<p>Written by the <a href="http://llvm.org">LLVM Team</a></p>
</div>
+<!--
<h1 style="color:red">These are in-progress notes for the upcoming LLVM 2.9
release.<br>
You may prefer the
<a href="http://llvm.org/releases/2.8/docs/ReleaseNotes.html">LLVM 2.8
Release Notes</a>.</h1>
+ -->
<!-- *********************************************************************** -->
<h1>
modular, library-based architecture that makes it suitable for creating or
integrating with other development tools. Clang is considered a
production-quality compiler for C, Objective-C, C++ and Objective-C++ on x86
-(32- and 64-bit), and for darwin-arm targets.</p>
+(32- and 64-bit), and for darwin/arm targets.</p>
<p>In the LLVM 2.9 time-frame, the Clang team has made many improvements in C,
C++ and Objective-C support. C++ support is now generally rock solid, has
-been exercised on a broad variety of code, and has several new C++'0x features
+been exercised on a broad variety of code, and has several new <a
+href="http://clang.llvm.org/cxx_status.html#cxx0x">C++'0x features</a>
implemented (such as rvalue references and variadic templates). LLVM 2.9 has
also brought in a large range of bug fixes and minor features (e.g. __label__
support), and is much more compatible with the Linux Kernel.</p>
-<p>If Clang rejects your code that is built with another compiler, please take a
+<p>If Clang rejects your code but another compiler accepts it, please take a
look at the <a href="http://clang.llvm.org/compatibility.html">language
-compatibility</a> guide to make sure the issue isn't intentional or a known
-issue.
+compatibility</a> guide to make sure this is not intentional or a known issue.
</p>
<ul>
The 2.9 release has the following notable changes:
<ul>
<li>The plugin is much more stable when compiling Fortran.</li>
-<li>Inline asm where an asm output is tied to an input of a different size is
-now supported in many more cases.</li>
+<li>Inline assembly where an asm output is tied to an input of a different size
+is now supported in many more cases.</li>
<li>Basic support for the __float128 type was added. It is now possible to
generate LLVM IR from programs using __float128 but code generation does not
work yet.</li>
this and other low-level routines (some are 3x faster than the equivalent
libgcc routines).</p>
-<p>
-All of the code in the compiler-rt project is available under the standard LLVM
-License, a "BSD-style" license.
-
-compiler_rt is now dual licensed under MIT and UIUC license
-
-Several minor changes for better ARM support.
-
-New in LLVM 2.9, UPDATE</p>
+<p>In the LLVM 2.9 timeframe, compiler_rt has had several minor changes for
+ better ARM support, and a fairly major license change. All of the code in the
+ compiler-rt project is now <a href="DeveloperPolicy.html#license">dual
+ licensed</a> under MIT and UIUC license, which allows you to use compiler-rt
+ in applications without the binary copyright reproduction clause. If you
+ prefer the LLVM/UIUC license, you are free to continue using it under that
+ license as well.</p>
</div>
LLVM disassembler and the LLVM JIT.</p>
<p>
-LLDB is in early development and not included as part of the LLVM 2.9 release,
-
-
-
-<!--
-but is mature enough to support basic debugging scenarios on Mac OS X in C,
-Objective-C and C++. We'd really like help extending and expanding LLDB to
-support new platforms, new languages, new architectures, and new features.-->
-</p>
+LLDB is has advanced by leaps and bounds in the 2.9 timeframe. It is
+dramatically more stable and useful, and includes both a new <a
+href="http://lldb.llvm.org/tutorial.html">tutorial</a> and a <a
+href="http://lldb.llvm.org/lldb-gdb.html">side-by-side comparison with
+GDB</a>.</p>
</div>
delivering great performance.</p>
<p>
-As of the LLVM 2.9 release, UPDATE!
