Kuba Brecka [Fri, 7 Nov 2014 22:27:12 +0000 (22:27 +0000)]
Use @rpath for ASan dylib in Makefiles
Reviewed at http://reviews.llvm.org/D6176
llvm-svn: 221556
Chad Rosier [Fri, 7 Nov 2014 22:12:57 +0000 (22:12 +0000)]
[Reassociate] Better preserve NSW/NUW flags.
Part of PR12985.
Phabricator Revision: http://reviews.llvm.org/D6172
llvm-svn: 221555
Saleem Abdulrasool [Fri, 7 Nov 2014 22:09:52 +0000 (22:09 +0000)]
Transforms: use typedef rather than using aliases
Visual Studio 2012 apparently does not support using alias declarations. Use
the more traditional typedef approach. This should let the Windows buildbots
pass. NFC.
llvm-svn: 221554
Nick Kledzik [Fri, 7 Nov 2014 22:00:26 +0000 (22:00 +0000)]
[mach-o] remove stray debug output
llvm-svn: 221553
Nick Kledzik [Fri, 7 Nov 2014 22:00:23 +0000 (22:00 +0000)]
[mach-o] Fix MachOFileNode to own archives same as ELFFileNode
My previous fix to have FileArchive own the member MemoryBuffers was not a
complete solution for darwin because nothing owned the FileArchive object.
Fixed MachOFileNode to be like ELFFileNode and have the graph node own the
archive object.
llvm-svn: 221552
Renato Golin [Fri, 7 Nov 2014 21:55:11 +0000 (21:55 +0000)]
PPC fix works for ARM, too
llvm-svn: 221551
David Peixotto [Fri, 7 Nov 2014 21:44:18 +0000 (21:44 +0000)]
Safely generate new loop metadata node
Polly was accidently modifying a debug info metadata node when
attempting to generate a new unique metadata node for the loop id.
The problem was that we had dwarf metadata that referred to a
metadata node with a null value, like this:
!6 = ... some dwarf metadata referring to !7 ...
!7 = {null}
When we attempt to generate a new metadata node, we reserve the
first space for self-referential node by setting the first argument
to null and then mutating the node later to refer to itself.
However, because the nodes are uniqued based on pointer values, when
we get the new metadata node it actually referred to an existing
node (!7 in the example). When we went to modify the metadata to
point to itself, we were accidently mutating the dwarf metatdata. We
ended up in this situation:
!6 = ... some dwarf metadata referring to !7 ...
!7 = {!7}
and this causes an assert when generating the debug info. The fix is
simple, we just need to use a unique value when getting a new
metadata node. The MDNode::getTemporary() provides exactly the API
we need (and it is used in clang to generate the unique nodes).
Differential Revision: http://reviews.llvm.org/D6174
llvm-svn: 221550
Rafael Espindola [Fri, 7 Nov 2014 21:33:09 +0000 (21:33 +0000)]
Remove unused variable.
llvm-svn: 221549
Saleem Abdulrasool [Fri, 7 Nov 2014 21:32:08 +0000 (21:32 +0000)]
Transform: add SymbolRewriter pass
This introduces the symbol rewriter. This is an IR->IR transformation that is
implemented as a CodeGenPrepare pass. This allows for the transparent
adjustment of the symbols during compilation.
It provides a clean, simple, elegant solution for symbol inter-positioning. This
technique is often used, such as in the various sanitizers and performance
analysis.
The control of this is via a custom YAML syntax map file that indicates source
to destination mapping, so as to avoid having the compiler to know the exact
details of the source to destination transformations.
llvm-svn: 221548
Michael J. Spencer [Fri, 7 Nov 2014 21:30:36 +0000 (21:30 +0000)]
Fix style.
llvm-svn: 221547
Michael J. Spencer [Fri, 7 Nov 2014 21:30:32 +0000 (21:30 +0000)]
Fix style.
llvm-svn: 221546
Nick Kledzik [Fri, 7 Nov 2014 21:01:21 +0000 (21:01 +0000)]
[mach-o] Add support for -order_file option
The darwin linker lets you rearrange functions and data for better locality
(less paging). You do this with the -order_file option which supplies a text
file containing one symbol per line.
