While making lld build under the tools directory I decided to refactor how this
works.

There is now a macro, add_llvm_external_project, which takes the name of the
expected subdirectory. This sets up two CMake options.

 * LLVM_EXTERNAL_${NAME}_SOURCE_DIR
     This is the path to the source. It defaults to
     ${CMAKE_CURRENT_SOURCE_DIR}/${name}.
 * LLVM_EXTERNAL_${NAME}_BUILD
     Enable and disable building the tool as part of LLVM.

I chose LLVM_EXTERNAL_${NAME} as a prefix so they all show up together in the
GUI.

llvm-svn: 155654

'REPLACEMENT CHARACTER' (U+FFFD) when getAsInteger fails.

llvm-svn: 155653

llvm-svn: 155652

Fixed SmallMap test. The order of items is undefined in DenseMap. So being checking the increment for big mode, we can only check that all items are in map.

llvm-svn: 155651

the declaration context as not having external visible storage any more.

This should improve performance as we won't needlessly reload the visible decls multiple times
and seems to fix the i386 crash in rdar://11327522.

llvm-svn: 155649

evaluate certain expressions involving invalidly-defined classes.

llvm-svn: 155645

llvm-svn: 155644

llvm-svn: 155642

llvm-svn: 155641

llvm-svn: 155640

llvm-svn: 155639

Save more memory by not parsing the symbol table for stand alone DWARF files. We currently have SymbolFile plug-ins which all get the chance to say what they can parse in a symbol file. Prior to this fix we would ask the SymbolFileDWARF plug-in what abilities it had, and it would answer with "everything", and then we would check the SymbolFileSymtab plug-in what abilities it had, in case it had more abilities. The checking that SymbolFileSymtab does is a bit expensive as it pulls in the entire symbol table just to see if it can offer a few scraps of debug information. This causes all stand along DWARF files to pull in their symbol tables even though those symbols will never be used. This fix will check all SymbolFile plug-ins for their abilities and if any plug-in responds with "everything", then we stop the search.

llvm-svn: 155638

with groups are unsupported.
rdar://11324283

llvm-svn: 155637

 2. Include the helper function and the helper class in the RegisterPasses.h into the polly namespace.

llvm-svn: 155636

llvm-svn: 155634

llvm-svn: 155632

llvm-svn: 155631

On some cores it's a bad idea for performance to mix VFP and NEON instructions
and since these patterns are NEON anyway, the NEON load should be used.

llvm-svn: 155630

llvm-svn: 155629

llvm-svn: 155628

llvm-svn: 155627

llvm-svn: 155626

"friend class Foo;" instead of just friend Foo;".

llvm-svn: 155625

llvm-svn: 155624

llvm-svn: 155622

itself a SFINAE context.

llvm-svn: 155621

Enable detection of AVX and AVX2 support through CPUID. Add AVX/AVX2 to corei7-avx, core-avx-i, and core-avx2 cpu names.

llvm-svn: 155618

trying to mmap a file fails so parent caller function can 
provide helpful information to use about what went wrong.

llvm-svn: 155617

elements to minimize the number of multiplies required to compute the
final result. This uses a heuristic to attempt to form near-optimal
binary exponentiation-style multiply chains. While there are some cases
it misses, it seems to at least a decent job on a very diverse range of
inputs.

Initial benchmarks show no interesting regressions, and an 8%
improvement on SPASS. Let me know if any other interesting results (in
either direction) crop up!

Credit to Richard Smith for the core algorithm, and helping code the
patch itself.

llvm-svn: 155616

[analyzer] check lazy bindings in RegionStore first before looking for default values.  Fixes <rdar://problem/11269741>.

llvm-svn: 155615

Teach RetainCountChecker that it doesn't quite understand pthread_setspecific and it should just give up when it sees it.  Fixes <rdar://problem/11282706>.

llvm-svn: 155613

which creates a temporary by calling a constructor.

llvm-svn: 155608

Linux and other (non-Darwin) platforms and have it use -fmath-errno by
default (for better or worse).

Darwin has seen the light here and uses -fno-math-errno by default, this
patch preserves that.

If any maintainers for a non-Linux platform would also like to opt-in to
-fno-math-errno by default, I'm happy to add folks, but we're currently
getting buts and misleading comparisons with GCC due to this difference
in behavior on Linux at least.

llvm-svn: 155607

arguments, and 'this' in exception-specifications.

llvm-svn: 155606

Automatically enabling the Cocoa formatter categories for command-line LLDB. Previously, the categories were filled in but disabled by default. Tweaking test cases appropriately to keep working and do the right thing

llvm-svn: 155605

llvm-svn: 155604

struct __attribute__((visibility("hidden"))) a;
struct __attribute__((visibility("default"))) b;

which gcc already rejects.

llvm-svn: 155603

the feature set of v7a. This comes about if the user specifies something like
-arch armv7 -mcpu=cortex-m3. We shouldn't be generating instructions such as
uxtab in this case.

rdar://11318438

llvm-svn: 155601

<rdar://problem/11285931>

Use the DWARRF end prologue markers when trying to skip prologue instructions instead of blindly using the second line table address entry.

llvm-svn: 155600

llvm-svn: 155599