going to be re-spilled again.

This also helps performance.  Pre-alloc-splitting now seems to be an overall win on SPEC.

llvm-svn: 62834

llvm-svn: 62832

 - Curiously, a number of the current translations gcc does appear to
   be useless?

llvm-svn: 62831

llvm-svn: 62830

llvm-svn: 62829

llvm-svn: 62828

llvm-svn: 62827

Objective-C declarations!) again. Fixes <rdar://problem/6517155>

llvm-svn: 62826

--assert, --classpath).
 - Requires providing some option parameters to over-ride rendering in
   order to match gcc. There may be a cleaner way to do this (probably
   by introducing a new option type for long JoinedOrSeparate forms).

llvm-svn: 62825

forwarding -s to linker, and was only taking last arg in some cases
when should have been taking all).

llvm-svn: 62824

extension. Addresses clang PR/3371.

llvm-svn: 62823

when there are multiple iterations of the loop.  This fixes PR3375.

llvm-svn: 62822

llvm-svn: 62821

 - Unlike groups (which gather distinct but related options), aliases
   are for options like '--all-warnings' which are effectively treated
   like some other option ('-Wall') both in the driver logic and when
   passing to tools.

llvm-svn: 62820

llvm-svn: 62819

llvm-svn: 62818

llvm-svn: 62817

For now, return UnknownVal() in RegionStore::getElementsSize() for AnonTypedRegions.  It wasn't really doing the right thing and was crashing on rdar-6442306-1.m.  This fix causes all path-sensitive test cases to pass with RegionStore.

llvm-svn: 62816

load is always aligned.

I verified that the bswap doesn't occur in the assembly code on x86.

llvm-svn: 62815

Add RegionStore support for the implicit object region that 'self' references.  This causes tests 'ObjCProperties.m' and 'refcnt_naming.m' to now pass with RegionStore.

llvm-svn: 62814

compatibility with VC++.  Patch by Max Burke!

llvm-svn: 62813

initializers, so that we are within the appropriate subobject after
we've processed a multi-designator designation. We're matching GCC and
EDG's behavior on all examples I've found thus far.

*Huge* thanks to Eli Friedman for pointing out my fundamental
misunderstanding of "current object" in the C99 spec.

llvm-svn: 62812

parent wasn't expecting it.

llvm-svn: 62811

llvm-svn: 62808

Handful of fixes to the VC project files.

llvm-svn: 62807

for the new meta-data.

llvm-svn: 62806

llvm-svn: 62805

llvm-svn: 62804

llvm-svn: 62803

llvm-svn: 62802

llvm-svn: 62801

llvm-svn: 62800

llvm-svn: 62799

llvm-svn: 62798

llvm-svn: 62796

llvm-svn: 62795

to "C ^ 1" is only valid when C is known to be either 0 or 1.  Most of the
similar foldings in this function only handle "i1" types, but this one appears
intentionally written to handle larger integer types.  If C has an integer
type larger than "i1", this needs to check if the high bits of a boolean
are known to be zero.  I also changed the comment to describe this folding as
"C ^ 1" instead of "~C", since that is what the code does and since the latter
would only be valid for "i1" types.  The good news is that most LLVM targets
use TargetLowering::ZeroOrOneBooleanContent so this change will not disable
the optimization; the bad news is that I've been unable to come up with a
testcase to demonstrate the problem.

I have also removed a "FIXME" comment for folding "select C, X, 0" to "C & X",
since the code looks correct to me.  It could be made more aggressive by not
limiting the type to "i1", but that would then require checking for
TargetLowering::ZeroOrNegativeOneBooleanContent.  Similar changes could be
done for the other SELECT foldings, but it was decided to be not worth the
trouble and complexity (see e.g., r44663).

llvm-svn: 62790

Simplify x+0 to x in unsafe-fp-math mode. This avoids a bunch of
redundant work in many cases, because in unsafe-fp-math mode,
ISD::FADD with a constant is considered free to negate, so the
DAGCombiner often negates x+0 to -0-x thinking it's free, when
in reality the end result is -x, which is more expensive than x.

Also, combine x*0 to 0.

This fixes PR3374.

llvm-svn: 62789

llvm-svn: 62788

llvm-svn: 62785