llvm-svn: 85994

Merge constant array and structures. This will create a global variables for arrays and structs that are constant and their initializer is constant. It is on by default but can be disable with the flag -fno-merge-all-constants.

llvm-svn: 85991

llvm-svn: 85990

overloaded operators, e.g.,

  p->template operator+<T>()

llvm-svn: 85989

so the optimizer can tailcall into the return value adjustment thunk.
This improves codesize for complex hierarchies.

llvm-svn: 85988

llvm-svn: 85986

llvm-svn: 85985

Refactor StoreManager::BindDecl() to take a VarRegion* instead of a VarDecl*, and modify GRExprEngine::EvalBind() to handle decl initialization as well.  This paves the way for adding "checker" visitation in EvalBind().

llvm-svn: 85983

llvm-svn: 85980

This introduces a new pass, SlotIndexes, which is responsible for numbering
instructions for register allocation (and other clients). SlotIndexes numbering
is designed to match the existing scheme, so this patch should not cause any
changes in the generated code.

For consistency, and to avoid naming confusion, LiveIndex has been renamed
SlotIndex.

The processImplicitDefs method of the LiveIntervals analysis has been moved
into its own pass so that it can be run prior to SlotIndexes. This was
necessary to match the existing numbering scheme.

llvm-svn: 85979

llvm-svn: 85978

an unconditional branch (possibly from tail merging), this code is
trying to redirect all of its predecessors to go directly to the branch
target, but that isn't feasible for indirect branches.  The other
predecessors (that don't end with indirect branches) could theoretically
still be handled, but that is not easily done right now.

The AnalyzeBranch interface doesn't currently let us distinguish jump table
branches from indirect branches, and this code is currently handling
jump tables.  To avoid punting on address-taken blocks, we would have to give
up handling jump tables.  That seems like a bad tradeoff.

llvm-svn: 85975

more aggressive an correct.  This survives building llvm in 64-bit
mode with optimizations and the built llvm passes make check.

llvm-svn: 85973

llvm-svn: 85972

Change GRTransferFuncs::RegisterChecks() to take a GRExprEngine& instead of a BugReporter&.  This paves the way for pulling some of the retain/release checker into a "Checker" class.

llvm-svn: 85971

type to look at the volatile specifier.  I found these all from just
hand auditing the code.

llvm-svn: 85967

operators, e.g., 

  operator+<int>

which now works in declarators, id-expressions, and member access
expressions. This commit only implements the non-dependent case, where
we can resolve the template-id to an actual declaration.

llvm-svn: 85966

Fix t2Int_eh_sjlj_setjmp. Immediate form of orr is a 32-bit instruction. So it should be 22 bytes instead of 20 bytes long.

llvm-svn: 85965

llvm-svn: 85963

llvm-svn: 85958

llvm-svn: 85957

one instance of a large problem.  assert for non-canoical types would
help track down these things.

llvm-svn: 85956

llvm-svn: 85955

llvm-svn: 85953

llvm-svn: 85952

llvm-svn: 85951

llvm-svn: 85947

llvm-svn: 85946

llvm-svn: 85945

llvm-svn: 85942

automatically shadowed by the ExtensionRAIIObject created by
ParseStructDeclaration.

llvm-svn: 85941

Do a scheduling pass ignoring anti-dependencies to identify candidate registers that should be renamed.

llvm-svn: 85939

llvm-svn: 85938

llvm-svn: 85937

Changes requested (avoid getFunction(), avoid Type creation via isVoidTy(), and avoid redundant isFreeCall cases) in feedback to r85176

llvm-svn: 85936

of value-initialization and trivial constructors.

llvm-svn: 85935

<rdar://problem/7352605>. When building schedule graph use mayAlias information to avoid chaining loads/stores of spill slots with non-aliased memory ops.

llvm-svn: 85934

Changes (* location in pointer variables, avoiding include, and using APInt::getLimitedValue) based on feedback to r85814

llvm-svn: 85933

Roman Divacky! PR5363

llvm-svn: 85932

llvm-svn: 85931