platform/upstream/llvm.git
11 years agoRemove PathV1.h from CompilerInvocation.cpp.
Rafael Espindola [Wed, 26 Jun 2013 05:40:30 +0000 (05:40 +0000)]
Remove PathV1.h from CompilerInvocation.cpp.

llvm-svn: 184918

11 years agoFix the build when __APPLE__ is defined.
Rafael Espindola [Wed, 26 Jun 2013 05:25:44 +0000 (05:25 +0000)]
Fix the build when __APPLE__ is defined.

llvm-svn: 184917

11 years agoRemove sys::GetMainExecutable.
Rafael Espindola [Wed, 26 Jun 2013 05:05:37 +0000 (05:05 +0000)]
Remove sys::GetMainExecutable.

llvm-svn: 184916

11 years agoUse llvm::sys::fs::getMainExecutable.
Rafael Espindola [Wed, 26 Jun 2013 05:03:40 +0000 (05:03 +0000)]
Use llvm::sys::fs::getMainExecutable.

llvm-svn: 184915

11 years agoPort GetMainExecutable over to PathV2.
Rafael Espindola [Wed, 26 Jun 2013 05:01:35 +0000 (05:01 +0000)]
Port GetMainExecutable over to PathV2.

I will remove the V1 version as soon as I change clang in the next commit.

llvm-svn: 184914

11 years agoRemove PathV1.h from CompilerInstance.cpp.
Rafael Espindola [Wed, 26 Jun 2013 04:32:59 +0000 (04:32 +0000)]
Remove PathV1.h from CompilerInstance.cpp.

llvm-svn: 184913

11 years agoDon't use PathV1.h in ASTUnit.cpp.
Rafael Espindola [Wed, 26 Jun 2013 04:26:38 +0000 (04:26 +0000)]
Don't use PathV1.h in ASTUnit.cpp.

llvm-svn: 184912

11 years agoRemove forward declarations of sys::Path.
Rafael Espindola [Wed, 26 Jun 2013 04:23:08 +0000 (04:23 +0000)]
Remove forward declarations of sys::Path.

llvm-svn: 184911

11 years agoRemove PathWithStatus.
Rafael Espindola [Wed, 26 Jun 2013 04:15:55 +0000 (04:15 +0000)]
Remove PathWithStatus.

llvm-svn: 184910

11 years agoRemove the last use of PathWithStatus.
Rafael Espindola [Wed, 26 Jun 2013 04:12:57 +0000 (04:12 +0000)]
Remove the last use of PathWithStatus.

llvm-svn: 184909

11 years agoUse llvm::sys::fs::unique_file.
Rafael Espindola [Wed, 26 Jun 2013 04:02:37 +0000 (04:02 +0000)]
Use llvm::sys::fs::unique_file.

llvm-svn: 184908

11 years agoRemove some uses of llvm::sys::Path.
Rafael Espindola [Wed, 26 Jun 2013 03:52:38 +0000 (03:52 +0000)]
Remove some uses of llvm::sys::Path.

llvm-svn: 184907

11 years agoRemove PathV1.h use from WindowsToolChain.cpp.
Rafael Espindola [Wed, 26 Jun 2013 03:39:10 +0000 (03:39 +0000)]
Remove PathV1.h use from WindowsToolChain.cpp.

llvm-svn: 184906

11 years agoFix a comment.
Nico Weber [Wed, 26 Jun 2013 02:42:46 +0000 (02:42 +0000)]
Fix a comment.

llvm-svn: 184905

11 years agoLazily deserialize the "first' friend declaration when deserializing a class
Richard Smith [Wed, 26 Jun 2013 02:41:25 +0000 (02:41 +0000)]
Lazily deserialize the "first' friend declaration when deserializing a class
declaration. This PCH a little lazier, and breaks a deserialization cycle that
causes crashes with modules enabled.

llvm-svn: 184904

11 years agoFix PCH bug with member templates of local classes in nontemplate functions.
Faisal Vali [Wed, 26 Jun 2013 02:34:24 +0000 (02:34 +0000)]
Fix PCH bug with member templates of local classes in nontemplate functions.

