Shankar Easwaran [Tue, 22 Jan 2013 04:12:03 +0000 (04:12 +0000)]
it looks like g++ 4.7.2 doesnot like bit fields when the type is a enumeration (Throws an error as bit-field with non integral type)
llvm-svn: 173125
NAKAMURA Takumi [Tue, 22 Jan 2013 04:11:33 +0000 (04:11 +0000)]
libclang: Update comment about USEDLIBS in c-*-test/Makefile.
"Note that 'USEDLIBS' must include all of the core clang libraries when -static is given to linker on cygming."
llvm-svn: 173124
NAKAMURA Takumi [Tue, 22 Jan 2013 04:02:41 +0000 (04:02 +0000)]
ErrorOrTest.cpp: Check existence of EXPECT_DEBUG_DEATH. It is not always available.
llvm-svn: 173123
John McCall [Tue, 22 Jan 2013 03:56:22 +0000 (03:56 +0000)]
Use the correct field to copy/dispose a __block variable.
We were previously hard-coding a particular field index. This was
fine before (because we were obviously guaranteed the presence
of a copy/dispose member) except for (1) alignment padding and
(2) future extensions adding extra members to the header, such
as the extended-layout pointer.
Note that we only introduce the extended-layout pointer in the
presence of structs. (We also seem to be introducing it even
in the presence of an all-non-object layout, but that's a
different potential issue.)
llvm-svn: 173122
NAKAMURA Takumi [Tue, 22 Jan 2013 03:49:16 +0000 (03:49 +0000)]
clang/test/Index/comment-to-html-xml-conversion.cpp: Mark this as XFAIL:valgrind, for now. Working in progress.
llvm-svn: 173121
Daniel Dunbar [Tue, 22 Jan 2013 03:42:49 +0000 (03:42 +0000)]
[MC/Mach-O] Load commands are supposed to 8-byte aligned on 64-bit.
llvm-svn: 173120
Enrico Granata [Tue, 22 Jan 2013 02:32:56 +0000 (02:32 +0000)]
lldb.thread is not to be used anymore
llvm-svn: 173119
Sean Callanan [Tue, 22 Jan 2013 02:20:20 +0000 (02:20 +0000)]
Quick fix to make LLDB TOT work with Clang TOT.
Avoids an error about an ambiguous constructor
call.
llvm-svn: 173118
Michael J. Spencer [Tue, 22 Jan 2013 02:15:30 +0000 (02:15 +0000)]
Add {,ELF,MachO}TargetInfo.
llvm-svn: 173117
NAKAMURA Takumi [Tue, 22 Jan 2013 01:52:04 +0000 (01:52 +0000)]
clang/test: [CMake] check-clang doesn't require llvm-dis any more.
llvm-svn: 173116
NAKAMURA Takumi [Tue, 22 Jan 2013 01:51:59 +0000 (01:51 +0000)]
clang/test/CodeGen: Nuke llvm's opt and llvm-dis, and FileCheck-ize two tests. -O1 is sufficient here.
llvm-svn: 173115
NAKAMURA Takumi [Tue, 22 Jan 2013 01:51:54 +0000 (01:51 +0000)]
clang/test/CodeGen/2006-01-13-StackSave.c: Nuke llvm's opt and llvm-dis, and FileCheck-ize.
@llvm.stacksave is emitted regardless of "opt -std-compile-opts". We have optimizers' tests in llvm/test/Transforms.
llvm-svn: 173114
NAKAMURA Takumi [Tue, 22 Jan 2013 01:51:48 +0000 (01:51 +0000)]
clang/test/CodeGen/blocks-seq.c: FileCheck-ize.
llvm-svn: 173113
NAKAMURA Takumi [Tue, 22 Jan 2013 01:44:00 +0000 (01:44 +0000)]
[CMake] bugpoint-passes should not be built in "all", when LLVM_BUILD_TOOLS is OFF.
llvm-svn: 173112
Enrico Granata [Tue, 22 Jan 2013 01:43:59 +0000 (01:43 +0000)]
Replacing a "no clue why" comment with the explanation it was demanding
llvm-svn: 173111
Andrew Trick [Tue, 22 Jan 2013 00:57:16 +0000 (00:57 +0000)]
Remove target triple from an LSR test.
