Commit Graph

520 Commits

Author SHA1 Message Date
Douglas Gregor 8997690ff1 Don't suppress the emission of available_externally functions marked
with always_inline attribute. Thanks to Howard for the tip.

llvm-svn: 108469
2010-07-15 22:58:18 +00:00
Douglas Gregor a700f68828 Reinstate the optimization suppressing available_externally functions
at -O0. The only change from the previous patch is that we don't try
to generate virtual method thunks for an available_externally
function.

llvm-svn: 108230
2010-07-13 06:02:28 +00:00
Douglas Gregor 553f3a9b30 Speculatively revert r108156; it appears to be breaking self-host.
llvm-svn: 108194
2010-07-12 21:08:32 +00:00
Douglas Gregor dbb2806a7b Do not generate LLVM IR for available_externally function bodies at
-O0, since we won't be using the definitions for anything anyway. For
lib/System/Path.o when built in Debug+Asserts mode, this leads to a 4%
improvement in compile time (and suppresses 440 function bodies).

<rdar://problem/7987644>

llvm-svn: 108156
2010-07-12 17:24:55 +00:00
Argyrios Kyrtzidis 36ea322579 Introduce Decl::hasBody() and FunctionDecl::hasBody() and use them instead of getBody() when we are just checking the existence of a body, to avoid de-serialization of the body from PCH.
Makes de-serialization of the function body even more "lazier".

llvm-svn: 107768
2010-07-07 11:31:19 +00:00
John McCall 09ae03299a Provide a hook for the benefit of clients using clang IR gen as a subroutine:
emit metadata associating allocas and global values with a Decl*.  This feature
is controlled by an option that (intentionally) cannot be enabled on the command
line.

To use this feature, simply set
  CodeGenOptions.EmitDeclMetadata = true;
and then interpret the completely underspecified metadata. :)

llvm-svn: 107739
2010-07-06 23:57:41 +00:00
Chris Lattner 5c740f1523 Reapply:
r107173, "fix PR7519: after thrashing around and remembering how all this stuff"
r107216, "fix PR7523, which was caused by the ABI code calling ConvertType instead"

This includes a fix to make ConvertTypeForMem handle the "recursive" case, and call
it as such when lowering function types which have an indirect result.

llvm-svn: 107310
2010-06-30 19:14:05 +00:00
Chris Lattner 749b8ed5af reduce nesting.
llvm-svn: 107292
2010-06-30 16:58:07 +00:00
Fariborz Jahanian 9b5528d278 Patch to correctly mangle block helper functions
when block literal is declared inside a ctor/dtor.
Fixes radr 8096995.

llvm-svn: 106700
2010-06-24 00:08:06 +00:00
Anders Carlsson ea836bc41c Switch over to the new caching version of getMangledName.
llvm-svn: 106549
2010-06-22 16:16:50 +00:00
Anders Carlsson 2e2f4d2e6d Add a new variant of getMangledName that caches the mangling for decls.
llvm-svn: 106547
2010-06-22 16:05:32 +00:00
Douglas Gregor 5dd34744c5 Instantiations subject to an explicit template instantiation
declaration have default visibility even under
-fvisibility=hidden. Fixes <rdar://problem/8109763>.

llvm-svn: 106440
2010-06-21 18:41:26 +00:00
Eli Friedman c96b2496fc Fix for PR7415: refactor CodeGenModule::MayDeferGeneration and make it less
conservative for static variables in templated classes.

llvm-svn: 106385
2010-06-19 06:24:06 +00:00
Chandler Carruth 8509824cdb Move CodeGenOptions.h *back* into Frontend. This should have been done when the
dependency edge was reversed such that CodeGen depends on Frontend.

llvm-svn: 106065
2010-06-15 23:19:56 +00:00
Douglas Gregor 0832963acd Implement -fvisibility-inlines-hidden. <rdar://problem/7819834>
llvm-svn: 106003
2010-06-15 17:05:35 +00:00
Charles Davis 95a546ee4d Add an option to specify the target C++ ABI to the frontend. Use it to
select either the default Itanium ABI or the new, experimental Microsoft ABI.

