Written by the LLVM team

This document contains the release notes for the LLVM compiler infrastructure, release 1.1. Here we describe the status of LLVM, including any known problems and bug fixes from the previous release. The most up-to-date version of this document can be found on the LLVM 1.1 web site. If you are not reading this on the LLVM web pages, you should probably go there because this document may be updated after the release.

For more information about LLVM, including information about potentially more current releases, please check out the main 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 CVS, this document applies to the next release, not the current one. To see the release notes for the current or previous releases, see the releases page.

This is the second public release of the LLVM compiler infrastructure. This release is primarily a bugfix release, dramatically improving the C/C++ front-end and improving support for C++ in the LLVM core. This release also includes a few new features, such as a simple profiler, support for Mac OS X, better interoperability with external source bases, a new example language front-end, and improvements in a few optimizations. The performance of several LLVM components has been improved, and several gratuitous type-safety issues in the C front-end have been fixed.

At this time, LLVM is known to correctly compile and run all non-unwinding C & C++ SPEC CPU2000 benchmarks, the Olden benchmarks, and the Ptrdist benchmarks. It has also been used to compile many other programs. LLVM now also works with a broad variety of C++ programs, though it has still received much less testing than the C front-end.

The LLVM native code generators are very stable but do not currently support unwinding (exception throwing or longjmping), which prevent them from working with programs like the 253.perlbmk in SPEC CPU2000. The C backend and the rest of LLVM supports these programs, so you can still use LLVM with them. Support for unwinding will be added in a future release.

1. A new LLVM profiler, similar to gprof, is available.
2. LLVM and the C/C++ front-end now compile on Mac OS X! Mac OS X users can now explore the LLVM optimizer with the C backend and interpreter. Note that LLVM requires GCC 3.3 on Mac OS X.
3. LLVM has been moved into an 'llvm' C++ namespace for easier integration with third-party code. Note that due to lack of namespace support in GDB 5.x, you will probably want to upgrade to GDB 6 or better to debug LLVM code.
4. The build system now copies Makefiles dynamically from the source tree to the object tree as subdirectories are built. This means that:
   1. New directories can be added to the source tree, and the build will automatically pick them up (i.e. no need to edit configure.ac and re-run configure).
   2. You will need to build LLVM from the top of the object tree once to ensure that all of the Makefiles are copied into the object tree subdirectories.
5. A front-end for "Stacker" (a simple Forth-like language) is now included in the main LLVM tree. Additionally, Reid Spencer, the author, contributed a document describing his experiences writing Stacker and the language itself. This document is invaluable for others writing front-ends targetting LLVM.
6. The configure script will now configure all projects placed in the llvm/projects directory.
7. The -tailcallelim pass can now introduce "accumulator" variables to transform functions in many common cases that it could not before.
8. The -licm pass can now sink instructions out the bottom of loops in addition to being able to hoist them out the top.
9. The -basicaa pass (the default alias analysis pass) has been upgraded to be significantly more precise.
10. LLVM 1.1 implements a simple size optimization for LLVM bytecode files. This means that the 1.1 files are smaller than 1.0, but LLVM 1.0 won't read 1.1 bytecode files.
11. The gccld program produces a runner script that includes command-line options to load the necessary shared objects.

1. The interpreter does not support invoke or unwind
2. Interpreter does not support the vaarg instruction
3. llvm-nm cannot read archive files
4. Interpreter does not handle setne constant expression

1. The C++ front-end now compiles functions to use the linkonce linkage type more, giving the optimizer more freedom.
2. The C front-end now generates type-safe code in several cases that it did not before, allowing optimization of code that could not be optimized previously.
3. The LLVM build system has been taught to catch some common configuration problems that caused it to get horribly confused.
4. The LLVM header files are now -Wold-style-cast clean.
5. The LLVM bytecode reader has been sped up a lot (up to 4x in some cases).
6. In C++, methods and functions in anonymous namespaces now get internal linkage.
7. Constant initializers now generate loops instead of potentially huge amounts of straight-line code.
8. Code for running C++ destructors is now properly shared when possible. Before, the C++ front-end generated N^2 amounts of duplicated cleanup code in some cases.
9. The JIT used to generate code for all functions pointed to by globals before the program started execution. Now, it waits until the first time the functions are called to compile them. This dramatically speeds up short runs of large C++ programs, which often have large numbers of functions pointed to by vtables.