-
-libc++ is now dual licensed under MIT and UIUC license
+In the LLVM 2.9 timeframe, libc++ has had numerous bugs fixed, and is now being
+co-developed with Clang's C++'0x mode.</p>
-<!--libc++ is virtually feature complete, but would
-benefit from more testing and better integration with Clang++. It is also
-looking forward to the C++ committee finalizing the C++'0x standard.-->
+<p>
+Like compiler_rt, libc++ is now <a href="DeveloperPolicy.html#license">dual
+ licensed</a> under the MIT and UIUC license, allowing it to be used more
+ permissively.
</p>
</div>
</div>
+<!--=========================================================================-->
+<h2>Crack Programming Language</h2>
+<div class="doc_text">
+<p>
+<a href="http://code.google.com/p/crack-language/">Crack</a> aims to provide the
+ease of development of a scripting language with the performance of a compiled
+language. The language derives concepts from C++, Java and Python, incorporating
+object-oriented programming, operator overloading and strong typing.</p>
+</div>
+
+
+<!--=========================================================================-->
+<h2>TTA-based Codesign Environment (TCE)</h2>
+
+<div class="doc_text">
+<p>TCE is a toolset for designing application-specific processors (ASP) based on
+the Transport triggered architecture (TTA). The toolset provides a complete
+co-design flow from C/C++ programs down to synthesizable VHDL and parallel
+program binaries. Processor customization points include the register files,
+function units, supported operations, and the interconnection network.</p>
+
+<p>TCE uses Clang and LLVM for C/C++ language support, target independent
+optimizations and also for parts of code generation. It generates new LLVM-based
+code generators "on the fly" for the designed TTA processors and loads them in
+to the compiler backend as runtime libraries to avoid per-target recompilation
+of larger parts of the compiler chain.</p>
+</div>
-<!-- *********************************************************************** -->
-<h1>
- <a name="whatsnew">What's New in LLVM 2.9?</a>
-</h1>
-<!-- *********************************************************************** -->
+
+<!--=========================================================================-->
+<h2>PinaVM</h2>
+
<div class="doc_text">
+<p><a href="http://gitorious.org/pinavm/pages/Home">PinaVM</a> is an open
+source, <a href="http://www.systemc.org/">SystemC</a> front-end. Unlike many
+other front-ends, PinaVM actually executes the elaboration of the
+program analyzed using LLVM's JIT infrastructure. It later enriches the
+bitcode with SystemC-specific information.</p>
+</div>
-<p>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.
-</p>
-
+<!--=========================================================================-->
+<h2>Pure</h2>
+
+<div class="doc_text">
+<p><a href="http://pure-lang.googlecode.com/">Pure</a> 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. The interpreter uses LLVM as a backend to JIT-compile Pure
+ programs to fast native code. 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 interface to C and other
+ programming languages (including the ability to load LLVM bitcode
+ modules, and inline C, C++, Fortran and Faust code in Pure programs if
+ the corresponding LLVM-enabled compilers are installed).</p>
+
+<p>Pure version 0.47 has been tested and is known to work with LLVM 2.9
+ (and continues to work with older LLVM releases >= 2.5).</p>
</div>
<!--=========================================================================-->
-<h2>
-<a name="majorfeatures">Major New Features</a>
-</h2>
+<h2 id="icedtea">IcedTea Java Virtual Machine Implementation</h2>
<div class="doc_text">
+<p>
+<a href="http://icedtea.classpath.org/wiki/Main_Page">IcedTea</a> provides a
+harness to build OpenJDK using only free software build tools and to provide
+replacements for the not-yet free parts of OpenJDK. One of the extensions that
+IcedTea provides is a new JIT compiler named <a
+href="http://icedtea.classpath.org/wiki/ZeroSharkFaq">Shark</a> which uses LLVM
+to provide native code generation without introducing processor-dependent
+code.