llvm-svn: 105804
2010-06-11 01:06:47 +00:00
Anders Carlsson 635186a8c4 Get rid of getMangledCXXCtorName and getMangledCXXDtorName.
llvm-svn: 105673
2010-06-09 02:36:32 +00:00
Anders Carlsson d4ce4e4bc0 Get rid of an unnecessary getMangledName overload.
llvm-svn: 105671
2010-06-09 02:20:01 +00:00
John McCall c2af939ab4 When deciding whether a deferred declaration has already been emitted,
aliases count as definitions regardless of whether their target has been
emitted yet.  Fixes PR 7142.

llvm-svn: 104796
2010-05-27 01:45:30 +00:00
Charles Davis 4e786ddccb IRgen: Add a stub class for generating ABI-specific C++ code.
This class only supports name mangling (which is apparently used during C/ObjC
codegen). For now only the Itanium C++ ABI is supported. Patches to add a
second C++ ABI are forthcoming.

llvm-svn: 104630
2010-05-25 19:52:27 +00:00
John McCall 7cb0220e53 If a function definition has any sort of weak linkage, its static local
variables should have that linkage.  Otherwise, its static local
variables should have internal linkage.  To avoid computing this excessively,
set a function's linkage before we emit code for it.

Previously we were assigning weak linkage to the static variables of
static inline functions in C++, with predictably terrible results.  This
fixes that and also gives better linkage than 'weak' when merging is required.

llvm-svn: 104581
2010-05-25 04:30:21 +00:00
Douglas Gregor 500d9f8221 Disable the available_externally optimization for inline virtual
methods for which the key function is guaranteed to be in another
translation unit. Unfortunately, this guarantee isn't the case when
dealing with shared libraries that fail to export these virtual method
definitions. 

I'm reopening PR6747 so we can consider this again at a later point in
time.

llvm-svn: 103741
2010-05-13 21:36:56 +00:00
Douglas Gregor 88d292ccb8 Rework when and how vtables are emitted, by tracking where vtables are
"used" (e.g., we will refer to the vtable in the generated code) and
when they are defined (i.e., because we've seen the key function
definition). Previously, we were effectively tracking "potential
definitions" rather than uses, so we were a bit too eager about emitting
vtables for classes without key functions. 

The new scheme:
  - For every use of a vtable, Sema calls MarkVTableUsed() to indicate
  the use. For example, this occurs when calling a virtual member
  function of the class, defining a constructor of that class type,
  dynamic_cast'ing from that type to a derived class, casting
  to/through a virtual base class, etc.
  - For every definition of a vtable, Sema calls MarkVTableUsed() to
  indicate the definition. This happens at the end of the translation
  unit for classes whose key function has been defined (so we can
  delay computation of the key function; see PR6564), and will also
  occur with explicit template instantiation definitions.
 - For every vtable defined/used, we mark all of the virtual member
 functions of that vtable as defined/used, unless we know that the key
 function is in another translation unit. This instantiates virtual
 member functions when needed.
  - At the end of the translation unit, Sema tells CodeGen (via the
  ASTConsumer) which vtables must be defined (CodeGen will define
  them) and which may be used (for which CodeGen will define the
  vtables lazily). 

From a language perspective, both the old and the new schemes are
permissible: we're allowed to instantiate virtual member functions
whenever we want per the standard. However, all other C++ compilers
were more lazy than we were, and our eagerness was both a performance
issue (we instantiated too much) and a portability problem (we broke
Boost test cases, which now pass).

Notes:
  (1) There's a ton of churn in the tests, because the order in which
  vtables get emitted to IR has changed. I've tried to isolate some of
  the larger tests from these issues.
  (2) Some diagnostics related to
  implicitly-instantiated/implicitly-defined virtual member functions
  have moved to the point of first use/definition. It's better this
  way.
  (3) I could use a review of the places where we MarkVTableUsed, to
  see if I missed any place where the language effectively requires a
  vtable.

