llvm-svn: 183136

This is necessary to avoid generating VGPR to SGPR copies in some
cases.

llvm-svn: 183135

llvm-svn: 183134

llvm-svn: 183133

The CopyToReg nodes will sometimes try to copy a value from a VGPR to an
SGPR.  This kind of copy is not possible, so we need to detect
VGPR->SGPR copies and do something else.  The current strategy is to
replace these copies with VGPR->VGPR copies and hope that all the users
of CopyToReg can accept VGPRs as arguments.

llvm-svn: 183132

llvm-svn: 183131

The lowering of stores is now mostly handled in the tablegen files.  No
more BUFFER_STORE nodes I generated during legalization.

llvm-svn: 183130

llvm-svn: 183116

llvm-svn: 183115

Contributed-by: Dmitry N. Mikushin <maemarcus@gmail.com>
llvm-svn: 183114

otherwise, use -polly-detect-scops-in-regions-without-loops to also detect scops
in regions without loops

llvm-svn: 183113

These lines almost never aid readability.

Before:
void f() {
  int i;  // some variable

}

After:
void f() {
  int i;  // some variable
}

llvm-svn: 183112

llvm-svn: 183111

Thanks to Daniel and valgrind.

llvm-svn: 183110

llvm-svn: 183109

llvm-svn: 183108

llvm-svn: 183107

llvm-svn: 183106

llvm-svn: 183105

llvm-svn: 183104

llvm-svn: 183103

llvm-svn: 183102

This patch ensures that APValues are deallocated with the ASTContext by
registering a deallocation function for APValues to the ASTContext.

Original version of the patch by James Dennett.

llvm-svn: 183101

This is needed in clang so one can check if the object needs the
destructor called after its memory was freed. This is useful when
creating many APInt/APFloat objects with placement new, where the
overhead of tracking the pointers for cleanup is significant.

llvm-svn: 183100

Objects allocated after a call to __lsan_disable() will be treated as
live memory. Also add a ScopedDisabler.

llvm-svn: 183099

Address issue reported by Greg Fitzgerald.

llvm-svn: 183098

An oversight in this detection made clang-format unable to format
the following nicely:
void aaaaaaaaaaaaaaaaaaa<aaaaaaaaaaaaaaaaaaaaaaaaaaa,
                         bbbbbbbbbbbbbbbbbbbbbbbbbb>(
    cccccccccccccccccccccccccccc);

llvm-svn: 183097

Before, clang-format would not find a solution for formatting:
if ((aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa ||
     bbbbbbbbbbbbbbbbbb) && // aaaaaaaaaaaaaaaa
    cccccc) {
}

llvm-svn: 183096

which is lifetime-extended by a reference binding. An additional temporary is
created for such a bitfield access (although we have no explicit AST
representation for it).

llvm-svn: 183095

llvm-svn: 183094

materialized temporary with the corresponding MaterializeTemporaryExpr. This is
groundwork for providing C++11's guaranteed static initialization for global
references bound to lifetime-extended temporaries (if the initialization is a
constant expression).

In passing, fix a couple of bugs where some evaluation failures didn't trigger
diagnostics, and a rejects-valid where potential constant expression testing
would assume that it knew the dynamic type of *this and would reject programs
which relied on it being some derived type.

llvm-svn: 183093

has an initializer.

llvm-svn: 183092

llvm-svn: 183091

       using two instructions (sethi and store).

llvm-svn: 183090

lifetime-extending temporaries in reference bindings.

llvm-svn: 183089

llvm-svn: 183088

Reviewers: Bigcheese

CC: llvm-commits

Differential Revision: http://llvm-reviews.chandlerc.com/D905

llvm-svn: 183087

llvm-svn: 183086

clang would incorrectly not allow the following:

int x = true ? (throw 1) : 2;

The problem exists because we don't see beyond the parens.
This, in turn, causes us to believe that we are choosing between void
and int which we diagnose as an error.

Instead, allow clang to see the 'throw' inside the parens.

llvm-svn: 183085

In some cases, clang applies the C++ rules for computing the range of a
value when said value is an enum.

Instead, apply C semantics when in C mode.

llvm-svn: 183084