As noted by Richard in the post:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20130624/082605.html, the following code should not add an entry
into PendingLocalImplicitInstantiations, since local instantiations
should only occur within the context of other instantiations:

int foo(double y) {
   struct Lambda {
      template<class T> T operator()(T t) const { return t; };
   } lambda;
   return lambda(y);
}

Hence the attached code does the following:
  1) In MarkFunctionReferenced, check if ActiveInstantiations.size()
      is non-zero before adding to PendingLocalImplicitInstantiations.
  2) In InstantiateFunctionDefinition, we swap out/in
      PendingLocalImplicitInstantiations so that only those
      pending local instantiations that are added during the instantiation
      of the current function are instantiated recursively.

llvm-svn: 184903

11 years agoRemove PathV1.h usage from ToolChains.cpp.
Rafael Espindola [Wed, 26 Jun 2013 02:13:00 +0000 (02:13 +0000)]
Remove PathV1.h usage from ToolChains.cpp.

llvm-svn: 184902

11 years agoRemove ifdef LLDB_CONFIGURATION_DEBUG directives around the formatter
Jason Molenda [Wed, 26 Jun 2013 01:51:03 +0000 (01:51 +0000)]
Remove ifdef LLDB_CONFIGURATION_DEBUG directives around the formatter
cache ivars/methods.

llvm-svn: 184901

11 years agoRemove the #define USE_CACHE since the formatters cache has been operational for...
Enrico Granata [Wed, 26 Jun 2013 01:03:38 +0000 (01:03 +0000)]
Remove the #define USE_CACHE since the formatters cache has been operational for a while now and has not caused issues that warrant disabling it
Also, print the cache hits statistics if the log is in debugging mode vs. LLDB being a debug build - this should make it easier to gather useful metrics on cache success rate for real users

llvm-svn: 184900

11 years ago<rdar://problem/14266578>
Enrico Granata [Wed, 26 Jun 2013 00:36:27 +0000 (00:36 +0000)]
<rdar://problem/14266578>

"command source" was not properly setting the stop-on-error option

llvm-svn: 184899

11 years ago<rdar://problem/14243761>
Enrico Granata [Wed, 26 Jun 2013 00:31:21 +0000 (00:31 +0000)]
<rdar://problem/14243761>

The argument to -w (--category) in type * list is a regular expression
This caused unhappiness with the gnu-libstdc++ category because of the double ++

Now we check for exact textual match as-well-as regexp matching

llvm-svn: 184898

11 years agodbgs() << Instruction doesn't print a newline on the end any more. Update these
Nick Lewycky [Wed, 26 Jun 2013 00:30:18 +0000 (00:30 +0000)]
dbgs() << Instruction doesn't print a newline on the end any more. Update these
debug statements to add a missing newline. Also canonicalize to '\n' instead of
"\n"; the latter calls a function with a loop the former does not.

llvm-svn: 184897

11 years agoRun clang-format on lib/Format code after r184894. No other changes.
Nico Weber [Wed, 26 Jun 2013 00:30:14 +0000 (00:30 +0000)]
Run clang-format on lib/Format code after r184894. No other changes.

llvm-svn: 184896

11 years agoFix a use after free I introduced and that Bill caught in code review
Chandler Carruth [Wed, 26 Jun 2013 00:26:16 +0000 (00:26 +0000)]
Fix a use after free I introduced and that Bill caught in code review
(thanks!) by deferring the free of the filename until we finish writing
the coverage data to that file.

Bill, let me know if you'd prefer a different approach!

llvm-svn: 184895

11 years agoFormatter: Don't put a space after parameter-naming block comments.
Nico Weber [Wed, 26 Jun 2013 00:15:19 +0000 (00:15 +0000)]
Formatter: Don't put a space after parameter-naming block comments.

