handling multiple clients.  However, occasionally an
expression must be run in the service of another
expression, and in this case two parsers need to access
the same list of persistent variables.

To allow this, persistent variables now provide state
for multiple parsers, and parsers must allocate, access,
and deallocate this state by providing their own ID
(at the moment, simply the value of the "this" pointer).

<rdar://problem/12914539>

llvm-svn: 172573

into which we can emit single instructions without fixups (which is most
instructions). This is an optimization required because MCDataFragment
is prety large (240 bytes on x64), with no change in functionality.

For large programs, this reduces memory usage overhead required for bundling
by 40%.

To make the code as palatable as possible, the MCEncodedFragment interface was
further fragmented (no pun intended) and MCEncodedFragmentWithFixups is used
as the interface to work against when the user expects fixups. MCDataFragment
and MCRelaxableFragment implement this interface, while the new
MCCompactEncodedInstFragment implements MCEncodeFragment.

llvm-svn: 172572

Do some casting to avoid a signed/unsigned compare.

llvm-svn: 172571

Cast arithmetic results to avoid comparison of an unsigned to an int.

llvm-svn: 172570

// FIXME: Constraints are hard coded to 'm', but we need an 'r'
// constraint for addressof.  This needs to be cleaned up!

Test cases are already in place.  Specifically,
clang/test/CodeGen/ms-inline-asm.c t15(), t16(), and t24().

llvm-svn: 172569

After discussing the refactoring with Jim and Daniel, the following changes were
made:

* All generic directive parsing is now done by AsmParser itself. The previous
  division between it and GenericAsmParser did not have clear boundaries and
  just produced unnatural code of GenericAsmParser juggling the internals of
  AsmParser through an interface. 
  The division of responsibilities is now clear: target-specific directives,
  other extensions (used by platform-specific parseres), and generic directives.
* Priority for directive parsing was reshuffled to ask extensions first and
  check the generic directives later.

No change in functionality.

llvm-svn: 172568

overriding and overridden method, allow the overridden method to have
a narrower contract (introduced earlier, deprecated/obsoleted later)
than the overriding method. Fixes <rdar://problem/12992023>.

llvm-svn: 172567

Fixed a compilation error where NULL was being assigned to a uint.

llvm-svn: 172566

Use const_cast<> to avoid a cast-away-const error.

llvm-svn: 172565

Use const_cast<> to avoid a cast-away-const error.

llvm-svn: 172564

Use const_cast<> to avoid a cast-away-const error.

llvm-svn: 172563

Use const_cast<> to avoid a cast-away-const error.

llvm-svn: 172562

Properly use const_cast to fix a cast-away-const error.

llvm-svn: 172561

Make the const_cast explicit to silence a compiler warning.

llvm-svn: 172560

Do proper casting to avoid a cast-away-const error.

llvm-svn: 172559

Avoid a cast-away-const error by properly using const_cast<>.

llvm-svn: 172558

Eliminate a cast and resulting cast-qual warning by using a temporary
as the target of memcpy.

llvm-svn: 172557

llvm-svn: 172556

llvm-svn: 172555

llvm-svn: 172554

This is done by creating a new non-weak anonymous symbol and creating
an atom with content for that symbol. This is added after the weak symbol
so that they have the same address.

llvm-svn: 172553

llvm-svn: 172552

   some optimization opportunities (in the enclosing supper-expressions).

   rule 1. (-0.0 - X ) * Y => -0.0 - (X * Y)
     if expression "-0.0 - X" has only one reference.

   rule 2. (0.0 - X ) * Y => -0.0 - (X * Y)
     if expression "0.0 - X" has only one reference, and
        the instruction is marked "noSignedZero".

2. Eliminate negation (The compiler was already able to handle these
    opt if the 0.0s are replaced with -0.0.)

   rule 3: (0.0 - X) * (0.0 - Y) => X * Y
   rule 4: (0.0 - X) * C => X * -C
   if the expr is flagged "noSignedZero".

3. 
  Rule 5: (X*Y) * X => (X*X) * Y
   if X!=Y and the expression is flagged with "UnsafeAlgebra".

   The purpose of this transformation is two-fold:
    a) to form a power expression (of X).
    b) potentially shorten the critical path: After transformation, the
       latency of the instruction Y is amortized by the expression of X*X,
       and therefore Y is in a "less critical" position compared to what it
      was before the transformation. 

4. Remove the InstCombine code about simplifiying "X * select".
   
   The reasons are following:
    a) The "select" is somewhat architecture-dependent, therefore the
       higher level optimizers are not able to precisely predict if
       the simplification really yields any performance improvement
       or not.

    b) The "select" operator is bit complicate, and tends to obscure
       optimization opportunities. It is btter to keep it as low as
       possible in expr tree, and let CodeGen to tackle the optimization.

llvm-svn: 172551

llvm-svn: 172550

llvm-svn: 172549

llvm-svn: 172548

This is temporarily necessary until styles are configurable through files
as it seems to be a contentious issue.

llvm-svn: 172546

llvm-svn: 172545

LoopVectorizer cost model. Honor the user command line flag that selects the vectorization factor even if the target machine does not have any vector registers.

llvm-svn: 172544

Some versions of gcc accept unsupported -W flags and run just fine if
there are no warnings, but die with an unsupported flag error if a
warning is encountered.  gcc 4.3 and gcc 4.4 both exhibit this
behavior for -Wno-maybe-uninitialized.  Therefore, if the flag check
for -Wno-maybe-uninitialized succeeds, only use
-Wno-maybe-uninitialized if we are using gcc version 4.7 or greater.
Use -Wno-uninitialized otherwise.

llvm-svn: 172543

llvm-svn: 172542

They are failing on the bots.

llvm-svn: 172540

Also improve test coveration of the handling of relational comparisons.

llvm-svn: 172539

Both underlying IR operations support vectors of pointers already.

llvm-svn: 172538

This is an optimization that djasper spottet. For now, we do not format
anything after the first token that belongs to such an implicit string
literal. All our state is not made for handling that anyway, so we'll
revisit this if we find a problem.

llvm-svn: 172537

Treat tokens inside <> for includes and everything from the second token
of a warning / error on as an implicit string literal, e.g. do not
change its whitespace at all.

Now correctly formats:
 #include < path with space >
 #error Leave     all         white!!!!! space* alone!

Note that for #error and #warning we still format the space up to the
first token of the text, so:
   #  error   Text
will become
 #error Text

llvm-svn: 172536

It was out of sync with the conditions in the other two constructors.

llvm-svn: 172535

Test was failing for clang-native-arm-cortex-a9 build-bot configuration.
The reason for the failure was the test was using hardcoded names.
The attached patch fixes this failure by replacing the hard-coded variables
names with pattern-matched variable names.

Patch by Manish Verma, ARM

llvm-svn: 172534

We used to incorrectly identify some operators (*, &, +, -, etc.) if
there were comments around them.

Example:
Before: int a = /**/ - 1;
After:  int a = /**/ -1;

llvm-svn: 172533

[asan] enable asan_allocator2 by default on Linux. Will enable it on other platforms and remove the old one later, after additional testing. The new allocator is much faster and uses less memory

llvm-svn: 172531