Manish already fixed this test to work with NoTTI.
llvm-svn: 173110
Bill Wendling [Tue, 22 Jan 2013 00:53:12 +0000 (00:53 +0000)]
Add a new method that adds the AttributeSet at the given index. No functional change.
llvm-svn: 173109
Bill Wendling [Tue, 22 Jan 2013 00:41:33 +0000 (00:41 +0000)]
Add the attributes that are passed in instead of the ones we're merging into.
llvm-svn: 173108
Greg Clayton [Mon, 21 Jan 2013 23:54:42 +0000 (23:54 +0000)]
Clear up the documentation for the "container-regs" and "invalidate-regs" key/value pair responses for qRegisterInfo with examples of single and multiple registers for each.
llvm-svn: 173107
Bill Wendling [Mon, 21 Jan 2013 23:41:50 +0000 (23:41 +0000)]
Remove unused method.
llvm-svn: 173106
Greg Clayton [Mon, 21 Jan 2013 23:32:42 +0000 (23:32 +0000)]
Changed the register number lists for the qRegisterInfo packet response to be raw hex to match all other register reading and writing APIs.
llvm-svn: 173105
Greg Clayton [Mon, 21 Jan 2013 23:25:18 +0000 (23:25 +0000)]
Added all of the 16 and 8 bit register variants for i386.
Modified the ARM register context to invalidate r8 - r14 when the CPSR register is modified.
llvm-svn: 173104
Joel Jones [Mon, 21 Jan 2013 23:20:47 +0000 (23:20 +0000)]
Fix spelling
llvm-svn: 173103
Greg Clayton [Mon, 21 Jan 2013 23:18:28 +0000 (23:18 +0000)]
Updated the LLDB qRegisterInfo documentation to contain information on the new "invalidate-regs" and "container-regs" keys that can be returned in response to a qRegisterInfo packet.
llvm-svn: 173102
Bill Wendling [Mon, 21 Jan 2013 23:03:18 +0000 (23:03 +0000)]
Use AttributeSet instead of Attribute to verify things.
llvm-svn: 173101
Daniel Jasper [Mon, 21 Jan 2013 22:49:20 +0000 (22:49 +0000)]
Remove "incorrect" aligning of trailing comments.
We used to align trailing comments belong to different things.
Before:
void f() { // some function..
}
int a; // some variable..
After:
void f() { // some function..
}
int a; // some variable..
llvm-svn: 173100
Bill Wendling [Mon, 21 Jan 2013 22:45:00 +0000 (22:45 +0000)]
Have AttributeSet::getRetAttributes() return an AttributeSet instead of Attribute.
This further restricts the use of the Attribute class to the Attribute family of
classes.
llvm-svn: 173099
Bill Wendling [Mon, 21 Jan 2013 22:44:49 +0000 (22:44 +0000)]
Have AttributeSet::getRetAttributes() return an AttributeSet instead of Attribute.
This further restricts the use of the Attribute class to the Attribute family of
classes.
llvm-svn: 173098
Fariborz Jahanian [Mon, 21 Jan 2013 22:32:29 +0000 (22:32 +0000)]
objectiveC: don't warn when in -Wselector mode and
an unimplemented selector is consumed by
"respondsToSelector:". // rdar://
12938616
llvm-svn: 173097
Greg Clayton [Mon, 21 Jan 2013 22:17:50 +0000 (22:17 +0000)]
<rdar://problem/
13020634>
Fixed the 32, 16, and 8 bit pseudo regs for x86_64 (real reg of "rax" which subvalues "eax", "ax", etc...) to correctly get updated when stepping. Also fixed it so actual registers can specify what other registers must be invalidated when a register is modified. Previously, only pseudo registers could invalidate other registers.
Modified the LLDB qRegisterInfo extension to the GDB remote interface to support specifying the containing registers with the new "container-regs" key whose value is a comma separated list of register numbers. Also added a "invalidate-regs" key whose value is also a comma separated list of register numbers.
