This document contains the release notes for the LLVM Compiler Infrastructure, release 2.5. Here we describe the status of LLVM, including major improvements from the previous release and significant known problems. All LLVM releases may be downloaded from the LLVM releases web site.

For more information about LLVM, including information about the latest release, please check out the main LLVM web site. If you have questions or comments, the LLVM Developer's Mailing List is a good place to send them.

Note that if you are reading this file from a Subversion checkout or the main LLVM web page, this document applies to the next release, not the current one. To see the release notes for a specific release, please see the releases page.

The LLVM 2.5 distribution currently consists of code from the core LLVM repository —which roughly includes the LLVM optimizers, code generators and supporting tools — and the llvm-gcc repository. In addition to this code, the LLVM Project includes other sub-projects that are in development. The two which are the most actively developed are the Clang Project and the VMKit Project.

The Clang project is an effort to build a set of new 'LLVM native' front-end technologies for the LLVM optimizer and code generator. While Clang is not included in the LLVM 2.5 release, it is continuing to make major strides forward in all areas. Its C and Objective-C parsing and code generation support is now very solid. For example, it is capable of successfully building many real-world applications for X86-32 andX86-64, including the FreeBSD kernel. C++ is also making incredible progress, and work on templates has recently started.

While Clang is not yet production quality, it is progressing very nicely and is quite usable for building many C and Objective-C applications. If you are interested in fast compiles and good diagnostics, we encourage you to try it out by building from mainline and reporting any issues you hit to the Clang front-end mailing list.

In the LLVM 2.5 time-frame, the Clang team has made many improvements:

• Clang now has a new driver, which is focused on providing a GCC-compatible interface.
• The X86-64 ABI is now supported.
• Precompiled header support is now implemented.
• Objective-C support is significantly improved beyond LLVM 2.4, supporting many features, such as Objective-C Garbage Collection.
• Many many bugs are fixed and many features have been added.

Previously announced in the last LLVM release, the Clang project also includes an early stage static source code analysis tool for automatically finding bugs in C and Objective-C programs. The tool performs a growing set of checks to find bugs that occur on a specific path within a program.

In the LLVM 2.5 time-frame there have been many significant improvements to the analyzer's core path simulation engine and machinery for generating path-based bug reports to end-users. Particularly noteworthy improvements include experimental support for full field-sensitivity and reasoning about heap objects as well as an improved value-constraints subengine that does a much better job of reasoning about inequality relationships (e.g., x > 2) between variables and constants.

The set of checks performed by the static analyzer continues to expand, and future plans for the tool include full source-level inter-procedural analysis and deeper checks such as buffer overrun detection. There are many opportunities to extend and enhance the static analyzer, and anyone interested in working on this project is encouraged to get involved!

The VMKit project is an implementation of a JVM and a CLI Virtual Machines (Microsoft .NET is an implementation of the CLI) using the Just-In-Time compiler of LLVM.

Following LLVM 2.5, VMKit has its second release that you can find on its webpage. The release includes bug fixes, cleanup and new features. The major changes are:

• Ahead of Time compiler: compiles .class files to llvm .bc. VMKit uses this functionality to native compile the standard classes (e.g. java.lang.String). Users can compile AoT .class files into dynamic libraries and run them with the help of VMKit.
• New exception model: the dwarf exception model is very slow for exception-intensive applications, so the JVM has had a new implementation of exceptions which check at each function call if an exception happened. There is a low performance penalty on applications without exceptions, but it is a big gain for exception-intensive applications. For example the jack benchmark in Spec JVM98 is 6x faster (performance gain of 83%).
• User-level management of thread stacks, so that thread local data access at runtime is fast and portable.
• Implementation of biased locking for faster object synchronizations at runtime.
• New support for OSX/X64, Linux/X64 (with the Boehm GC) and Linux/ppc32.

Pure is an algebraic/functional programming language based on term rewriting. Programs are collections of equations which are used to evaluate expressions in a symbolic fashion. Pure offers dynamic typing, eager and lazy evaluation, lexical closures, a hygienic macro system (also based on term rewriting), built-in list and matrix support (including list and matrix comprehensions) and an easy-to-use C interface. The interpreter uses LLVM as a backend to JIT-compile Pure programs to fast native code.

In addition to the usual algebraic data structures, Pure also has MATLAB-style matrices in order to support numeric computations and signal processing in an efficient way. Pure is mainly aimed at mathematical applications right now, but it has been designed as a general purpose language. The dynamic interpreter environment and the C interface make it possible to use it as a kind of functional scripting language for many application areas.

