt> argument. If <tt>Force</tt> is set to
850 <tt>false</tt>, an error will be reported if a file with the given name
851 already exists. If <tt>Force</tt> is set to <tt>true</tt>, the file will
852 be silently truncated (which is the behavior before this flag was
854 <li><tt>SCEVHandle</tt> no longer exists, because reference counting is no
855 longer done for <tt>SCEV*</tt> objects, instead <tt>const SCEV*</tt> should be
858 <li>Many APIs, notably <tt>llvm::Value</tt>, now use the <tt>StringRef</tt>
859 and <tt>Twine</tt> classes instead of passing <tt>const char*</tt>
860 or <tt>std::string</tt>, as described in
861 the <a href="ProgrammersManual.html#string_apis">Programmer's Manual</a>. Most
862 clients should be unaffected by this transition, unless they are used to <tt>Value::getName()</tt> returning a string. Here are some tips on updating to 2.6:
864 <li><tt>getNameStr()</tt> is still available, and matches the old
865 behavior. Replacing <tt>getName()</tt> calls with this is an safe option,
866 although more efficient alternatives are now possible.</li>
868 <li>If you were just relying on <tt>getName()</tt> being able to be sent to
869 a <tt>std::ostream</tt>, consider migrating
870 to <tt>llvm::raw_ostream</tt>.</li>
872 <li>If you were using <tt>getName().c_str()</tt> to get a <tt>const
873 char*</tt> pointer to the name, you can use <tt>getName().data()</tt>.
874 Note that this string (as before), may not be the entire name if the
875 name containts embedded null characters.</li>
877 <li>If you were using operator plus on the result of <tt>getName()</tt> and
878 treating the result as an <tt>std::string</tt>, you can either
879 uses <tt>Twine::str</tt> to get the result as an <tt>std::string</tt>, or
880 could move to a <tt>Twine</tt> based design.</li>
882 <li><tt>isName()</tt> should be replaced with comparison
883 against <tt>getName()</tt> (this is now efficient).
887 <li>The registration interfaces for backend Targets has changed (what was
888 previously TargetMachineRegistry). For backend authors, see the <a href="WritingAnLLVMBackend.html#TargetRegistration">Writing An LLVM Backend</a> guide. For clients, the notable API changes are:
890 <li><tt>TargetMachineRegistry</tt> has been renamed
891 to <tt>TargetRegistry</tt>.</li>
893 <li>Clients should move to using the <tt>TargetRegistry::lookupTarget()</tt>
894 function to find targets.</li>
898 <li>llvm-dis now fails if output file exists, instead of dumping to stdout.
899 FIXME: describe any other tool changes due to the raw_fd_ostream change. FIXME:
900 This is not an API change, maybe there should be a tool changes section?</li>
901 <li>temporarely due to Context API change passes should call doInitialization()
902 method of the pass they inherit from, otherwise Context is NULL.
903 FIXME: remove this entry when this is no longer needed.<li>
910 <!-- *********************************************************************** -->
911 <div class="doc_section">
912 <a name="portability">Portability and Supported Platforms</a>
914 <!-- *********************************************************************** -->
916 <div class="doc_text">
918 <p>LLVM is known to work on the following platforms:</p>
921 <li>Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat
922 Linux, Fedora Core, FreeBSD and AuroraUX (and probably other unix-like systems).</li>
923 <li>PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit
924 and 64-bit modes.</li>
925 <li>Intel and AMD machines running on Win32 using MinGW libraries (native).</li>
926 <li>Intel and AMD machines running on Win32 with the Cygwin libraries (limited
927 support is available for native builds with Visual C++).</li>
928 <li>Sun UltraSPARC workstations running Solaris 10.</li>
929 <li>Alpha-based machines running Debian GNU/Linux.</li>
932 <p>The core LLVM infrastructure uses GNU autoconf to adapt itself
933 to the machine and operating system on which it is built. However, minor
934 porting may be required to get LLVM to work on new platforms. We welcome your
935 portability patches and reports of successful builds or error messages.</p>
939 <!-- *********************************************************************** -->
940 <div class="doc_section">
941 <a name="knownproblems">Known Problems</a>
943 <!-- *********************************************************************** -->