Removed the hack GDBRemoteDynamicRegisterInfo::Addx86_64ConvenienceRegisters() function and modified "debugserver" to specify the registers correctly using the new "container-regs" and "invalidate-regs" keys.
llvm-svn: 173096
Bill Wendling [Mon, 21 Jan 2013 21:57:40 +0000 (21:57 +0000)]
Make AttributeSet::getFnAttributes() return an AttributeSet instead of an Attribute.
This is more code to isolate the use of the Attribute class to that of just
holding one attribute instead of a collection of attributes.
llvm-svn: 173095
Bill Wendling [Mon, 21 Jan 2013 21:57:28 +0000 (21:57 +0000)]
Make AttributeSet::getFnAttributes() return an AttributeSet instead of an Attribute.
This is more code to isolate the use of the Attribute class to that of just
holding one attribute instead of a collection of attributes.
llvm-svn: 173094
Paul Redmond [Mon, 21 Jan 2013 21:57:20 +0000 (21:57 +0000)]
Transform (sub 0, (zext bool to A)) to (sext bool to A) and
(sub 0, (sext bool to A)) to (zext bool to A).
Patch by Muhammad Ahmad
Reviewed by Duncan Sands
llvm-svn: 173093
Michael Ilseman [Mon, 21 Jan 2013 21:46:32 +0000 (21:46 +0000)]
Docs for SparseMultiSet
llvm-svn: 173092
Richard Osborne [Mon, 21 Jan 2013 21:12:30 +0000 (21:12 +0000)]
Fix some incorrectly named u10 / lu10 instructions.
llvm-svn: 173090
Jakub Staszak [Mon, 21 Jan 2013 21:02:47 +0000 (21:02 +0000)]
Remove unneeded #include.
llvm-svn: 173088
Richard Osborne [Mon, 21 Jan 2013 20:50:54 +0000 (20:50 +0000)]
Remove unused multiclass.
llvm-svn: 173087
Richard Osborne [Mon, 21 Jan 2013 20:44:17 +0000 (20:44 +0000)]
Add instruction encodings / disassembly support for u6 / lu6 instructions.
llvm-svn: 173086
Richard Osborne [Mon, 21 Jan 2013 20:42:16 +0000 (20:42 +0000)]
Add instruction encoding / disassembly support for ru6 / lru6 instructions.
llvm-svn: 173085
Howard Hinnant [Mon, 21 Jan 2013 20:39:41 +0000 (20:39 +0000)]
Implement the ATOMIC_*_LOCK_FREE macros.
llvm-svn: 173084
Richard Osborne [Mon, 21 Jan 2013 20:32:54 +0000 (20:32 +0000)]
Use correct format for the LDAWCP instruction (u6).
llvm-svn: 173083
Shankar Easwaran [Mon, 21 Jan 2013 20:09:55 +0000 (20:09 +0000)]
no functionality changes(split WriterELF into multiple files)
llvm-svn: 173082
Douglas Gregor [Mon, 21 Jan 2013 20:07:12 +0000 (20:07 +0000)]
Give ModuleFiles an index, so that we can use indexed vectors rather
than DenseMaps and SmallPtrSets for module-visitation data. ~2.6%
speedup for modules.
llvm-svn: 173081
Douglas Gregor [Mon, 21 Jan 2013 19:42:21 +0000 (19:42 +0000)]
Eliminate Sema::CompareProperties(), which was walking over a pile of
lexical declarations looking for properties when we could more
efficiently check for property mismatches at property declaration
time. Good for ~1% of -fsyntax-only time when most of the properties
we're checking against come from an AST file.
llvm-svn: 173079
Nico Weber [Mon, 21 Jan 2013 19:35:06 +0000 (19:35 +0000)]
Formatter: Set MatchingParen for [], to match <>, (), {}. No functionality change.
llvm-svn: 173078
Nico Weber [Mon, 21 Jan 2013 19:29:31 +0000 (19:29 +0000)]
Formatter: Rename LSquare to Left to make parseSquare() more consistent with the other paren parsing methods.