LDC is an implementation of the D Programming Language using the LLVM optimizer and code generator. The LDC project works great with the LLVM 2.5 release. General improvements in this cycle have included new inline asm constraint handling, better debug info support, general bugfixes, and better x86-64 support. This has allowed some major improvements in LDC, getting us much closer to being as fully featured as the original DMD compiler from DigitalMars.

Roadsend PHP (rphp) is an open source implementation of the PHP programming language that uses LLVM for its optimizer, JIT, and static compiler. This is a reimplementation of an earlier project that is now based on LLVM.

This release includes a huge number of bug fixes, performance tweaks, and minor improvements. Some of the major improvements and new features are listed in this section.

LLVM 2.5 includes several major new capabilities:

• LLVM 2.5 includes a brand new XCore backend.
• llvm-gcc now generally supports the GFortran front-end, and the precompiled release binaries now support Fortran, even on Mac OS/X.
• CMake is now used by the LLVM build process on Windows. It automatically generates Visual Studio project files (and more) from a set of simple text files. This makes it much easier to maintain. In time, we'd like to standardize on CMake for everything.
• LLVM 2.5 now uses (and includes) Google Test for unit testing.
• The LLVM native code generator now supports arbitrary precision integers. Types like i33 have long been valid in the LLVM IR, but were previously only supported by the interpreter. Note that the C backend still does not support these.
• LLVM 2.5 no longer uses 'bison,' so it is easier to build on Windows.

LLVM fully supports the llvm-gcc 4.2 front-end, which marries the GCC front-ends and driver with the LLVM optimizer and code generator. It currently includes support for the C, C++, Objective-C, Ada, and Fortran front-ends.

• In this release, the GCC inliner is completely disabled. Previously the GCC inliner was used to handle always-inline functions and other cases. This caused problems with code size growth, and it is completely disabled in this release.
• llvm-gcc (and LLVM in general) now support code generation for stack canaries, which is an effective form of buffer overflow protection. llvm-gcc supports this with the -fstack-protector command line option (just like GCC). In LLVM IR, you can request code generation for stack canaries with function attributes.

LLVM IR has several new features that are used by our existing front-ends and can be useful if you are writing a front-end for LLVM:

• The shufflevector instruction has been generalized to allow different shuffle mask width than its input vectors. This allows you to use shufflevector to combine two "<4 x float>" vectors into a "<8 x float>" for example.
• LLVM IR now supports new intrinsics for computing and acting on overflow of integer operations. This allows efficient code generation for languages that must trap or throw an exception on overflow. While these intrinsics work on all targets, they only generate efficient code on X86 so far.
• LLVM IR now supports a new private linkage type to produce labels that are stripped by the assembler before it produces a .o file (thus they are invisible to the linker).
• LLVM IR supports two new attributes for better alias analysis. The noalias attribute can now be used on the return value of a function to indicate that it returns new memory (e.g. 'malloc', 'calloc', etc). The new nocapture attribute can be used on pointer arguments to indicate that the function does not return the pointer, store it in an object that outlives the call, or let the value of the pointer escape from the function in any other way. Note that it is the pointer itself that must not escape, not the value it points to: loading a value out of the pointer is perfectly fine. Many standard library functions (e.g. 'strlen', 'memcpy') have this property.
• The parser for ".ll" files in lib/AsmParser is now completely rewritten as a recursive descent parser. This parser produces better error messages (including caret diagnostics), is less fragile (less likely to crash on strange things), does not leak memory, is more efficient, and eliminates LLVM's last use of the 'bison' tool.
• Debug information representation and manipulation internals have been consolidated to use a new set of classes in llvm/Analysis/DebugInfo.h. These routines are more efficient, robust, and extensible and replace the older mechanisms. llvm-gcc, clang, and the code generator now use them to create and process debug information.

In addition to a large array of bug fixes and minor performance tweaks, this release includes a few major enhancements and additions to the optimizers:

• The loop optimizer now improves floating point induction variables in several ways, including adding shadow induction variables to avoid "integer <-> floating point" conversions in loops when safe.
• The "-mem2reg" pass is now much faster on code with large basic blocks.
• The "-jump-threading" pass is more powerful: it is iterative and handles threading based on values with fully and partially redundant loads.
• The "-memdep" memory dependence analysis pass (used by GVN and memcpyopt) is both faster and more aggressive.
• The "-scalarrepl" scalar replacement of aggregates pass is more aggressive about promoting unions to registers.