Implementing this required a small change to LayoutPass to add a custom sorter
hook.
llvm-svn: 221545
Nick Kledzik [Fri, 7 Nov 2014 20:52:38 +0000 (20:52 +0000)]
Fix FileArchive member MemoryBuffer early destruction
When FileArchive loads a member, it instantiates a temporary MemoryBuffer
which points to the member range of the archive file. The problem is that the
object file parsers call getBufferIndentifer() on that temporary MemoryBuffer
and store that StringRef as the _path data member for that lld::File. When
FileArchive::instantiateMember() goes out of scope the MemoryBuffer is deleted
and the File::._path becomes a dangling reference.
The fix adds a vector<> to FileArchive to own the instantiated MemoryBuffers.
In addition it fixes member's path to be the standard format
(e.g. "/path/libfoo.a(foo.o)") instead of just the leaf name.
llvm-svn: 221544
Matt Arsenault [Fri, 7 Nov 2014 20:45:00 +0000 (20:45 +0000)]
R600: Remove unused define
llvm-svn: 221543
Jay Foad [Fri, 7 Nov 2014 20:42:11 +0000 (20:42 +0000)]
[ASan] Fix zero_page_pc test for PowerPC
Summary:
In the Power architecture, all branch instructions ignore the 2 least
significant bits of the target address. Consequently if you branch to an
invalid address, the address reported along with the SIGSEGV will have
been rounded down to a multiple of 4. Tweak this test accordingly.
This may fix the test for ARM too, in which case we could remove
the XFAIL, but I have no way of testing that.
Reviewers: kcc, willschm, glider, samsonov
Reviewed By: samsonov
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6167
llvm-svn: 221542
Enrico Granata [Fri, 7 Nov 2014 20:37:17 +0000 (20:37 +0000)]
This was meant to be count, not m_count
llvm-svn: 221541
Enrico Granata [Fri, 7 Nov 2014 20:22:18 +0000 (20:22 +0000)]
SBValue::Cast is deprecated. This does not work as reliably as free-form casting via the expression evaluator. Some of our clients depend on it, so we can't remove it. But any new adopters should favor the expression parser as the way forward
llvm-svn: 221540
Simon Atanasyan [Fri, 7 Nov 2014 19:32:08 +0000 (19:32 +0000)]
[Mips] Do not hard-code the paired relocation type
Request `getPairRelocation()` function to get paired relocation type.
That allows us to look up another pairs like R_MICROMIPS_HI16/LO16
in the future.
llvm-svn: 221539
Rafael Espindola [Fri, 7 Nov 2014 18:07:04 +0000 (18:07 +0000)]
Don't redeclare a pure virtual method.
I.E., there is no value is having
void foo() override = 0;
If it is override it is already present in a base class. Since it is pure,
some other class will have to implement it.
llvm-svn: 221537
Rafael Espindola [Fri, 7 Nov 2014 17:59:05 +0000 (17:59 +0000)]
Use StringRefMemoryObject in llvm-mc. NFC.
llvm-svn: 221536
Samuel Antao [Fri, 7 Nov 2014 17:48:03 +0000 (17:48 +0000)]
Fix clash of gcc toolchains in hexagon driver regression tests.
If clang was configured with a custom gcc toolchain (either by using GCC_INSTALL_PREFIX in cmake or the equivalent configure command), the path to the custom gcc toolchain path takes precedence to the one specified by -ccc-install-dir. This causes several regression tests to fail as they will be using an unexpected path. Adding the switch --gcc-toolchain="" in each test command is not enough as the hexagon toolchain implementation in the driver is not evaluating this argument. This commit modifies the hexagon toolchain to take the --gcc-toolchain="" argument into account when deciding the toolchain path, similarly to what is already done for other targets toolchains. Additionally, the faulty regression tests are modified in order to --gcc-toolchain="" be passed to the commands.
llvm-svn: 221535
Daniel Sanders [Fri, 7 Nov 2014 16:54:21 +0000 (16:54 +0000)]
[mips] Promote i32 arguments to i64 for the N32/N64 ABI and fix <64-bit structs...