Bugs in the LLVM Core:

1. [inliner] Inlining invoke with PHI in unwind target is broken
2. [linker] linkonce globals should link successfully to external globals
3. [constmerge] Constant merging pass merges constants with external linkage
4. [scalarrepl] Scalar Replacement of aggregates is decimating structures it shouldn't be
5. [instcombine] Resolving invoke inserts cast after terminator
6. llvm-as crashes when labels are used in phi nodes
7. [build problem] Callgraph.cpp not pulled in from libipa.a
8. Variables in scope of output setjmp calls should be volatile (Note that this does not affect correctness on many platforms, such as X86).
9. [X86] Emission of global bool initializers broken
10. [gccld] The -r (relinking) option does not work correctly
11. [bcreader] Cannot read shift constant expressions from bytecode file
12. [lowersetjmp] Lowersetjmp pass breaks dominance properties!
13. SymbolTable::getUniqueName is very inefficient
14. bugpoint must not pass -R<directory> to Mach-O linker
15. [buildscripts] Building into objdir with .o in it fails
16. [setjmp/longjmp] Linking C programs which use setjmp/longjmp sometimes fail with references to the C++ runtime library!
17. AsmParser Misses Symbol Redefinition Error
18. gccld -Lfoo -lfoo fails to find ./foo/libfoo.a
19. [bcreader] Incorrect cast causes misread forward constant references
20. [adce] ADCE considers blocks without postdominators to be unreachable
21. [X86] div and rem constant exprs invalidate iterators!
22. [vmcore] Symbol table doesn't rename colliding variables during type resolution
23. Archive reader does not understand 4.4BSD/Mac OS X long filenames
24. [llvm-ar] Command line arguments have funny syntax

Bugs in the C/C++ front-end:

1. C++ frontend can crash when compiling virtual base classes
2. C backend fails on constant cast expr to ptr-to-anonymous struct
3. #ident is not recognized by C frontend
4. C front-end miscompiles the builtin_expect intrinsic!
5. 1.0 precompiled libstdc++ does not include wchar_t support
6. llvmgcc asserts when compiling functions renamed with asm's
7. C frontend crashes on some programs with lots of types.
8. llvm-gcc crashes compiling global union initializer
9. C front-end crash on empty structure
10. CFrontend crashes when compiling C99 compound expressions
11. llvm-gcc infinite loops on "case MAXINT:"
12. [C++] Catch blocks make unparsable labels
13. [C++] Initializing array with constructible objects fail
14. llvm-gcc tries to add bools
15. [c++] C++ Frontend lays out superclasses like anonymous bitfields!
16. C front-end miscompiles unsigned enums whose LLVM types are signed
17. Casting a string constant to void crashes llvm-gcc
18. [llvmg++] Enum types are incorrectly shrunk to smaller than 'int' size
19. [llvmg++] Cannot use pointer to member to initialize global
20. [llvm-gcc] ?: operator as lvalue not implemented
21. [C/C++] Bogus warning about taking the address of 'register' variable
22. crash assigning into an array in a struct which contains a bitfield.
23. Oversized integer bitfields cause crash
24. [llvm-gcc] Bitfields & large array don't mix well
25. [llvm-gcc] Complex division is not supported
26. [llvm-gcc] Illegal union field reference
27. [llvmg++] Front-end attempts to return structure by value
28. [llvmg++] Pointer to member initializers not supported in constructors
29. [llvm-gcc] crash on union initialization
30. [llvm-g++] ?: expressions do not run correct number of destructors!
31. [llvm-gcc] Pointer & constant results in invalid shift
32. [llvmg++] call through array of pointers to member functions causes assertion

LLVM has been extensively tested on Intel and AMD machines running Red Hat Linux and has been tested on Sun UltraSPARC workstations running Solaris 8. Additionally, LLVM works on Mac OS X 10.3 and above, but only with the C backend or interpreter (no native backend for the PowerPC is available yet).

NOTE: You may have to reset the LLVMGCCDIR variable in your Makefile.config to use the target triplet powerpc-apple-darwin7.0.0 instead of whatever configure detected; otherwise you will get errors trying to build crtend. This is fixed in 1.2.

The core LLVM infrastructure uses "autoconf" for portability, so hopefully we work on more platforms than that. However, it is likely that we missed something and that minor porting is required to get LLVM to work on new platforms. We welcome portability patches and error messages.

This section contains all known problems with the LLVM system, listed by component. As new problems are discovered, they will be added to these sections. If you run into a problem, please check the LLVM bug database and submit a bug if there isn't already one.