+</p>
-<p>LLVM 2.9 includes several major new capabilities:</p>
+<p> OpenJDK 7 b112, IcedTea6 1.9 and IcedTea7 1.13 and later have been tested
+and are known to work with LLVM 2.9 (and continue to work with older LLVM
+releases >= 2.6 as well).</p>
+</div>
-<ul>
-<li><pre>
- last release for llvm-gcc
-TBAA: On by default in clang. Disable it with -fno-strict-aliasing.
- Could be more aggressive for structs.
+<!--=========================================================================-->
+<h2>Glasgow Haskell Compiler (GHC)</h2>
-Triple::normalize is new, llvm triples are always stored in normalized form internally.
-
-Triple x86_64--mingw64 is obsoleted. Use x86_64--mingw32 instead.
+<div class="doc_text">
+<p>GHC is an open source, state-of-the-art programming suite for Haskell,
+a standard lazy functional programming language. It includes an
+optimizing static compiler generating good code for a variety of
+platforms, together with an interactive system for convenient, quick
+development.</p>
+
+<p>In addition to the existing C and native code generators, GHC 7.0 now
+supports an LLVM code generator. GHC supports LLVM 2.7 and later.</p>
+</div>
-MC Assembler: X86 now generates much better diagnostics for common errors,
- is much faster at matching instructions, is much more bug-compatible with
- the GAS assembler, and is now generally useful for a broad range of X86
- assembly.
+<!--=========================================================================-->
+<h2>Polly - Polyhedral optimizations for LLVM</h2>
-New Nvidia PTX backend, not generally useful in 2.9 though.
-
-Much better debug info generated, particularly in optimized code situations.
-
-ARM Fast ISel
+<div class="doc_text">
+<p>Polly is a project that aims to provide advanced memory access optimizations
+to better take advantage of SIMD units, cache hierarchies, multiple cores or
+even vector accelerators for LLVM. Built around an abstract mathematical
+description based on Z-polyhedra, it provides the infrastructure to develop
+advanced optimizations in LLVM and to connect complex external optimizers. In
+its first year of existence Polly already provides an exact value-based
+dependency analysis as well as basic SIMD and OpenMP code generation support.
+Furthermore, Polly can use PoCC(Pluto) an advanced optimizer for data-locality
+and parallelism.</p>
+</div>
-ELF MC support: on by default in clang. There are still known missing features
- for human written assembly.
+<!--=========================================================================-->
+<h2>Rubinius</h2>
-X86: Reimplemented all of MMX to introduce a new LLVM IR x86_mmx type. Now
- random types like <2 x i32> are not iseld to mmx without emms. The
- -disable-mmx flag is gone now.
+<div class="doc_text">
+ <p><a href="http://github.com/evanphx/rubinius">Rubinius</a> is an environment
+ for running Ruby code which strives to write as much of the implementation in
+ Ruby as possible. Combined with a bytecode interpreting VM, it uses LLVM to
+ optimize and compile ruby code down to machine code. Techniques such as type
+ feedback, method inlining, and deoptimization are all used to remove dynamism
+ from ruby execution and increase performance.</p>
+</div>
-Some basic <a href="CodeGenerator.html#mc">internals documentation</a> for MC.
-
-MC Assembler support for .file and .loc.
-
-
-inline asm multiple alternative constraint support.
-
-LoopIdiom: memset/memcpy formation and memset_pattern on darwin. Build with
- -ffreestanding or -fno-builtin if your memcpy is being compiled into infinite
- recursion.
-
-TargetLibraryInfo
-
-X86 support for FS/GS relative loads and stores using address space 256/257 are
- reliable now.
-
-ARM: New code placement pass.
-
-unnamed_addr + PR8927
-
-PointerTracking has been removed from mainline, moved to ClamAV.
-
-EarlyCSE pass.
-LoopInstSimplify pass.
-- DIBuilder provides simpler interface for front ends like Clang to encode debug info in LLVM IR.
- - This interface hides implementation details (e.g. DIDerivedType, existence of compile unit etc..) that any front end should not know about.