llvm-svn: 173077
Enrico Granata [Mon, 21 Jan 2013 19:20:50 +0000 (19:20 +0000)]
<rdar://problem/
12437929>
Providing a special mode of operator for "memory read -f c-str" which actually works in most common cases
Where the old behavior would provide:
(lldb) mem read --format s `foo`
0x100000f5d: NULL
Now we do:
(lldb) mem read --format s `foo`
0x100000f5d: "hello world"
You can also specify a count and that many strings will be showed starting at the initial address:
(lldb) mem read -c 2 -f c-str `foo`
0x100000f1d: "hello world"
0x100000f29: "short"
llvm-svn: 173076
Manuel Klimek [Mon, 21 Jan 2013 19:17:52 +0000 (19:17 +0000)]
Fixes various problems around enum parsing.
Very similar to what we do for record definitions:
- tighten down what is an enum definition, so that we don't mistake a
function for an enum
- allow common idioms around declarations (we'll want to handle that
more centrally in the future)
We now correctly format:
enum X f() {
a();
return 42;
}
llvm-svn: 173075
Chris Lattner [Mon, 21 Jan 2013 19:08:15 +0000 (19:08 +0000)]
r173072 is causing some regressions on big endian hosts, I don't have time to debug it
so revert it for now.
llvm-svn: 173074
Douglas Gregor [Mon, 21 Jan 2013 19:05:22 +0000 (19:05 +0000)]
Eliminate the oddly-named Sema::ComparePropertiesInBaseAndSuper, which
did a redundant traversal of the lexical declarations in the
superclass. Instead, when we declare a new property, look into the
superclass to see whether we're redeclaring the property. Goot for 1%
of -fsyntax-only time on Cocoa.h and a little less than 3% on my
modules test case.
llvm-svn: 173073
Chris Lattner [Mon, 21 Jan 2013 18:48:26 +0000 (18:48 +0000)]
rework the Bitstream reader to actually work a machine word at a time, instead of 32-bits at a time.
This cuts in half the number of virtual methods called to refill that word when compiling on a 64-bit
host, and will make 64-bit read operations faster.
llvm-svn: 173072
Argyrios Kyrtzidis [Mon, 21 Jan 2013 18:45:42 +0000 (18:45 +0000)]
[ASTUnit] Unlike LoadFromCommandLine, LoadFromCompilerInvocation causes a crash if Precompilepreamble
is set to true because there is no FileManager at that point.
Patch by Hurcan Solter!
llvm-svn: 173071
Jordan Rose [Mon, 21 Jan 2013 18:41:05 +0000 (18:41 +0000)]
[analyzer] Fix test for r173067.
Note to self: don't remove comments /after/ updating the line-sensitive
part of a test.
llvm-svn: 173070
Douglas Gregor [Mon, 21 Jan 2013 18:35:55 +0000 (18:35 +0000)]
Replace some unnecessary O(N^2) lookups for properties with
DeclContext lookups. The performance win is negligible in my tests,
but it's the Right Thing To Do (TM).
llvm-svn: 173068
Jordan Rose [Mon, 21 Jan 2013 18:28:30 +0000 (18:28 +0000)]
[analyzer] Show notes inside implicit calls at the last explicit call site.
Before:
struct Wrapper { <-- 2. Calling default constructor for 'NonTrivial'.
NonTrivial m;
};
Wrapper w; <-- 1. Calling implicit default constructor for 'Wrapper'.
After:
struct Wrapper {
NonTrivial m;
};
Wrapper w; <-- 1. Calling implicit default constructor for 'Wrapper'.
^-- 2. Calling default constructor for 'NonTrivial'.
llvm-svn: 173067
Chris Lattner [Mon, 21 Jan 2013 18:28:26 +0000 (18:28 +0000)]
ReadSourceManagerBlock is skipping over records that can contain Blobs. Not passing
in a StringRef to bind to them forces them to be unpacked into the Record as individual
bytes. This is wasteful, but not likely to be measurable in this instance.
llvm-svn: 173066
Chris Lattner [Mon, 21 Jan 2013 18:24:49 +0000 (18:24 +0000)]
Fix a heinous inefficiency introduced in r149918, wherein reading each byte of a
BLOB (i.e., large, performance intensive data) in a bitcode file was switched to
invoking one virtual method call per byte read. Now we do one virtual call per
BLOB.
llvm-svn: 173065
Michael Ilseman [Mon, 21 Jan 2013 18:18:53 +0000 (18:18 +0000)]
Introduce a new data structure, the SparseMultiSet, and changes to the MI scheduler to use it.