Before: f(a, b, /*doFoo=*/ false);
Now: f(a, b, /*doFoo=*/false);

This style is a lot more common:
$ ack -H '=\*\/\w' lib | wc -l
    1281
$ ack -H '=\*\/ \w' lib | wc -l
      70

llvm-svn: 184894

11 years ago<rdar://problem/14266411>
Enrico Granata [Tue, 25 Jun 2013 23:43:28 +0000 (23:43 +0000)]
<rdar://problem/14266411>

The semi-unofficial way of returning a status from a Python command was to return a string (e.g. return "no such variable was found") that LLDB would pick as a clue of an error having happened

This checkin changes that:
- SBCommandReturnObject now exports a SetError() call, which can take an SBError or a plain C-string
- script commands now drop any return value and expect the SBCommandReturnObject ("return object") to be filled in appropriately - if you do nothing, a success will be assumed

If your commands were relying on returning a value and having LLDB pick that up as an error, please change your commands to SetError() through the return object or expect changes in behavior

llvm-svn: 184893

11 years agos/C++0x/C++11/
Adrian Prantl [Tue, 25 Jun 2013 23:42:37 +0000 (23:42 +0000)]
s/C++0x/C++11/

llvm-svn: 184892

11 years agoMerge isReachable into isBackedge.
Jakob Stoklund Olesen [Tue, 25 Jun 2013 23:32:10 +0000 (23:32 +0000)]
Merge isReachable into isBackedge.

Prefer using RPO.lookup() instead of RPO[] which can mutate the map.

llvm-svn: 184891

11 years agoDon't check for triviality on fields of templated records. We can't know the
Nick Lewycky [Tue, 25 Jun 2013 23:22:23 +0000 (23:22 +0000)]
Don't check for triviality on fields of templated records. We can't know the
answer until after instantiation. Fixes PR16061!

llvm-svn: 184890

11 years agoImplement DR136
David Majnemer [Tue, 25 Jun 2013 23:09:30 +0000 (23:09 +0000)]
Implement DR136

Friend declarations that specify a default argument must be a definition
and the only declaration in the translation unit.

llvm-svn: 184889

11 years agoSLPVectorizer: support slp-vectorization of PHINodes between basic blocks
Nadav Rotem [Tue, 25 Jun 2013 23:04:09 +0000 (23:04 +0000)]
SLPVectorizer: support slp-vectorization of PHINodes between basic blocks

llvm-svn: 184888

11 years agoMake this test properly test both things it's trying to test.
Richard Smith [Tue, 25 Jun 2013 22:37:05 +0000 (22:37 +0000)]
Make this test properly test both things it's trying to test.

llvm-svn: 184887

11 years agoFixed a bug in ClangASTSource where we would return
Sean Callanan [Tue, 25 Jun 2013 22:36:17 +0000 (22:36 +0000)]
Fixed a bug in ClangASTSource where we would return
the target of a typedef when asked for a typedef.

llvm-svn: 184886

11 years agoMissing test update from r184884.
Richard Smith [Tue, 25 Jun 2013 22:22:42 +0000 (22:22 +0000)]
Missing test update from r184884.

llvm-svn: 184885

11 years agoPR8302: Check for shadowing a template parameter when declaring a template
Richard Smith [Tue, 25 Jun 2013 22:21:36 +0000 (22:21 +0000)]
PR8302: Check for shadowing a template parameter when declaring a template
template parameter.

llvm-svn: 184884

11 years agoDon't try to get the layout of an invalid decl in getDeclAlign.
Matt Beaumont-Gay [Tue, 25 Jun 2013 22:19:15 +0000 (22:19 +0000)]
Don't try to get the layout of an invalid decl in getDeclAlign.

When the decl that we're getting alignment for is a FieldDecl, and the field's
parent record is invalid, skip the actual field alignment calculation (and
return 1-byte alignment in the general case).

Also, assert in in getASTRecordLayout that the decl is valid. This was
inspired by PR16292; see also r184581 and r184751.

llvm-svn: 184883

11 years agoImplement DR21
David Majnemer [Tue, 25 Jun 2013 22:08:55 +0000 (22:08 +0000)]
Implement DR21

A default template-argument shall not be specified in a friend template
declaration.

Interestingly, we properly handled default template arguments on friend
class members but not on just friend classes.

llvm-svn: 184882

11 years agoPrint block frequencies in decimal form.
Jakob Stoklund Olesen [Tue, 25 Jun 2013 21:57:38 +0000 (21:57 +0000)]
Print block frequencies in decimal form.

This is easier to read than the internal fixed-point representation.

