Commit Graph

125473 Commits

Author SHA1 Message Date
Francois Pichet 7ebc4c1910 ext_reserved_user_defined_literal must not default to Error in MicrosoftMode. Hence create ext_ms_reserved_user_defined_literal that doesn't default to Error; otherwise MSVC headers won't parse.
Fixes PR12383.

llvm-svn: 154273
2012-04-07 23:09:23 +00:00
Craig Topper d024cef233 Turn avx2 vinserti128 intrinsic calls into INSERT_SUBVECTOR DAG nodes and remove patterns for selecting the intrinsic. Similar was already done for avx1.
llvm-svn: 154272
2012-04-07 22:32:29 +00:00
Simon Atanasyan 571d7bde3c MIPS: Pass -mabi option to the assmbler when compile MIPS targets.
llvm-svn: 154270
2012-04-07 22:31:29 +00:00
Simon Atanasyan 3b7589a6c3 MIPS: Move code calculates CPU and ABI names to the separate function to reuse this function later.
llvm-svn: 154269
2012-04-07 22:09:23 +00:00
Craig Topper aa9aab5ad2 Move vinsertf128 patterns near the instruction definitions. Add AddedComplexity to AVX2 vextracti128 patterns to give them priority over the integer versions of vextractf128 patterns.
llvm-svn: 154268
2012-04-07 21:57:43 +00:00
Craig Topper e09d1c5c48 Remove 'else' after 'if' that ends in return.
llvm-svn: 154267
2012-04-07 21:23:41 +00:00
Nadav Rotem 71d07ae5cb 1. Remove the part of r153848 which optimizes shuffle-of-shuffle into a new
shuffle node because it could introduce new shuffle nodes that were not
   supported efficiently by the target.

2. Add a more restrictive shuffle-of-shuffle optimization for cases where the
   second shuffle reverses the transformation of the first shuffle.

llvm-svn: 154266
2012-04-07 21:19:08 +00:00
Duncan Sands 5f8397a934 Convert floating point division by a constant into multiplication by the
reciprocal if converting to the reciprocal is exact.  Do it even if inexact
if -ffast-math.  This substantially speeds up ac.f90 from the polyhedron
benchmarks.

llvm-svn: 154265
2012-04-07 20:04:00 +00:00
Chandler Carruth 75a1cf327a Perform partial SROA on the helper hashing structure. I really wish the
optimizers could do this for us, but expecting partial SROA of classes
with template methods through cloning is probably expecting too much
heroics. With this change, the begin/end pointer pairs which indicate
the status of each loop iteration are actually passed directly into each
layer of the combine_data calls, and the inliner has a chance to see
when most of the combine_data function could be deleted by inlining.
Similarly for 'length'.

We have to be careful to limit the places where in/out reference
parameters are used as those will also defeat the inliner / optimizers
from properly propagating constants.

With this change, LLVM is able to fully inline and unroll the hash
computation of small sets of values, such as two or three pointers.
These now decompose into essentially straight-line code with no loops or
function calls.

There is still one code quality problem to be solved with the hashing --
LLVM is failing to nuke the alloca. It removes all loads from the
alloca, leaving only lifetime intrinsics and dead(!!) stores to the
alloca. =/ Very unfortunate.

llvm-svn: 154264
2012-04-07 20:01:31 +00:00
Chandler Carruth 28192c9398 Fix ValueTracking to conclude that debug intrinsics are safe to
speculate. Without this, loop rotate (among many other places) would
suddenly stop working in the presence of debug info. I found this
looking at loop rotate, and have augmented its tests with a reduction
out of a very hot loop in yacr2 where failing to do this rotation costs
sometimes more than 10% in runtime performance, perturbing numerous
downstream optimizations.