- For example, DIFactory DebugFactory;
- Ty = DebugFactory.CreateDerivedType(DW_TAG_volatile_type,
- findRegion(TYPE_CONTEXT(type)),
- StringRef(),
- getOrCreateFile(main_input_filename),
- 0 /*line no*/,
- NodeSizeInBits(type),
- NodeAlignInBits(type),
- 0 /*offset */,
- 0 /* flags */,
- MainTy);
- can be replaced by
- DbgTy = DBuilder.createQualifiedType(DW_TAG_volatile_type, MainTy);
-DIFactory is gone now.
-
-PPC: Switched to MCInstPrinter, and MCCodeEmitter. Ready to implement support
- for directly writing out mach-o object files, but noone seems interested.
+<!-- *********************************************************************** -->
+<h1>
+ <a name="whatsnew">What's New in LLVM 2.9?</a>
+</h1>
+<!-- *********************************************************************** -->
-ARM: Improved code generation for Cortex-A8 and Cortex-A9 CPUs.
-
-Scheduler now models operand latency and pipeline forwarding.
-
-Can optimize printf to iprintf when no floating point is used, for embedded
- targets with smaller iprintf implementation.
-
-error_code + libsystem + PathV2 changes
- The system_error header from C++0x was added.
- * Use if (error_code ec = function()) to check for error conditions
- from functions which return it.
- * error_code::message returns a human readable description of the error.
-
- PathV1 has been deprecated in favor of PathV2 (sorry I didn't finish
- this before the release).
- * No Path class, use a r-value convertible to a twine instead.
- * Assumes all paths are UTF-8.
-
-new macho-dump tool
+<div class="doc_text">
-Major regalloc rewrite, not on by default for 2.9 and not advised to use it.
- * New basic register allocator that can be used as a safe fallback when
- debugging. Enable with -regalloc=basic.
- * New infrastructure for live range splitting. SplitKit can break a live
- interval into smaller pieces while preserving SSA form, and SpillPlacement
- can help find the best split points. This is a work in progress so the API
- is changing quickly.
- * The inline spiller has learned to clean up after live range splitting. It
- can hoist spills out of loops, and it can eliminate redundant spills.
- Rematerialization works with live range splitting.
- * New greedy register allocator using live range splitting. This will be the
- default register allocator in the next LLVM release, but it is not turned on
- by default in 2.9.
+<p>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.
+</p>
-ARM: __builtin_prefetch turns into prefetch instructions.
-
-MC assembler support for 3dNow! and 3DNowA instructions.
-
-tblgen support for assembler aliases: <a
- href="CodeGenerator.html#na_instparsing">MnemonicAlias and InstAlias</a>
-
-LoopIndexSplit pass was removed, unmaintained.
-LiveValues, SimplifyHalfPowrLibCalls, and GEPSplitter were removed.
-
-include/llvm/System merged into include/llvm/Support.
-
-Win32 PE-COFF support in the MC assembler has made a lot of progress in the 2.9
- timeframe, but is still not generally useful. Please see
- "http://llvm.org/bugs/showdependencytree.cgi?id=9100&hide_resolved=1" for open bugs?
+</div>
-New <a href="WritingAnLLVMPass.html#RegionPass">RegionPass</a> infrastructure
- for region-based optimizations.
-
-MicroBlaze: major updates for aggressive delay slot filler, MC-based assembly
- printing, assembly instruction parsing, ELF .o file emission, and MC
- instruction disassembler.
-
-Countless ARM microoptimizations.
+<!--=========================================================================-->
+<h2>
+<a name="majorfeatures">Major New Features</a>
+</h2>
-Speedups to various mid-level passes:
- GVN is much faster on functions with deep dominator trees / lots of BBs.
- DomTree and DominatorFrontier are much faster to compute, and preserved by
- more passes (so they are computed less often)
- SRoA is also much faster and doesn't use DominanceFrontier.