If anybody knows the correct algorithm for converting fixed-point
numbers to base 10, feel free to fix it.

llvm-svn: 184881

11 years agoR600: Use new getNamedOperandIdx function generated by TableGen
Tom Stellard [Tue, 25 Jun 2013 21:22:18 +0000 (21:22 +0000)]
R600: Use new getNamedOperandIdx function generated by TableGen

llvm-svn: 184880

11 years agoTableGen: Generate a function for getting operand indices based on their defined...
Tom Stellard [Tue, 25 Jun 2013 21:22:09 +0000 (21:22 +0000)]
TableGen: Generate a function for getting operand indices based on their defined names

This patch modifies TableGen to generate a function in
${TARGET}GenInstrInfo.inc called getNamedOperandIdx(), which can be used
to look up indices for operands based on their names.

In order to activate this feature for an instruction, you must set the
UseNamedOperandTable bit.

For example, if you have an instruction like:

def ADD : TargetInstr <(outs GPR:$dst), (ins GPR:$src0, GPR:$src1)>;

You can look up the operand indices using the new function, like this:

Target::getNamedOperandIdx(Target::ADD, Target::OpName::dst)  => 0
Target::getNamedOperandIdx(Target::ADD, Target::OpName::src0) => 1
Target::getNamedOperandIdx(Target::ADD, Target::OpName::src1) => 2

The operand names are case sensitive, so $dst and $DST are considered
different operands.

This change is useful for R600 which has instructions with a large number
of operands, many of which model single bit instruction configuration
values.  These configuration bits are common across most instructions,
but may have a different operand index depending on the instruction type.
It is useful to have a convenient way to look up the operand indices,
so these bits can be generically set on any instruction.

llvm-svn: 184879

11 years agoDon't use 'errno.h' on Apple just yet. This breaks for some of our buildbots.
Bill Wendling [Tue, 25 Jun 2013 21:08:40 +0000 (21:08 +0000)]
Don't use 'errno.h' on Apple just yet. This breaks for some of our buildbots.

llvm-svn: 184878

11 years agoFix undefined behaviour in data formatter test -- ensure char* null-terminated
Daniel Malea [Tue, 25 Jun 2013 20:54:24 +0000 (20:54 +0000)]
Fix undefined behaviour in data formatter test -- ensure char* null-terminated
so LLDB does not read off the end of the array.

llvm-svn: 184877

11 years agoUpdate version sscanf to match FreeBSD release info
Ed Maste [Tue, 25 Jun 2013 20:50:07 +0000 (20:50 +0000)]
Update version sscanf to match FreeBSD release info

Release strings are of the form 9.1-RELEASE-p3 or 10.0-CURRENT.

llvm-svn: 184876

11 years agoFix null pointer dereference if we redeclare an unprototyped function. Patch by
Richard Smith [Tue, 25 Jun 2013 20:34:17 +0000 (20:34 +0000)]
Fix null pointer dereference if we redeclare an unprototyped function. Patch by
WenHan Gu!

llvm-svn: 184875

11 years agoSkip dsym tests also on FreeBSD
Ed Maste [Tue, 25 Jun 2013 19:59:56 +0000 (19:59 +0000)]
Skip dsym tests also on FreeBSD

llvm-svn: 184874

11 years agoFix a comment.
Nico Weber [Tue, 25 Jun 2013 19:25:12 +0000 (19:25 +0000)]
Fix a comment.

llvm-svn: 184873

11 years agoX86 cost model: Vectorizing integer division is a bad idea
Arnold Schwaighofer [Tue, 25 Jun 2013 19:14:09 +0000 (19:14 +0000)]
X86 cost model: Vectorizing integer division is a bad idea

radar://14057959

llvm-svn: 184872

11 years agoSkip tests that hang on FreeBSD
Ed Maste [Tue, 25 Jun 2013 19:11:36 +0000 (19:11 +0000)]
Skip tests that hang on FreeBSD

There's still significant work to do in the FreeBSD port, so no point
in a pr for these yet.

llvm-svn: 184871

11 years agoFix SROA to avoid unnecessary scalar conversions for 1-element vectors.
Bob Wilson [Tue, 25 Jun 2013 19:09:50 +0000 (19:09 +0000)]
Fix SROA to avoid unnecessary scalar conversions for 1-element vectors.