Fixes PR7114 and PR6564.

llvm-svn: 103718
2010-05-13 16:44:06 +00:00
Douglas Gregor d8bb3aff76 Do not give implicitly-defined virtual members functions
available_externally linkage, since they may not have been given a
strong definition in another translation unit. Without this patch, the
following test case fails to link with a GCC-compiled libstdc++:

  #include <sstream>
  int main() { std::basic_stringbuf<char> bs; }

Fixes the last problem with the Boost.IO library.

llvm-svn: 103208
2010-05-06 23:13:35 +00:00
Douglas Gregor d450f06ef4 When we emit a non-constant initializer for a global variable of
reference type, make sure that the initializer we build is the
of the appropriate type for the *reference*, not for the thing that it
refers to. Fixes PR7050.

llvm-svn: 103115
2010-05-05 20:15:55 +00:00
Fariborz Jahanian 0dec1e0d56 IRGen for initialization/destruction of
ivar class objects (NeXt runtime).
(radar 7900343).

llvm-svn: 102533
2010-04-28 21:28:56 +00:00
John McCall d06fb865eb Properly pass the address of a lazily-generated function declaration with
incomplete type.  Fixes PR6911.

llvm-svn: 102473
2010-04-28 00:00:30 +00:00
Fariborz Jahanian d3fa70181d More -fno-constant-cfstrings API work.
llvm-svn: 102219
2010-04-23 22:33:39 +00:00
Fariborz Jahanian e804c28745 More work toward implementing
NeXt's -fno-constant-cfstrings - wip.

llvm-svn: 102189
2010-04-23 17:41:07 +00:00
Fariborz Jahanian 63408e848b Support for -fno-constant-cfstrings option - wip.
llvm-svn: 102112
2010-04-22 20:26:39 +00:00
Rafael Espindola 47cf5b581a Add comment explaning the use of c99 inline in c++.
llvm-svn: 101787
2010-04-19 16:58:20 +00:00
Dan Gohman 145f3f1e83 Fix -Wcast-qual warnings.
llvm-svn: 101786
2010-04-19 16:39:44 +00:00
Rafael Espindola 683fe4fc4c If a method is virtual and the class key function is in another file, emit the method as available_externally.
Fixes PR6747

llvm-svn: 101757
2010-04-19 00:44:22 +00:00
Anders Carlsson 11e5140db9 Vtable -> VTable renames across the board.
llvm-svn: 101666
2010-04-17 20:15:18 +00:00
Chris Lattner d8d760ce66 unbreak tests.
llvm-svn: 101153
2010-04-13 17:57:11 +00:00
Chris Lattner 7a4a29f89a minor cleanups
llvm-svn: 101151
2010-04-13 17:39:09 +00:00
Benjamin Kramer 2e8ca0b86d Fix use after free. Incrementing an use_iterator after its user is erased is unsafe.
llvm-svn: 100926
2010-04-10 11:02:40 +00:00
Gabor Greif d394aec87d rely even less on CallInst internals
llvm-svn: 100919
2010-04-10 03:45:50 +00:00
Gabor Greif d0ef1346f3 do not rely on CallInst interna, use CallSite to access arguments
llvm-svn: 100918
2010-04-10 02:56:12 +00:00
John McCall 6936c863b0 Provide an extremely unsatisfactory diagnostic (instead of crashing) when
mangling an unknown expression kind.  Also conveniently tells the user what
kind of expression they should add to the mangler!

llvm-svn: 100907
2010-04-09 22:26:14 +00:00
Douglas Gregor eadd3cace9 Eliminate excessive PCH deserialization caused by the search for
__cxxabiv1::__fundamental_type_info in every translation
unit. Previously, we would perform name lookup for
__cxxabiv1::__fundamental_type_info at the end of IRGen for a each
translation unit, to determine whether it was present. If so, we we
produce type information for all of the fundamental types. However,
this name lookup causes PCH deserialization of a significant part of the
translation unit, which has a woeful impact on performance.

With this change, we now look at each record type after we've
generated its vtable to see if it is
__cxxabiv1::__fundamental_type_info. If so, we generate type info for
all of the fundamental types. This works because
__cxxabiv1::__fundamental_type_info should always have a key function
(typically the virtual destructor), that will be defined once in the
support library. The fundamental type information will end up there.