A SparseMultiSet adds multiset behavior to SparseSet, while retaining SparseSet's desirable properties. Essentially, SparseMultiSet provides multiset behavior by storing its dense data in doubly linked lists that are inlined into the dense vector. This allows it to provide good data locality as well as vector-like constant-time clear() and fast constant time find(), insert(), and erase(). It also allows SparseMultiSet to have a builtin recycler rather than keeping SparseSet's behavior of always swapping upon removal, which allows it to preserve more iterators. It's often a better alternative to a SparseSet of a growable container or vector-of-vector.
llvm-svn: 173064
Chris Lattner [Mon, 21 Jan 2013 18:18:25 +0000 (18:18 +0000)]
wean Blob handling logic off of banging on NextChar directly. Instead, make
it reason about the current bit position, which is always independent of the
underlying cursors word size.
llvm-svn: 173063
Chris Lattner [Mon, 21 Jan 2013 18:04:19 +0000 (18:04 +0000)]
rename "SkipToWord" to "SkipToFourByteBoundary" since a word is not always 4 bytes.
llvm-svn: 173062
Nadav Rotem [Mon, 21 Jan 2013 17:59:18 +0000 (17:59 +0000)]
Fix a comment. Induction vars dont need to start at zero.
llvm-svn: 173061
Howard Hinnant [Mon, 21 Jan 2013 17:26:55 +0000 (17:26 +0000)]
Donated anonymously: This enables GCC 4.8.0 to build libc++.
llvm-svn: 173060
Douglas Gregor [Mon, 21 Jan 2013 16:52:34 +0000 (16:52 +0000)]
Revert r173056; it breaks one of the CodeGen-with-PCH tests.
llvm-svn: 173058
Douglas Gregor [Mon, 21 Jan 2013 16:47:11 +0000 (16:47 +0000)]
When loading an identifier from an AST file solely for the purpose of
forming the identifier, e.g., as part of a selector or a declaration
name, don't actually deserialize any information about the
identifier. Instead, simply mark it "out-of-date" and we'll load the
the information on demand. 2% speedup on the modules testcase I'm
looking at; should also help PCH.
llvm-svn: 173056
Manuel Klimek [Mon, 21 Jan 2013 16:42:44 +0000 (16:42 +0000)]
Fixes formatting of empty blocks.
We now only put empty blocks into a single line, if all of:
- all tokens of the structural element fit into a single line
- we're not in a control flow statement
Note that we usually don't put record definitions into a single line, as
there's usually at least one more token (the semicolon) after the
closing brace. This doesn't hold when we are in a context where there is
no semicolon, like "enum E {}".
There were some missing tests around joining lines around the corner
cases of the allowed number of columns, so this patch adds some.
llvm-svn: 173055
Douglas Gregor [Mon, 21 Jan 2013 16:16:40 +0000 (16:16 +0000)]
When deserializing a declaration, don't look for redeclarations if its
kind indicates that it can never be redeclared. Good for a 1% speedup,
and redeclaration searching drops off the profile.
llvm-svn: 173054
Tom Stellard [Mon, 21 Jan 2013 15:40:48 +0000 (15:40 +0000)]
R600/SI: Use unnormalized coordinates for sampling with the RECT target.
Patch by: Michel Dänzer
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 173053
Tom Stellard [Mon, 21 Jan 2013 15:40:47 +0000 (15:40 +0000)]
R600/SI: Take target parameter for sample intrinsics.
Patch by: Michel Dänzer
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 173052
Tom Stellard [Mon, 21 Jan 2013 15:40:46 +0000 (15:40 +0000)]
R600/SI: Derive all sample intrinsics from a single class.
