Alexander Kornienko [Wed, 13 Apr 2016 08:46:32 +0000 (08:46 +0000)]
[clang-tidy] add_new_check.py should fail if check name starts with the module name
+ updated formatting
llvm-svn: 266181
Alexey Bader [Wed, 13 Apr 2016 08:33:41 +0000 (08:33 +0000)]
[OpenCL] Move OpenCLImageTypes.def from clangAST to clangBasic library.
Putting OpenCLImageTypes.def to clangAST library violates layering requirement: "It's not OK for a Basic/ header to include an AST/ header".
This fixes the modules build.
Differential revision: http://reviews.llvm.org/D18954
Reviewers: Richard Smith, Vassil Vassilev.
llvm-svn: 266180
Zlatko Buljan [Wed, 13 Apr 2016 08:02:26 +0000 (08:02 +0000)]
[mips][microMIPS] Add CodeGen support for DIV, MOD, DIVU, MODU, DDIV, DMOD, DDIVU and DMODU instructions
Differential Revision: http://reviews.llvm.org/D17137
This patch was reverted after the revertion of dependant patch http://reviews.llvm.org/D17068.
There was the problem with test-suite failure.
The problem is hopefully solved with dependant patch so this patch is commited again.
llvm-svn: 266179
Richard Smith [Wed, 13 Apr 2016 07:47:38 +0000 (07:47 +0000)]
constexpr -> const to appease MSVC bots.
llvm-svn: 266178
NAKAMURA Takumi [Wed, 13 Apr 2016 07:45:10 +0000 (07:45 +0000)]
ASTWriterDecl.cpp: Prune a couple of \param(s), corresponding to r266160. [-Wdocumentation]
llvm-svn: 266177
Richard Smith [Wed, 13 Apr 2016 07:41:35 +0000 (07:41 +0000)]
[modules] Add some missing blockinfo records. No functionality change except to llvm-bcanalyzer output.
llvm-svn: 266176
David Majnemer [Wed, 13 Apr 2016 06:55:52 +0000 (06:55 +0000)]
[InstCombine] We folded an fcmp to an i1 instead of a vector of i1
Remove an ad-hoc transform in InstCombine and replace it with more
general machinery (ValueTracking, InstructionSimplify and VectorUtils).
This fixes PR27332.
llvm-svn: 266175
Mehdi Amini [Wed, 13 Apr 2016 06:32:46 +0000 (06:32 +0000)]
Simplify LTOInternalize into UpdateLLVMCompilerUsed
It is now only doing the update to the llvm.compiler_used global.
The client has to call separately the internalization stage.
Hopefully the code is simpler to understand this way.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266174
Mehdi Amini [Wed, 13 Apr 2016 06:32:29 +0000 (06:32 +0000)]
Minor cleanup in Internalize, hide helper class using anonymous namespace (NFC)
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266173
Mehdi Amini [Wed, 13 Apr 2016 06:32:04 +0000 (06:32 +0000)]
LTOInternalize: Use a StringSet instead of a sorted vector and a binary search query for each function
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266172
Hrvoje Varga [Wed, 13 Apr 2016 06:17:21 +0000 (06:17 +0000)]
[mips][microMIPS] Fix for "Cannot copy registers" assertion
Differential Revision: http://reviews.llvm.org/D17068
This changes contains fix for failing test-suite. So, this patch should hopefully work now.
llvm-svn: 266171
Mehdi Amini [Wed, 13 Apr 2016 05:36:06 +0000 (05:36 +0000)]
Move "ExternalSymbols" out of LTOInternalize (NFC)
This is not really related to internalization per se.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266170
Mehdi Amini [Wed, 13 Apr 2016 05:25:16 +0000 (05:25 +0000)]
Really return whether Internalize did change the Module or not.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266169
Mehdi Amini [Wed, 13 Apr 2016 05:25:12 +0000 (05:25 +0000)]
Modernize Internalizer with for-range loop (NFC)
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266168
Mehdi Amini [Wed, 13 Apr 2016 05:25:08 +0000 (05:25 +0000)]
Refactor the InternalizePass into a helper class, and expose it through a public free function (NFC)
There is really no reason to require to instanciate a pass manager to
internalize.
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266167
Chuang-Yu Cheng [Wed, 13 Apr 2016 05:16:31 +0000 (05:16 +0000)]
[PPC64][VSX] Add a couple of new data types for vec_vsx_ld and vec_vsx_st intrinsics and fix incorrect testcases with minor refactoring
New added data types:
vector double vec_vsx_ld (int, const double *);
vector float vec_vsx_ld (int, const float *);
vector bool short vec_vsx_ld (int, const vector bool short *);
vector bool int vec_vsx_ld (int, const vector bool int *);
vector signed int vec_vsx_ld (int, const signed int *);
vector unsigned int vec_vsx_ld (int, const unsigned int *);
void vec_vsx_st (vector double, int, double *);
void vec_vsx_st (vector float, int, float *);
void vec_vsx_st (vector bool short, int, vector bool short *);
void vec_vsx_st (vector bool short, int, signed short *);
void vec_vsx_st (vector bool short, int, unsigned short *);
void vec_vsx_st (vector bool int, int, vector bool int *);
void vec_vsx_st (vector bool int, int, signed int *);
void vec_vsx_st (vector bool int, int, unsigned int *);
Also fix testcases which use non-vector argument version of vec_vsx_ld or
vec_vsx_st, but pass incorrect parameter.
