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">
7 <link rel="stylesheet" href="llvm.css" type="text/css">
8 <title>LLVM 2.8 Release Notes</title>
12 <div class="doc_title">LLVM 2.8 Release Notes</div>
14 <img align=right src="http://llvm.org/img/DragonSmall.png"
15 width="136" height="136" alt="LLVM Dragon Logo">
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>
27 <div class="doc_author">
28 <p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
32 <h1 style="color:red">These are in-progress notes for the upcoming LLVM 2.8
35 <a href="http://llvm.org/releases/2.7/docs/ReleaseNotes.html">LLVM 2.7
36 Release Notes</a>.</h1>
39 <!-- *********************************************************************** -->
40 <div class="doc_section">
41 <a name="intro">Introduction</a>
43 <!-- *********************************************************************** -->
45 <div class="doc_text">
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>
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>
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>
69 include/llvm/Analysis/LiveValues.h => Dan
70 lib/Transforms/IPO/MergeFunctions.cpp => consider for 2.8.
75 <!-- Features that need text if they're finished for 2.9:
78 loop dependence analysis
80 CorrelatedValuePropagation
83 <!-- Announcement, lldb, libc++ -->
86 <!-- *********************************************************************** -->
87 <div class="doc_section">
88 <a name="subproj">Sub-project Status Update</a>
90 <!-- *********************************************************************** -->
92 <div class="doc_text">
94 The LLVM 2.8 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.
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><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>
120 <p>In the LLVM 2.8 time-frame, the Clang team has made many improvements:</p>
123 <li>Surely these guys have done something</li>
124 <li>X86-64 abi improvements? Did they make it in?</li>
128 <!--=========================================================================-->
129 <div class="doc_subsection">
130 <a name="clangsa">Clang Static Analyzer</a>
133 <div class="doc_text">
135 <p>The <a href="http://clang-analyzer.llvm.org/">Clang Static Analyzer</a>
136 project is an effort to use static source code analysis techniques to
137 automatically find bugs in C and Objective-C programs (and hopefully <a
138 href="http://clang-analyzer.llvm.org/dev_cxx.html">C++ in the
139 future</a>!). The tool is very good at finding bugs that occur on specific
140 paths through code, such as on error conditions.</p>
142 <p>The LLVM 2.8 release fixes a number of bugs and slightly improves precision
143 over 2.7, but there are no major new features in the release.
148 <!--=========================================================================-->
149 <div class="doc_subsection">
150 <a name="vmkit">VMKit: JVM/CLI Virtual Machine Implementation</a>
153 <div class="doc_text">
155 The <a href="http://vmkit.llvm.org/">VMKit project</a> is an implementation of
156 a Java Virtual Machine (Java VM or JVM) that uses LLVM for static and
157 just-in-time compilation. As of LLVM 2.8, VMKit now supports copying garbage
158 collectors, and can be configured to use MMTk's copy mark-sweep garbage
159 collector. In LLVM 2.8, the VMKit .NET VM is no longer being maintained.
164 <!--=========================================================================-->
165 <div class="doc_subsection">
166 <a name="compiler-rt">compiler-rt: Compiler Runtime Library</a>
169 <div class="doc_text">
171 The new LLVM <a href="http://compiler-rt.llvm.org/">compiler-rt project</a>
172 is a simple library that provides an implementation of the low-level
173 target-specific hooks required by code generation and other runtime components.
174 For example, when compiling for a 32-bit target, converting a double to a 64-bit
175 unsigned integer is compiled into a runtime call to the "__fixunsdfdi"
176 function. The compiler-rt library provides highly optimized implementations of
177 this and other low-level routines (some are 3x faster than the equivalent
178 libgcc routines).</p>
181 All of the code in the compiler-rt project is available under the standard LLVM
182 License, a "BSD-style" license. New in LLVM 2.8, compiler_rt now supports
183 soft floating point (for targets that don't have a real floating point unit),
184 and includes an extensive testsuite for the "blocks" language feature and the
185 blocks runtime included in compiler_rt.</p>
189 <!--=========================================================================-->
190 <div class="doc_subsection">
191 <a name="dragonegg">DragonEgg: llvm-gcc ported to gcc-4.5</a>
194 <div class="doc_text">
196 <a href="http://dragonegg.llvm.org/">DragonEgg</a> is a port of llvm-gcc to
197 gcc-4.5. Unlike llvm-gcc, dragonegg in theory does not require any gcc-4.5
198 modifications whatsoever (currently one small patch is needed) thanks to the
199 new <a href="http://gcc.gnu.org/wiki/plugins">gcc plugin architecture</a>.
200 DragonEgg is a gcc plugin that makes gcc-4.5 use the LLVM optimizers and code
201 generators instead of gcc's, just like with llvm-gcc.
