start with a cv-qualifier. DeclaratorDecl::getTypeSpecStartLoc() does not
produce the location of the first type-specifier (the cv-qualifier) in this
case, because we don't track source locations for cv-qualifiers.

No test here: I've not found a way to test this with a lit-style test, and
introducing a gtest test for this seems unwarranted. Suggestions welcome!

Patch by Daniel Jasper!

llvm-svn: 157311

pointer, but such folding encounters side-effects, ignore the side-effects
rather than performing them at runtime: CodeGen generates wrong code for
__builtin_object_size in that case.

llvm-svn: 157310

llvm-svn: 157309

llvm-svn: 157308

llvm-svn: 157306

llvm-svn: 157305

Part of rdar://11496790

llvm-svn: 157304

Part of rdar://11496790

llvm-svn: 157303

        self.expect("expr -- $ax == (($ah << 8) | $al)",
            substrs = ['true'])

llvm-svn: 157302

teach computeAllocSize about realloc, reallocf, and valloc

llvm-svn: 157298

llvm-svn: 157292

(0xd 0xd 0xa0 0xf4, or "vld2hs.8 {d0[], d1[]}, [r0]!")

llvm-svn: 157289

on the RecordDecl.  Persist the MS portability type attributes and
ignore them in Sema rather than the parser.

Patch by João Matos!

llvm-svn: 157288

llvm-svn: 157287

llvm-svn: 157278

Plus fix the test class name as well as wrong directory path.

llvm-svn: 157277

The RegisterInfo descriptors for the convenience registers can specify an offset to be added to the offset as derived from
the value_regs field, which is useful for future expansion purposes.  As of now, we have:

    calculated_offset_of_eax = offset_of_rax + (offset_of_eax_from_the_descriptor which is 0)

llvm-svn: 157275

llvm-svn: 157274

llvm-svn: 157273

rdar://problem/11487457

llvm-svn: 157272

CodeGenRegisterClass has two constructors. Both need to compute the
TopoSigs BitVector.

llvm-svn: 157271

llvm-svn: 157270

Part of rdar://11457152

llvm-svn: 157269

llvm-svn: 157268

llvm-svn: 157266

llvm-svn: 157264

Fixes PR12898: SCEVExpander crash.

llvm-svn: 157263

llvm-svn: 157262

move EmitGEPOffset from InstCombine to Transforms/Utils/Local.h

(a draft of this) patch reviewed by Andrew, thanks.

llvm-svn: 157261

first writing the changed files to a temporary location and then overwriting
the original files atomically.

Also adds a RewriterTestContext to aid unit testing rewrting logic in general.

llvm-svn: 157260

allow to build custom configurations

llvm-svn: 157259

llvm-svn: 157256

After a lot of discussion, we realized it's not the best option for run-time bounds checking

llvm-svn: 157255

Also make sure registers aren't erased twice if the dead def mentions
the register twice.

This fixes PR12911.

llvm-svn: 157254

llvm-svn: 157253

llvm-svn: 157252

leader table.  That's because it wasn't expecting instructions to turn up as
leader for a value number that is not its own, but equality propagation could
create this situation.  One solution is to have the leader table use a WeakVH
but this slows down GVN by about 5%.  Instead just have equality propagation not
add instructions to the leader table, only constants and arguments.  In theory
this might cause GVN to run more (each time it changes something it runs again)
but it doesn't seem to occur enough to cause a slow down.

llvm-svn: 157251

llvm-svn: 157250

[asan] increase the stack size limit to 256M (yes, that happens); also CHECK that the stack size is less than that on a non-main thread

llvm-svn: 157249

llvm-svn: 157248