This should have no impact on performance without debug info, but the
change in performance when debug info is enabled can be extreme. As
a consequence (and this how I got to this yak) any profiling of
performance problems should be treated with deep suspicion -- they may
have been wildly innacurate of debug info was enabled for profiling. =/
Just a heads up.

llvm-svn: 154263
2012-04-07 19:22:18 +00:00
Benjamin Kramer e1f4ca1b0f SCEV: When expanding a GEP the final addition to the base pointer has NUW but not NSW.
Found by inspection.

llvm-svn: 154262
2012-04-07 17:19:26 +00:00
Bob Wilson 6f9be7e2c6 Fix Thumb __builtin_longjmp with integrated assembler. <rdar://problem/11203543>
The tLDRr instruction with the last register operand set to the zero register
prints in assembly as if no register was specified, and the assembler encodes
it as a tLDRi instruction with a zero immediate.  With the integrated assembler,
that zero register gets emitted as "r0", so we get "ldr rx, [ry, r0]" which
is broken.  Emit the instruction as tLDRi with a zero immediate.  I don't
know if there's a good way to write a testcase for this.  Suggestions welcome.

Opportunities for follow-up work:
1) The asm printer should complain if a non-optional register operand is set
   to the zero register, instead of silently dropping it.
2) The integrated assembler should complain in the same situation, instead of
   silently emitting the operand as "r0".

llvm-svn: 154261
2012-04-07 16:51:59 +00:00
Hongbin Zheng ed986ab6a4 Rewritten expandRegion to clarify the intention and improve
performance, patched by Johannes Doerfert <johannes@jdoerfert.de>.

llvm-svn: 154260
2012-04-07 15:14:28 +00:00
Hongbin Zheng 3a2d6035d2 ScopDetection: Add some comments to function "expandRegion".
llvm-svn: 154259
2012-04-07 12:29:27 +00:00
Hongbin Zheng 94868e6cc6 Speed up SCoP detection time by checking the exit of the region first,
patched by Johannes Doerfert <johannes@jdoerfert.de>.

llvm-svn: 154258
2012-04-07 12:29:17 +00:00
Benjamin Kramer c2b5c67df3 Linux/ProcessMonitor: include sys/user.h for user_regs_struct and user_fpregs_struct.
llvm-svn: 154255
2012-04-07 09:13:49 +00:00
NAKAMURA Takumi a7d49883bb [Cygwin] Work around to flush stdout in a thread, or stdout in threads won't be flushed at exit.
llvm-svn: 154254
2012-04-07 06:59:28 +00:00
Jason Molenda 7a2f433367 Version bump to lldb-138.
llvm-svn: 154252
2012-04-07 06:18:20 +00:00
Tobias Grosser 84ecc47e1c CodeGen: Allow Polly to do 'grouped unrolling', but no vector generation.
Grouped unrolling means that we unroll a loop such that the different instances
of a certain statement are scheduled right after each other, but we do
not generate any vector code. The idea here is that we can schedule the
bb vectorizer right afterwards and use it heuristics to decide when
vectorization should be performed.

llvm-svn: 154251
2012-04-07 06:16:08 +00:00
Jason Molenda b9e88d4186 Fix a integer trauction issue - calculating the current time in
nanoseconds in 32-bit expression would cause pthread_cond_timedwait
to time out immediately.  Add explicit casts to the TimeValue::TimeValue
ctor that takes a struct timeval and change the NanoSecsPerSec etc
constants defined in TimeValue to be uint64_t so any other calculations
involving these should be promoted to 64-bit even when lldb is built
for 32-bit.

<rdar://problem/11204073>, <rdar://problem/11179821>, <rdar://problem/11194705>.

llvm-svn: 154250
2012-04-07 04:55:02 +00:00
Hongbin Zheng 5758f495da Refactor: Use positive field names in VectorizeConfig.
llvm-svn: 154249
2012-04-07 03:56:23 +00:00
John McCall 5dadb65e07 Fix several problems with protected access control:
- The [class.protected] restriction is non-trivial for any instance
    member, even if the access lacks an object (for example, if it's
    a pointer-to-member constant).  In this case, it is equivalent to
    requiring the naming class to equal the context class.
  - The [class.protected] restriction applies to accesses to constructors
    and destructors.  A protected constructor or destructor can only be
    used to create or destroy a base subobject, as a direct result.
  - Several places were dropping or misapplying object information.

