LLVM 1.9 is available now! Download it here: http://llvm.org/
releases/ or view the release here: http://llvm.org/releases/1.9/docs/
ReleaseNotes.html
This is a huge new release with many improvements, new features,
better codegen, faster compiles, and many bugs fixed. One telling
feature is that LLVM now correctly builds itself and passes all its
regression tests when self-hosted: an impressive feat, considering
that it is ~400,000 lines of complex C/C++ code.
While LLVM 1.9 stuck to our planned 3-month release cycle, we plan to
extend this cycle to 6 months for the 2.0 release. This longer
release cycle will allow us to focus on cleaning up LLVM APIs and
removing some old backward compatibility code we have maintained
since 1.0. As such, we plan for LLVM 2.0 to drop support for some
really old .bc and .ll files (LLVM 1.4 and earlier), and LLVM 2.0
will probably drop support for llvm-gcc3. I plan to send out a
status update in 3 months to update everyone on the progress of LLVM
2.0.
Here are some of the most notable aspects of the LLVM 1.9 Release:
Major new Features:
1. Evan contributed a new X86-64 backend. Currently it supports
Darwin/x86-64, and other operating systems in -static mode.
Support for PIC codegen on other target should be straight-forward
to add if someone is interested.
2. Devang contributed liblto, a library which allows a system
linker to efficiently and tightly integrate LLVM Link-Time-
Optimization: http://llvm.org/docs/LinkTimeOptimization.html
3. Bill and Nate integrated support for SPECCPU2006 into the
llvm-test makefile system, which allows it to work with our
nightly tester and automatic testcase reduction tools.
4. Reid wrote a new FAQ for the oft-misunderstood getelementptr
instruction: http://llvm.org/docs/GetElementPtr.html
Mid-Level Optimizer Enhancements:
5. Nick Lewycky contributed a new 'predicate simplifier' pass, which
currently performs dominator tree-based optimizations.
6. Owen extended the loop unroll pass to support unrolling of
multiple basic block loops.
7. The 'globalopt' pass can now perform the scalar replacement of
aggregates transformation on some heap allocations.
8. The globalsmodref-aa alias analysis can now track 'indirect
pointer globals' more accurately.
9. The instruction combiner now performs element propagation
analysis of vector expressions, eliminating computation of vector
elements that are not used.
10. The Loop Closed SSA pass is significantly faster at updating SSA.
11. Several LLVM transformations which had compile time explosions
have been fixed.
12. Devang fixed several cases of nondeterministic behavior and
tracked down several bugs blocking LLVM from building itself.
Target-Independent Code Generator Enhancements:
13. LLVM now includes a late branch folding pass which optimizes code
layout, performs several branch optimizations, and deletes
unreachable code.
14. Evan contributed common code generator support for targets that
support pre/post-increment addressing modes.
15. Jim added support for dynamically-loadable register allocators and
schedulers.
16. Evan modified the DAG instruction selector to be iterative instead
of recursive, fixing "out of stack space" problems with very
large
basic blocks when LLVM is compiled with GCC 4.x.
17. LLVM 1.9 includes several improvements to inline asm support,
including support for new constraints and modifiers.
18. The register coalescer is now more aggressive than before,
allowing it to eliminate more copies.
19. The core SelectionDAG data structures are now much more efficient,
which yields faster compiles.
20. Bill sped up PHI elimination and live variables significantly on
large functions by eliminating N^2 algorithms.
21. The target-independent switch lowering code now produces better
code in several situations for common idioms.
22. Jim added several enhancements to speed up debug info generation
and reduce the size of the emitted debug info.
Target-specific Enhancements:
23. Anton contributed support for Win32 dllimport/dllexport linkage
and stdcall/fastcall calling conventions.
24. Jim contributed support for Darwin/x86 debug information.
25. Reid contributed basic Linux/X86 support for LLVM debug info.
26. Anton contributed basic debug support for Cygwin and MingW when
those environments are configured to use DWARF.
27. Anton contributed several patches to improve compatibility with
MingW.
28. Evan added code generator support for machine-specific constant
pool entries.
29. Rafael enhanced the LLVM ARM backend to support more instructions
and the use of a frame pointer. It is now possible to build
libgcc and a simple cross compiler, but it is not considered
"complete" yet.
30. Nate contributed initial support for a direct Mach-o object file
writer, but it is not yet complete.
Target Description Enhancements:
31. tblgen now allows definition of 'multiclasses' which can be used
to factor instruction patterns more aggressively in .td files.
32. Jim added a new TargetAsmInfo class which captures a variety of
information about the target assembly language format.
33. .td files now support "${:foo}" syntax for encoding subtarget-
specific assembler syntax into instruction descriptions.
Major LLVM IR / API Changes:
34. Reid merged the ConstantUInt and ConstantSInt classes together
into a single ConstantInt class.
35. Reid split the 'div' instruction into fdiv/sdiv/udiv operations
and 'rem' into frem/srem/urem. We hope that LLVM 2.0 will split
all instructions that vary behavior based on the signedness of
their operands. See http://llvm.org/PR950 for more details.
36. ConstantBool::True and False have been renamed to
ConstantBool::getTrue() and ConstantBool::getFalse().
LLVM Build System Changes:
37. The LLVM library dependency graph is now acyclic, allowing
llvm-config to always work. The LLVM build system itself now
internally uses llvm-config.
38. Reid updated autoconf/libtool to newer versions.
39. Reid contributed changes to eliminate C++ exception handling use
from several libraries. Almost all of LLVM now builds with
-fno-exceptions, which shrinks executables significantly.
40. Reid contributed changes to make LLVM -Woverloaded-virtual and
-Wunused clean. These are now enabled by default.
41. Jeff Cohen contributed changes to ensure LLVM builds with MSVC++
7.1 and updated project files to work with it.
42. Jim updated the Xcode project files.
Other Changes:
43. Patrick contributed a new mode to bugpoint: -find-bugs. This asks
bugpoint to permute pass sequences to try to expose bugs due to
pass sequencing.
44. The JIT now supports lazily streaming code from multiple modules
at a time, implicitly linking the code as it goes.
45. Bill and Fernando contributed new sections describing the code
generator: http://llvm.org/docs/CodeGenerator.html
46. Reid merged the functionality of the 'analyze' tool into the
'opt'
tool, cutting LLVM's disk footprint significantly.
47. Domagoj Babic contributed patches to make bugpoint converge faster
when reducing programs.
48. LLVM now uses far fewer static constructors than it used to: it
now lazily initializes global datastructures and provides an
explicit 'teardown' API for releasing the memory associated with
them.
49. Devang changed llvm-gcc to obey several GCC optimization options,
including -funroll-loops, -fno-unit-at-a-time,
-fno-inline-functions, etc.
In addition to the new features and infrastructure we have built, we
also fixed many minor bugs and made many diverse optimization
improvements. LLVM 1.9 is the clearly our best release yet, and
upgrading from a previous release is highly recommended.
This release wouldn't be possible without the LLVM community building
new features, reporting bugs, testing the pre-release bits, and
contributing in many other ways. Tanya (our release manager)
deserves a lot of credit for this being the smoothest and best
release so far!
Here is the previous status report, the LLVM 1.8 Release:
http://lists.cs.uiuc.edu/pipermail/llvm-announce/2006-August/000019.html
If you have any questions or comments, please contact the LLVMdev
mailing list (llvmdev at cs.uiuc.edu)!
-Chris
--
http://nondot.org/sabre/
http://llvm.org/
