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 fun

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

show more ...

Revert changes r97693, r97700, and r97718.

Our testing framework can't deal with disabled targets yet.

llvm-svn: 97719

Create a TargetMachine whenever we create a CodeGenAction. The codegen of
some builtins will rely on target knowledge.

llvm-svn: 97693

Remove ';' after method definition. Noticed by clang++, which one would think
would have a higher respect for its own code. This is getting old, is this
warning really adding value?

llvm-svn: 91779

Remove remaining VISIBILITY_HIDDEN from anonymous namespaces.

llvm-svn: 90044

Move CompileOptions -> CodeGenOptions, and sink it into the CodeGen library.
This resolves the layering violation where CodeGen depended on Frontend.

llvm-svn: 86998

Remove tabs, and whitespace cleanups.

llvm-svn: 81346

Switch TargetInfo to store an llvm::Triple.
- Primarily to discourage clients form making decisions based on the string.

llvm-svn: 79901

Update for changes in LLVM. Hopefully this is the last one for a while.

llvm-svn: 74657

Hold the LLVMContext by reference instead of by pointer.

llvm-svn: 74642

Update for LLVMContext+Module change.

llvm-svn: 74615

Explictly track tentative definitions within Sema, then hand those
tentative definitions off to the ASTConsumer at the end of the
translation unit.

Eliminate CodeGen's internal tracking of tentativ

Explictly track tentative definitions within Sema, then hand those
tentative definitions off to the ASTConsumer at the end of the
translation unit.

Eliminate CodeGen's internal tracking of tentative definitions, and
instead hook into ASTConsumer::CompleteTentativeDefinition. Also,
tweak the definition-deferal logic for C++, where there are no
tentative definitions.

Fixes <rdar://problem/6808352>, and will make it much easier for
precompiled headers to cope with tentative definitions in the future.

llvm-svn: 69681

show more ...

Push DeclGroup much farther throughout the compiler. Now the various
productions (except the already broken ObjC cases like @class X,Y;) in
the parser that can produce more than one Decl return a D

Push DeclGroup much farther throughout the compiler. Now the various
productions (except the already broken ObjC cases like @class X,Y;) in
the parser that can produce more than one Decl return a DeclGroup instead
of a Decl, etc.

This allows elimination of the Decl::NextDeclarator field, and exposes
various clients that should look at all decls in a group, but which were
only looking at one (such as the dumper, printer, etc). These have been
fixed.

Still TODO:

1) there are some FIXME's in the code about potentially using
DeclGroup for better location info.
2) ParseObjCAtDirectives should return a DeclGroup due to @class etc.
3) I'm not sure what is going on with StmtIterator.cpp, or if it can
be radically simplified now.
4) I put a truly horrible hack in ParseTemplate.cpp.

I plan to bring up #3/4 on the mailing list, but don't plan to tackle
#1/2 in the short term.

llvm-svn: 68002

show more ...

change HandleTranslationUnit to take an ASTContext instead of TranslationUnit

llvm-svn: 67910

most of this is plumbing to get CompileOptions down into
CodeGenModule. Once there, add a new NoCommon option to
it and implement -fno-common.

llvm-svn: 67735

Remove some redundant Decl -> Decl castings.

llvm-svn: 64804

Remove ScopedDecl, collapsing all of its functionality into Decl, so
that every declaration lives inside a DeclContext.

Moved several things that don't have names but were ScopedDecls (and,
therefor

Remove ScopedDecl, collapsing all of its functionality into Decl, so
that every declaration lives inside a DeclContext.

Moved several things that don't have names but were ScopedDecls (and,
therefore, NamedDecls) to inherit from Decl rather than NamedDecl,
including ObjCImplementationDecl and LinkageSpecDecl. Now, we don't
store empty DeclarationNames for these things, nor do we try to insert
them into DeclContext's lookup structure.

The serialization tests are temporarily disabled. We'll re-enable them
once we've sorted out the remaining ownership/serialiazation issues
between DeclContexts and TranslationUnion, DeclGroups, etc.

llvm-svn: 62562

show more ...

Add GetModule accessor to ModuleBuilder

llvm-svn: 57924

Clean up CodeGenModule interface.
- Add CodeGenModule::EmitTopLevelDecl which uses switch on kind
instead of ugly & slow dyn_cast chains.

- Drop some simple forwarding methods into the ObjC run

Clean up CodeGenModule interface.
- Add CodeGenModule::EmitTopLevelDecl which uses switch on kind
instead of ugly & slow dyn_cast chains.

- Drop some simple forwarding methods into the ObjC runtime.

- Privatize appropriate methods.

llvm-svn: 54827

show more ...

Change ObjCRuntime GenerateProtocol[Ref] methods to take
ObjCProtocolDecl directly.

Implement CodeGen support for forward protocol decls (no-ops are so
nice to implement).

Also moved CGObjC

Change ObjCRuntime GenerateProtocol[Ref] methods to take
ObjCProtocolDecl directly.

Implement CodeGen support for forward protocol decls (no-ops are so
nice to implement).

Also moved CGObjCRuntime.h out of CodeGenModule.h

llvm-svn: 54709

show more ...

Add LangOptions::NeXTRuntime.
- Wired to -fnext-runtime and -fgnu-runtime options.
- Defaults to GNU, no autoselection for NeXT.

Emit NeXT OBJC_IMAGE_INFO marker.

llvm-svn: 54651

More #include cleaning
- Kill unnecessary #includes in .cpp files. This is an automatic
sweep so some things removed are actually used, but happen to be
included by a previous header. I tried

More #include cleaning
- Kill unnecessary #includes in .cpp files. This is an automatic
sweep so some things removed are actually used, but happen to be
included by a previous header. I tried to get rid of the obvious
examples and this was the easiest way to trim the #includes in one
fell swoop.
- We now return to regularly scheduled development.

llvm-svn: 54632

show more ...

More #include cleaning
- Drop Expr.h,RecordLayout.h from ASTContext.h (for DeclBase.h and
SourceLocation.h)
- Move ASTContext constructor into implementation

llvm-svn: 54627

Add dummy Mac Objective-C runtime interface.
- Not currently accessible and completely non-functional.

llvm-svn: 54624

80 cols

llvm-svn: 54599