same so we put in the comment field an indicator when we think we are
emitting the 16 bit version. For the direct object emitter, the difference is 
important as well as for other passes which need an accurate count of 
program size. There will be other similar putbacks to this for various
instructions.

llvm-svn: 174747

Previously, even when a pre-increment load or store was generated,
we often needed to keep a copy of the original base register for use
with other offsets. If all of these offsets are constants (including
the offset which was combined into the addressing mode), then this is
clearly unnecessary. This change adjusts these other offsets to use the
new incremented address.

llvm-svn: 174746

read another one, just as we do for types.

llvm-svn: 174745

llvm-svn: 174744

This is a follow-up to the cost-model change in r174713 which splits
the cost of a memory operation between the address computation and the
actual memory access. In r174713, this cost is always added to the
memory operation cost, and so BBVectorize will do the same.

Currently, this new cost function is used only by ARM, and I don't
have any ARM test cases for BBVectorize. Assistance in generating some
good ARM test cases for BBVectorize would be greatly appreciated!

llvm-svn: 174743

llvm-svn: 174742

Aside from the question of whether we report a warning or an error when we
can't satisfy a requested stack object alignment, the current implementation
of this is not good.  We're not providing any source location in the diagnostics
and the current warning is not connected to any warning group so you can't
control it.  We could improve the source location somewhat, but we can do a
much better job if this check is implemented in the front-end, so let's do that
instead.  <rdar://problem/13127907>

llvm-svn: 174741

This updates the current references to links that work for me.
In the future, we should update the list of references itself to provide
information on newer architecture variants.

Thanks to Sean Silva for pointing out that the current links were broken!

llvm-svn: 174739

llvm-svn: 174738

This correctly formats:
  {
    a;
  }
where { is incorrectly indented by 2, but is at level 0, when
reformatting only 'a;'.

llvm-svn: 174737

Fixes <rdar://problem/12322528>.

llvm-svn: 174736

llvm-svn: 174735

'override' on the method.

This was fixed in a previous commit, generally handling attributes that are at the
end of the declaration.

rdar://13140589

llvm-svn: 174734

llvm-svn: 174733

llvm-svn: 174732

llvm-svn: 174731

llvm-svn: 174730

involving property getter expressions on rhs of property setter.
// rdar://13138459

llvm-svn: 174729

llvm-svn: 174726

Thanks to help from Nadav and Hal, I have a more reasonable (and even
correct!) approach.  This specifically penalizes the insertelement
and extractelement operations for the performance hit that will occur
on PowerPC processors.

llvm-svn: 174725

isn't using the default calling convention.  However, if the transformation is
from a call to inline IR, then the calling convention doesn't matter.
rdar://13157990

llvm-svn: 174724

llvm-svn: 174723

 - Patch by Michael van der Westhuizen:
--
r174404 accidentally removed stdc format, limit and constant macros from the Linux test runner logic.  This small patch re-adds the macros.

Making this change fixes the following tests on Linux:
 - depr/depr.c.headers/inttypes_h.pass.cpp
 - depr/depr.c.headers/stdint_h.pass.cpp
 - input.output/file.streams/c.files/cinttypes.pass.cpp
 - language.support/cstdint/cstdint.syn/cstdint.pass.cpp
--

llvm-svn: 174722

 - Basically I just ran the thread tests many many times on a busy machine and
   bumped the timeouts whenever I hit a test failure.

 - This is obviously subpar, but is the best I can do without the tests being
   rewritten to not depend on arbitrary timeouts.

llvm-svn: 174721

Fixes llvm.org/PR14916.

llvm-svn: 174720

rewriting bug where #ifdef ended up on the same
line as the attribute declaration.

llvm-svn: 174719

They are much easier to handle when attached to the previous token.