• In the JIT, dlsym on a symbol compiled by the JIT will not work.
• The JIT does not use mutexes to protect its internal data structures. As such, execution of a threaded program could cause these data structures to be corrupted.
• It is not possible to dlopen an LLVM bytecode file in the JIT.
• Linking in static archive files (.a files) is very slow (there is no symbol table in the archive).
• LLVM cannot handle structures with more than 256 elements.
• The gccld program does not link objects/archives in the order specified on the command line.

• [build] Usage of autoconf is awkward
• [licm] LICM promotes volatile memory locations to registers
• [licm] Memory read after free causes infrequent crash
• [indvars] Induction variable canonicalization always makes 32-bit indvars
• [constantmerge] Merging globals can cause use of invalid pointers!
• [build] Makefiles break if C frontend target string has unexpected value
• [bcreader] Bytecode reader misreads 'long -9223372036854775808'!
• Tail duplication does not update SSA form correctly.
• VMCore mishandles double -0.0
• make tools-only doesn't make lib/Support
• [loopsimplify] Loopsimplify incorrectly updates dominator information
• [pruneeh] -pruneeh pass removes invoke instructions it shouldn't
• [interpreter] va_list values silently corrupted by function calls
• [inliner] Error inlining intrinsic calls into invoke instructions
• Linking weak and strong global variables is dependent on link order
• [X86] Missing cast from ULong -> Double, cast FP -> bool and support for -9223372036854775808
• [JIT] Programs cannot resolve the fstat function
• [indvars] Induction variable analysis violates LLVM invariants
• [execution engines] Unhandled cast constant expression

• C99 Variable sized arrays do not release stack memory when they go out of scope. Thus, the following program may run out of stack space:

      for (i = 0; i != 1000000; ++i) {
        int X[n];
        foo(X);
      }
  

• [llvm-gcc] Error when an implicitly external function is re-declared as static
• Initialization of global union variables can only be done with the largest union member.

• [llvm-gcc] asserts when an extern inline function is redefined
• Functions marked "extern inline" are not compiled into LLVM with linkonce linkage.
• The memory management functions in the libc runtime need weak linkage so that they can be overridden.
• [llvmgcc] floating-point unary minus is incorrect for +0.0

• Inline assembly is not yet supported.
• "long double" is transformed by the front-end into "double". There is no support for floating point data types of any size other than 32 and 64 bits.
• The following Unix system functionality has not been tested and may not work:
  1. sigsetjmp, siglongjmp - These are not turned into the appropriate invoke/unwind instructions. Note that setjmp and longjmp are compiled correctly.
  2. getcontext, setcontext, makecontext - These functions have not been tested.
• Although many GCC extensions are supported, some are not. In particular, the following extensions are known to not be supported:
  1. Local Labels: Labels local to a block.
  2. Labels as Values: Getting pointers to labels and computed gotos.
  3. Nested Functions: As in Algol and Pascal, lexical scoping of functions.
  4. Constructing Calls: Dispatching a call to another function.
  5. Extended Asm: Assembler instructions with C expressions as operands.
  6. Constraints: Constraints for asm operands.
  7. Asm Labels: Specifying the assembler name to use for a C symbol.
  8. Explicit Reg Vars: Defining variables residing in specified registers.
  9. Return Address: Getting the return or frame address of a function.
  10. Vector Extensions: Using vector instructions through built-in functions.
  11. Target Builtins: Built-in functions specific to particular targets.
  12. Thread-Local: Per-thread variables.
  13. Pragmas: Pragmas accepted by GCC.

  The following GCC extensions are partially supported. An ignored attribute means that the LLVM compiler ignores the presence of the attribute, but the code should still work. An unsupported attribute is one which is ignored by the LLVM compiler and will cause a different interpretation of the program.

  1. Variable Length: Arrays whose length is computed at run time.
     Supported, but allocated stack space is not freed until the function returns (noted above).
  2. Function Attributes: Declaring that functions have no side effects or that they can never return.
     Supported: format, format_arg, non_null, constructor, destructor, unused, deprecated, warn_unused_result, weak
     Ignored: noreturn, noinline, always_inline, pure, const, nothrow, malloc, no_instrument_function, cdecl
     Unsupported: used, section, alias, visibility, regparm, stdcall, fastcall, all other target specific attributes
  3. Variable Attributes: Specifying attributes of variables.
     Supported: cleanup, common, nocommon, deprecated, transparent_union, unused, weak
     Unsupported: aligned, mode, packed, section, shared, tls_model, vector_size, dllimport, dllexport, all target specific attributes.
  4. Type Attributes: Specifying attributes of types.
     Supported: transparent_union, unused, deprecated, may_alias
     Unsupported: aligned, packed, all target specific attributes.
  5. Other Builtins: Other built-in functions.
     We support all builtins which have a C language equivalent (e.g., __builtin_cos), __builtin_alloca, __builtin_types_compatible_p, __builtin_choose_expr, __builtin_constant_p, and __builtin_expect (ignored).