llvm-svn: 266166
Jason Molenda [Wed, 13 Apr 2016 04:32:49 +0000 (04:32 +0000)]
Update Symtab::InitAddressIndexes so that computed symbol sizes
will not exceed the bounds of their Section. This is addressing a
problem where a file had a large space between two sections that
were not used by this module - the last symbol in the text section
had an enormous size because the distance between that and the first
symbol in the data section were used to compute the size.
http://reviews.llvm.org/D19004
<rdar://problem/
25227945>
llvm-svn: 266165
Oleksiy Vyalov [Wed, 13 Apr 2016 04:21:05 +0000 (04:21 +0000)]
Attempt to fix TestCPPBreakpointLocations on Linux/Android.
llvm-svn: 266164
Mehdi Amini [Wed, 13 Apr 2016 04:20:32 +0000 (04:20 +0000)]
Refactor Internalization pass to use as a callback instead of a StringSet (NFC)
This will save a bunch of copies / initialization of intermediate
datastructure, and (hopefully) simplify the code.
This also abstract the symbol preservation mechanism outside of the
Internalization pass into the client code, which is not forced
to keep a map of strings for instance (ThinLTO will prefere hashes).
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266163
Wei Mi [Wed, 13 Apr 2016 03:08:27 +0000 (03:08 +0000)]
Recommit r265547, and r265610,r265639,r265657 on top of it, plus
two fixes with one about error verify-regalloc reported, and
another about live range update of phi after rematerialization.
r265547:
Replace analyzeSiblingValues with new algorithm to fix its compile
time issue. The patch is to solve PR17409 and its duplicates.
analyzeSiblingValues is a N x N complexity algorithm where N is
the number of siblings generated by reg splitting. Although it
causes siginificant compile time issue when N is large, it is also
important for performance since it removes redundent spills and
enables rematerialization.
To solve the compile time issue, the patch removes analyzeSiblingValues
and replaces it with lower cost alternatives containing two parts. The
first part creates a new spill hoisting method in postOptimization of
register allocation. It does spill hoisting at once after all the spills
are generated instead of inside every instance of selectOrSplit. The
second part queries the define expr of the original register for
rematerializaiton and keep it always available during register allocation
even if it is already dead. It deletes those dead instructions only in
postOptimization. With the two parts in the patch, it can remove
analyzeSiblingValues without sacrificing performance.
Patches on top of r265547:
r265610 "Fix the compare-clang diff error introduced by r265547."
r265639 "Fix the sanitizer bootstrap error in r265547."
r265657 "InlineSpiller.cpp: Escap \@ in r265547. [-Wdocumentation]"
Differential Revision: http://reviews.llvm.org/D15302
Differential Revision: http://reviews.llvm.org/D18934
Differential Revision: http://reviews.llvm.org/D18935
Differential Revision: http://reviews.llvm.org/D18936
llvm-svn: 266162
Peter Collingbourne [Wed, 13 Apr 2016 02:24:48 +0000 (02:24 +0000)]
ELF: Do not create relative relocations for undefined symbols.
We need to ensure that the address of an undefined weak symbol evaluates to
zero. We were getting this right for non-PIC executables (where the symbol
can be evaluated directly) and for DSOs (where we emit a symbolic relocation
for these symbols, as they are preemptible). But we weren't getting it right
for PIEs. Probably the simplest way to ensure that these symbols evaluate
to zero is by not creating a relocation in .got for them.
Differential Revision: http://reviews.llvm.org/D19044
llvm-svn: 266161
Richard Smith [Wed, 13 Apr 2016 02:12:03 +0000 (02:12 +0000)]
[modules] Refactor handling of cases where we write an offset to a prior record into the bitstream and simplify a little, in preparation for doing this in more cases.
llvm-svn: 266160
Mehdi Amini [Wed, 13 Apr 2016 01:52:32 +0000 (01:52 +0000)]
Fix FunctionImport export list computation: need to take a reference to a map entry to actually modify it
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266159
Rafael Espindola [Wed, 13 Apr 2016 01:40:19 +0000 (01:40 +0000)]
Change how we apply relocations.
With this patch we use the first scan over the relocations to remember
the information we found about them: will them be relaxed, will a plt be
used, etc.
With that the actual relocation application becomes much simpler. That
is particularly true for the interfaces in Target.h.
This unfortunately means that we now do two passes over relocations for
non SHF_ALLOC sections. I think this can be solved by factoring out the
code that scans a single relocation. It can then be used both as a scan
that record info and for a dedicated direct relocation of non SHF_ALLOC
sections.
I also think it is possible to reduce the number of enum values by
representing a target with just an OutputSection and an offset (which
can be from the start or end).