Fixes <rdar://problem/7840011>.

llvm-svn: 100772
2010-04-08 15:52:03 +00:00
Mon P Wang cc2ab0cdc9 Reapply patch for adding support for address spaces and added a isVolatile field to memcpy, memmove, and memset.
llvm-svn: 100305
2010-04-04 03:10:52 +00:00
Mon P Wang f7f3bff646 Revert r100193 since it causes failures in objc in clang
llvm-svn: 100200
2010-04-02 18:43:42 +00:00
Mon P Wang 4b82a88764 Reapply patch for adding support for address spaces and added a isVolatile field to memcpy, memmove, and memset.
llvm-svn: 100193
2010-04-02 18:04:30 +00:00
Chris Lattner e675d0f924 adjust to IRBuilder change and use faster DebugLoc apis.
llvm-svn: 100093
2010-04-01 06:31:43 +00:00
Chris Lattner be49c03ff4 use the new optimized debug info metadata accessors. In
addition to the inherent win, this eliminates the pointless
cost of going through the name -> mdkind stringmap that we
were paying.

llvm-svn: 99983
2010-03-31 03:36:01 +00:00
Bob Wilson adb58e32cc Revert Mon Ping's 99930 due to broken llvm-gcc buildbots.
llvm-svn: 99949
2010-03-30 22:28:46 +00:00
Mon P Wang 231e99743a Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset
llvm-svn: 99930
2010-03-30 21:02:45 +00:00
Rafael Espindola adcc1d18e2 When given the magic class __cxxabiv1::__fundamental_type_info, produce
the typeinfo for the fundamental types.

Fixes PR6685.

llvm-svn: 99701
2010-03-27 02:52:14 +00:00
Rafael Espindola fa1708fdea revert 99311. Looks like it broke darwin bootstrap.
llvm-svn: 99317
2010-03-23 19:55:22 +00:00
Rafael Espindola 4d3745ade6 Avoid producing implicit methods when we have a explicit template instantiation
declaration.

llvm-svn: 99311
2010-03-23 18:56:16 +00:00
Anders Carlsson af82f63b47 Always emit associated thunks when emitting the function itself. Remove getVtableAddressPoint, it's not used.
llvm-svn: 99252
2010-03-23 04:31:31 +00:00
Anders Carlsson 8841024a7c Rename MaybeEmitVtable to EmitVTableRelatedData in preparation for making it emit thunks as well.
llvm-svn: 99251
2010-03-23 04:15:00 +00:00
Anders Carlsson a864caff8c Rename CGVtableInfo to CodeGenVTables in preparation of adding another VTableInfo class.
llvm-svn: 99250
2010-03-23 04:11:45 +00:00
Daniel Dunbar fe06df4bf3 C++: Add support for -fno-use-cxa-atexit.
- So much typing, so little gain...

Also, rename the __cxx_global_initialization function just to match llvm-gcc.

llvm-svn: 99039
2010-03-20 04:15:41 +00:00
John McCall 7ec5043c2c Change CodeGenModule to rely on the Module's symbol table instead of
shadowing it in the GlobalDeclMap.  Eliminates the string-uniquing
requirement for mangled names, which should help C++ codegen times a little.
Forces us to do string lookups instead of pointer lookups, which might hurt
codegen times a little across the board.  We'll see how it plays out.

Removing the string-uniquing requirement implicitly fixes any bugs like
PR6635 which arose from the fact that we had multiple uniquing tables for
different kinds of identifiers.

llvm-svn: 99012
2010-03-19 23:29:14 +00:00
Douglas Gregor b14d123774 Give explicit template instantiations weak ODR linkage. Former
iterations of this patch gave explicit template instantiation
link-once ODR linkage, which permitted the back end to eliminate
unused symbols. Weak ODR linkage still requires the symbols to be
generated.

llvm-svn: 98441
2010-03-13 18:23:07 +00:00
Douglas Gregor 17b76185f1 Re-revert the explicit template instantiation linkage patch. I am beginning to look incompetent
llvm-svn: 98425
2010-03-13 03:49:57 +00:00
Douglas Gregor 89cae0f224 Reinstate patch to turn explicit template instantiations into weak symbols
llvm-svn: 98424
2010-03-13 03:14:19 +00:00
Douglas Gregor 539bc40906 Revert the linkage change for explicit template instantiations; something is amiss
llvm-svn: 98332
2010-03-12 05:13:59 +00:00
Douglas Gregor ee3f72299c Give explicit template instantiations weak linkage (but don't defer
them). Fixes PR6578.