The standard could really be much clearer about what the object type is
supposed to be in some of these accesses.  Usually it's easy enough to
find a reasonable answer, but still, the standard makes a very confident
statement about accesses to instance members only being possible in
either pointer-to-member literals or member access expressions, which
just completely ignores concepts like constructor and destructor
calls, using declarations, unevaluated field references, etc.

llvm-svn: 154248
2012-04-07 03:04:20 +00:00
NAKAMURA Takumi b95f64134e Target/X86/MCTargetDesc/X86MCAsmInfo.cpp: Enable DwarfCFI (aka DW2) on Cygming.
Cygwin-1.7 supports dw2. Some recent mingw distros support one, too.
I have confirmed test-suite/SingleSource/Benchmarks/Shootout-C++/except.cpp can pass on Cygwin.

llvm-svn: 154247
2012-04-07 02:24:20 +00:00
Nick Kledzik 5a1528f49d Add to-do lists
llvm-svn: 154246
2012-04-07 01:54:17 +00:00
Alexis Hunt 78fce432b7 Make the test for r154235 more platform-independent with a shorter
string.

llvm-svn: 154243
2012-04-07 01:33:14 +00:00
Nick Kledzik b334be1ed2 First implementation of Darwin Platform. It is rich enough to generate
a hello world executable from atoms.  There is still much to be flushed out.
Added one test case, test/darwin/hello-world.objtxt, which exercises the
darwin platform.

Added -platform option to lld-core tool to dynamically select platform.

llvm-svn: 154242
2012-04-07 01:31:00 +00:00
NAKAMURA Takumi c4d558a43e test/lit.cfg: Please pass %INCLUDE% to clang.exe on Win32. MS-compatible clang may refer to %INCLUDE%. It fixes r154188.
llvm-svn: 154240
2012-04-07 01:02:53 +00:00
Greg Clayton 7b70be3970 Removed redundant isxdigit checks and added the ability to GetHexU8() so it can extract an 8 bit hex value if one is available. It will set EOF if "set_eof_on_fail" is true or if out of data. This allows a string decoder to grab a string without losing the last part of the packet with a packet like "414243,abc" (it can extract "ABC" and leave the file position set to the comma).
llvm-svn: 154239
2012-04-07 00:42:53 +00:00
Alexis Hunt 0235f684f0 Output UTF-8-encoded characters as identifier characters into assembly
by default.

This is a behaviour configurable in the MCAsmInfo. I've decided to turn
it on by default in (possibly optimistic) hopes that most assemblers are
reasonably sane. If this proves a problem, switching to default seems
reasonable.

I'm not sure if this is the opportune place to test, but it seemed good
to make sure it was tested somewhere.

llvm-svn: 154235
2012-04-07 00:37:53 +00:00
Sean Callanan bf81ea5a9f Hardened the struct layout code to eliminate a
potential crash if the underlying type couldn't
be completed.

llvm-svn: 154234
2012-04-07 00:06:00 +00:00
Chad Rosier 706c235013 [driver] In general, the driver claims redundant args and uses the last arg.
However, the '-x' option has special handling and wasn't following this
paradigm.  Fix it to do so by claiming the arg as we parse the '-x' option.
rdar://11203340

llvm-svn: 154231
2012-04-07 00:01:31 +00:00
Jim Ingham 372787fc19 We sometimes need to be able to call functions (via Process::RunThreadPlan) from code run on the private state thread. To do that we have to
spin up a temporary "private state thread" that will respond to events from the lower level process plugins.  This check-in should work to do
that, but it is still buggy.  However, if you don't call functions on the private state thread, these changes make no difference.