Summary:
... and after all that refactoring, it's possible to distinguish softfloat
floating point values from integers so this patch no longer breaks softfloat to
do it.
Remove direct handling of i32's in the N32/N64 ABI by promoting them to
i64. This more closely reflects the ABI documentation and also fixes
problems with stack arguments on big-endian targets.
We now rely on signext/zeroext annotations (already generated by clang) and
the Assert[SZ]ext nodes to avoid the introduction of unnecessary sign/zero
extends.
It was not possible to convert three tests to use signext/zeroext. These tests
are bswap.ll, ctlz-v.ll, ctlz-v.ll. It's not possible to put signext on a
vector type so we just accept the sign extends here for now. These tests don't
pass the vectors the same way clang does (clang puts multiple elements in the
same argument, these map 1 element to 1 argument) so we don't need to worry too
much about it.
With this patch, all known N32/N64 bugs should be fixed and we now pass the
first 10,000 tests generated by ABITest.py.
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6117
llvm-svn: 221534
Fariborz Jahanian [Fri, 7 Nov 2014 16:47:05 +0000 (16:47 +0000)]
Revert r221404 which caused lldb to not display
vararg expressions.
llvm-svn: 221533
Renato Golin [Fri, 7 Nov 2014 16:33:58 +0000 (16:33 +0000)]
Make sure only NEON enabled devices save/restore D16+ registers
llvm-svn: 221532
NAKAMURA Takumi [Fri, 7 Nov 2014 16:08:19 +0000 (16:08 +0000)]
[CMake] LLVMSupport: Give system_libs PRIVATE scope when LLVMSupport is built as SHARED. Users of LLVMSupport won't inherit ${system_libs}.
unittests/SupporTests is another user of libpthreads. Apply LLVM_SYSTEM_LIBS for him explicitly.
llvm-svn: 221531
Rafael Espindola [Fri, 7 Nov 2014 15:33:56 +0000 (15:33 +0000)]
Pass PRIVATE to target_link_libraries if using shared libraries.
A shared library (unlike a .a), has its dependencies recorded in the library and
we can pass PRIVATE to target_link_libraries.
This patch then removes some bogus dependencies when using
BUILD_SHARED_LIBS=ON. For example, we go from
build lib/CodeGen/CMakeFiles/LLVMCodeGen.dir/AggressiveAntiDepBreaker.cpp.o:
CXX_COMPILER /home/espindola/llvm/llvm/lib/CodeGen/AggressiveAntiDepBreaker.cpp
|| include/llvm/IR/intrinsics_gen lib/libLLVMSupport.so
lib/libLLVMCore.so lib/libLLVMBitReader.so
lib/libLLVMTransformUtils.so lib/libLLVMInstCombine.so
lib/libLLVMScalarOpts.so lib/libLLVMipa.so lib/libLLVMAnalysis.so
lib/libLLVMMCParser.so lib/libLLVMMC.so lib/libLLVMObject.so
lib/libLLVMTarget.so lib/libLLVMProfileData.so
to
build lib/CodeGen/CMakeFiles/LLVMCodeGen.dir/AggressiveAntiDepBreaker.cpp.o:
CXX_COMPILER /home/espindola/llvm/llvm/lib/CodeGen/AggressiveAntiDepBreaker.cpp
|| include/llvm/IR/intrinsics_gen lib/libLLVMSupport.so
lib/libLLVMCore.so lib/libLLVMTransformUtils.so
lib/libLLVMScalarOpts.so lib/libLLVMAnalysis.so lib/libLLVMMC.so
lib/libLLVMTarget.so
In fact, build.ninja goes from 5231028 bytes to 4896759 bytes.
