1======================
2LLVM 3.4 Release Notes
3======================
4
5.. contents::
6    :local:
7
8Introduction
9============
10
11This document contains the release notes for the LLVM Compiler Infrastructure,
12release 3.4.  Here we describe the status of LLVM, including major improvements
13from the previous release, improvements in various subprojects of LLVM, and
14some of the current users of the code.  All LLVM releases may be downloaded
15from the `LLVM releases web site <http://llvm.org/releases/>`_.
16
17For more information about LLVM, including information about the latest
18release, please check out the `main LLVM web site <http://llvm.org/>`_.  If you
19have questions or comments, the `LLVM Developer's Mailing List
20<http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev>`_ is a good place to send
21them.
22
23Note that if you are reading this file from a Subversion checkout or the main
24LLVM web page, this document applies to the *next* release, not the current
25one.  To see the release notes for a specific release, please see the `releases
26page <http://llvm.org/releases/>`_.
27
28Non-comprehensive list of changes in this release
29=================================================
30
31.. NOTE
32   For small 1-3 sentence descriptions, just add an entry at the end of
33   this list. If your description won't fit comfortably in one bullet
34   point (e.g. maybe you would like to give an example of the
35   functionality, or simply have a lot to talk about), see the `NOTE` below
36   for adding a new subsection.
37
38* This is expected to be the last release of LLVM which compiles using a C++98
39  toolchain. We expect to start using some C++11 features in LLVM and other
40  sub-projects starting after this release. That said, we are committed to
41  supporting a reasonable set of modern C++ toolchains as the host compiler on
42  all of the platforms. This will at least include Visual Studio 2012 on
43  Windows, and Clang 3.1 or GCC 4.7.x on Mac and Linux. The final set of
44  compilers (and the C++11 features they support) is not set in stone, but we
45  wanted users of LLVM to have a heads up that the next release will involve
46  a substantial change in the host toolchain requirements.
47
48* The regression tests now fail if any command in a pipe fails. To disable it in
49  a directory, just add ``config.pipefail = False`` to its ``lit.local.cfg``.
50  See :doc:`Lit <CommandGuide/lit>` for the details.
51
52* Support for exception handling has been removed from the old JIT. Use MCJIT
53  if you need EH support.
54
55* The R600 backend is not marked experimental anymore and is built by default.
56
57* APFloat::isNormal() was renamed to APFloat::isFiniteNonZero() and
58  APFloat::isIEEENormal() was renamed to APFloat::isNormal(). This ensures that
59  APFloat::isNormal() conforms to IEEE-754R-2008.
60
61* The library call simplification pass has been removed.  Its functionality
62  has been integrated into the instruction combiner and function attribute
63  marking passes.
64
65* Support for building using Visual Studio 2008 has been dropped. Use VS 2010
66  or later instead. For more information, see the `Getting Started using Visual
67  Studio <GettingStartedVS.html>`_ page.
68
69* The Loop Vectorizer that was previously enabled for -O3 is now enabled for
70  -Os and -O2.
71
72* The new SLP Vectorizer is now enabled by default.
73
74* llvm-ar now uses the new Object library and produces archives and
75  symbol tables in the gnu format.
76
77* FileCheck now allows specifing -check-prefix multiple times. This
78  helps reduce duplicate check lines when using multiple RUN lines.
79
80* The bitcast instruction no longer allows casting between pointers
81   with different address spaces. To achieve this, use the new
82   addrspacecast instruction.
83
84* Different sized pointers for different address spaces should now
85  generally work. This is primarily useful for GPU targets.
86
87* ... next change ...
88
89.. NOTE
90   If you would like to document a larger change, then you can add a
91   subsection about it right here. You can copy the following boilerplate
92   and un-indent it (the indentation causes it to be inside this comment).
93
94   Special New Feature
95   -------------------
96
97   Makes programs 10x faster by doing Special New Thing.
98
99Mips Target
100-----------
101
102Support for the MIPS SIMD Architecture (MSA) has been added. MSA is supported
103through inline assembly, intrinsics with the prefix '__builtin_msa', and normal
104code generation.