We have put a significant amount of work into the code generator infrastructure, which allows us to implement more aggressive algorithms and make it run faster:

• The Writing an LLVM Compiler Backend document has been greatly expanded and is substantially more complete.
• The SelectionDAG type legalization logic has been completely rewritten, is now more powerful (it supports arbitrary precision integer types for example), and is more correct in several corner cases. The type legalizer converts operations on types that are not natively supported by the target machine into equivalent code sequences that only use natively supported types. The old type legalizer is still available (for now) and will be used if -disable-legalize-types is passed to the code generator.
• The code generator now supports widening illegal vectors to larger legal ones (for example, converting operations on <3 x float> to work on <4 x float>) which is very important for common graphics applications.
• The assembly printers for each target are now split out into their own libraries that are separate from the main code generation logic. This reduces the code size of JIT compilers by not requiring them to be linked in.
• The 'fast' instruction selection path (used at -O0 and for fast JIT compilers) now supports accelerating codegen for code that uses exception handling constructs.
• The optional PBQP register allocator now supports register coalescing.

New features of the X86 target include:

• The llvm.returnaddress intrinsic (which is used to implement __builtin_return_address) now supports non-zero stack depths on X86.
• The X86 backend now supports code generation of vector shift operations using SSE instructions.
• X86-64 code generation now takes advantage of red zone, unless the -mno-red-zone option is specified.
• The X86 backend now supports using address space #256 in LLVM IR as a way of performing memory references off the GS segment register. This allows a front-end to take advantage of very low-level programming techniques when targeting X86 CPUs. See test/CodeGen/X86/movgs.ll for a simple example.
• The X86 backend now supports a -disable-mmx command line option to prevent use of MMX even on chips that support it. This is important for cases where code does not contain the proper llvm.x86.mmx.emms intrinsics.
• The X86 JIT now detects the new Intel Core i7 and Atom chips and auto-configures itself appropriately for the features of these chips.
• The JIT now supports exception handling constructs on Linux/X86-64 and Darwin/x86-64.
• The JIT supports Thread Local Storage (TLS) on Linux/X86-32 but not yet on X86-64.

New features of the PIC16 target include:

• Both direct and indirect load/stores work now.
• Logical, bitwise and conditional operations now work for integer data types.
• Function calls involving basic types work now.
• Support for integer arrays.
• The compiler can now emit libcalls for operations not supported by m/c instructions.
• Support for both data and ROM address spaces.

Things not yet supported:

• Floating point.
• Passing/returning aggregate types to and from functions.
• Variable arguments.
• Indirect function calls.
• Interrupts/programs.
• Debug info.

New features include:

• Beginning with LLVM 2.5, llvmc2 is known as just llvmc. The old llvmc driver was removed.
• The Clang plugin was substantially improved and is now enabled by default. The command llvmc --clang can be now used as a synonym to ccc.
• There is now a --check-graph option, which is supposed to catch common errors like multiple default edges, mismatched output/input language names and cycles. In general, these checks can't be done at compile-time because of the need to support plugins.
• Plugins are now more flexible and can refer to compilation graph nodes and options defined in other plugins. To manage dependencies, a priority-sorting mechanism was introduced. This change affects the TableGen file syntax. See the documentation for details.
• Hooks can now be provided with arguments. The syntax is "$CALL(MyHook, 'Arg1', 'Arg2', 'Arg3')".
• A new option type: multi-valued option, for options that take more than one argument (for example, "-foo a b c").
• New option properties: 'one_or_more', 'zero_or_more', 'hidden' and 'really_hidden'.
• The 'case' expression gained an 'error' action and an 'empty' test (equivalent to "(not (not_empty ...))").
• Documentation now looks more consistent to the rest of the LLVM docs. There is also a man page now.

If you're already an LLVM user or developer with out-of-tree changes based on LLVM 2.4, this section lists some "gotchas" that you may run into upgrading from the previous release.

• llvm-gcc defaults to -fno-math-errno on all X86 targets.

In addition, many APIs have changed in this release. Some of the major LLVM API changes are:

• Some deprecated interfaces to create Instruction subclasses, that were spelled with lower case "create," have been removed.

LLVM is known to work on the following platforms:

• Intel and AMD machines (IA32, X86-64, AMD64, EMT-64) running Red Hat Linux, Fedora Core and FreeBSD (and probably other unix-like systems).
• PowerPC and X86-based Mac OS X systems, running 10.3 and above in 32-bit and 64-bit modes.
• Intel and AMD machines running on Win32 using MinGW libraries (native).
• Intel and AMD machines running on Win32 with the Cygwin libraries (limited support is available for native builds with Visual C++).
• Sun UltraSPARC workstations running Solaris 10.
• Alpha-based machines running Debian GNU/Linux.
• Itanium-based (IA64) machines running Linux and HP-UX.

The core LLVM infrastructure uses GNU autoconf to adapt itself to the machine and operating system on which it is built. However, minor porting may be required to get LLVM to work on new platforms. We welcome your portability patches and reports of successful builds or error messages.

This section contains significant known problems with the LLVM system, listed by component. If you run into a problem, please check the LLVM bug database and submit a bug if there isn't already one.