205 DragonEgg is still a work in progress, but it is able to compile a lot of code,
206 for example all of gcc, LLVM and clang. Currently Ada, C, C++ and Fortran work
207 well, while all other languages either don't work at all or only work poorly.
208 For the moment only the x86-32 and x86-64 targets are supported, and only on
209 linux and darwin (darwin may need additional gcc patches).
213 The 2.8 release has the following notable changes:
215 <li>The plugin loads faster due to exporting fewer symbols.</li>
216 <li>Additional vector operations such as addps256 are now supported.</li>
217 <li>Ada global variables with no initial value are no longer zero initialized,
218 resulting in better optimization.</li>
219 <li>The '-fplugin-arg-dragonegg-enable-gcc-optzns' flag now runs all gcc
220 optimizers, rather than just a handful.</li>
221 <li>Fortran programs using common variables now link correctly.</li>
222 <li>GNU OMP constructs no longer crash the compiler.</li>
229 <!--=========================================================================-->
230 <div class="doc_subsection">
231 <a name="lldb">LLDB: Low Level Debugger</a>
234 <div class="doc_text">
236 <a href="http://lldb.llvm.org/">LLDB</a> is</p>
247 <!--=========================================================================-->
248 <div class="doc_subsection">
249 <a name="libc++">libc++: C++ Standard Library</a>
252 <div class="doc_text">
254 <a href="http://libc++.llvm.org/">libc++</a> is</p>
266 <!-- *********************************************************************** -->
267 <div class="doc_section">
268 <a name="externalproj">External Open Source Projects Using LLVM 2.8</a>
270 <!-- *********************************************************************** -->
272 <div class="doc_text">
274 <p>An exciting aspect of LLVM is that it is used as an enabling technology for
275 a lot of other language and tools projects. This section lists some of the
276 projects that have already been updated to work with LLVM 2.8.</p>
279 <!--=========================================================================-->
280 <div class="doc_subsection">
281 <a name="tce">TTA-based Codesign Environment (TCE)</a>
284 <div class="doc_text">
286 <a href="http://tce.cs.tut.fi/">TCE</a> is a toolset for designing
287 application-specific processors (ASP) based on the Transport triggered
288 architecture (TTA). The toolset provides a complete co-design flow from C/C++
289 programs down to synthesizable VHDL and parallel program binaries. Processor
290 customization points include the register files, function units, supported
291 operations, and the interconnection network.</p>
293 <p>TCE uses llvm-gcc/Clang and LLVM for C/C++ language support, target
294 independent optimizations and also for parts of code generation. It generates
295 new LLVM-based code generators "on the fly" for the designed TTA processors and
296 loads them in to the compiler backend as runtime libraries to avoid per-target
297 recompilation of larger parts of the compiler chain.</p>
301 <!--=========================================================================-->
302 <div class="doc_subsection">
303 <a name="Horizon">Horizon Bytecode Compiler</a>
306 <div class="doc_text">
308 <a href="http://www.quokforge.org/projects/horizon">Horizon</a> is a bytecode
309 language and compiler written on top of LLVM, intended for producing
310 single-address-space managed code operating systems that
311 run faster than the equivalent multiple-address-space C systems.
312 More in-depth blurb is available on <a
313 href="http://www.quokforge.org/projects/horizon/wiki/Wiki">the wiki</a>.</p>
317 <!--=========================================================================-->
318 <div class="doc_subsection">
319 <a name="clamav">Clam AntiVirus</a>
322 <div class="doc_text">
324 <a href=http://www.clamav.net>Clam AntiVirus</a> is an open source (GPL)
325 anti-virus toolkit for UNIX, designed especially for e-mail scanning on mail
326 gateways. Since version 0.96 it has <a
327 href="http://vrt-sourcefire.blogspot.com/2010/09/introduction-to-clamavs-low-level.html">bytecode
