llvm-svn: 57673

should work for intel long double, but ppc long double aborts
in convert.

llvm-svn: 57672

RVal => SVal
LVal => Loc
NonLVal => NonLoc
lval => loc
nonlval => nonloc

llvm-svn: 57671

llvm-svn: 57670

constants that don't fit in an int. This fixes
"this decimal constant is unsigned only in ISO C90"
warnings.

llvm-svn: 57668

floating point. This is only correct for x86-32 at the moment.

llvm-svn: 57667

llvm-svn: 57666

type T.

llvm-svn: 57665

struct s {};
void f() {
  int a[10];
  int (*p)[10];
  p = &a;
  (*p)[3] =1;

  struct s d;
  struct s *q;
  q = &d;
}

We return the corresponding MemRegionVal for them.

llvm-svn: 57664

use a SUB instruction instead of an ADD, because -128 can be
encoded in an 8-bit signed immediate field, while +128 can't be.
This avoids the need for a 32-bit immediate field in this case.

A similar optimization applies to 64-bit adds with 0x80000000,
with the 32-bit signed immediate field.

To support this, teach tablegen how to handle 64-bit constants.

llvm-svn: 57663

shift counts, and patterns that match dynamic shift counts
when the subtract is obscured by a truncate node.

Add DAGCombiner support for recognizing rotate patterns
when the shift counts are defined by truncate nodes.

Fix and simplify the code for commuting shld and shrd
instructions to work even when the given instruction doesn't
have a parent, and when the caller needs a new instruction.

These changes allow LLVM to use the shld, shrd, rol, and ror
instructions on x86 to replace equivalent code using two
shifts and an or in many more cases.

llvm-svn: 57662

structure.
 - I'm not sure yet about the behavior, but this at least prevents the
   crash.

Add some asserts on RegionStack usage.

llvm-svn: 57661

llvm-svn: 57660

llvm-svn: 57659

llvm-svn: 57658

Remove GRExprEngine::getLVal and RValues::MakeVal.
Enhance StoreManager "GetLValue" methods to dispatch for specific kinds of lvalue queries, as opposed to interogating the expression tree (GRExprEngine already does this).

Added FIXMEs.  In particular, we no longer "assume" that a base pointer in a field/array access is null (this logic was removed).  Perhaps we should do this when fetching the lvalue for fields and array elements?

llvm-svn: 57657

llvm-svn: 57654

llvm-svn: 57653

llvm-svn: 57649

llvm-svn: 57646

llvm-svn: 57645

llvm-svn: 57644

 - Disables the freeing of the ASTContext and the TranslationUnit
   after parsing & sema.
 - Primarily for timing the impact on -fsyntax-only timings.

llvm-svn: 57643

llvm-svn: 57642

using the 'volatile' qualifier. This should not have any operational consequences
on code, because tags should always be stripped off (giving a non-volatile pointer)
before dereferencing. The new qualification is there to catch some attempts to use
tagged pointers in a context where an untagged pointer is appropriate.

Notably this approach does not catch dereferencing of tagged pointers, but helps
in separating the two concepts a bit.

llvm-svn: 57641

and
http://llvm.org/bugs/show_bug.cgi?id=2893

llvm-svn: 57640

llvm-svn: 57636

uninitialized in these functions with gcc-4.3.

llvm-svn: 57635

warning on x86-64 with gcc-4.3.

llvm-svn: 57634

llvm-svn: 57633

 - Implement fast path when value easily fits in a uint64.
 - ~6x faster, translates to 1-2% on Cocoa.h

llvm-svn: 57632

llvm-svn: 57631

x86 backend.  These will all be answered with "patches welcome", so
a PR doesn't help drive them along.

llvm-svn: 57630

llvm-svn: 57629

basically working, feel free to remove the tag.  The other targets have
really basic things that break them.

llvm-svn: 57628

new VisitLValue method is added to replace the old VisitLVal. The semantics
model becomes more explicit to separate rvalue evaluation from lvalue
evaluation.  

llvm-svn: 57627

llvm-svn: 57626

llvm-svn: 57625

floats.
 - With testcase, which also has some other things GCC folds but we
   don't commented out in it.

llvm-svn: 57624

is to encode the state of the #pragma pack stack as an attribute when
the structure is declared. 

 - Extend PackedAttr to take an alignment (in bits), and reuse for
   both __attribute__((packed)) (which takes no argument, instead
   packing tightly (to "minimize the memory required") and for #pragma
   pack (which allows specification of the maximum alignment in
   bytes). __attribute__((packed)) is just encoded as Alignment=1.

   This conflates two related but different mechanisms, but it didn't
   seem worth another attribute.

 - I have attempted to follow the MSVC semantics as opposed to the gcc
   ones, since if I understand correctly #pragma pack originated with
   MSVC. The semantics are generally equivalent except when the stack
   is altered during the definition of a structure; its not clear if
   anyone does this in practice. See testcase if curious.

llvm-svn: 57623