When a 1-element vector alloca is promoted, a store instruction can often be
rewritten without converting the value to a scalar and using an insertelement
instruction to stuff it into the new alloca.  This patch just adds a check
to skip that conversion when it is unnecessary.  This turns out to be really
important for some ARM Neon operations where <1 x i64> is used to get around
the fact that i64 is not a legal type.

llvm-svn: 184870

11 years agocpp11-migrate: Really fixing doxygen warning
Edwin Vane [Tue, 25 Jun 2013 19:01:18 +0000 (19:01 +0000)]
cpp11-migrate: Really fixing doxygen warning

Last attempt at this fix was bogus.

llvm-svn: 184869

11 years agoCorrect use of __FreeBSD_kernel__
Ed Maste [Tue, 25 Jun 2013 18:58:11 +0000 (18:58 +0000)]
Correct use of __FreeBSD_kernel__

It is defined on recent FreeBSD versions, so must not be mutually
exclusive with an #elif FreeBSD block.

Patch submitted by Robert Millan.

Fixes PR#16447.

llvm-svn: 184867

11 years agoRemove unused code. No functionality change.
Manman Ren [Tue, 25 Jun 2013 18:49:55 +0000 (18:49 +0000)]
Remove unused code. No functionality change.

llvm-svn: 184866

11 years agoMore of N3652: don't add an implicit 'const' to 'constexpr' member functions when...
Richard Smith [Tue, 25 Jun 2013 18:46:26 +0000 (18:46 +0000)]
More of N3652: don't add an implicit 'const' to 'constexpr' member functions when checking for overloads in C++1y.

llvm-svn: 184865

11 years agoThe GCDA 402 format won't have a second checksum either.
Bill Wendling [Tue, 25 Jun 2013 18:13:52 +0000 (18:13 +0000)]
The GCDA 402 format won't have a second checksum either.

llvm-svn: 184864

11 years agoObjective-C: Warn when IBOutletCollection property
Fariborz Jahanian [Tue, 25 Jun 2013 17:34:50 +0000 (17:34 +0000)]
Objective-C: Warn when IBOutletCollection property
is declared to have 'assign' attribute.
// rdar://14212998

llvm-svn: 184863

11 years agoRemove misplaced comment
Eli Bendersky [Tue, 25 Jun 2013 17:07:56 +0000 (17:07 +0000)]
Remove misplaced comment

llvm-svn: 184862

11 years agoIn tooling, reenable freeing of datastructures in codegen, just as we do for
Nick Lewycky [Tue, 25 Jun 2013 17:01:21 +0000 (17:01 +0000)]
In tooling, reenable freeing of datastructures in codegen, just as we do for
the frontend. We don't want to respect the -disable-free flag here.

llvm-svn: 184861

11 years ago[PowerPC] Support @got modifier
Ulrich Weigand [Tue, 25 Jun 2013 16:49:50 +0000 (16:49 +0000)]
[PowerPC] Support @got modifier

Add VK_... values and relocation types necessary to support
the @got family of modifiers.  Used by the asm parser only.

llvm-svn: 184860

11 years agoImplement full support for non-pointer pointers in custom allocators for list.
Howard Hinnant [Tue, 25 Jun 2013 16:08:47 +0000 (16:08 +0000)]
Implement full support for non-pointer pointers in custom allocators for list.

llvm-svn: 184859

11 years agotsan: add missing __attribute__((visibility("default"))) to interface functions
Dmitry Vyukov [Tue, 25 Jun 2013 15:36:25 +0000 (15:36 +0000)]
tsan: add missing __attribute__((visibility("default"))) to interface functions

llvm-svn: 184858

11 years agoUse simpler version of llvm::sys::fs::exists.
Rafael Espindola [Tue, 25 Jun 2013 15:14:22 +0000 (15:14 +0000)]
Use simpler version of llvm::sys::fs::exists.

llvm-svn: 184857

11 years agoDon't use PathV1.h in Tools.cpp.
Rafael Espindola [Tue, 25 Jun 2013 15:03:59 +0000 (15:03 +0000)]
Don't use PathV1.h in Tools.cpp.