945 <div class="doc_text">
947 <p>This section contains significant known problems with the LLVM system,
948 listed by component. If you run into a problem, please check the <a
949 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
950 there isn't already one.</p>
953 <li>LLVM will not correctly compile on Solaris and/or OpenSolaris
954 using the stock GCC 3.x.x series 'out the box',
955 See: <a href="#brokengcc">Broken versions of GCC and other tools</a>.
956 However, A <a href="http://pkg.auroraux.org/GCC">Modern GCC Build</a>
957 for x86/x64 has been made available from the third party AuroraUX Project
958 that has been meticulously tested for bootstrapping LLVM & Clang.</li>
963 <!-- ======================================================================= -->
964 <div class="doc_subsection">
965 <a name="experimental">Experimental features included with this release</a>
968 <div class="doc_text">
970 <p>The following components of this LLVM release are either untested, known to
971 be broken or unreliable, or are in early development. These components should
972 not be relied on, and bugs should not be filed against them, but they may be
973 useful to some people. In particular, if you would like to work on one of these
974 components, please contact us on the <a
975 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
978 <li>The MSIL, Alpha, SPU, MIPS, and PIC16 backends are experimental.</li>
979 <li>The <tt>llc</tt> "<tt>-filetype=asm</tt>" (the default) is the only
980 supported value for this option.</li>
985 <!-- ======================================================================= -->
986 <div class="doc_subsection">
987 <a name="x86-be">Known problems with the X86 back-end</a>
990 <div class="doc_text">
993 <li>The X86 backend does not yet support
994 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
995 floating point stack</a>. It supports the 'f' and 't' constraints, but not
997 <li>The X86 backend generates inefficient floating point code when configured
998 to generate code for systems that don't have SSE2.</li>
999 <li>Win64 code generation wasn't widely tested. Everything should work, but we
1000 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
1001 runtime currently due
1002 to <a href="http://llvm.org/PR2255">several</a>
1003 <a href="http://llvm.org/PR2257">bugs</a> and due to lack of support for
1005 'u' inline assembly constraint and for X87 floating point inline assembly.</li>
1006 <li>The X86-64 backend does not yet support the LLVM IR instruction
1007 <tt>va_arg</tt>. Currently, the llvm-gcc and front-ends support variadic
1008 argument constructs on X86-64 by lowering them manually.</li>
1013 <!-- ======================================================================= -->
1014 <div class="doc_subsection">
1015 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1018 <div class="doc_text">
1021 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1022 compilation, and lacks support for debug information.</li>
1027 <!-- ======================================================================= -->
1028 <div class="doc_subsection">
1029 <a name="arm-be">Known problems with the ARM back-end</a>
1032 <div class="doc_text">
1035 <li>Support for the Advanced SIMD (Neon) instruction set is still incomplete
1036 and not well tested. Some features may not work at all, and the code quality
1037 may be poor in some cases.</li>
1038 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1039 processors, thumb programs can crash or produce wrong
1040 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1041 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1047 <!-- ======================================================================= -->
1048 <div class="doc_subsection">
1049 <a name="sparc-be">Known problems with the SPARC back-end</a>
1052 <div class="doc_text">
1055 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1056 support the 64-bit SPARC ABI (-m64).</li>
1061 <!-- ======================================================================= -->
1062 <div class="doc_subsection">
1063 <a name="mips-be">Known problems with the MIPS back-end</a>
1066 <div class="doc_text">
1069 <li>The O32 ABI is not fully supported.</li>
1070 <li>64-bit MIPS targets are not supported yet.</li>
1075 <!-- ======================================================================= -->
1076 <div class="doc_subsection">
1077 <a name="alpha-be">Known problems with the Alpha back-end</a>
1080 <div class="doc_text">
1084 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1085 appropriate nops inserted to ensure restartability.</li>
1090 <!-- ======================================================================= -->
1091 <div class="doc_subsection">
1092 <a name="c-be">Known problems with the C back-end</a>
1095 <div class="doc_text">
1098 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1099 inline assembly code</a>.</li>
1100 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1101 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1102 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1103 <li>The C backend does not support all exception handling constructs.</li>
1104 <li>The C backend does not support arbitrary precision integers.</li>
1110 <!-- ======================================================================= -->
1111 <div class="doc_subsection">
1112 <a name="c-fe">Known problems with the llvm-gcc C front-end</a>
1115 <div class="doc_text">
1117 <p>llvm-gcc does not currently support <a href="http://llvm.org/PR869">Link-Time
1118 Optimization</a> on most platforms "out-of-the-box". Please inquire on the
1119 LLVMdev mailing list if you are interested.</p>
1121 <p>The only major language feature of GCC not supported by llvm-gcc is
1122 the <tt>__builtin_apply</tt> family of builtins. However, some extensions
1123 are only supported on some targets. For example, trampolines are only
1124 supported on some targets (these are used when you take the address of a
1125 nested function).</p>
1127 <p>If you run into GCC extensions which are not supported, please let us know.