llvm-svn: 98328
2010-03-12 04:49:06 +00:00
Rafael Espindola e7113ca907 Delay codegen of vtables when handling implicit instantiations.
This fixes PR6474.

llvm-svn: 98123
2010-03-10 02:19:29 +00:00
Benjamin Kramer f0a0f68c79 Add a message to these asserts.
llvm-svn: 97873
2010-03-06 09:07:19 +00:00
John McCall beec5a080f Implement __builtin_dwarf_sp_column for i386 (Darwin and not), x86-64 (all),
and ARM.  Implement __builtin_init_dwarf_reg_size_table for i386 (both) and
x86-64 (all).

llvm-svn: 97859
2010-03-06 00:35:14 +00:00
Rafael Espindola 2e42fec3a0 Fix PR6473.
Clang's support for weakref is now better than llvm-gcc's :-)

We don't introduce a new symbol and we correctly mark undefined references weak only if there is no
definition or regular undefined references in the same file.

llvm-svn: 97733
2010-03-04 18:17:24 +00:00
John McCall 731be6620c Revert changes r97693, r97700, and r97718.
Our testing framework can't deal with disabled targets yet.

llvm-svn: 97719
2010-03-04 04:29:44 +00:00
Fariborz Jahanian 9eba9dfb0c Refactor local class name mangling and make it
ABI conforming.

llvm-svn: 97702
2010-03-04 01:02:03 +00:00
John McCall 6abe39c61e Create a TargetMachine whenever we create a CodeGenAction. The codegen of
some builtins will rely on target knowledge.

llvm-svn: 97693
2010-03-04 00:23:29 +00:00
Rafael Espindola 70e040d552 During codegen assert that any copy assignment, destructor or constructor that
we need to synthesize has been marked as used by Sema.

Change Sema to avoid these asserts.

llvm-svn: 97589
2010-03-02 21:28:26 +00:00
Rafael Espindola c18086ae17 Add support for the weakref attribute. We still produce "alias weak" as llvm-gcc does, but are more strict on what uses of weakref we accept.
llvm-svn: 96992
2010-02-23 22:00:30 +00:00
John McCall f8ff7b9fd1 Perform two more constructor/destructor code-size optimizations:
1) emit base destructors as aliases to their unique base class destructors
under some careful conditions.  This is enabled for the same targets that can
support complete-to-base aliases, i.e. not darwin.

2) Emit non-variadic complete constructors for classes with no virtual bases
as calls to the base constructor.  This is enabled on all targets and in
theory can trigger in situations that the alias optimization can't (mostly
involving virtual bases, mostly not yet supported).

These are bundled together because I didn't think it worthwhile to split them,
not because they really need to be.

llvm-svn: 96842
2010-02-23 00:48:20 +00:00
John McCall d4324148d7 Re-introduce the ctor/dtor alias optimization, this time hidden behind a
command-line option which defaults off.

llvm-svn: 96649
2010-02-19 01:32:20 +00:00
John McCall 334ce7c1c4 Revert the ctor/dtor alias optimization for now; the buildbots can detect
some failure here that I can't.

llvm-svn: 96612
2010-02-18 21:31:48 +00:00
John McCall 67cea74745 Emit complete constructors and destructors as aliases to base constructors
and destructors when the two entities are semantically identical, i.e. when
the class has no virtual base classes.  We only do this for linkage types
for which aliases are supported, i.e. internal and external, i.e. not linkonce.

llvm-svn: 96451
2010-02-17 03:52:49 +00:00
Ted Kremenek 7f4945aa9c Remove use of 'std::string' from Attr objects, using instead a byte
array allocated using the allocator in ASTContext.  This addresses
these strings getting leaked when using a BumpPtrAllocator (in
ASTContext).