328 signatures</a> that allow writing detections for complex malware. It
329 uses LLVM's JIT to speed up the execution of bytecode on
330 X86,X86-64,PPC32/64, falling back to its own interpreter otherwise.
331 The git version was updated to work with LLVM 2.8
335 href="http://git.clamav.net/gitweb?p=clamav-bytecode-compiler.git;a=blob_plain;f=docs/user/clambc-user.pdf">
336 ClamAV bytecode compiler</a> uses Clang and LLVM to compile a C-like
337 language, insert runtime checks, and generate ClamAV bytecode.</p>
341 <!--=========================================================================-->
342 <div class="doc_subsection">
343 <a name="pure">Pure</a>
346 <div class="doc_text">
348 <a href="http://pure-lang.googlecode.com/">Pure</a>
349 is an algebraic/functional
350 programming language based on term rewriting. Programs are collections
351 of equations which are used to evaluate expressions in a symbolic
352 fashion. Pure offers dynamic typing, eager and lazy evaluation, lexical
353 closures, a hygienic macro system (also based on term rewriting),
354 built-in list and matrix support (including list and matrix
355 comprehensions) and an easy-to-use C interface. The interpreter uses
356 LLVM as a backend to JIT-compile Pure programs to fast native code.</p>
358 <p>Pure versions 0.44 and later have been tested and are known to work with
359 LLVM 2.8 (and continue to work with older LLVM releases >= 2.5).</p>
363 <!--=========================================================================-->
364 <div class="doc_subsection">
365 <a name="GHC">Glasgow Haskell Compiler (GHC)</a>
368 <div class="doc_text">
370 <a href="http://www.haskell.org/ghc/">GHC</a> is an open source,
371 state-of-the-art programming suite for
372 Haskell, a standard lazy functional programming language. It includes
373 an optimizing static compiler generating good code for a variety of
374 platforms, together with an interactive system for convenient, quick
377 <p>In addition to the existing C and native code generators, GHC 7.0 now
379 href="http://hackage.haskell.org/trac/ghc/wiki/Commentary/Compiler/Backends/LLVM">LLVM
380 code generator</a>. GHC supports LLVM 2.7 and later.</p>
384 <!--=========================================================================-->
385 <div class="doc_subsection">
386 <a name="Clay">Clay Programming Language</a>
389 <div class="doc_text">
391 <a href="http://tachyon.in/clay/">Clay</a> is a new systems programming
392 language that is specifically designed for generic programming. It makes
393 generic programming very concise thanks to whole program type propagation. It
394 uses LLVM as its backend.</p>
398 <!--=========================================================================-->
399 <div class="doc_subsection">
400 <a name="llvm-py">llvm-py Python Bindings for LLVM</a>
403 <div class="doc_text">
405 <a href="http://www.mdevan.org/llvm-py/">llvm-py</a> has been updated to work
406 with LLVM 2.8. llvm-py provides Python bindings for LLVM, allowing you to write a
407 compiler backend or a VM in Python.</p>
412 <!--=========================================================================-->
413 <div class="doc_subsection">
414 <a name="FAUST">FAUST Real-Time Audio Signal Processing Language</a>
417 <div class="doc_text">
419 <a href="http://faust.grame.fr">FAUST</a> is a compiled language for real-time
420 audio signal processing. The name FAUST stands for Functional AUdio STream. Its
421 programming model combines two approaches: functional programming and block
422 diagram composition. In addition with the C, C++, JAVA output formats, the
423 Faust compiler can now generate LLVM bitcode, and works with LLVM 2.7 and
428 <!--=========================================================================-->
429 <div class="doc_subsection">
430 <a name="jade">Jade Just-in-time Adaptive Decoder Engine</a>
433 <div class="doc_text">
435 href="http://sourceforge.net/apps/trac/orcc/wiki/JadeDocumentation">Jade</a>
436 (Just-in-time Adaptive Decoder Engine) is a generic video decoder engine using
437 LLVM for just-in-time compilation of video decoder configurations. Those
438 configurations are designed by MPEG Reconfigurable Video Coding (RVC) committee.
439 MPEG RVC standard is built on a stream-based dataflow representation of
440 decoders. It is composed of a standard library of coding tools written in
441 RVC-CAL language and a dataflow configuration &emdash; block diagram &emdash;
444 <p>Jade project is hosted as part of the <a href="http://orcc.sf.net">Open
445 RVC-CAL Compiler</a> and requires it to translate the RVC-CAL standard library
446 of video coding tools into an LLVM assembly code.</p>
450 <!--=========================================================================-->
451 <div class="doc_subsection">
452 <a name="neko_llvm_jit">LLVM JIT for Neko VM</a>
455 <div class="doc_text">
456 <p><a href="http://github.com/vava/neko_llvm_jit">Neko LLVM JIT</a>
457 replaces the standard Neko JIT with an LLVM-based implementation. While not
458 fully complete, it is already providing a 1.5x speedup on 64-bit systems.