This should unblock adding features like relocation optimizations.
llvm-svn: 266158
Devin Coughlin [Wed, 13 Apr 2016 00:41:54 +0000 (00:41 +0000)]
[analyzer] Nullability: Treat nil _Nonnull ivar as invariant violation.
Treat a _Nonnull ivar that is nil as an invariant violation in a similar
fashion to how a nil _Nonnull parameter is treated as a precondition violation.
This avoids warning on defensive returns of nil on defensive internal
checks, such as the following common idiom:
@class InternalImplementation
@interface PublicClass {
InternalImplementation * _Nonnull _internal;
}
-(id _Nonnull)foo;
@end
@implementation PublicClass
-(id _Nonnull)foo {
if (!_internal)
return nil; // no-warning
return [_internal foo];
}
@end
rdar://problem/
24485171
llvm-svn: 266157
Matt Arsenault [Wed, 13 Apr 2016 00:39:52 +0000 (00:39 +0000)]
AMDGPU: Add test for m0 initialization in basic loop
Initialization of m0 is emitted for each LDS operation, so
every block with LDS usage ends up with one. MachineLICM
used to fail to hoist this out of the loop, so every loop
iteration with LDS usage in it would re-initialize it.
This seems to be fixed now, so add a test to make sure that
it stays this way.
llvm-svn: 266156
Matt Arsenault [Wed, 13 Apr 2016 00:39:48 +0000 (00:39 +0000)]
AMDGPU: Remove leftover ShaderType attributes in tests
llvm-svn: 266155
NAKAMURA Takumi [Wed, 13 Apr 2016 00:22:25 +0000 (00:22 +0000)]
clang/test/Driver/cl-options.c: Fix an expression to recognize dos r'\\'.
llvm-svn: 266154
Mehdi Amini [Tue, 12 Apr 2016 23:58:30 +0000 (23:58 +0000)]
LTOInternalize: Fix member type, should be a reference and not a copy
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266153
Tom Stellard [Tue, 12 Apr 2016 23:57:30 +0000 (23:57 +0000)]
AMDGPU/SI: Fix spilling of 96-bit registers
Summary:
It seems like this was broken in r252327. I thought we had test cases
for this, but it's really hard to tirgger spills of this exact register
size since they aren't used very much.
Reviewers: arsenm, nhaehnle
Subscribers: nhaehnle, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19021
llvm-svn: 266152
Mehdi Amini [Tue, 12 Apr 2016 23:25:11 +0000 (23:25 +0000)]
Fix mismatch on returned type between header and implementation for createNameAnonFunctionPass()
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266151
Justin Bogner [Tue, 12 Apr 2016 23:21:53 +0000 (23:21 +0000)]
CodeGen: Clear the MFI's save and restore point after PrologEpilogInserter
This state is no longer useful and not guaranteed to be valid in later
codegen passes. For example, see the added test, which would print a
savepoint of %bb.-1 without this change, and crashes with a
use-after-free error under ASan if you apply the recycling allocator
patch from llvm.org/PR26808.
llvm-svn: 266150
Jeroen Ketema [Tue, 12 Apr 2016 23:21:46 +0000 (23:21 +0000)]
Add space between words in verify-scev-maps option help message
llvm-svn: 266149
Sanjay Patel [Tue, 12 Apr 2016 23:16:23 +0000 (23:16 +0000)]
[x86, InstCombine] fix masked load pass-through operand to be a zero vector
This bug was introduced with:
http://reviews.llvm.org/rL262269
AVX masked loads are specified to set vector lanes to zero when the high bit of the mask
element for that lane is zero:
"If the mask is 0, the corresponding data element is set to zero in the load form of these
instructions, and unmodified in the store form." --Intel manual
Differential Revision: http://reviews.llvm.org/D19017
llvm-svn: 266148
Akira Hatanaka [Tue, 12 Apr 2016 23:10:58 +0000 (23:10 +0000)]
[ObjC] Pop all cleanups created in EmitObjCForCollectionStmt before
exiting the for-in loop.
This commit fixes a bug where EmitObjCForCollectionStmt didn't pop
cleanups for captures.
For example, in the following for-in loop, a block which captures self
is passed to foo1:
for (id x in [self foo1:^{ use(self); }]) {
use(x);
break;
}
Previously, the code in EmitObjCForCollectionStmt wouldn't pop the
cleanup for the captured self before exiting the loop, which caused
code-gen to generate an IR in which objc_release was called twice on
the captured self.
This commit fixes the bug by entering a RunCleanupsScope before the
loop condition is evaluated and forcing its cleanup before exiting the
loop.
rdar://problem/
16865751
Differential Revision: http://reviews.llvm.org/D18618
llvm-svn: 266147
Manman Ren [Tue, 12 Apr 2016 23:01:55 +0000 (23:01 +0000)]
ObjC class properties: add diagnostics for unimplemented class properties.
rdar://
24711047
llvm-svn: 266146
Adrian McCarthy [Tue, 12 Apr 2016 22:45:03 +0000 (22:45 +0000)]
Fix breakpoint_set_restart test for Windows
When run with the multiprocess test runner, the getchar() trick doesn't work, so ninja check-lldb would fail on this test, but running the test directly worked fine.