-
-
-new 'hotpatch' attribute: LangRef.html#fnattrs
+<div class="doc_text">
-APInt API changes, see PR5207.
+<p>LLVM 2.9 includes several major new capabilities:</p>
-DSE is more aggressive with stores of different types: e.g. a large store
- following a small one to the same address.
-
-New naming rules in coding standards: CodingStandards.html#ll_naming
-
-LiveDebugVariables is a new pass that keeps track of debugging information for
- user variables that are kept in registers in optimized builds.
+<ul>
-We now optimize various idioms for overflow detection into check of the flag
- register on various CPUs, e.g.:
- unsigned long t = a+b;
- if (t < a) ...
- into:
- addq %rdi, %rbx
- jno LBB0_2
+<li>Type Based Alias Analysis (TBAA) is now implemented and turned on by default
+ in Clang. This allows substantially better load/store optimization in some
+ cases. TBAA can be disabled by passing -fno-strict-aliasing.
+</li>
-X86: Much better codegen for several cases using adc/sbb instead of cmovs for
- conditional increment and other idioms.
-
-MVT::Flag renamed to MVT::Glue
-
-Removed the PartialSpecialization pass, it was unmaintained and buggy.
-
-SPARC: Many improvements, including using the Y registers for multiplications
- and addition of a simple delay slot filler.
-
-
-udiv, ashr, lshr, shl now have exact and nuw/nsw bits: PR8862 / LangRef.html
-
-lib/Object and llvm-objdump
+<li>This release has seen a continued focus on quality of debug information.
+ LLVM now generates much higher fidelity debug information, particularly when
+ debugging optimized code.</li>
-
- Target Independent Code Gen:
- The pre-register-allocation (preRA) instruction scheduler models register pressure
- much more accurately in some cases. This allows the adoption of more
- aggressive scheduling heuristics.
-
- The X86 backend has adopted a new preRA scheduling
- mode, "list-ilp", to shorten the height of instruction schedules
- without inducing register spills.
-
- The ARM backend preRA scheduler now models machine resources at cycle
- granularity. This allows the scheduler to both accurately model
- instruction latency and avoid overcommitting functional units.
+<li>Inline assembly now supports multiple alternative constraints.</li>
+<li>A new backend for the NVIDIA PTX virtual ISA (used to target its GPUs) is
+ under rapid development. It is not generally useful in 2.9, but is making
+ rapid progress.</li>
-</pre></li>
</ul>
</div>
expose new optimization opportunities:</p>
<ul>
+<li>The <a href="LangRef.html#bitwiseops">udiv, ashr, lshr, and shl</a>
+ instructions now have support exact and nuw/nsw bits to indicate that they
+ don't overflow or shift out bits. This is useful for optimization of <a
+ href="http://llvm.org/PR8862">pointer differences</a> and other cases.</li>
+
+<li>LLVM IR now supports the <a href="LangRef.html#globalvars">unnamed_addr</a>
+ attribute to indicate that constant global variables with identical
+ initializers can be merged. This fixed <a href="http://llvm.org/PR8927">an
+ issue</a> where LLVM would incorrectly merge two globals which were supposed
+ to have distinct addresses.</li>
+
+<li>The new <a href="LangRef.html#fnattrs">hotpatch attribute</a> has been added
+ to allow runtime patching of functions.</li>
</ul>
</div>
release includes a few major enhancements and additions to the optimizers:</p>
<ul>
- <li>TBAA.</li>
- <li>LTO has been improved to use MC for parsing inline asm and now
- can build large programs like Firefox 4 on both OS X and Linux.</li>
-</ul>
+<li>Link Time Optimization (LTO) has been improved to use MC for parsing inline
+ assembly and now can build large programs like Firefox 4 on both Mac OS X and
+ Linux.</li>
+
+<li>The new -loop-idiom pass recognizes memset/memcpy loops (and memset_pattern
+ on darwin), turning them into library calls, which are typically better
+ optimized than inline code. If you are building a libc and notice that your
+ memcpy and memset functions are compiled into infinite recursion, please build
+ with -ffreestanding or -fno-builtin to disable this pass.</li>
+
+<li>A new -early-cse pass does a fast pass over functions to fold constants,
+ simplify expressions, perform simple dead store elimination, and perform
+ common subexpression elimination. It does a good job at catching some of the
+ trivial redundancies that exist in unoptimized code, making later passes more
+ effective.<,/li>
+
+<li>A new -loop-instsimplify pass is used to clean up loop bodies in the loop
+ optimizer.</li>
+
+<li>The new TargetLibraryInfo interface allows mid-level optimizations to know
+ whether the current target's runtime library has certain functions. For
+ example, the optimizer can now transform integer-only printf calls to call
+ iprintf, allowing reduced code size for embedded C libraries (e.g. newlib).