459 Neko LLVM JIT requires LLVM 2.8 or later.</p>
463 <!--=========================================================================-->
464 <div class="doc_subsection">
465 <a name="crack">Crack Scripting Language</a>
468 <div class="doc_text">
470 <a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide
471 the ease of development of a scripting language with the performance of a
472 compiled language. The language derives concepts from C++, Java and Python,
473 incorporating object-oriented programming, operator overloading and strong
474 typing. Crack 0.2 works with LLVM 2.7, and the forthcoming Crack 0.2.1 release
475 builds on LLVM 2.8.</p>
479 <!--=========================================================================-->
480 <div class="doc_subsection">
481 <a name="DresdenTM">Dresden TM Compiler (DTMC)</a>
484 <div class="doc_text">
486 <a href="http://tm.inf.tu-dresden.de">DTMC</a> provides support for
487 Transactional Memory, which is an easy-to-use and efficient way to synchronize
488 accesses to shared memory. Transactions can contain normal C/C++ code (e.g.,
489 __transaction { list.remove(x); x.refCount--; }) and will be executed
490 virtually atomically and isolated from other transactions.</p>
494 <!--=========================================================================-->
495 <div class="doc_subsection">
496 <a name="Kai">Kai Interpreter</a>
499 <div class="doc_text">
501 <a href="http://www.oriontransfer.co.nz/research/kai">Kai</a> (Japanese 会 for
502 meeting/gathering) is an experimental interpreter that provides a highly
503 extensible runtime environment and explicit control over the compilation
504 process. Programs are defined using nested symbolic expressions, which are all
505 parsed into first-class values with minimal intrinsic semantics. Kai can
506 generate optimised code at run-time (using LLVM) in order to exploit the nature
507 of the underlying hardware and to integrate with external software libraries.
508 It is a unique exploration into world of dynamic code compilation, and the
509 interaction between high level and low level semantics.</p>
514 <!-- *********************************************************************** -->
515 <div class="doc_section">
516 <a name="whatsnew">What's New in LLVM 2.8?</a>
518 <!-- *********************************************************************** -->
520 <div class="doc_text">
522 <p>This release includes a huge number of bug fixes, performance tweaks and
523 minor improvements. Some of the major improvements and new features are listed
529 <!--=========================================================================-->
530 <div class="doc_subsection">
531 <a name="orgchanges">LLVM Community Changes</a>
534 <div class="doc_text">
536 <p>In addition to changes to the code, between LLVM 2.7 and 2.8, a number of
537 organization changes have happened:
541 <li>libc++ and lldb are new</li>
542 <li>Debugging optimized code support.</li>
546 <!--=========================================================================-->
547 <div class="doc_subsection">
548 <a name="majorfeatures">Major New Features</a>
551 <div class="doc_text">
553 <p>LLVM 2.8 includes several major new capabilities:</p>
557 <li>Direct .o file writing support for darwin/x86[64].</li>
562 <!--=========================================================================-->
563 <div class="doc_subsection">
564 <a name="coreimprovements">LLVM IR and Core Improvements</a>
567 <div class="doc_text">
568 <p>LLVM IR has several new features for better support of new targets and that
569 expose new optimization opportunities:</p>
573 memcpy, memmove, and memset now take address space qualified pointers + volatile.
574 per-instruction debug info metadata is much faster and uses less space (new DebugLoc class).
575 New "trap values" concept: http://llvm.org/docs/LangRef.html#trapvalues
576 New linker_private_weak and linker_private_weak_def_auto linkage types
577 Triples are now stored in normalized form. Triple::normalize.
583 <!--=========================================================================-->
584 <div class="doc_subsection">
585 <a name="optimizer">Optimizer Improvements</a>
588 <div class="doc_text">
590 <p>In addition to a large array of minor performance tweaks and bug fixes, this
591 release includes a few major enhancements and additions to the optimizers:</p>
596 Preliminary work on TBAA but not usable in 2.8.
597 New CorrelatedValuePropagation pass, not on by default in 2.8 yet.
598 JumpThreading much more aggressive about implied value relations.
599 New RegionInfo pass "opt -regions analyze" or "opt -view-regions".
600 Improved trip count analysis for <= and >= loops, and uses sign overflow info.
601 llvm.dbg.value: variable debug info for optimized code
602 Now iterate function passes when a cgsccpassmanager detects a devirtualization
603 Atomic lowering patch: -loweratomic (see Passes.html#loweratomic)
610 <!--=========================================================================-->
611 <div class="doc_subsection">
612 <a name="executionengine">Interpreter and JIT Improvements</a>
615 <div class="doc_text">
624 <!--=========================================================================-->
625 <div class="doc_subsection">
626 <a name="mc">MC Level Improvements</a>
629 <div class="doc_text">
631 The LLVM Machine Code (aka MC) subsystem was created to solve a number
632 of problems in the realm of assembly, disassembly, object file format handling,
633 and a number of other related areas that CPU instruction-set level tools work
636 <p>The MC subproject has made great leaps in LLVM 2.8. For example, support for
637 directly writing .o files from LLC (and clang) now works reliably for
638 darwin/x86[-64] (including inline assembly support) and the integrated
639 assembler is turned on by default in Clang for these targets. This provides
640 improved compile times among other things.</p>
643 <li>The entire compiler has converted over to using the MCStreamer assembler API
644 instead of writing out a .s file textually.</li>
645 <li>The "assembler parser" is far more mature than in 2.7, supporting a full
646 complement of directives, now supports assembler macros, etc.</li>
647 <li>The "assembler backend" has been completed, including support for relaxation
648 relocation processing and all the other things that an assembler does.</li>
649 <li>The MachO file format support is now fully functional and works.</li>
650 <li>The MC disassembler now fully supports ARM and Thumb. ARM assembler support
651 is still in early development though.</li>
652 <li>The X86 MC assembler now supports the X86 AES and AVX instruction set.</li>
653 <li>Work on ELF and COFF support is well underway, but isn't useful yet in LLVM
654 2.8. Please contact the llvmdev mailing list if you're interested in
658 <p>For more information, please see the <a
659 href="http://blog.llvm.org/2010/04/intro-to-llvm-mc-project.html">Intro to the