Differential Revision: http://reviews.llvm.org/D19035
llvm-svn: 266145
Evandro Menezes [Tue, 12 Apr 2016 22:42:36 +0000 (22:42 +0000)]
[AArch64] Fuse AES{D,E}/AESMC for Exynos M1. (NFC)
llvm-svn: 266144
Rafael Espindola [Tue, 12 Apr 2016 22:36:05 +0000 (22:36 +0000)]
Add test for the interaction of copy relocations and non alloc sections.
llvm-svn: 266143
James Y Knight [Tue, 12 Apr 2016 22:32:47 +0000 (22:32 +0000)]
Pre-fill LibcallRoutineNames with nullptr.
And rearrange InitLibcallNames slightly.
llvm-svn: 266142
Rafael Espindola [Tue, 12 Apr 2016 22:08:23 +0000 (22:08 +0000)]
Add a few more cases to this test.
It now documents how we currently handle relocations for
alloc/non-alloc.
llvm-svn: 266141
Davide Italiano [Tue, 12 Apr 2016 22:05:38 +0000 (22:05 +0000)]
[COFF] Simplify the code leveraging implicit conversion.
Suggested by: David Blaikie!
llvm-svn: 266140
Greg Clayton [Tue, 12 Apr 2016 22:02:37 +0000 (22:02 +0000)]
Fixed being able to set breakpoints on destructors when we don't fully specify the demangled name. So all of the following now work:
(lldb) b ~Foo
(lldb) b Foo::~Foo
(lldb) b Bar::Foo::~Foo
Improved out C++ breakpoint locations tests as well to cover this issue.
<rdar://problem/
25577252>
llvm-svn: 266139
Enrico Granata [Tue, 12 Apr 2016 21:57:11 +0000 (21:57 +0000)]
Use the FormatEntity work for great good - parse summary strings before accepting them, and fail to add any strings that fail parsing
llvm-svn: 266138
James Y Knight [Tue, 12 Apr 2016 21:54:57 +0000 (21:54 +0000)]
Update psabi link for x86-64. Add link to linux gabi supplement.
llvm-svn: 266137
David Blaikie [Tue, 12 Apr 2016 21:45:53 +0000 (21:45 +0000)]
[MC/ELFObjectWriter] Fix indentation of class body.
llvm-svn: 266136
David L Kreitzer [Tue, 12 Apr 2016 21:45:09 +0000 (21:45 +0000)]
Fixed a few typos and formatting problems. NFCI.
llvm-svn: 266135
Davide Italiano [Tue, 12 Apr 2016 21:44:16 +0000 (21:44 +0000)]
[DebugInfo] Add error message to test.
Suggested by Rafael as post-commit review (r266102).
llvm-svn: 266134
Konstantin Zhuravlyov [Tue, 12 Apr 2016 21:42:15 +0000 (21:42 +0000)]
[AMDGPU] Add debugger related target options
Differential Revision: http://reviews.llvm.org/D18748
llvm-svn: 266133
Mehdi Amini [Tue, 12 Apr 2016 21:35:28 +0000 (21:35 +0000)]
Add a pass to name anonymous/nameless function
Summary:
For correct handling of alias to nameless
function, we need to be able to refer them through a GUID in the summary.
Here we name them using a hash of the non-private global names in the module.
Reviewers: tejohnson
Subscribers: joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D18883
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266132
Mehdi Amini [Tue, 12 Apr 2016 21:35:18 +0000 (21:35 +0000)]
Move summary creation out of llvm-as into opt
Summary:
Let keep llvm-as "dumb": it converts textual IR to bitcode. This
commit removes the dependency from llvm-as to libLLVMAnalysis.
We'll add back summary in llvm-as if we get to a textual
representation for it at some point. In the meantime, opt seems
like a better place for that.
Reviewers: tejohnson
Subscribers: joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D19032
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266131
Justin Bogner [Tue, 12 Apr 2016 21:34:24 +0000 (21:34 +0000)]
X86: Avoid accessing SDValues after they've been RAUW'd
This fixes two use-after-frees in selectLEA64_32Addr. If matchAddress
matches an ADD with an AND as an operand, and that AND hits one of the
"heroic transforms" that folds masks and shifts, we end up with N
pointing to an SDNode that was deleted. Make sure we're done accessing
it before that.
Found by ASan with the recycling allocator changes in llvm.org/PR26808.
llvm-svn: 266130
Enrico Granata [Tue, 12 Apr 2016 21:26:48 +0000 (21:26 +0000)]
Cleanup the arguments for 'memory find' such that the help system reflects the real way to invoke it
llvm-svn: 266129
JF Bastien [Tue, 12 Apr 2016 21:23:05 +0000 (21:23 +0000)]
NFC: MergeFunctions return early
Same effect, easier to read.
llvm-svn: 266128
David Blaikie [Tue, 12 Apr 2016 21:22:48 +0000 (21:22 +0000)]
Add a fixme for an old patch I had lying around that I'm not going to finish any time so n
llvm-svn: 266127
Nicolai Haehnle [Tue, 12 Apr 2016 21:18:10 +0000 (21:18 +0000)]
AMDGPU: add llvm.amdgcn.buffer.load/store intrinsics
Summary:
They correspond to BUFFER_LOAD/STORE_DWORD[_X2,X3,X4] and mostly behave like
llvm.amdgcn.buffer.load/store.format. They will be used by Mesa for SSBO and
atomic counters at least when robust buffer access behavior is desired.