+</li>
+
+<li>LLVM has a new <a href="WritingAnLLVMPass.html#RegionPass">RegionPass</a>
+ infrastructure for region-based optimizations.</li>
+
+<li>Several optimizer passes have been substantially sped up:
+ GVN is much faster on functions with deep dominator trees and lots of basic
+ blocks. The dominator tree and dominance frontier passes are much faster to
+ compute, and preserved by more passes (so they are computed less often). The
+ -scalar-repl pass is also much faster and doesn't use DominanceFrontier.
+</li>
-<!--
-<p>In addition to these features that are done in 2.8, there is preliminary
- support in the release for Type Based Alias Analysis
- Preliminary work on TBAA but not usable in 2.8.
- New CorrelatedValuePropagation pass, not on by default in 2.8 yet.
--->
+<li>The Dead Store Elimination pass is more aggressive optimizing stores of
+ different types: e.g. a large store following a small one to the same address.
+ The MemCpyOptimizer pass handles several new forms of memcpy elimination.</li>
+
+<li>LLVM now optimizes various idioms for overflow detection into check of the
+ flag register on various CPUs. For example, we now compile:
+
+ <pre>
+ unsigned long t = a+b;
+ if (t < a) ...
+ </pre>
+ into:
+ <pre>
+ addq %rdi, %rbx
+ jno LBB0_2
+ </pre>
+</li>
+
+</ul>
</div>
in.</p>
<ul>
- <li>MC is now used by default for ELF systems on x86 and
- x86-64.</li>
- <li>MC supports and CodeGen uses the <tt>.loc</tt> directives for
- producing line number debug info. This produces more compact line
- tables.</li>
- <li>MC supports the <tt>.cfi_*</tt> directives for producing DWARF
+<li>ELF MC support has matured enough for the integrated assembler to be turned
+ on by default in Clang on X86-32 and X86-64 ELF systems.</li>
+
+<li>MC supports and CodeGen uses the <tt>.file</tt> and <tt>.loc</tt> directives
+ for producing line number debug info. This produces more compact line
+ tables and easier to read .s files.</li>
+
+<li>MC supports the <tt>.cfi_*</tt> directives for producing DWARF
frame information, but it is still not used by CodeGen by default.</li>
- <li>COFF support?</li>
+
+
+<li>The MC assembler now generates much better diagnostics for common errors,
+ is much faster at matching instructions, is much more bug-compatible with
+ the GAS assembler, and is now generally useful for a broad range of X86
+ assembly.</li>
+
+<li>We now have some basic <a href="CodeGenerator.html#mc">internals
+ documentation</a> for MC.</li>
+
+<li>.td files can now specify assembler aliases directly with the <a
+ href="CodeGenerator.html#na_instparsing">MnemonicAlias and InstAlias</a>
+ tblgen classes.</li>
+
+<li>LLVM now has an experimental format-independent object file manipulation
+ library (lib/Object). It supports both PE/COFF and ELF. The llvm-nm tool has
+ been extended to work with native object files, and the new llvm-objdump tool
+ supports disassembly of object files (but no relocations are displayed yet).