Fixes: <rdar://problem/7636765>
llvm-svn: 95853
2010-02-11 05:28:37 +00:00
John McCall 340aafaac6 Emit global references with constant initializers as constants. Fixes PR5585.
The standard actually says that such references should have internal linkage,
but gcc doesn't do that, so we probably can't get away with it.

llvm-svn: 95577
2010-02-08 21:46:50 +00:00
Anders Carlsson 10d369d1a2 Improved handling of the visibility attribute. Declarations now inherit their parent's visibility.
(This is kind of a risky change, but I did a self-host build and everything appears to work fine!)

llvm-svn: 95511
2010-02-07 01:44:36 +00:00
Douglas Gregor 61f6db54e1 Switch CodeGen's "is this variable declaration a definition?" logic
over to VarDecl::isThisDeclarationADefinition(), which handles
variables declared with linkage specifications better (among other
things). CMake 2.9 (from CVS) now builds with clang++ and is somewhat
functional.

llvm-svn: 95486
2010-02-06 05:15:45 +00:00
Anders Carlsson 6710c5351e Use the correct function info for constructors when applying function attributes. Fixes PR6245.
llvm-svn: 95474
2010-02-06 02:44:09 +00:00
Douglas Gregor 7dc5c17d92 When a function or variable somehow depends on a type or declaration
that is in an anonymous namespace, give that function or variable
internal linkage.

This change models an oddity of the C++ standard, where names declared
in an anonymous namespace have external linkage but, because anonymous
namespace are really "uniquely-named" namespaces, the names cannot be
referenced from other translation units. That means that they have
external linkage for semantic analysis, but the only sensible
implementation for code generation is to give them internal
linkage. We now model this notion via the UniqueExternalLinkage
linkage type. There are several changes here:

  - Extended NamedDecl::getLinkage() to produce UniqueExternalLinkage
    when the declaration is in an anonymous namespace.
  - Added Type::getLinkage() to determine the linkage of a type, which
    is defined as the minimum linkage of the types (when we're dealing
    with a compound type that is not a struct/class/union).
  - Extended NamedDecl::getLinkage() to consider the linkage of the
    template arguments and template parameters of function template
    specializations and class template specializations.
  - Taught code generation to rely on NamedDecl::getLinkage() when
    determining the linkage of variables and functions, also
    considering the linkage of the types of those variables and
    functions (C++ only). Map UniqueExternalLinkage to internal
    linkage, taking out the explicit checks for
    isInAnonymousNamespace().

This fixes much of PR5792, which, as discovered by Anders Carlsson, is
actually the reason behind the pass-manager assertion that causes the
majority of clang-on-clang regression test failures. With this fix,
Clang-built-Clang+LLVM passes 88% of its regression tests (up from
67%). The specific numbers are:

LLVM:
  Expected Passes    : 4006
  Expected Failures  : 32
  Unsupported Tests  : 40
  Unexpected Failures: 736

Clang:
  Expected Passes    : 1903
  Expected Failures  : 14
  Unexpected Failures: 75

Overall:
  Expected Passes    : 5909
  Expected Failures  : 46
  Unsupported Tests  : 40
  Unexpected Failures: 811

Still to do:
  - Improve testing
  - Check whether we should allow the presence of types with
  InternalLinkage (in addition to UniqueExternalLinkage) given
  variables/functions internal linkage in C++, as mentioned in
  PR5792. 
  - Determine how expensive the getLinkage() calls are in practice;
  consider caching the result in NamedDecl.
  - Assess the feasibility of Chris's idea in comment  of PR5792.

llvm-svn: 95216
2010-02-03 09:33:45 +00:00
Chris Lattner 5b5d2db3f6 Don't explicitly force utf strings into the __TEXT,__ustring
by setting the section of the generated global.  This is an
optimization done by the code generator, and the code being
removed didn't handle the case when the string contained an
embedded nul (which the code generator does correctly 
handle).  This is rdar://7589850

llvm-svn: 95003
2010-02-01 20:59:08 +00:00
Sebastian Redl 5ca7984bb4 In C++, an initializer on a variable doesn't necessarily mean it's the definition. With that in mind, rename getDefinition to getAnyInitializer (to distinguish it from getInit) and reimplement it in terms of isThisDeclarationADefinition. Update all code to use this new function.
llvm-svn: 94999
2010-02-01 20:16:42 +00:00
Ken Dyck 160146eba2 Change the return type of ASTContext::getDeclAlignInBytes() to CharUnits and,
now that the "InBytes" part of the name is implied by the return type, rename
it to getDeclAlign().