(These instructions perform no format conversion and do buffer range checking
per component.)
As a side effect of sharing patterns with llvm.amdgcn.buffer.store.format,
it has become trivial to add support for the f32 and v2f32 variants of that
intrinsic, so the patch does so.
Also DAG-ify (and fix) some tests that I noticed intermittent failures in
while developing this patch.
Some tests were (temporarily) adjusted for the required mayLoad/hasSideEffects
changes to the BUFFER_STORE_DWORD* instructions. See also
http://reviews.llvm.org/D18291.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D18292
llvm-svn: 266126
Teresa Johnson [Tue, 12 Apr 2016 21:13:11 +0000 (21:13 +0000)]
[ThinLTO] Only compute imports for current module in FunctionImport pass
Summary:
The function import pass was computing all the imports for all the
modules in the index, and only using the imports for the current module.
Change this to instead compute only for the given module. This means
that the exports list can't be populated, but they weren't being used
anyway.
Longer term, the linker can collect all the imports and export lists
and serialize them out for consumption by the distributed backend
processes which use this pass.
Reviewers: joker.eph
Subscribers: llvm-commits, joker.eph
Differential Revision: http://reviews.llvm.org/D18945
llvm-svn: 266125
JF Bastien [Tue, 12 Apr 2016 21:13:01 +0000 (21:13 +0000)]
NFC: MergeFunctions update more comments
They are wordy. Some words were wrong.
llvm-svn: 266124
Ed Maste [Tue, 12 Apr 2016 21:11:46 +0000 (21:11 +0000)]
Always use --eh-frame-hdr on FreeBSD, even for -static
FreeBSD uses LLVM's libunwind on FreeBSD/arm64 today (and is expected to
use it more widely in the future), and it requires the EH frame segment
in static binaries.
This is the same as r203742 for NetBSD.
Differential Revision: http://reviews.llvm.org/D19029
llvm-svn: 266123
Eric Christopher [Tue, 12 Apr 2016 21:08:54 +0000 (21:08 +0000)]
Add a couple of missing vsx load and store intrinsics.
Patch by Jing Yu!
llvm-svn: 266122
Peter Collingbourne [Tue, 12 Apr 2016 20:41:42 +0000 (20:41 +0000)]
ELF: Give automatically generated __start_* and __stop_* symbols hidden visibility.
These symbols describe a property of a linkage unit, so it seems reasonable
to limit their visibility to the linkage unit. Furthermore the use cases I
am aware of do not require more than hidden visibility.
This is a departure from the behavior of the bfd and gold linkers. However,
it is unclear that the decision to give these symbols default visibility
in those linkers was made deliberately. The __start_*/__stop_* feature
was added to the bfd linker in 1994 [1], while the visibility feature was
added about five years later [2], so it may have been that the visibility
of these symbols was not considered. The feature was implemented in gold
[3] in the same way; the behavior may have simply been copied from bfd.
The only related discussion I could find on the binutils mailing list [4]
was a user issue which would most likely not have occurred if the symbols
had hidden visibility.
[1] https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=
5efddb2e7c3229b569a862205f61d42860af678b
[2] https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=
0fc731e447cd01e7fc35197b487ff0e4fd25afca
[3] https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=
bfd58944a64b0997a310b95fbe0423338961e71c
[4] https://sourceware.org/ml/binutils/2014-05/msg00011.html
Differential Revision: http://reviews.llvm.org/D19024
llvm-svn: 266121
Jonathan Peyton [Tue, 12 Apr 2016 20:37:18 +0000 (20:37 +0000)]
Add declarations of OpenMP 4.5 target/offload routines to headers
All these routines are implemented in the offload library.
llvm-svn: 266120
Simon Atanasyan [Tue, 12 Apr 2016 20:37:00 +0000 (20:37 +0000)]
[ELF][MIPS] Add test case to check using GP0 value in relocation calculation. NFC.
llvm-svn: 266119
Greg Clayton [Tue, 12 Apr 2016 20:26:41 +0000 (20:26 +0000)]
Revert to using libdispatch to reap threads on MacOSX. Code was accidentally checked in that is now reverted.
<rdar://problem/
25643874>
llvm-svn: 266118
Yaxun Liu [Tue, 12 Apr 2016 20:22:32 +0000 (20:22 +0000)]
Pass -backend-option to LLVM when there is no target machine.
Clang should pass -backend-option to LLVM even though there is no target machine, since LLVM passes are used when emitting LLVM IR.