1132 <!-- ======================================================================= -->
1133 <div class="doc_subsection">
1134 <a name="c++-fe">Known problems with the llvm-gcc C++ front-end</a>
1137 <div class="doc_text">
1139 <p>The C++ front-end is considered to be fully
1140 tested and works for a number of non-trivial programs, including LLVM
1141 itself, Qt, Mozilla, etc.</p>
1144 <li>Exception handling works well on the X86 and PowerPC targets. Currently
1145 only Linux and Darwin targets are supported (both 32 and 64 bit).</li>
1150 <!-- ======================================================================= -->
1151 <div class="doc_subsection">
1152 <a name="fortran-fe">Known problems with the llvm-gcc Fortran front-end</a>
1155 <div class="doc_text">
1157 <li>Fortran support generally works, but there are still several unresolved bugs
1158 in Bugzilla. Please see the tools/gfortran component for details.</li>
1162 <!-- ======================================================================= -->
1163 <div class="doc_subsection">
1164 <a name="ada-fe">Known problems with the llvm-gcc Ada front-end</a>
1167 <div class="doc_text">
1168 The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature
1169 technology, and problems should be expected.
1171 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1172 to lack of trampoline support (pointers to nested functions) on other platforms.
1173 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1174 which does support trampolines.</li>
1175 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1176 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1177 exception handling, which is used internally by the compiler.
1178 Workaround: configure with --disable-bootstrap.</li>
1179 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1180 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1181 (c380004 also fails with gcc-4.2 mainline).
1182 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1183 causes the compiler to go into an infinite loop, using up all system memory.</li>
1184 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1185 <li>The -E binder option (exception backtraces)
1186 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1187 crashing if an exception is raised. Workaround: do not use -E.</li>
1188 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1189 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1190 or use representation clauses that result in a field of a non-discrete type
1191 starting or finishing in the middle of a byte.</li>
1192 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1193 'main' as generated by the Ada binder to be invalid</a>.
1194 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1195 <tt>envp</tt> rather than integers.</li>
1196 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1201 <!-- ======================================================================= -->
1202 <div class="doc_subsection">
1203 <a name="ocaml-bindingse">Known problems with the O'Caml bindings</a>
1206 <div class="doc_text">
1208 <p>The Llvm.Linkage module is broken, and has incorrect values. Only
1209 Llvm.Linkage.External, Llvm.Linkage.Available_externally, and
1210 Llvm.Linkage.Link_once will be correct. If you need any of the other linkage
1211 modes, you'll have to write an external C library in order to expose the
1212 functionality. This has been fixed in the trunk.</p>
1215 <!-- *********************************************************************** -->
1216 <div class="doc_section">
1217 <a name="additionalinfo">Additional Information</a>
1219 <!-- *********************************************************************** -->
1221 <div class="doc_text">
1223 <p>A wide variety of additional information is available on the <a
1224 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1225 href="http://llvm.org/docs/">documentation</a> section. The web page also
1226 contains versions of the API documentation which is up-to-date with the
1227 Subversion version of the source code.
1228 You can access versions of these documents specific to this release by going
1229 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1231 <p>If you have any questions or comments about LLVM, please feel free to contact
1232 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1237 <!-- *********************************************************************** -->
1241 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1242 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1243 <a href="http://validator.w3.org/check/referer"><img
1244 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1246 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1247 Last modified: $Date$