llvm-svn: 184856

11 years agoUse the simpler version of llvm::sys::fs::exists.
Rafael Espindola [Tue, 25 Jun 2013 14:48:00 +0000 (14:48 +0000)]
Use the simpler version of llvm::sys::fs::exists.

llvm-svn: 184855

11 years agoMatch printf format specifiers and arguments
Ed Maste [Tue, 25 Jun 2013 14:47:45 +0000 (14:47 +0000)]
Match printf format specifiers and arguments

llvm-svn: 184854

11 years agoMove GetEXESuffix to the one place it is used.
Rafael Espindola [Tue, 25 Jun 2013 14:42:30 +0000 (14:42 +0000)]
Move GetEXESuffix to the one place it is used.

llvm-svn: 184853

11 years agoRemove sys::PathSeparator.
Rafael Espindola [Tue, 25 Jun 2013 14:32:45 +0000 (14:32 +0000)]
Remove sys::PathSeparator.

llvm-svn: 184852

11 years agoUse the new llvm::sys::EnvPathSeparator.
Rafael Espindola [Tue, 25 Jun 2013 14:29:51 +0000 (14:29 +0000)]
Use the new llvm::sys::EnvPathSeparator.

llvm-svn: 184851

11 years agoUpdate comment after Linux->POSIX rename
Ed Maste [Tue, 25 Jun 2013 14:29:15 +0000 (14:29 +0000)]
Update comment after Linux->POSIX rename

llvm-svn: 184850

11 years ago[lsan] Define interceptors more correctly. Also, always clear allocated memory.
Sergey Matveev [Tue, 25 Jun 2013 14:05:52 +0000 (14:05 +0000)]
[lsan] Define interceptors more correctly. Also, always clear allocated memory.

llvm-svn: 184849

11 years agoR600: Consolidate expansion of v2i32/v4i32 ops for EG/SI
Aaron Watry [Tue, 25 Jun 2013 13:55:57 +0000 (13:55 +0000)]
R600: Consolidate expansion of v2i32/v4i32 ops for EG/SI

By default, we expand these operations for both EG and SI. Move the
duplicated code into a common space for now. If the targets ever actually
implement these operations as instructions, we can override that in the relevant
target.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184848

11 years agoR600: Add v2i32 test for vselect
Aaron Watry [Tue, 25 Jun 2013 13:55:54 +0000 (13:55 +0000)]
R600: Add v2i32 test for vselect

Note: Only adding test for evergreen, not SI yet.

When I attempted to expand vselect for SI, I got the following:
llc: /home/awatry/src/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp:522:
llvm::SDValue llvm::DAGTypeLegalizer::PromoteIntRes_SETCC(llvm::SDNode*):
Assertion `SVT.isVector() == N->getOperand(0).getValueType().isVector() &&
"Vector compare must return a vector result!"' failed.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184847

11 years agoR600/SI: Expand xor v2i32/v4i32
Aaron Watry [Tue, 25 Jun 2013 13:55:52 +0000 (13:55 +0000)]
R600/SI: Expand xor v2i32/v4i32

Add test cases for both vector sizes on SI and also add v2i32 test for EG.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184846

11 years agoR600: Add v2i32 test for setcc on evergreen
Aaron Watry [Tue, 25 Jun 2013 13:55:49 +0000 (13:55 +0000)]
R600: Add v2i32 test for setcc on evergreen

No test/expansion for SI has been added yet. Attempts to expand this
operation for SI resulted in a stacktrace in (IIRC) LegalizeIntegerTypes
which was complaining about vector comparisons being required to return
a vector type.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184845

11 years agoR600/SI: Expand urem of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:46 +0000 (13:55 +0000)]
R600/SI: Expand urem of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Note: I followed the guidance of the v4i32 EG check... UREM produces really
complex code, so let's just check that the instruction was lowered
successfully.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184844

11 years agoR600/SI: Expand udiv v[24]i32 for SI and v2i32 for EG
Aaron Watry [Tue, 25 Jun 2013 13:55:43 +0000 (13:55 +0000)]
R600/SI: Expand udiv v[24]i32 for SI and v2i32 for EG

Also add lit test for both cases on SI, and v2i32 for evergreen.