Differential Revision: http://reviews.llvm.org/D17552
llvm-svn: 266117
Richard Smith [Tue, 12 Apr 2016 20:20:33 +0000 (20:20 +0000)]
[modules] Extend r266113 to cope with submodules.
llvm-svn: 266116
James Y Knight [Tue, 12 Apr 2016 20:18:48 +0000 (20:18 +0000)]
Add __atomic_* lowering to AtomicExpandPass.
(Recommit of r266002, with r266011, r266016, and not accidentally
including an extra unused/uninitialized element in LibcallRoutineNames)
AtomicExpandPass can now lower atomic load, atomic store, atomicrmw, and
cmpxchg instructions to __atomic_* library calls, when the target
doesn't support atomics of a given size.
This is the first step towards moving all atomic lowering from clang
into llvm. When all is done, the behavior of __sync_* builtins,
__atomic_* builtins, and C11 atomics will be unified.
Previously LLVM would pass everything through to the ISelLowering
code. There, unsupported atomic instructions would turn into __sync_*
library calls. Because of that behavior, Clang currently avoids emitting
llvm IR atomic instructions when this would happen, and emits __atomic_*
library functions itself, in the frontend.
This change makes LLVM able to emit __atomic_* libcalls, and thus will
eventually allow clang to depend on LLVM to do the right thing.
It is advantageous to do the new lowering to atomic libcalls in
AtomicExpandPass, before ISel time, because it's important that all
atomic operations for a given size either lower to __atomic_*
libcalls (which may use locks), or native instructions which won't. No
mixing and matching.
At the moment, this code is enabled only for SPARC, as a
demonstration. The next commit will expand support to all of the other
targets.
Differential Revision: http://reviews.llvm.org/D18200
llvm-svn: 266115
Derek Schuff [Tue, 12 Apr 2016 20:12:05 +0000 (20:12 +0000)]
[WebAssembly] Fix debug info in reg-stackify.ll test
It lacked a CU and thus became invalid with r266102
llvm-svn: 266114
Richard Smith [Tue, 12 Apr 2016 19:58:30 +0000 (19:58 +0000)]
[modules] When an incompatible module file is explicitly provided for a module,
and we fall back to textual inclusion, don't require the module as a whole to
be marked available; it's OK if some other file in the same module is missing,
just as it would be if the header were explicitly marked textual.
llvm-svn: 266113
JF Bastien [Tue, 12 Apr 2016 19:44:40 +0000 (19:44 +0000)]
Delete mergefunctions.clang.svn.patch
The patch doesn't apply, and was removed from zorg by rL266094.
llvm-svn: 266112
Yaxun Liu [Tue, 12 Apr 2016 19:43:36 +0000 (19:43 +0000)]
PR19957: [OpenCL] Incorrectly accepts implicit address space conversion with ternary operator.
Generates addrspacecast instead of bitcast for ternary operator when necessary, and diagnose ternary operator with incompatible second and third operands.
https://llvm.org/bugs/show_bug.cgi?id=19957
Differential Revision: http://reviews.llvm.org/D17412
llvm-svn: 266111
Devin Coughlin [Tue, 12 Apr 2016 19:29:52 +0000 (19:29 +0000)]
[analyzer] Nullability: Suppress return diagnostics in inlined functions.
The nullability checker can sometimes miss detecting nullability precondition
violations in inlined functions because the binding for the parameter
that violated the precondition becomes dead before the return:
int * _Nonnull callee(int * _Nonnull p2) {
if (!p2)
// p2 becomes dead here, so binding removed.
return 0; // warning here because value stored in p2 is symbolic.
else
return p2;
}
int *caller(int * _Nonnull p1) {
return callee(p1);
}
The fix, which is quite blunt, is to not warn about null returns in inlined
methods/functions. This won’t lose much coverage for ObjC because the analyzer
always analyzes each ObjC method at the top level in addition to inlined. It
*will* lose coverage for C — but there aren’t that many codebases with C
nullability annotations.
rdar://problem/
25615050
llvm-svn: 266109
Nico Weber [Tue, 12 Apr 2016 19:04:37 +0000 (19:04 +0000)]
clang-cl: Remove -isystem, add -imsvc.
r260990 exposed -isystem in clang-cl. -isystem adds a directory to the front of
the system include search path. The idea was to use this to point to a hermetic
msvc install, but as it turns out this doesn't work: -isystem then adds the
hermetic headers in front of clang's builtin headers, and clang's headers that
are supposed to wrap msvc headers (say, stdarg.h) aren't picked up at all
anymore.
So revert that, and instead expose -imsvc which works as if the passed
directory was part of %INCLUDE%: The header is treated as a system header, but
it is searched after clang's lib/Header headers.
Fixes half of PRPR26751.
llvm-svn: 266108
Yaxun Liu [Tue, 12 Apr 2016 19:03:49 +0000 (19:03 +0000)]
[OpenCL] Handle AddressSpaceConversion when target address space does not change.
In codegen different address spaces may be mapped to the same address
space for a target, e.g. in x86/x86-64 all address spaces are mapped
to 0. Therefore AddressSpaceConversion should be translated by
CreatePointerBitCastOrAddrSpaceCast instead of CreateAddrSpaceCast.