+</li>
+
+<li>Win32 PE-COFF support in the MC assembler has made a lot of progress in the
+ 2.9 timeframe, but is still not generally useful.</li>
+
</ul>
<p>For more information, please see the <a
it run faster:</p>
<ul>
-<!-- SplitKit -->
-FastISel for ARM.
+<li>The pre-register-allocation (preRA) instruction scheduler models register
+ pressure much more accurately in some cases. This allows the adoption of more
+ aggressive scheduling heuristics without causing spills to be generated.
+</li>
+
+<li>LiveDebugVariables is a new pass that keeps track of debugging information
+ for user variables that are promoted to registers in optimized builds.</li>
+
+<li>The scheduler now models operand latency and pipeline forwarding.</li>
+
+<li>A major register allocator infrastructure rewrite is underway. It is not on
+ by default for 2.9 and you are not advised to use it, but it has made
+ substantial progress in the 2.9 timeframe:
+ <ul>
+ <li>A new -regalloc=basic "basic" register allocator can be used as a simple
+ fallback when debugging. It uses the new infrastructure.</li>
+ <li>New infrastructure is in place for live range splitting. "SplitKit" can
+ break a live interval into smaller pieces while preserving SSA form, and
+ SpillPlacement can help find the best split points. This is a work in
+ progress so the API is changing quickly.</li>
+ <li>The inline spiller has learned to clean up after live range splitting. It
+ can hoist spills out of loops, and it can eliminate redundant spills.</li>
+ <li>Rematerialization works with live range splitting.</li>
+ <li>The new "greedy" register allocator using live range splitting. This will
+ be the default register allocator in the next LLVM release, but it is not
+ turned on by default in 2.9.</li>
+ </ul>
+</li>
</ul>
</div>
</p>
<ul>
- <li>Several bugs have been fixed for Windows x64 code generator.</li>
+<li>LLVM 2.9 includes a complete reimplementation of the MMX instruction set.
+ The reimplementation uses a new LLVM IR <a
+ href="LangRef.html#t_x86mmx">x86_mmx</a> type to ensure that MMX operations
+ are <em>only</em> generated from source that uses MMX builtin operations. With
+ this, random types like <2 x i32> are not turned into MMX operations
+ (which can be catastrophic without proper "emms" insertion). Because the X86
+ code generator always generates reliable code, the -disable-mmx flag is now
+ removed.
+</li>
+
+<li>X86 support for FS/GS relative loads and stores using <a
+ href="CodeGenerator.html#x86_memory">address space 256/257</a> works reliably
+ now.</li>
+
+<li>LLVM 2.9 generates much better code in several cases by using adc/sbb to
+ avoid generation of conditional move instructions for conditional increment
+ and other idioms.</li>
+
+<li>The X86 backend has adopted a new preRA scheduling mode, "list-ilp", to
+ shorten the height of instruction schedules without inducing register spills.