llvm-svn: 94681
2010-01-27 17:10:57 +00:00
Ken Dyck a0f99ff230 Use CharUnits for alignment of UTF16 string in GetAddrOfConstantCFString().
llvm-svn: 94564
2010-01-26 18:46:23 +00:00
Anders Carlsson ca4a5459d8 Use the right definition when emitting a global variable. Fixes PR5564.
llvm-svn: 94555
2010-01-26 17:43:42 +00:00
Ken Dyck 98ca79435a Introduce CodeGenModule::GetTargetTypeStoreSize() to calculate the store size
of LLVM types in character units.

llvm-svn: 94542
2010-01-26 13:48:07 +00:00
Anders Carlsson d6b21e4842 If a global variable has an initializer with side effects, it can never be deferred (even if it's in an anonymous namespace).
llvm-svn: 94525
2010-01-26 06:15:16 +00:00
David Chisnall 481e3a87fe Created __builtin___NSStringMakeConstantString() builtin, which generates constant Objective-C strings.
llvm-svn: 94274
2010-01-23 02:40:42 +00:00
Anton Korobeynikov 55bcea1e65 Generalize target weirdness handling having proper layering in mind:
1. Add helper class for sema checks for target attributes
 2. Add helper class for codegen of target attributes

As a proof-of-concept - implement msp430's 'interrupt' attribute.

llvm-svn: 93118
2010-01-10 12:58:08 +00:00
Chris Lattner 34c72dd683 revert r92749, which is just dead code.
llvm-svn: 93074
2010-01-09 21:14:47 +00:00
Eli Friedman 5866fe306d Fix for PR5967: Make const-marking for LLVM globals correct for cases requiring
run-time initialization, and emit run-time initializers aggresively to avoid
ordering issues with deferred globals.

llvm-svn: 92976
2010-01-08 00:50:11 +00:00
Douglas Gregor 2a34df3264 Fix linkage for RTTI names by re-using the logic for computing the
linkage of vtables. Before this, we were emitting RTTI names for
template instantiations with strong external linkage rather than with
weak ODR linkage.

llvm-svn: 92857
2010-01-06 22:00:56 +00:00
Douglas Gregor ccecc1bb43 Fix marking of virtual members for nested classes whose first non-pure virtual function has a body inlined in the class
llvm-svn: 92855
2010-01-06 20:27:16 +00:00
Douglas Gregor a318efd1f2 Improve key-function computation for templates. In particular:
- All classes can have a key function; templates don't change that.
  non-template classes when computing the key function.
  - We always mark all of the virtual member functions of class
  template instantiations. 
  - The vtable for an instantiation of a class template has weak
  linkage. 

We could probably use available_externally linkage for vtables of
classes instantiated by explicit instantiation declarations (extern
templates), but GCC doesn't do this and I'm not 100% that the ABI
permits it.

llvm-svn: 92753
2010-01-05 19:06:31 +00:00
Mike Stump 79ed590c85 Add code to skip the emission of available externally functions at -O0. WIP.
llvm-svn: 92749
2010-01-05 18:48:04 +00:00
Chris Lattner 3f8e5dd898 strength reduce this call away.
llvm-svn: 92253
2009-12-29 07:49:13 +00:00
Chris Lattner 5e124bf913 this form of SetDebugLocation is about to go away, add some #includes that
are about to not come in implicitly.

llvm-svn: 92228
2009-12-28 21:44:41 +00:00
Eli Friedman 2e06e8bbcc Some small improvements to dead code elimination; helps a bit on
LLVM-Code-Symbols test.

llvm-svn: 92152
2009-12-25 05:29:40 +00:00
Jeffrey Yasskin 1615d45daa Un-namespace-qualify llvm_unreachable. It's a macro, so the qualification gave
no extra safety anyway.

llvm-svn: 91207
2009-12-12 05:05:38 +00:00
Eli Friedman b095e15af4 Fix for PR5714: make sure globals that will be modified aren't marked const.
llvm-svn: 91156
2009-12-11 21:23:03 +00:00