Differential Revision: http://reviews.llvm.org/D18713
llvm-svn: 266107
Evgeniy Stepanov [Tue, 12 Apr 2016 18:48:41 +0000 (18:48 +0000)]
[safestack] Fix stack canary test on Mac.
Disable FORTIFY_SOURCE and explicitly disable stack protector in the
no-stack-protector run.
llvm-svn: 266106
Tom Stellard [Tue, 12 Apr 2016 18:40:43 +0000 (18:40 +0000)]
AMDGPU/SI: Insert wait states required after v_readfirstlane on SI
Summary:
We will be able to handle this case much better once the hazard recognizer
is finished, but this conservative implementation fixes a hang with the piglit
test:
spec/arb_arrays_of_arrays/execution/sampler/fs-nested-struct-arrays-nonconst-nested-arra
Reviewers: arsenm, nhaehnle
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D18988
llvm-svn: 266105
Matt Arsenault [Tue, 12 Apr 2016 18:24:38 +0000 (18:24 +0000)]
AMDGPU: Eliminate half of i64 or if one operand is zero_extend from i32
This helps clean up some of the mess when expanding unaligned 64-bit
loads when changed to be promote to v2i32, and fixes situations
where or x, 0 was emitted after splitting 64-bit ors during moveToVALU.
I think this could be a generic combine but I'm not sure.
llvm-svn: 266104
Enrico Granata [Tue, 12 Apr 2016 18:23:18 +0000 (18:23 +0000)]
Initialize the Python script interpreter lazily (i.e. not at debugger startup)
This time it should also pass the gtests
llvm-svn: 266103
Davide Italiano [Tue, 12 Apr 2016 18:22:33 +0000 (18:22 +0000)]
[IR/Verifier] Each DISubprogram with isDefinition: true must belong to a CU.
Add a check to catch violations. ~60 tests were broken and prevented
this change to be committed. Adrian and I (thanks Adrian!) went
through them in the last week or so updating. The check can be
done more efficiently but I'd still like to get this in ASAP to
avoid more broken tests to be checked in (if any).
PR: 27101
llvm-svn: 266102
Matt Arsenault [Tue, 12 Apr 2016 18:17:23 +0000 (18:17 +0000)]
APInt: Add overload of isMask
This mimics the version in MathExtras.h which isn't testing for a
specific mask size.
llvm-svn: 266101
Ahmed Bougacha [Tue, 12 Apr 2016 18:15:39 +0000 (18:15 +0000)]
[CodeGen] Remove constant-folding dead code. NFC.
This code was specific to vector operations with scalar operands:
all the opcodes in FoldValue (via FoldConstantArithmetic) can't
match those criteria.
Replace it with an assert if that ever changes: at that point,
we might need to add back a splat BUILD_VECTOR.
llvm-svn: 266100
JF Bastien [Tue, 12 Apr 2016 18:06:55 +0000 (18:06 +0000)]
Check alloca's special state
Following up to a similar fix in MergeFunctions: r266022. This patch keeps both in sync, it would be nice to not have to do this. It doesn't look like there's an easy way to test this code directly at the moment: AFAICT all currect uses of isSameOperationAs are looking at instructions deep inside a function. IndVarSimplify/pr24952.ll and InstMerge/st_sink_* look at alloca inadvertently but are brittle tests.
llvm-svn: 266099
Philip Reames [Tue, 12 Apr 2016 18:05:10 +0000 (18:05 +0000)]
Introduce an GCRelocateInst class [NFC]
Previously, we were using isGCRelocate predicates. Using a subclass of IntrinsicInst is far more idiomatic. The refactoring also enables a couple of minor simplifications and code sharing.
llvm-svn: 266098
Sanjay Patel [Tue, 12 Apr 2016 18:01:48 +0000 (18:01 +0000)]
fix indentation; NFC
llvm-svn: 266097
Johannes Doerfert [Tue, 12 Apr 2016 17:54:29 +0000 (17:54 +0000)]
[FIX] Check the invalid context agains the context to rule out SCoPs
llvm-svn: 266096
Evgeniy Stepanov [Tue, 12 Apr 2016 17:51:59 +0000 (17:51 +0000)]
Stricter checks in the stack-protector codegen test.
llvm-svn: 266095
Jim Ingham [Tue, 12 Apr 2016 17:17:35 +0000 (17:17 +0000)]
Breakpoint conditions were making result variables, which they should not do.
The result variables aren't useful, and if you have a breakpoint on a
common function you can generate a lot of these. So I changed the
code that checks the condition to set ResultVariableIsInternal in the
EvaluateExpressionOptions that we pass to the execution.
