Teach SimpleSValBuilder that (in the absence of more information) stack memory doesn't alias symbolic memory.  This is a heuristic/hack, but works well in practice.  Fixes <rdar://problem/10978247>.

llvm-svn: 152065

C++ method.

llvm-svn: 152064

Use the new composite physical registers.

llvm-svn: 152063

bool. Ugh.

llvm-svn: 152062

llvm-svn: 152061

from the ASTContext.
 - Doing so requires the compiler to generate null checks against the returned
   result, but the BumpPtrAllocator never returns null pointers.
 - The optimizer can usually eliminate such checks, but not always, so this
   gives us tighter codegen in some places.
 - It would be really nice if we could just use __builtin_unreachable or
   something to tell the optimizer that the allocator never returns null, but
   LLVM isn't currently that smart.

llvm-svn: 152060

llvm-svn: 152059

llvm-svn: 152058

the stub SDK to try and ensure we always pick up the right archs when they
should be able to be used.
 - Also, warn when we drop an architecture.

This is a reapplication of r150232 that should stick now that we have the
integrated assembler on for ARM.

llvm-svn: 152057

llvm-svn: 152054

llvm-svn: 152053

 (a) the SystemParameters object is now passed around to the formatters; doing so enables the formatters to reuse computed values for things such as pointer-size and endianness
     instead of repeatedly computing these on their own
 (b) replacing the global ISA cache with a per-process one
 (c) providing a per-process types cache where each formatter can store the types it needs to operate, and be sure to find them the next time without recalculating them
     this also enables formatters to share types if they agree on a local naming convention
 (d) lazy fetching of data from Objective-C runtime data structures
     data is fetched as needed and we stop reading as soon as we determine that an ISA is actually garbage

llvm-svn: 152052

Add SBFrame::IsEqual(const SBFrame &that) method and export it to the Python binding.
Alos add a test case test_frame_api_IsEqual() to TestFrames.py file.

llvm-svn: 152050

llvm-svn: 152049

llvm-svn: 152048

return types that return non-void values. // rdar://10735698

llvm-svn: 152047

llvm-svn: 152046

With the new composite physical registers to represent arbitrary pairs
of DPR registers, we don't need the pseudo-registers anymore. Get rid of
a bunch of them that use DPR register pairs and just use the real
instructions directly instead.

llvm-svn: 152045

llvm-svn: 152044

Used to allow context sensitive printing of super-register or sub-register
references.

llvm-svn: 152043

Patch by Sean Silva!

llvm-svn: 152042

Specifically, remove the magic number when checking to see if the copy has a 
glue operand and simplify the checking logic.

rdar://10930395

llvm-svn: 152041

This was testing the handling of sub-register coalescing followed by
remat.  The original problem was caused by the extra <imp-def> operands
added by sub-register coalescing.  Those <imp-def> operands are not
added any longer, and the test case passes even when the original patch
is reverted.

llvm-svn: 152040

calls with self as a parameter.

llvm-svn: 152039

llvm-svn: 152038

calling an ObjC method ending with 'NoCopy'.

llvm-svn: 152037

In this update:
- I assumed neon2 does not imply vfpv4, but neon and vfpv4 imply neon2.
- I kept setting .fpu=neon-vfpv4 code attribute because that is what the
assembler understands.

Patch by Ana Pazos <apazos@codeaurora.org>

llvm-svn: 152036

llvm-svn: 152035

llvm-svn: 152034

llvm-svn: 152033

ASTContext::getUnaryTransformType.  This can happen if, for example,
an enumeration's underlying type is a typedef.

llvm-svn: 152031

Patch from Sean Silva <silvas@purdue.edu>!

llvm-svn: 152030

llvm-svn: 152029

llvm-svn: 152028

llvm-svn: 152027

llvm-svn: 152026

This implicitly fixes a nasty bug in the GVN hashing (that thankfully
could only manifest as a performance bug): actually include the opcode
in the hash. The old code started the hash off with the opcode, but then
overwrote it with the type pointer.

Since this is likely to be pretty hot (GVN being already pretty
expensive) I've included a micro-optimization to just not bother with
the varargs hashing if they aren't present. I can't measure any change
in GVN performance due to this, even with a big test case like Duncan's
sqlite one. Everything I see is in the noise floor. That said, this
closes a loop hole for a potential scaling problem due to collisions if
the opcode were the differentiating aspect of the expression.

llvm-svn: 152025

In the included testcase, soma thinks that we already have a definition after we
see the out of line decl. Codegen puts it in a deferred list, to be output if
a use is seen. This would break when we saw an explicit template instantiation
definition, since codegen would not be notified.

This patch adds a method to the consumer interface so that soma can notify
codegen that this decl is now required.

llvm-svn: 152024

hashing infrastructure. I wonder why we don't just use StringMap here,
and I may revisit the issue if I have time, but for now I'm just trying
to consolidate.

llvm-svn: 152023

Fix a bug in the mangler where in 'namespace std { extern "C" {X;} }', X would not be seen to be in ::std::.

Migrate two other places where the same logic is used to use the helper function that already exists.

llvm-svn: 152022