Before:
unsigned Indent =
    formatFirstToken(TheLine.First, IndentForLevel[TheLine.Level] >=
                                    0 ? IndentForLevel[TheLine.Level]
: TheLine.Level * 2, TheLine.InPPDirective, PreviousEndOfLineColumn);

After:
unsigned Indent = formatFirstToken(
    TheLine.First, IndentForLevel[TheLine.Level] >= 0
                       ? IndentForLevel[TheLine.Level] : TheLine.Level * 2,
    TheLine.InPPDirective, PreviousEndOfLineColumn);

llvm-svn: 174718

of lines which weren't being explicitly looked at and were printing incorrect results. These
values clearly must lie within 32 bits, so the casts are definitely safe.

llvm-svn: 174717

With this patch, the formatter introduces 'fake' parenthesis according
to the operator precedence of binary operators.

Before:
return aaaa & AAAAAAAAAAAAAAAAAAAAAAAAAAAAA || bbbb &
       BBBBBBBBBBBBBBBBBBBBBBBBBBBBB || cccc & CCCCCCCCCCCCCCCCCCCCCCCCCC ||
       dddd & DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD;
f(aaaaaaaaaaaaaaaaaaaa && aaaaaaaaaaaaaaaaaaaa, aaaaaaaaaaaaaaaaaaaa &&
  aaaaaaaaaaaaaaaaaaaa, aaaaaaaaaaaaaaaaaaaa && aaaaaaaaaaaaaaaaaaaa);

After:
return aaaa & AAAAAAAAAAAAAAAAAAAAAAAAAAAAA ||
       bbbb & BBBBBBBBBBBBBBBBBBBBBBBBBBBBB ||
       cccc & CCCCCCCCCCCCCCCCCCCCCCCCCC ||
       dddd & DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD;
f(aaaaaaaaaaaaaaaaaaaa && aaaaaaaaaaaaaaaaaaaa,
  aaaaaaaaaaaaaaaaaaaa && aaaaaaaaaaaaaaaaaaaa,
  aaaaaaaaaaaaaaaaaaaa && aaaaaaaaaaaaaaaaaaaa);

Future improvements:
- Get rid of some of the hacky ways to nicely format certain constructs.
- Merge this parser and the AnnotatingParser as we now have several parsers
  that analyze (), [], etc.

llvm-svn: 174714

Adds a function to target transform info to query for the cost of address
computation. The cost model analysis pass now also queries this interface.
The code in LoopVectorize adds the cost of address computation as part of the
memory instruction cost calculation. Only there, we know whether the instruction
will be scalarized or not.
Increase the penality for inserting in to D registers on swift. This becomes
necessary because we now always assume that address computation has a cost and
three is a closer value to the architecture.

radar://13097204

llvm-svn: 174713

Update tests for DWARF parser: store sources next to pre-built object files and provide build instructions

llvm-svn: 174711

llvm-svn: 174710

llvm-svn: 174709

llvm-svn: 174708

llvm-svn: 174707

llvm-svn: 174706

llvm-svn: 174705

This a rewrite of the scanf parser. The new implementation is pretty close to
the spec, with a few shortcuts taken here and there. It is conservative, i.e.
it gives up parsing if it does not understand some part of the format string,
or runs into an ambiguous % spec. It does not handle some rarely used parts of
the spec, like %n$ - for now.

I'm also moving parser call to after the original *scanf function completes,
so that we can find out the store size of %s directive by the use of strlen()
on the target buffer.

llvm-svn: 174704

Before:
  [pboard setData:[NSData dataWithBytes:&button
                                 length:sizeof(button)]
          forType:kBookmarkButtonDragType];
After:
  [pboard setData:[NSData dataWithBytes:&button length:sizeof(button)]
          forType:kBookmarkButtonDragType];

llvm-svn: 174701

My previous attempt was extremely deficient, allowing more volatiles
to be introduced and not even checking all of the ones that are
present.

This attempt doesn't try to keep track of the values stored or offsets
within particular objects, just that the correct objects are accessed
in a correctly volatile manner throughout.

llvm-svn: 174700