Unfortunately, the check for this variable was done in the wrong place
(the static UserExpression::Evaluate) which is not how breakpoint
conditions execute expressions (UserExpression::Execute). So I moved
the check to UserExpression::Execute (which Evaluate also calls) and made the
overridden method DoExecute.
llvm-svn: 266093
Jim Ingham [Tue, 12 Apr 2016 17:04:12 +0000 (17:04 +0000)]
'int' is reported as an exception on OS X not as a signal. I don't think
this test ever succeeded on OS X.
llvm-svn: 266092
Nico Weber [Tue, 12 Apr 2016 16:52:30 +0000 (16:52 +0000)]
Revert 266090, needs more testing first.
llvm-svn: 266091
Nico Weber [Tue, 12 Apr 2016 16:38:07 +0000 (16:38 +0000)]
clang-cl: Expose -nostdlibinc.
llvm-svn: 266090
David Majnemer [Tue, 12 Apr 2016 16:33:53 +0000 (16:33 +0000)]
[FileManager] Don't crash if reading from stdin and stat(".") fails
addAncestorsAsVirtualDirs("<stdin>") quickly returns without doing work
because "<stdin>" has no parent_path. This violates the expectation
that a subsequent call to getDirectoryFromFile("<stdin>") would succeed.
Instead, it fails because it uses the "." if the file has no path
component.
Fix this by keeping the behavior between addAncestorsAsVirtualDirs and
getDirectoryFromFile symmetric.
llvm-svn: 266089
Nicolai Haehnle [Tue, 12 Apr 2016 16:10:38 +0000 (16:10 +0000)]
AMDGPU/SI: Fix a mis-compilation of multi-level breaks
Summary:
Under certain circumstances, multi-level breaks (or what is understood by
the control flow passes as such) could be miscompiled in a way that causes
infinite loops, by emitting incorrect control flow intrinsics.
This fixes a hang in
dEQP-GLES3.functional.shaders.loops.while_dynamic_iterations.conditional_continue_vertex
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D18967
llvm-svn: 266088
Johannes Doerfert [Tue, 12 Apr 2016 16:09:44 +0000 (16:09 +0000)]
Do not by default minimize remarks
We used checks to minimize the number of remarks we present to a user
but these checks can become expensive, especially since all wrapping
assumptions are emitted separately. Because there is not benefit for a
"headless" run we put these checks under a command line flag. Thus, if
the flag is not given we will emit "non-effective" remarks, e.g.,
duplicates and revert to the old behaviour if it is given. As this
also changes the internal representation of some sets we set the flag
by default for our unit tests.
llvm-svn: 266087
Artur Pilipenko [Tue, 12 Apr 2016 15:58:04 +0000 (15:58 +0000)]
Support arbitrary addrspace pointers in masked load/store intrinsics
This is a resubmittion of 263158 change.
This patch fixes the problem which occurs when loop-vectorize tries to use @llvm.masked.load/store intrinsic for a non-default addrspace pointer. It fails with "Calling a function with a bad signature!" assertion in CallInst constructor because it tries to pass a non-default addrspace pointer to the pointer argument which has default addrspace.
The fix is to add pointer type as another overloaded type to @llvm.masked.load/store intrinsics.
Reviewed By: reames
Differential Revision: http://reviews.llvm.org/D17270
llvm-svn: 266086
Davide Italiano [Tue, 12 Apr 2016 15:51:23 +0000 (15:51 +0000)]
[Bitcode] Fix + regenerate old test so that it includes a DICompileUnit.
llvm-svn: 266085
Geoff Berry [Tue, 12 Apr 2016 15:50:19 +0000 (15:50 +0000)]
[ScheduleDAGInstrs] Handle instructions with multiple MMOs
Summary:
In getUnderlyingObjectsForInstr(): Don't give up on instructions with
multiple MMOs, instead look through all the MMOs and if they all meet
the conservative criteria previously used for single MMO instructions,
then return all of the underlying objects derived from the MMOs.
The change to ScheduleDAGInstrs::buildSchedGraph() is needed to avoid
the case where multiple underlying objects are present and are related
in such a way that successive iterations of the loop end up adding a
dependency from an instruction to itself.
Reviewers: atrick, hfinkel
Subscribers: MatzeB, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D18093
llvm-svn: 266084
Yaxun Liu [Tue, 12 Apr 2016 15:46:24 +0000 (15:46 +0000)]
Verify commit right by adding a blank line to test/CodeGenOpenCL/address-spaces-conversions.cl.
llvm-svn: 266083
Than McIntosh [Tue, 12 Apr 2016 15:35:05 +0000 (15:35 +0000)]
Test commit, NFC.
Adds a blank line.
llvm-svn: 266082
Petar Jovanovic [Tue, 12 Apr 2016 15:28:16 +0000 (15:28 +0000)]
[mips] add assembler support for .set arch=octeon
This patch enables assembler support for .set arch=octeon.
It will fix issues with inline assembler when this directive is used.
Patch by Strahinja Petrovic.
Differential Revision: http://reviews.llvm.org/D18548
llvm-svn: 266081
Kuba Brecka [Tue, 12 Apr 2016 15:18:11 +0000 (15:18 +0000)]
[tsan] Fix a crash with dispatch_source_set_cancel_handler(NULL) on OS X
We need to handle the case when handler is NULL in dispatch_source_set_cancel_handler and similar interceptors.
Differential Revision: http://reviews.llvm.org/D18968
llvm-svn: 266080