The following components of this LLVM release are either untested, known to be broken or unreliable, or are in early development. These components should not be relied on, and bugs should not be filed against them, but they may be useful to some people. In particular, if you would like to work on one of these components, please contact us on the LLVMdev list.

• The MSIL, IA64, Alpha, SPU, MIPS, and PIC16 backends are experimental.
• The llc "-filetype=asm" (the default) is the only supported value for this option.

• The X86 backend does not yet support all inline assembly that uses the X86 floating point stack. It supports the 'f' and 't' constraints, but not 'u'.
• The X86 backend generates inefficient floating point code when configured to generate code for systems that don't have SSE2.
• Win64 code generation wasn't widely tested. Everything should work, but we expect small issues to happen. Also, llvm-gcc cannot build the mingw64 runtime currently due to several bugs and due to lack of support for the 'u' inline assembly constraint and for X87 floating point inline assembly.
• The X86-64 backend does not yet support the LLVM IR instruction va_arg. Currently, the llvm-gcc and front-ends support variadic argument constructs on X86-64 by lowering them manually.

• The Linux PPC32/ABI support needs testing for the interpreter and static compilation, and lacks support for debug information.

• Thumb mode works only on ARMv6 or higher processors. On sub-ARMv6 processors, thumb programs can crash or produce wrong results (PR1388).
• Compilation for ARM Linux OABI (old ABI) is supported but not fully tested.
• There is a bug in QEMU-ARM (<= 0.9.0) which causes it to incorrectly execute programs compiled with LLVM. Please use more recent versions of QEMU.

• The SPARC backend only supports the 32-bit SPARC ABI (-m32); it does not support the 64-bit SPARC ABI (-m64).

• The O32 ABI is not fully supported.
• 64-bit MIPS targets are not supported yet.

• On 21164s, some rare FP arithmetic sequences which may trap do not have the appropriate nops inserted to ensure restartability.

• The Itanium backend is highly experimental and has a number of known issues. We are looking for a maintainer for the Itanium backend. If you are interested, please contact the LLVMdev mailing list.

• The C backend has only basic support for inline assembly code.
• The C backend violates the ABI of common C++ programs, preventing intermixing between C++ compiled by the CBE and C++ code compiled with llc or native compilers.
• The C backend does not support all exception handling constructs.
• The C backend does not support arbitrary precision integers.

llvm-gcc does not currently support Link-Time Optimization on most platforms "out-of-the-box". Please inquire on the LLVMdev mailing list if you are interested.

The only major language feature of GCC not supported by llvm-gcc is the __builtin_apply family of builtins. However, some extensions are only supported on some targets. For example, trampolines are only supported on some targets (these are used when you take the address of a nested function).

If you run into GCC extensions which are not supported, please let us know.

The C++ front-end is considered to be fully tested and works for a number of non-trivial programs, including LLVM itself, Qt, Mozilla, etc.

• Exception handling works well on the X86 and PowerPC targets. Currently only Linux and Darwin targets are supported (both 32 and 64 bit).

• Fortran support generally works, but there are still several unresolved bugs in Bugzilla. Please see the tools/gfortran component for details.

The llvm-gcc 4.2 Ada compiler works fairly well; however, this is not a mature technology, and problems should be expected.

• The Ada front-end currently only builds on X86-32. This is mainly due to lack of trampoline support (pointers to nested functions) on other platforms. However, it also fails to build on X86-64 which does support trampolines.
• The Ada front-end fails to bootstrap. This is due to lack of LLVM support for setjmp/longjmp style exception handling, which is used internally by the compiler. Workaround: configure with --disable-bootstrap.
• The c380004, c393010 and cxg2021 ACATS tests fail (c380004 also fails with gcc-4.2 mainline). If the compiler is built with checks disabled then c393010 causes the compiler to go into an infinite loop, using up all system memory.
• Some GCC specific Ada tests continue to crash the compiler.
• The -E binder option (exception backtraces) does not work and will result in programs crashing if an exception is raised. Workaround: do not use -E.
• Only discrete types are allowed to start or finish at a non-byte offset in a record. Workaround: do not pack records or use representation clauses that result in a field of a non-discrete type starting or finishing in the middle of a byte.
• The lli interpreter considers 'main' as generated by the Ada binder to be invalid. Workaround: hand edit the file to use pointers for argv and envp rather than integers.
• The -fstack-check option is ignored.

A wide variety of additional information is available on the LLVM web page, in particular in the documentation section. The web page also contains versions of the API documentation which is up-to-date with the Subversion version of the source code. You can access versions of these documents specific to this release by going into the "llvm/doc/" directory in the LLVM tree.

If you have any questions or comments about LLVM, please feel free to contact us via the mailing lists.