Note: I followed the guidance of the v4i32 EG check... UDIV produces really
complex code, so let's just check that the instruction was lowered
successfully.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184843

11 years agoR600/SI: Expand ashr of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:40 +0000 (13:55 +0000)]
R600/SI: Expand ashr of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184842

11 years agoR600/SI: Expand srl of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:37 +0000 (13:55 +0000)]
R600/SI: Expand srl of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184841

11 years agoR600/SI: Expand shl of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:32 +0000 (13:55 +0000)]
R600/SI: Expand shl of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184840

11 years agoR600/SI: Expand or of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:29 +0000 (13:55 +0000)]
R600/SI: Expand or of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184839

11 years agoR600/SI: Expand mul of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:26 +0000 (13:55 +0000)]
R600/SI: Expand mul of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184838

11 years agoR600/SI: Expand and of v2i32/v4i32 for SI
Aaron Watry [Tue, 25 Jun 2013 13:55:23 +0000 (13:55 +0000)]
R600/SI: Expand and of v2i32/v4i32 for SI

Also add lit test for both cases on SI, and v2i32 for evergreen.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 184837

11 years ago[sanitizer] Move log_path to common flag and use it in MSan.
Evgeniy Stepanov [Tue, 25 Jun 2013 13:50:44 +0000 (13:50 +0000)]
[sanitizer] Move log_path to common flag and use it in MSan.

llvm-svn: 184836

11 years agoBlockFrequency: Bump up the entry frequency a bit.
Benjamin Kramer [Tue, 25 Jun 2013 13:34:40 +0000 (13:34 +0000)]
BlockFrequency: Bump up the entry frequency a bit.

This is a band-aid to fix the most severe regressions we're seeing from basing
spill decisions on block frequencies, until we have a better solution.

llvm-svn: 184835

11 years ago[PowerPC] Add extended rotate/shift mnemonics
Ulrich Weigand [Tue, 25 Jun 2013 13:17:41 +0000 (13:17 +0000)]
[PowerPC] Add extended rotate/shift mnemonics

This adds all missing extended rotate/shift mnemonics to the asm parser.

llvm-svn: 184834

11 years ago[PowerPC] Add rldcr/rldic instructions
Ulrich Weigand [Tue, 25 Jun 2013 13:17:10 +0000 (13:17 +0000)]
[PowerPC] Add rldcr/rldic instructions

This adds pattern for the rldcr and rldic instructions (the last instruction
from the rotate/shift family that were missing).  They are currently used
only by the asm parser.

llvm-svn: 184833

11 years ago[PowerPC] Add extended subtract mnemonics
Ulrich Weigand [Tue, 25 Jun 2013 13:16:48 +0000 (13:16 +0000)]
[PowerPC] Add extended subtract mnemonics

This adds support for the extended subtract mnemonics to the asm parser:
   subi
   subis
   subic
   subic.
   sub
   sub.
   subc
   subc.

llvm-svn: 184832

11 years ago[NVPTX] Default pointer type doesn't make sense for getParamSymbol()
Justin Holewinski [Tue, 25 Jun 2013 12:22:21 +0000 (12:22 +0000)]
[NVPTX] Default pointer type doesn't make sense for getParamSymbol()

llvm-svn: 184831

11 years agoFix a couple of PPC predefined macros that I spotted while driving by
Chandler Carruth [Tue, 25 Jun 2013 11:13:47 +0000 (11:13 +0000)]
Fix a couple of PPC predefined macros that I spotted while driving by
this code. These aren't technically standard predefines for the platform
but apparantly lots of folks use them as they show up within LLVM's own
codebase. ;] This may even fix some self host issues w/ the JIT!!!

llvm-svn: 184830

11 years agoIf debugserver fails to interrogate the inferior process CPU type
Jason Molenda [Tue, 25 Jun 2013 06:42:09 +0000 (06:42 +0000)]
If debugserver fails to interrogate the inferior process CPU type
for any reason, use debugserver own's cputype as a best guess when
we reply to the debugger's qProcessInfo packet or when initializing
our register tables.
<rdar://problem/13406879>

llvm-svn: 184829

11 years agoRemove some unnecessary uses of nub_break_t in the arm specific
Jason Molenda [Tue, 25 Jun 2013 06:01:20 +0000 (06:01 +0000)]
Remove some unnecessary uses of nub_break_t in the arm specific
support files for debugserver to fix a build failure for arm.  Also
remove some of the code used for software-driven single instruction
stepping; this is slowly being yanked out and these particular bits
overlap with the nub_break_t going away.