Patch by: Michel Dänzer
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 173051
Douglas Gregor [Mon, 21 Jan 2013 15:25:38 +0000 (15:25 +0000)]
Introduce a fast path for the ASTReader's name lookup within a
DeclContext. When the DeclContext is of a kind that can only be
defined once and never updated, we limit the search to the module file
that conatins the lookup table. Provides a 15% speedup in one
modules-heavy source file.
llvm-svn: 173050
Kostya Serebryany [Mon, 21 Jan 2013 15:04:36 +0000 (15:04 +0000)]
[asan] fix linux build
llvm-svn: 173049
Alexander Potapenko [Mon, 21 Jan 2013 14:49:55 +0000 (14:49 +0000)]
[ASan] Move Mac-specific tests to asan_mac_test.cc
llvm-svn: 173048
Daniel Jasper [Mon, 21 Jan 2013 14:39:22 +0000 (14:39 +0000)]
Fix parsing of templated declarations.
Before: template <template <typename T>, typename P > class X;
After: template <template <typename T>, typename P> class X;
More importantly, the token annotations for the second ">" are now computed
correctly.
llvm-svn: 173047
Edwin Vane [Mon, 21 Jan 2013 14:35:54 +0000 (14:35 +0000)]
Removing empty loop-convert directory
llvm-svn: 173046
Manuel Klimek [Mon, 21 Jan 2013 14:32:05 +0000 (14:32 +0000)]
Fixes indent in linkage specification blocks.
We now indent:
extern "C" {
int a;
}
without additional indent inside the extern "C" block.
llvm-svn: 173045
Alexey Samsonov [Mon, 21 Jan 2013 14:31:45 +0000 (14:31 +0000)]
CMake: generalize checking for target availability and add initial support for PowerPC native arch. With this patch, building LLVM on PowerPC native arch produces a working ASan runtime.
llvm-svn: 173044
Daniel Jasper [Mon, 21 Jan 2013 14:18:28 +0000 (14:18 +0000)]
Fix bug discovered by valgrind.
When trying to merge lines, we should not touch lines that are invalid,
as we don't know how long they might be.
llvm-svn: 173043
Manuel Klimek [Mon, 21 Jan 2013 14:16:56 +0000 (14:16 +0000)]
Add regression test.
llvm-svn: 173042
NAKAMURA Takumi [Mon, 21 Jan 2013 14:15:24 +0000 (14:15 +0000)]
AST/VTableBuilder.h: Suppress a warning. [-Wunused-private-field]
llvm-svn: 173041
NAKAMURA Takumi [Mon, 21 Jan 2013 14:06:48 +0000 (14:06 +0000)]
R600/SILowerControlFlow.cpp: Fix a warning. [-Wunused-variable]
llvm-svn: 173040
Manuel Klimek [Mon, 21 Jan 2013 13:58:54 +0000 (13:58 +0000)]
Fixes detection of class template specializations.
Now correctly formats:
template <> class A<int> {} a;
llvm-svn: 173038
Evgeniy Stepanov [Mon, 21 Jan 2013 13:30:08 +0000 (13:30 +0000)]
[sanitizer] Disable scanf interceptor on windows.
llvm-svn: 173037
Chandler Carruth [Mon, 21 Jan 2013 13:04:33 +0000 (13:04 +0000)]
Switch CodeMetrics itself over to use TTI to determine if an instruction
is free. The whole CodeMetrics API should probably be reworked more, but
this is enough to allow deleting the duplicate code there for computing
whether an instruction is free.
All of the passes using this have been updated to pull in TTI and hand
it to the CodeMetrics stuff. Further, a dead CodeMetrics API
(analyzeFunction) is nuked for lack of users.
llvm-svn: 173036
Timur Iskhodzhanov [Mon, 21 Jan 2013 13:02:41 +0000 (13:02 +0000)]
First step towards vftable generation with -cxx-abi microsoft PR13231
llvm-svn: 173035
Chandler Carruth [Mon, 21 Jan 2013 12:14:42 +0000 (12:14 +0000)]
Fix indentation and formatting.
