Support constructor and destructor attributes in CodeGen

- There is an miscompilation issue remaining due to a poor
interaction between the delayed emission of static functions and
the emissi

Support constructor and destructor attributes in CodeGen

- There is an miscompilation issue remaining due to a poor
interaction between the delayed emission of static functions and
the emission of constructors, but that already existed prior to
this change.

llvm-svn: 54258

show more ...

Change CodeGenModule GlobalDeclMap to directly reference globals
instead of mapping the decl to a bitcast of the global to the correct
type.

- GetAddrOf{Function,GlobalVar} introduce the bitcast o

Change CodeGenModule GlobalDeclMap to directly reference globals
instead of mapping the decl to a bitcast of the global to the correct
type.

- GetAddrOf{Function,GlobalVar} introduce the bitcast on every use now.

- This solves a problem where a dangling pointer could be introduced
by the RAUW done when replacing a forward or tentative
definition. See testcase for more details.

- Fixes <rdar://problem/6108358>

llvm-svn: 54211

show more ...

Rework codegen emission of globals
- No (intended) functionality change.

- Primary purpose is to clearly separate (lazy) construction of
globals that are a forward declaration or tentative defi

Rework codegen emission of globals
- No (intended) functionality change.

- Primary purpose is to clearly separate (lazy) construction of
globals that are a forward declaration or tentative definition from
those that are the final definition.

- Lazy construction is now encapsulated in
GetAddrOf{Function,GlobalVar} while final definitions are
constructed in EmitGlobal{Function,Var}Definition.

- External interface for dealing with globals is now limited to
EmitGlobal and GetAddrOf{Function,GlobalVar}.

- Also updated helper functions dealing with statics, annotations,
and ctors to be private.

llvm-svn: 54179

show more ...

Fix codegen of chained declarations
- Killed useless CodeGenModule::EmitGlobalVarDeclarator, instead just
recurse on any ScopedDecl.
- Fix for <rdar://problem/6093838>

llvm-svn: 54162

move two more lists of protocols over to use ObjCList<ObjCProtocolDecl>,
simplifying code along the way and fixing a problem and memory leak or two.

llvm-svn: 53876

introduce a new ObjCList templated class and start moving
various objc lists over to it. First up, the protocol list
on ObjCInterfaceDecl.

llvm-svn: 53856

refactor more objc codegen interfaces to pass around selectors so
we don't push strings into LLVM IR and then have to read them back out.

llvm-svn: 52765

indenting and other minor things.

llvm-svn: 52764

give CreateObjCRuntime a full CGM so it can get whatever state it needs,
instead of passing in just a couple random things it currently
happens to use.

llvm-svn: 52756

implement the alias attirbute (in both Sema and Codegen)

llvm-svn: 52092

Emit debug information for global and static variables when -g is specified.

llvm-svn: 51993

For setting attributes, don't assume there are ParamVarDecls available,
because trying to access non-existent ParamVarDecls can crash.
Testcase from the original source for PR2414.

llvm-svn: 51960

First cut at setting attributes for functions and calls; this puts us
much closer to passing the gcc struct layout tests.

It might be possible to refactor this a bit, but I'm not sure there's
actu

First cut at setting attributes for functions and calls; this puts us
much closer to passing the gcc struct layout tests.

It might be possible to refactor this a bit, but I'm not sure there's
actually enough common code for that to be useful.

To get the calling convention completely correct, a bit of
platform-specific code is necessary even for x86-Linux. On x86-Linux, the
alignment of function parameters is extremely strange; as far as I can tell,
it's always 4 except for SSE vectors or structs containing SSE vectors. I'm
continuing to investigate this.

llvm-svn: 51839

show more ...

Support for code generation of Objective-C top-level language constructs.
Implemented by David Chisnall!

llvm-svn: 51835

PR1893: Fix up the type of tentative definitions of incomplete array
types so that they end up the correct size.

llvm-svn: 51787

Allow the type of a global to be different from the type of its
associated declaration. This is a prerequisite to handling
general union initializations; for example, an array of unions involving
p

Allow the type of a global to be different from the type of its
associated declaration. This is a prerequisite to handling
general union initializations; for example, an array of unions involving
pointers has to be turned into a struct because the elements can have
incompatible types.

I refactored the code a bit to make it more readable; now, the logic for
definitions is all in EmitGlobalVarInit.

The second parameter for GetAddrOfGlobalVar is now dead; I'll remove it
separately.

By itself, this patch should not cause any visible changes.

llvm-svn: 51783

show more ...

Always check that the definition of a function has the correct type.
This fixes a crash on the included testcase (found in NetHack).

llvm-svn: 51767

Add CodeGen support for alignment on globals, both for unusual natural
alignment and alignment attributes.

llvm-svn: 51676

Tentative declarations are supposed to have common linkage, not weak.
This change makes clang generate the same thing as llvm-gcc.

llvm-svn: 51674

The cache is working again; re-enable it.

llvm-svn: 51587

Add a more reliable check for whether a static declaration has already
been used. In preparation for the fix to PR2360, but also a minor bug
in its own right.

llvm-svn: 51583

Emit memmove, not memcpy, for structure copies; this is unfortunately
required for correctness in cases of copying a struct to itself or to
an overlapping struct (itself for cases like *a = *a, and

Emit memmove, not memcpy, for structure copies; this is unfortunately
required for correctness in cases of copying a struct to itself or to
an overlapping struct (itself for cases like *a = *a, and overlapping
is possible with unions).

Hopefully, this won't end up being a perf issue; LLVM *should* be able
to optimize memmove to memcpy in a lot of cases, and for small copies
the generated code *should* be mostly comparable. (In reality, LLVM
is currently horrible at optimizing memmove, but that's a bug, not a
fundamental issue.)

gcc currently generates wrong code; that's
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32667.

llvm-svn: 51566

show more ...

Move getAccessedFieldNo out of lib/AST/Expr.cpp into
lib/CodeGen/CGExpr.cpp and to change include/clang/AST/Attr.h to
use its own enum for visibility types instead of using
llvm::GlobalValue::Visibil

Move getAccessedFieldNo out of lib/AST/Expr.cpp into
lib/CodeGen/CGExpr.cpp and to change include/clang/AST/Attr.h to
use its own enum for visibility types instead of using
llvm::GlobalValue::VisibilityTypes. These changes eliminate
dependencies in the AST library on LLVM's VMCore library.

llvm-svn: 51398

show more ...

Disable the function address cache; this doesn't have any ill effects,
as far as I can tell, and it fixes code like test/CodeGen/functions.c.

(Whatever performance effect it might have, crashing on

Disable the function address cache; this doesn't have any ill effects,
as far as I can tell, and it fixes code like test/CodeGen/functions.c.

(Whatever performance effect it might have, crashing on a
construct like this isn't really acceptable; I've run into this
multiple times.)

llvm-svn: 51312

show more ...

Added -g command line options to clang for generating source level debug information. This patch currently enables generation of line number debug information (stoppoints) and region begin/end debug

Added -g command line options to clang for generating source level debug information. This patch currently enables generation of line number debug information (stoppoints) and region begin/end debug information. The new files CGDebugInfo.h and CGDebugInfo.cpp implements the debug info manager class CGDebugInfo.

llvm-svn: 50848

show more ...