llvm-svn: 184828

11 years agoFix a typo in the code that collected the costs recursively.
Nadav Rotem [Tue, 25 Jun 2013 05:30:56 +0000 (05:30 +0000)]
Fix a typo in the code that collected the costs recursively.

llvm-svn: 184827

11 years agokeep only the StringRef version of getFileOrSTDIN.
Rafael Espindola [Tue, 25 Jun 2013 05:28:34 +0000 (05:28 +0000)]
keep only the StringRef version of getFileOrSTDIN.

llvm-svn: 184826

11 years agoAdd r184803 back now that the bug in unique_file has been fixed.
Rafael Espindola [Tue, 25 Jun 2013 04:26:55 +0000 (04:26 +0000)]
Add r184803 back now that the bug in unique_file has been fixed.

Original message:

Use the new PathV2 instead of implementing the logic in clang.

llvm-svn: 184825

11 years agoDon't assume ResultPath is null terminated.
Rafael Espindola [Tue, 25 Jun 2013 04:23:46 +0000 (04:23 +0000)]
Don't assume ResultPath is null terminated.

llvm-svn: 184824

11 years agoRevert "Temporarily enable MI-Sched on X86."
Andrew Trick [Tue, 25 Jun 2013 02:48:58 +0000 (02:48 +0000)]
Revert "Temporarily enable MI-Sched on X86."

This reverts commit 98a9b72e8c56dc13a2617de84503a3d78352789c.

llvm-svn: 184823

11 years agoR600/SI: Report unaligned memory accesses as legal for > 32-bit types
Tom Stellard [Tue, 25 Jun 2013 02:39:35 +0000 (02:39 +0000)]
R600/SI: Report unaligned memory accesses as legal for > 32-bit types

In reality, some unaligned memory accesses are legal for 32-bit types and
smaller too, but it all depends on the address space.  Allowing
unaligned loads/stores for > 32-bit types is mainly to prevent the
legalizer from splitting one load into multiple loads of smaller types.

https://bugs.freedesktop.org/show_bug.cgi?id=65873

llvm-svn: 184822

11 years agoR600: Add support for i32 loads from the constant address space on Cayman
Tom Stellard [Tue, 25 Jun 2013 02:39:30 +0000 (02:39 +0000)]
R600: Add support for i32 loads from the constant address space on Cayman

Tested-By: Aaron Watry <awatry@gmail.com>
llvm-svn: 184821

11 years agoR600/SI: Add support for v4i32 and v4f32 kernel args
Tom Stellard [Tue, 25 Jun 2013 02:39:25 +0000 (02:39 +0000)]
R600/SI: Add support for v4i32 and v4f32 kernel args

Tested-By: Aaron Watry <awatry@gmail.com>
llvm-svn: 184820

11 years agoR600: Fix typo in R600Schedule.td
Tom Stellard [Tue, 25 Jun 2013 02:39:20 +0000 (02:39 +0000)]
R600: Fix typo in R600Schedule.td

This should only make a difference in programs that use a lot of the
vector ALU instructions like BFI_INT and BIT_ALIGN.  There is a slight
improvement in the phatk bitcoin mining kernel with this patch on
Evergreen (vector size == 1):

Before:
1173 Instruction Groups / 9520 dwords

After:
1167 Instruction Groups / 9510 dwords

Reviewed-by: Reviewed-by: Vincent Lejeune<vljn at ovi.com>
llvm-svn: 184819

11 years agoRevert "Use the new PathV2 instead of implementing the logic in clang."
Rafael Espindola [Tue, 25 Jun 2013 02:35:32 +0000 (02:35 +0000)]
Revert "Use the new PathV2 instead of implementing the logic in clang."

This reverts commit 184803 while I debug the failures on the bots.

llvm-svn: 184818