660 LLVM MC Project Blog Post</a>.
667 <!--=========================================================================-->
668 <div class="doc_subsection">
669 <a name="codegen">Target Independent Code Generator Improvements</a>
672 <div class="doc_text">
674 <p>We have put a significant amount of work into the code generator
675 infrastructure, which allows us to implement more aggressive algorithms and make
681 MachineCSE tuned and on by default.
683 Rewrote tblgen's type inference for backends to be more consistent and
684 diagnose more target bugs. This also allows limited support for writing
685 patterns for instructions that return multiple results, e.g. a virtual
686 register and a flag result. Stuff that used 'parallel' before should use
689 New -regalloc=fast, =local got removed
690 New -regalloc=default option that chooses a register allocator based on the -O optimization level.
691 New SubRegIndex tblgen class for targets -> jakob
693 Bottom up fast isel. Simple Load reuse. No more machinedce.
694 IR ABI: <3 x float> is passed as <4 x float> instead of 3 floats.
696 New COPY instruction. copyRegToReg -> copyPhysReg, isMoveInstr is gone.
697 RenderMachineFunction: -rendermf
699 Evan: Teach bottom up pre-ra scheduler to track register pressure. Work in progress.
700 Evan: Add an ILP scheduler. On x86_64, this is a win for all tests in CFP2000. It also sped up 256.bzip2 by 16%.
702 New OptimizeExts+OptimizeCmps -> PeepholeOptimizer pass
703 New LocalStackSlotAllocation.cpp pass (jimg)
704 Atomics now get legalized when not natively supported (jim g)
706 -ffunction-sections and -fdata-sections are supported on ELF targets.
707 -momit-leaf-frame-pointer now supported.
712 <!--=========================================================================-->
713 <div class="doc_subsection">
714 <a name="x86">X86-32 and X86-64 Target Improvements</a>
717 <div class="doc_text">
718 <p>New features of the X86 target include:
722 <li>The X86 backend now supports holding X87 floating point stack values
723 in registers across basic blocks, dramatically improving performance of code
724 that uses long double, and when targetting CPUs that don't support SSE.</li>
726 New SSEDomainFix pass:
727 On Nehalem and newer CPUs there is a 2 cycle latency penalty on using a
728 register in a different domain than where it was defined. Some instructions
729 have equvivalents for different domains, like por/orps/orpd. The
730 SSEDomainFix pass tries to minimize the number of domain crossings by
731 changing between equvivalent opcodes where possible.
733 X86 backend attempts to promote 16-bit integer operations to 32-bits to avoid
734 0x66 prefixes, which are slow on some microarchitectures and bloat the code
737 New support for X86 "thiscall" calling convention (x86_thiscallcc in IR) for windows.
739 New llvm.x86.int intrinsic (for int $42 and int3)
741 Verbose assembly decodes X86 shuffle instructions, e.g.:
742 insertps $113, %xmm3, %xmm0 ## xmm0 = zero,xmm0[1,2],xmm3[1]
743 unpcklps %xmm1, %xmm0 ## xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
744 pshufd $1, %xmm1, %xmm1 ## xmm1 = xmm1[1,0,0,0]
746 X86 ABI: <2 x float> in IR no longer maps onto MMX, it turns into <4 x float>
748 new GHC calling convention
754 <!--=========================================================================-->
755 <div class="doc_subsection">
756 <a name="ARM">ARM Target Improvements</a>
759 <div class="doc_text">
760 <p>New features of the ARM target include:
765 NEON: Better performance for QQQQ (4-consecutive Q register) instructions. New reg sequence abstraction?
766 ARM: Better scheduling (list-hybrid, hybrid?)
767 ARM: Tail call support.
768 ARM: General performance work and tuning.
770 ARM: Half float support through intrinsics LangRef.html#int_fp16
771 <li>ARMGlobalMerge: <!-- Anton --> </li>
773 <li>The ARM NEON intrinsics have been substantially reworked to reduce
774 redundancy and improve code generation. Some of the major changes are:
777 All of the NEON load and store intrinsics (llvm.arm.neon.vld* and
778 llvm.arm.neon.vst*) take an extra parameter to specify the alignment in bytes
779 of the memory being accessed.