This change brought to by clang-format. =]
llvm-svn: 173034
Chandler Carruth [Mon, 21 Jan 2013 12:09:41 +0000 (12:09 +0000)]
Sink InlineCost.cpp into IPA -- it is now officially an interprocedural
analysis. How cute that it wasn't previously. ;]
Part of this confusion stems from the flattened header file tree. Thanks
to Benjamin for pointing out the goof on IRC, and we're considering
un-flattening the headers, so speak now if that would bug you.
llvm-svn: 173033
Chandler Carruth [Mon, 21 Jan 2013 12:05:16 +0000 (12:05 +0000)]
Move the inline cost analysis's primary cost query to TTI instead of the
old CodeMetrics system. TTI has the specific advantage of being
extensible and customizable by targets to reflect target-specific cost
metrics.
llvm-svn: 173032
Chandler Carruth [Mon, 21 Jan 2013 11:55:09 +0000 (11:55 +0000)]
Now that the inline cost analysis is a pass, we can easily have it
depend on and use other analyses (as long as they're either immutable
passes or CGSCC passes of course -- nothing in the pass manager has been
fixed here). Leverage this to thread TargetTransformInfo down through
the inline cost analysis.
No functionality changed here, this just threads things through.
llvm-svn: 173031
Chandler Carruth [Mon, 21 Jan 2013 11:39:18 +0000 (11:39 +0000)]
Make the inline cost a proper analysis pass. This remains essentially
a dynamic analysis done on each call to the routine. However, now it can
use the standard pass infrastructure to reference other analyses,
instead of a silly setter method. This will become more interesting as
I teach it about more analysis passes.
This updates the two inliner passes to use the inline cost analysis.
Doing so highlights how utterly redundant these two passes are. Either
we should find a cheaper way to do always inlining, or we should merge
the two and just fiddle with the thresholds to get the desired behavior.
I'm leaning increasingly toward the latter as it would also remove the
Inliner sub-class split.
llvm-svn: 173030
Chandler Carruth [Mon, 21 Jan 2013 11:39:16 +0000 (11:39 +0000)]
Formatting and comment fixes to the always inliner.
Formatting fixes brought to you by clang-format.
llvm-svn: 173029
Chandler Carruth [Mon, 21 Jan 2013 11:39:14 +0000 (11:39 +0000)]
Clean up the formatting and doxygen for the simple inliner a bit. No
functionality changed.
llvm-svn: 173028
Chandler Carruth [Mon, 21 Jan 2013 11:39:12 +0000 (11:39 +0000)]
Fix an old-style doxygen comment.
llvm-svn: 173027
Alexey Samsonov [Mon, 21 Jan 2013 11:36:38 +0000 (11:36 +0000)]
Fixup for r173021: build tests with zero-based shadow on Android only, fix condition in checking for shadow gap in asan_mapping.h
llvm-svn: 173026
Evgeniy Stepanov [Mon, 21 Jan 2013 11:32:50 +0000 (11:32 +0000)]
[sanitizer] Define va_copy on win32.
llvm-svn: 173025
Dmitri Gribenko [Mon, 21 Jan 2013 11:25:03 +0000 (11:25 +0000)]
Add a fixit for _Noreturn main,
add tests for fixits removing static and inline from main
llvm-svn: 173024
Kostya Serebryany [Mon, 21 Jan 2013 11:20:49 +0000 (11:20 +0000)]
[tsan] fix thread_name.cc test to work with older versions of libc
llvm-svn: 173023
NAKAMURA Takumi [Mon, 21 Jan 2013 10:51:28 +0000 (10:51 +0000)]
CGDebugInfo.cpp: Fix a warning. [-Wunused-variable]
llvm-svn: 173022
Alexey Samsonov [Mon, 21 Jan 2013 10:51:18 +0000 (10:51 +0000)]
ASan: build unit tests with -fsanitize-address-zero-base-shadow on Linux and Android
llvm-svn: 173021
Evgeniy Stepanov [Mon, 21 Jan 2013 10:33:14 +0000 (10:33 +0000)]
[asan] Move *scanf declarations out of mac-only section.
llvm-svn: 173020
Manuel Klimek [Mon, 21 Jan 2013 10:17:14 +0000 (10:17 +0000)]
Allow for nested name specifiers in record declarations.
Now correctly formats:
class A::B {} n;
llvm-svn: 173019