+</li>
+
+<li>The MC assembler supports 3dNow! and 3DNowA instructions.</li>
+
+<li>Several bugs have been fixed for Windows x64 code generator.</li>
</ul>
</div>
</p>
<ul>
+<li>The ARM backend now has a fast instruction selector, which dramatically
+ improves -O0 compile times.</li>
+<li>The ARM backend has new tuning for Cortex-A8 and Cortex-A9 CPUs.</li>
+<li>The __builtin_prefetch builtin (and llvm.prefetch intrinsic) is compiled
+ into prefetch instructions instead of being discarded.</li>
+
+<li> The ARM backend preRA scheduler now models machine resources at cycle
+ granularity. This allows the scheduler to both accurately model
+ instruction latency and avoid overcommitting functional units.</li>
+
+<li>Countless ARM microoptimizations have landed in LLVM 2.9.</li>
</ul>
</div>
+
+<!--=========================================================================-->
+<h2>
+<a name="OtherTS">Other Target Specific Improvements</a>
+</h2>
+<div class="doc_text">
+<ul>
+<li>MicroBlaze: major updates for aggressive delay slot filler, MC-based
+ assembly printing, assembly instruction parsing, ELF .o file emission, and MC
+ instruction disassembler have landed.</li>
+
+<li>SPARC: Many improvements, including using the Y registers for
+ multiplications and addition of a simple delay slot filler.</li>
+
+<li>PowerPC: The backend has been largely MC'ized and is ready to support
+ directly writing out mach-o object files. No one seems interested in finishing
+ this final step though.</li>
+
+</ul>
+</div>
<!--=========================================================================-->
<h2>
from the previous release.</p>
<ul>
-</ul>
+<li><b>This is the last release to support the llvm-gcc frontend.</b></li>
+<li>LLVM has a new <a href="CodingStandards.html#ll_naming">naming
+ convention standard</a>, though the codebase hasn't fully adopted it yet.</li>
+
+<li>The new DIBuilder class provides a simpler interface for front ends to
+ encode debug info in LLVM IR, and has replaced DIFactory.</li>
+<li>LLVM IR and other tools always work on normalized target triples (which have
+ been run through <tt>Triple::normalize</tt>).</li>
-<p>In addition, many APIs have changed in this release. Some of the major LLVM
-API changes are:</p>
-<ul>
+<li>The target triple x86_64--mingw64 is obsoleted. Use x86_64--mingw32
+ instead.</li>
+
+<li>The PointerTracking pass has been removed from mainline, and moved to The
+ ClamAV project (its only client).</li>
+
+<li>The LoopIndexSplit, LiveValues, SimplifyHalfPowrLibCalls, GEPSplitter, and
+ PartialSpecialization passes were removed. They were unmaintained,
+ buggy, or deemed to be a bad idea.</li>
</ul>
</div>
<!--=========================================================================-->
<h2>
-<a name="devtree_changes">Development Infrastructure Changes</a>
+<a name="api_changes">Internal API Changes</a>
</h2>
<div class="doc_text">
-<p>This section lists changes to the LLVM development infrastructure. This
-mostly impacts users who actively work on LLVM or follow development on
-mainline, but may also impact users who leverage the LLVM build infrastructure
-or are interested in LLVM qualification.</p>
+<p>In addition, many APIs have changed in this release. Some of the major
+ LLVM API changes are:</p>
<ul>
+<li>include/llvm/System merged into include/llvm/Support.</li>
+<li>The <a href="http://llvm.org/PR5207">llvm::APInt API</a> was significantly
+ cleaned up.</li>
+
+<li>In the code generator, MVT::Flag was renamed to MVT::Glue to more accurately
+ describe its behavior.</li>
+
+<li>The system_error header from C++0x was added, and is now pervasively used to
+ capture and handle i/o and other errors in LLVM.</li>
+
+<li>The old sys::Path API has been deprecated in favor of the new PathV2 API,
+ which is more efficient and flexible.</li>
</ul>
</div>
<li>The Alpha, Blackfin, CellSPU, MicroBlaze, MSP430, MIPS, PTX, SystemZ
and XCore backends are experimental.</li>
<li><tt>llc</tt> "<tt>-filetype=obj</tt>" is experimental on all targets
- other than darwin-i386 and darwin-x86_64. FIXME: Not true on ELF anymore?</li>
+ other than darwin and ELF X86 systems.</li>
</ul>
<div class="doc_text">
+<p><b>LLVM 2.9 will be the last release of llvm-gcc.</b></p>
+
<p>llvm-gcc is generally very stable for the C family of languages. The only
major language feature of GCC not supported by llvm-gcc is the
<tt>__builtin_apply</tt> family of builtins. However, some extensions