782 The llvm.arm.neon.vaba intrinsic (vector absolute difference and
783 accumulate) has been removed. This operation is now represented using
784 the llvm.arm.neon.vabd intrinsic (vector absolute difference) followed by a
788 The llvm.arm.neon.vabdl and llvm.arm.neon.vabal intrinsics (lengthening
789 vector absolute difference with and without accumlation) have been removed.
790 They are represented using the llvm.arm.neon.vabd intrinsic (vector absolute
791 difference) followed by a vector zero-extend operation, and for vabal,
795 The llvm.arm.neon.vmovn intrinsic has been removed. Calls of this intrinsic
796 are now replaced by vector truncate operations.
799 The llvm.arm.neon.vmovls and llvm.arm.neon.vmovlu intrinsics have been
800 removed. They are now represented as vector sign-extend (vmovls) and
801 zero-extend (vmovlu) operations.
804 The llvm.arm.neon.vaddl*, llvm.arm.neon.vaddw*, llvm.arm.neon.vsubl*, and
805 llvm.arm.neon.vsubw* intrinsics (lengthening vector add and subtract) have
806 been removed. They are replaced by vector add and vector subtract operations
807 where one (vaddw, vsubw) or both (vaddl, vsubl) of the operands are either
808 sign-extended or zero-extended.
811 The llvm.arm.neon.vmulls, llvm.arm.neon.vmullu, llvm.arm.neon.vmlal*, and
812 llvm.arm.neon.vmlsl* intrinsics (lengthening vector multiply with and without
813 accumulation and subtraction) have been removed. These operations are now
814 represented as vector multiplications where the operands are either
815 sign-extended or zero-extended, followed by a vector add for vmlal or a
816 vector subtract for vmlsl. Note that the polynomial vector multiply
817 intrinsic, llvm.arm.neon.vmullp, remains unchanged.
824 <!--=========================================================================-->
825 <div class="doc_subsection">
826 <a name="newapis">New Useful APIs</a>
829 <div class="doc_text">
831 <p>This release includes a number of new APIs that are used internally, which
832 may also be useful for external clients.
842 <!--=========================================================================-->
843 <div class="doc_subsection">
844 <a name="otherimprovements">Other Improvements and New Features</a>
847 <div class="doc_text">
848 <p>Other miscellaneous features include:</p>
857 <!--=========================================================================-->
858 <div class="doc_subsection">
859 <a name="changes">Major Changes and Removed Features</a>
862 <div class="doc_text">
864 <p>If you're already an LLVM user or developer with out-of-tree changes based
865 on LLVM 2.7, this section lists some "gotchas" that you may run into upgrading
866 from the previous release.</p>
869 renamed "Release" -> "Release+Asserts"; "Release-Asserts" -> "Release etc.
873 <li>.ll file doesn't produce #uses comments anymore, to get them, run a .bc file
874 through "llvm-dis --show-annotations".</li>
875 <li>MSIL Backend removed.</li>
876 <li>ABCD and SSI passes removed.</li>
877 <li>'Union' LLVM IR feature removed.</li>
878 <li>SCCVN pass removed.</li>
881 <p>In addition, many APIs have changed in this release. Some of the major LLVM
885 RegisterPass<> -> INTIALIZE_PASS()
889 <li>LLVM 2.8 changes the internal order of operands in <a
890 href="http://llvm.org/doxygen/classllvm_1_1InvokeInst.html"><tt>InvokeInst</tt></a>
891 and <a href="http://llvm.org/doxygen/classllvm_1_1CallInst.html"><tt>CallInst</tt></a>.
892 To be portable across releases, resort to <tt>CallSite</tt> and the
893 high-level accessors, such as <tt>getCalledValue</tt> and <tt>setUnwindDest</tt>.
896 You can no longer pass use_iterators directly to cast<> (and similar), because
897 these routines tend to perform costly dereference operations more than once. You
898 have to dereference the iterators yourself and pass them in.
901 llvm.memcpy.*, llvm.memset.*, llvm.memmove.* (and possibly other?) intrinsics
902 take an extra parameter now (i1 isVolatile), totaling 5 parameters.
903 If you were creating these intrinsic calls and prototypes yourself (as opposed
904 to using Intrinsic::getDeclaration), you can use UpgradeIntrinsicFunction/UpgradeIntrinsicCall
905 to be portable accross releases.
906 Note that you cannot use Intrinsic::getDeclaration() in a backwards compatible
907 way (needs 2/3 types now, in 2.7 it needed just 1).
