Give the rdrand instructions a SideEffect flag and a chain so MachineCSE and MachineLICM don't touch it.

I already had the necessary things in place for IR-level passes but missed the machine passes.

llvm-svn: 160137

Fixed a bug in wstring_convert concerning zero-length inputs.  Thanks to Jonathan Coxhead for reporting this bug.

llvm-svn: 160136

Mac OS X, because the UUID is the same for all
OpenCL-generated dylibs and therefore would
conflict.

<rdar://problem/11620586>

llvm-svn: 160135

llvm-svn: 160134

Always initialize variable.  If opening quote is the last character in comment, we will be using an uninitialized value.  There is already a test for that in unittests/AST/CommentLexer.cpp, but it seems like we were lucky with the value that got loaded into the variable so the test passed.

llvm-svn: 160133

Summary: Provide more information on usage in -help

Test Plan: ran once

Reviewers: klimek, chandlerc, djasper

Reviewed By: klimek

CC: cfe-commits

Differential Revision: http://llvm-reviews.chandlerc.com/D5

llvm-svn: 160132

llvm-svn: 160131

path passed with -w

Test this functionality.

llvm-svn: 160130

The LIT tests below do not specify the exact cpu model and fail on AVX2 machines, because we select different instructions such as vbroadcast, new shuffles, etc.
Patch by Michael Liao.

llvm-svn: 160129

llvm-svn: 160128

llvm-svn: 160127

Rawr!

llvm-svn: 160125

llvm-svn: 160124

llvm-svn: 160123

llvm-svn: 160122

llvm-svn: 160121

llvm-svn: 160120

llvm-svn: 160118

The rdrand/cmov sequence is the same that is emitted by both
GCC and ICC.

Fixes PR13284.

llvm-svn: 160117

the input vector, it can be bigger (this is helpful for powerpc where <2 x i16>
is a legal vector type but i16 isn't a legal type, IIRC).  However this wasn't
being taken into account by ExpandRes_EXTRACT_VECTOR_ELT, causing PR13220.
Lightly tweaked version of a patch by Michael Liao.

llvm-svn: 160116

One adds matchers for the various parts of a for loop (initializer, condition,
increment), as well as extending the hasBody matcher to work for while and
do-while loops. The second patch adds an isInteger matcher for types.
The third patch fixes a bug in allOf, where a few of the name chages
(AllOf --> allOf) had been missed.

All matchers come with unit tests.

Patches by Sam Panzer!

llvm-svn: 160115

If a non-unicode locale is used, the unicode character is escaped and any
byte that is in the escaped representation but not the semicolon will
become whitespace.

llvm-svn: 160113

llvm-svn: 160112

llvm-svn: 160111

llvm-svn: 160110

llvm-svn: 160107

llvm-svn: 160106

pr13338.

llvm-svn: 160105

behavior and is the first step in fixing pr13338.

llvm-svn: 160104

Failing Tests (3):
    Clang :: Index/complete-cxx-inline-methods.cpp
    Clang :: Index/recursive-cxx-member-calls.cpp
    Clang :: SemaTemplate/inject-templated-friend-post.cpp

llvm-svn: 160103

llvm-svn: 160102

  %shr = lshr i64 %key, 3
  %0 = load i64* %val, align 8
  %sub = add i64 %0, -1
  %and = and i64 %sub, %shr
  ret i64 %and

to:
  %shr = lshr i64 %key, 3
  %0 = load i64* %val, align 8
  %sub = add i64 %0, 2305843009213693951
  %and = and i64 %sub, %shr
  ret i64 %and

The demanded bit optimization is actually a pessimization because add -1 would
be codegen'ed as a sub 1. Teach the demanded constant shrinking optimization
to check for negated constant to make sure it is actually reducing the width
of the constant.

rdar://11793464

llvm-svn: 160101

Add a command channel to wait on along with the file descriptor the ConnectionFileDescriptor class is managing, so we can always pop ourselves out of our select call regardless of how well behaved the channel we are talking to is.

<rdar://problem/11448282>

llvm-svn: 160100

generation of the FDE index.

<rdar://problem/11813705>

llvm-svn: 160099

def Pat<...>;

Results in 'record name is not a string!' diagnostic. Not the best,
but the lack of location information moves it from not very helpful
into completely useless. We're in the Record class when throwing the
error, so just add the location info directly.

llvm-svn: 160098

AST/CommentSema.cpp: Fix signess in abs() to appease msvc. It would not make sense to pass (unsigned)-(unsigned) to abs().

llvm-svn: 160097

llvm-svn: 160096

UnwindPlans for a function.  This specifically does not use any
previously-generated UnwindPlans so if any logging is performed
while creating the UnwindPlans, it will be repeated.  This is
useful for when an lldb stack trace is not correct and you want
to gather diagnostic information from the user -- they can do 
log enable -v lldb unwind, image show-unwind of the function, and
you'll get the full logging as the UnwindPlans are recreated.

llvm-svn: 160095

Previously we were using the static type of the base object to inline
methods, whether virtual or non-virtual. Now, we try to see if the base
object has a known type, and if so ask for its implementation of the method.

llvm-svn: 160094

llvm-svn: 160093