This patch also moves the code in the AppleObjCRuntime step-through-trampoline handler that might call functions (in the case where the debug
server doesn't support the memory allocate/deallocate packet) out to a safe place to do that call.

llvm-svn: 154230
2012-04-07 00:00:41 +00:00
Daniel Dunbar e84f5ca2ea [docs] Note that 'make.bat' provides the same interface as the Makefile for
Windows users.

llvm-svn: 154229
2012-04-06 23:55:50 +00:00
Daniel Dunbar eaa88c8f71 [docs] Start a development guide, and write an introduction to Sphinx based
documentation.

llvm-svn: 154228
2012-04-06 23:47:34 +00:00
Michael J. Spencer 6f71434a3b [docs] Don't duplicate the intro and fix wording.
llvm-svn: 154227
2012-04-06 23:44:58 +00:00
Jim Grosbach 0c509fa6bf Tidy up. 80 columns.
llvm-svn: 154226
2012-04-06 23:43:50 +00:00
David Blaikie 89f13cb5bf Remove "parse error" in favor of more descriptive diagnostics.
In a few cases clang emitted a rather content-free diagnostic: 'parse error'.
This change replaces two actual cases (template parameter parsing and K&R
parameter declaration parsing) with more specific diagnostics and removes a
third dead case of this in the BalancedDelimiterTracker (the ctor already
checked the invariant necessary to ensure that the diag::parse_error was never
actually used).

llvm-svn: 154224
2012-04-06 23:33:59 +00:00
Sean Callanan d13c23c393 Changed some tabs to spaces to make Python like
this file.

llvm-svn: 154222
2012-04-06 23:02:44 +00:00
Sean Callanan 1106434be3 The remaining time calculation didn't reflect the
--start argument.  Fixed that.

llvm-svn: 154221
2012-04-06 23:00:31 +00:00
Fariborz Jahanian 668b9e9c65 more testing of objc's dictionary literal translation.
llvm-svn: 154220
2012-04-06 22:51:48 +00:00
Douglas Gregor 31f55dced5 Implement support for null non-type template arguments for non-type
template parameters of pointer, pointer-to-member, or nullptr_t
type in C++11. Fixes PR9700 / <rdar://problem/11193097>.

llvm-svn: 154219
2012-04-06 22:40:38 +00:00
Fariborz Jahanian 4460e0f805 modern objective-c translation: support for
dictionary literals. This concludes // rdar://10803676

llvm-svn: 154218
2012-04-06 22:29:36 +00:00
Daniel Dunbar 1eafbaad32 [docs] Remove old HTML docs.
llvm-svn: 154217
2012-04-06 22:27:01 +00:00
Ted Kremenek a85f38ba3a Rework ExprEngine::evalLoad and clients (e.g. VisitBinaryOperator) so that when we generate a new ExplodedNode
we use the same Expr* as the one being currently visited.  This is preparation for transitioning to having
ProgramPoints refer to CFGStmts.

This required a bit of trickery.  We wish to keep the old Expr* bindings in the Environment intact,
as plenty of logic relies on it and there is no reason to change it, but we sometimes want the Stmt* for
the ProgramPoint to be different than the Expr* being used for bindings.  This requires adding an extra
argument for some functions (e.g., evalLocation).  This looks a bit strange for some clients, but
it will look a lot cleaner when were start using CFGStmt* in the appropriate places.

As some fallout, the diagnostics arrows are a bit difference, since some of the node locations have changed.
I have audited these, and they look reasonable.

llvm-svn: 154214
2012-04-06 22:10:18 +00:00
Daniel Dunbar feee554e3b [docs] Test commit.
llvm-svn: 154213
2012-04-06 22:06:07 +00:00
Daniel Dunbar 462217eab3 docs: Add the standard LLVM favicon.
llvm-svn: 154212
2012-04-06 21:49:53 +00:00
Johnny Chen 7bba2c6d64 The final batch of adding @dsym_test/@dwarf_test decorators.
llvm-svn: 154211
2012-04-06 21:33:58 +00:00
Jakob Stoklund Olesen baa3566091 ARMPat is equivalent to Requires<[IsARM]>.
llvm-svn: 154210
2012-04-06 21:21:59 +00:00
Jakob Stoklund Olesen b4bd3880ba Eliminate iOS-specific tail call instructions.
After register masks were introdruced to represent the call clobbers, it
is no longer necessary to have duplicate instruction for iOS.

llvm-svn: 154209
2012-04-06 21:17:42 +00:00
Sean Callanan afc7b13e56 Removed a call to truncate() which slowed down
the stress test by a LOT.

llvm-svn: 154208
2012-04-06 21:04:20 +00:00