910 SetCurrentDebugLocation takes a DebugLoc now instead of a MDNode.
911 Change your code to use
912 SetCurrentDebugLocation(DebugLoc::getFromDILocation(...)).
915 VISIBILITY_HIDDEN is gone.
918 The <tt>RegisterPass</tt> and <tt>RegisterAnalysisGroup</tt> templates are
919 considered deprecated, but continue to function in LLVM 2.8. Clients are
920 strongly advised to use the upcoming <tt>INITIALIZE_PASS()</tt> and
921 <tt>INITIALIZE_AG_PASS()</tt> macros instead.
923 SMDiagnostic takes different parameters now. //FIXME: how to upgrade?
926 The constructor for the Triple class no longer tries to understand odd triple
927 specifications. Frontends should ensure that they only pass valid triples to
928 LLVM. The Triple::normalize utility method has been added to help front-ends
929 deal with funky triples.
931 Some APIs got renamed:
933 <li>llvm_report_error -> report_fatal_error</li>
934 <li>llvm_install_error_handler -> install_fatal_error_handler</li>
935 <li>llvm::DwarfExceptionHandling -> llvm::JITExceptionHandling</li>
944 <!-- *********************************************************************** -->
945 <div class="doc_section">
946 <a name="knownproblems">Known Problems</a>
948 <!-- *********************************************************************** -->
950 <div class="doc_text">
952 <p>This section contains significant known problems with the LLVM system,
953 listed by component. If you run into a problem, please check the <a
954 href="http://llvm.org/bugs/">LLVM bug database</a> and submit a bug if
955 there isn't already one.</p>
959 <!-- ======================================================================= -->
960 <div class="doc_subsection">
961 <a name="experimental">Experimental features included with this release</a>
964 <div class="doc_text">
966 <p>The following components of this LLVM release are either untested, known to
967 be broken or unreliable, or are in early development. These components should
968 not be relied on, and bugs should not be filed against them, but they may be
969 useful to some people. In particular, if you would like to work on one of these
970 components, please contact us on the <a
971 href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev">LLVMdev list</a>.</p>
974 <li>The Alpha, SPU, MIPS, PIC16, Blackfin, MSP430, SystemZ and MicroBlaze
975 backends are experimental.</li>
976 <li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets
977 other than darwin-i386 and darwin-x86_64.</li>
982 <!-- ======================================================================= -->
983 <div class="doc_subsection">
984 <a name="x86-be">Known problems with the X86 back-end</a>
987 <div class="doc_text">
990 <li>The X86 backend does not yet support
991 all <a href="http://llvm.org/PR879">inline assembly that uses the X86
992 floating point stack</a>. It supports the 'f' and 't' constraints, but not
994 <li>Win64 code generation wasn't widely tested. Everything should work, but we
995 expect small issues to happen. Also, llvm-gcc cannot build the mingw64
996 runtime currently due to lack of support for the 'u' inline assembly
997 constraint and for X87 floating point inline assembly.</li>
998 <li>The X86-64 backend does not yet support the LLVM IR instruction
999 <tt>va_arg</tt>. Currently, front-ends support variadic
1000 argument constructs on X86-64 by lowering them manually.</li>
1005 <!-- ======================================================================= -->
1006 <div class="doc_subsection">
1007 <a name="ppc-be">Known problems with the PowerPC back-end</a>
1010 <div class="doc_text">
1013 <li>The Linux PPC32/ABI support needs testing for the interpreter and static
1014 compilation, and lacks support for debug information.</li>
1019 <!-- ======================================================================= -->
1020 <div class="doc_subsection">
1021 <a name="arm-be">Known problems with the ARM back-end</a>
1024 <div class="doc_text">
1027 <li>Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6
1028 processors, thumb programs can crash or produce wrong
1029 results (<a href="http://llvm.org/PR1388">PR1388</a>).</li>
1030 <li>Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
1036 <!-- ======================================================================= -->
1037 <div class="doc_subsection">
1038 <a name="sparc-be">Known problems with the SPARC back-end</a>
1041 <div class="doc_text">
1044 <li>The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not
1045 support the 64-bit SPARC ABI (-m64).</li>
1050 <!-- ======================================================================= -->
1051 <div class="doc_subsection">
1052 <a name="mips-be">Known problems with the MIPS back-end</a>
1055 <div class="doc_text">
1058 <li>64-bit MIPS targets are not supported yet.</li>
1063 <!-- ======================================================================= -->
1064 <div class="doc_subsection">
1065 <a name="alpha-be">Known problems with the Alpha back-end</a>
1068 <div class="doc_text">
1072 <li>On 21164s, some rare FP arithmetic sequences which may trap do not have the
1073 appropriate nops inserted to ensure restartability.</li>
1078 <!-- ======================================================================= -->
1079 <div class="doc_subsection">
1080 <a name="c-be">Known problems with the C back-end</a>
1083 <div class="doc_text">
1085 <p>The C backend has numerous problems and is not being actively maintained.