105
106For more information on MSA (including documentation for the instruction set),
107see the `MIPS SIMD page at Imagination Technologies
108<http://imgtec.com/mips/mips-simd.asp>`_
109
110External Open Source Projects Using LLVM 3.4
111============================================
112
113An exciting aspect of LLVM is that it is used as an enabling technology for
114a lot of other language and tools projects. This section lists some of the
115projects that have already been updated to work with LLVM 3.4.
116
117DXR
118---
119
120`DXR <https://wiki.mozilla.org/DXR>`_ is Mozilla's code search and navigation
121tool, aimed at making sense of large projects like Firefox. It supports
122full-text and regex searches as well as structural queries like "Find all the
123callers of this function." Behind the scenes, it uses a custom trigram index,
124the re2 library, and structural data collected by a clang compiler plugin.
125
126LDC - the LLVM-based D compiler
127-------------------------------
128
129`D <http://dlang.org>`_ is a language with C-like syntax and static typing. It
130pragmatically combines efficiency, control, and modeling power, with safety and
131programmer productivity. D supports powerful concepts like Compile-Time Function
132Execution (CTFE) and Template Meta-Programming, provides an innovative approach
133to concurrency and offers many classical paradigms.
134
135`LDC <http://wiki.dlang.org/LDC>`_ uses the frontend from the reference compiler
136combined with LLVM as backend to produce efficient native code. LDC targets
137x86/x86_64 systems like Linux, OS X, FreeBSD and Windows and also Linux/PPC64.
138Ports to other architectures like ARM and AArch64 are underway.
139
140Likely
141------
142
143`Likely <http://www.liblikely.org/>`_ is an open source domain specific
144language for image recognition.  Algorithms are just-in-time compiled using
145LLVM's MCJIT infrastructure to execute on single or multi-threaded CPUs as well
146as OpenCL SPIR or CUDA enabled GPUs. Likely exploits the observation that while
147image processing and statistical learning kernels must be written generically
148to handle any matrix datatype, at runtime they tend to be executed repeatedly
149on the same type.
150
151Portable Computing Language (pocl)
152----------------------------------
153
154In addition to producing an easily portable open source OpenCL
155implementation, another major goal of `pocl <http://portablecl.org/>`_
156is improving performance portability of OpenCL programs with
157compiler optimizations, reducing the need for target-dependent manual
158optimizations. An important part of pocl is a set of LLVM passes used to
159statically parallelize multiple work-items with the kernel compiler, even in
160the presence of work-group barriers. This enables static parallelization of
161the fine-grained static concurrency in the work groups in multiple ways.
162
163TTA-based Co-design Environment (TCE)
164-------------------------------------
165
166`TCE <http://tce.cs.tut.fi/>`_ is a toolset for designing new
167exposed datapath processors based on the Transport triggered architecture (TTA).
168The toolset provides a complete co-design flow from C/C++
169programs down to synthesizable VHDL/Verilog and parallel program binaries.
170Processor customization points include the register files, function units,
171supported operations, and the interconnection network.
172
173TCE uses Clang and LLVM for C/C++/OpenCL C language support, target independent
174optimizations and also for parts of code generation. It generates
175new LLVM-based code generators "on the fly" for the designed processors and
176loads them in to the compiler backend as runtime libraries to avoid
177per-target recompilation of larger parts of the compiler chain.
178
179WebCL Validator
180---------------
181
182`WebCL Validator <https://github.com/KhronosGroup/webcl-validator>`_ implements
183validation for WebCL C language which is a subset of OpenCL ES 1.1. Validator
184checks the correctness of WebCL C, and implements memory protection for it as a
185source-2-source transformation. The transformation converts WebCL to memory
186protected OpenCL. The protected OpenCL cannot access any memory ranges which
187were not allocated for it, and its memory is always initialized to prevent
188information leakage from other programs.
189
190
191Additional Information
192======================
193
194A wide variety of additional information is available on the `LLVM web page
195<http://llvm.org/>`_, in particular in the `documentation
196<http://llvm.org/docs/>`_ section.  The web page also contains versions of the
197API documentation which is up-to-date with the Subversion version of the source
198code.  You can access versions of these documents specific to this release by
199going into the ``llvm/docs/`` directory in the LLVM tree.
200
201If you have any questions or comments about LLVM, please feel free to contact
202us via the `mailing lists <http://llvm.org/docs/#maillist>`_.
203
204