  The following extensions are known to be supported:

  1. Statement Exprs: Putting statements and declarations inside expressions.
  2. Typeof: typeof: referring to the type of an expression.
  3. Lvalues: Using ?:, "," and casts in lvalues.
  4. Conditionals: Omitting the middle operand of a ?: expression.
  5. Long Long: Double-word integers.
  6. Complex: Data types for complex numbers.
  7. Hex Floats:Hexadecimal floating-point constants.
  8. Zero Length: Zero-length arrays.
  9. Empty Structures: Structures with no members.
  10. Variadic Macros: Macros with a variable number of arguments.
  11. Escaped Newlines: Slightly looser rules for escaped newlines.
  12. Subscripting: Any array can be subscripted, even if not an lvalue.
  13. Pointer Arith: Arithmetic on void-pointers and function pointers.
  14. Initializers: Non-constant initializers.
  15. Compound Literals: Compound literals give structures, unions, or arrays as values.
  16. Designated Inits: Labeling elements of initializers.
  17. Cast to Union: Casting to union type from any member of the union.
  18. Case Ranges: `case 1 ... 9' and such.
  19. Mixed Declarations: Mixing declarations and code.
  20. Function Prototypes: Prototype declarations and old-style definitions.
  21. C++ Comments: C++ comments are recognized.
  22. Dollar Signs: Dollar sign is allowed in identifiers.
  23. Character Escapes: \e stands for the character <ESC>.
  24. Alignment: Inquiring about the alignment of a type or variable.
  25. Inline: Defining inline functions (as fast as macros).
  26. Alternate Keywords:__const__, __asm__, etc., for header files.
  27. Incomplete Enums: enum foo;, with details to follow.
  28. Function Names: Printable strings which are the name of the current function.
  29. Unnamed Fields: Unnamed struct/union fields within structs/unions.
  30. Attribute Syntax: Formal syntax for attributes.

If you run into GCC extensions which have not been included in any of these lists, please let us know (also including whether or not they work).

For this release, the C++ front-end is considered to be fully functional but has not been tested as thoroughly as the C front-end. It has been tested and works for a number of non-trivial programs, but there may be lurking bugs. Please report any bugs or problems.

• The C++ front-end inherits all problems afflicting the C front-end.

• [llvmg++] Dynamically initialized constants cannot be marked 'constant'

• The C++ front-end is based on a pre-release of the GCC 3.4 C++ parser. This parser is significantly more standards compliant (and picky) than prior GCC versions. For more information, see the C++ section of the GCC 3.4 release notes.
• Destructors for local objects are not always run when a longjmp is performed. In particular, destructors for objects in the longjmping function and in the setjmp receiver function may not be run. Objects in intervening stack frames will be destroyed, however (which is better than most compilers).
• The LLVM C++ front-end follows the Itanium C++ ABI. This document, which is not Itanium specific, specifies a standard for name mangling, class layout, v-table layout, RTTI formats, and other C++ representation issues. Because we use this API, code generated by the LLVM compilers should be binary compatible with machine code generated by other Itanium ABI C++ compilers (such as G++, the Intel and HP compilers, etc). However, the exception handling mechanism used by LLVM is very different from the model used in the Itanium ABI, so exceptions will not interact correctly.

• [X86] X86 backend code generates -0.0 as +0.0
• The X86 code generator does not currently support the unwind instruction, so code that throws a C++ exception or calls the C longjmp function will abort.

• The Sparc code generator does not currently support the unwind instruction, so code that throws a C++ exception or calls the C longjmp function will abort.
• [sparc] Boolean constants are emitted as true and false
• The llc program can crash on legal code.

• The C back-end produces code that violates the ANSI C Type-Based Alias Analysis rules. As such, special options may be necessary to compile the code (for example, GCC requires the -fno-strict-aliasing option). This problem probably cannot be fixed.
• Initializers for global variables cannot include special floating point numbers like Not-A-Number or Infinity.
• Zero arg vararg functions are not supported. This should not affect LLVM produced by the C or C++ frontends.

A wide variety of additional information is available on the LLVM web page, including mailing lists and publications describing algorithms and components implemented in LLVM. The web page also contains versions of the API documentation which is up-to-date with the CVS 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.