1086 Depending on it for anything serious is not advised.</p>
1089 <li><a href="http://llvm.org/PR802">The C backend has only basic support for
1090 inline assembly code</a>.</li>
1091 <li><a href="http://llvm.org/PR1658">The C backend violates the ABI of common
1092 C++ programs</a>, preventing intermixing between C++ compiled by the CBE and
1093 C++ code compiled with <tt>llc</tt> or native compilers.</li>
1094 <li>The C backend does not support all exception handling constructs.</li>
1095 <li>The C backend does not support arbitrary precision integers.</li>
1101 <!-- ======================================================================= -->
1102 <div class="doc_subsection">
1103 <a name="llvm-gcc">Known problems with the llvm-gcc front-end</a>
1106 <div class="doc_text">
1108 <p>llvm-gcc is generally very stable for the C family of languages. The only
1109 major language feature of GCC not supported by llvm-gcc is the
1110 <tt>__builtin_apply</tt> family of builtins. However, some extensions
1111 are only supported on some targets. For example, trampolines are only
1112 supported on some targets (these are used when you take the address of a
1113 nested function).</p>
1115 <p>Fortran support generally works, but there are still several unresolved bugs
1116 in <a href="http://llvm.org/bugs/">Bugzilla</a>. Please see the
1117 tools/gfortran component for details. Note that llvm-gcc is missing major
1118 Fortran performance work in the frontend and library that went into GCC after
1119 4.2. If you are interested in Fortran, we recommend that you consider using
1120 <a href="#dragonegg">dragonegg</a> instead.</p>
1122 <p>The llvm-gcc 4.2 Ada compiler has basic functionality. However, this is not a
1123 mature technology, and problems should be expected. For example:</p>
1125 <li>The Ada front-end currently only builds on X86-32. This is mainly due
1126 to lack of trampoline support (pointers to nested functions) on other platforms.
1127 However, it <a href="http://llvm.org/PR2006">also fails to build on X86-64</a>
1128 which does support trampolines.</li>
1129 <li>The Ada front-end <a href="http://llvm.org/PR2007">fails to bootstrap</a>.
1130 This is due to lack of LLVM support for <tt>setjmp</tt>/<tt>longjmp</tt> style
1131 exception handling, which is used internally by the compiler.
1132 Workaround: configure with <tt>--disable-bootstrap</tt>.</li>
1133 <li>The c380004, <a href="http://llvm.org/PR2010">c393010</a>
1134 and <a href="http://llvm.org/PR2421">cxg2021</a> ACATS tests fail
1135 (c380004 also fails with gcc-4.2 mainline).
1136 If the compiler is built with checks disabled then <a href="http://llvm.org/PR2010">c393010</a>
1137 causes the compiler to go into an infinite loop, using up all system memory.</li>
1138 <li>Some GCC specific Ada tests continue to crash the compiler.</li>
1139 <li>The <tt>-E</tt> binder option (exception backtraces)
1140 <a href="http://llvm.org/PR1982">does not work</a> and will result in programs
1141 crashing if an exception is raised. Workaround: do not use <tt>-E</tt>.</li>
1142 <li>Only discrete types <a href="http://llvm.org/PR1981">are allowed to start
1143 or finish at a non-byte offset</a> in a record. Workaround: do not pack records
1144 or use representation clauses that result in a field of a non-discrete type
1145 starting or finishing in the middle of a byte.</li>
1146 <li>The <tt>lli</tt> interpreter <a href="http://llvm.org/PR2009">considers
1147 'main' as generated by the Ada binder to be invalid</a>.
1148 Workaround: hand edit the file to use pointers for <tt>argv</tt> and
1149 <tt>envp</tt> rather than integers.</li>
1150 <li>The <tt>-fstack-check</tt> option <a href="http://llvm.org/PR2008">is
1155 <!-- *********************************************************************** -->
1156 <div class="doc_section">
1157 <a name="additionalinfo">Additional Information</a>
1159 <!-- *********************************************************************** -->
1161 <div class="doc_text">
1163 <p>A wide variety of additional information is available on the <a
1164 href="http://llvm.org">LLVM web page</a>, in particular in the <a
1165 href="http://llvm.org/docs/">documentation</a> section. The web page also
1166 contains versions of the API documentation which is up-to-date with the
1167 Subversion version of the source code.
1168 You can access versions of these documents specific to this release by going
1169 into the "<tt>llvm/doc/</tt>" directory in the LLVM tree.</p>
1171 <p>If you have any questions or comments about LLVM, please feel free to contact
1172 us via the <a href="http://llvm.org/docs/#maillist"> mailing
1177 <!-- *********************************************************************** -->
1181 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1182 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
1183 <a href="http://validator.w3.org/check/referer"><img
1184 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
1186 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
1187 Last modified: $Date$