With this, old verisons of bfd ld (2.24 is OK, 2.23 warns) will print a bogus
warning when building with BUILD_SHARED_LIBS.
llvm-svn: 221530
Daniel Sanders [Fri, 7 Nov 2014 15:33:08 +0000 (15:33 +0000)]
[mips] Removed the remainder of MipsCC. NFC.
Summary:
One of the calls to AllocateStack (the one in LowerCall) doesn't look like
it should be there but it was there before and removing it breaks the
frame size calculation.
Reviewers: vmedic, theraven
Reviewed By: theraven
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6116
llvm-svn: 221529
Daniel Sanders [Fri, 7 Nov 2014 15:03:53 +0000 (15:03 +0000)]
[mips] Remove MipsCC::reservedArgArea() in favour of MipsABIInfo::GetCalleeAllocdArgSizeInBytes(). NFC.
Summary:
Reviewers: theraven, vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6115
llvm-svn: 221528
NAKAMURA Takumi [Fri, 7 Nov 2014 14:56:31 +0000 (14:56 +0000)]
MipsCCState.h: Use LLVM_DELETED_FUNCTION for msc17.
llvm-svn: 221527
NAKAMURA Takumi [Fri, 7 Nov 2014 14:50:10 +0000 (14:50 +0000)]
llvm/test/tools/llvm-symbolizer/ppc64.test: Avoid subshell.
llvm-svn: 221526
Daniel Sanders [Fri, 7 Nov 2014 14:24:31 +0000 (14:24 +0000)]
[mips] Move MipsCCState to a separate file and clang-formatted it.
Summary: Depends on D6113
Reviewers: theraven, vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6114
llvm-svn: 221525
Timur Iskhodzhanov [Fri, 7 Nov 2014 14:11:14 +0000 (14:11 +0000)]
Improve comments in vtordisp tests: fix a typo, add a bit more clarity
llvm-svn: 221524
Shankar Easwaran [Fri, 7 Nov 2014 14:08:43 +0000 (14:08 +0000)]
[ELF] Remove is64bits() and isLittlEndian().
ELFLinkingContext had these two functions, which is really not needed since
the Writer uses a llvm::object template composed of Endianness, Alignment,
Is32bit/64bit. We could just use that and not duplicate functionality.
No Change In Functionality.
llvm-svn: 221523
Daniel Sanders [Fri, 7 Nov 2014 12:43:01 +0000 (12:43 +0000)]
[mips] Fix unused variable warnings introduced in r221521
llvm-svn: 221522
Daniel Sanders [Fri, 7 Nov 2014 12:21:37 +0000 (12:21 +0000)]
[mips] Remove remaining use of MipsCC::intArgRegs() in favour of MipsABIInfo::GetByValArgRegs() and MipsABIInfo::GetVarArgRegs()
Summary: Depends on D6112
Reviewers: theraven, vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6113
llvm-svn: 221521
Evgeniy Stepanov [Fri, 7 Nov 2014 12:03:07 +0000 (12:03 +0000)]
[sanitizer] Never remove the last frame off the stack trace.
It can only make it worse.
llvm-svn: 221520
Daniel Sanders [Fri, 7 Nov 2014 12:02:59 +0000 (12:02 +0000)]
[mips] Remove MipsCC::getRegVT(). NFC
Summary: It's no longer used.
Reviewers: vmedic, theraven
Reviewed By: theraven
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6112
llvm-svn: 221519
Daniel Sanders [Fri, 7 Nov 2014 11:43:49 +0000 (11:43 +0000)]
[mips] Remove MipsCC::analyzeCallOperands in favour of CCState::AnalyzeCallOperands. NFC
Summary:
In addition to the usual f128 workaround, it was also necessary to provide
a means of accessing ArgListEntry::IsFixed.
Reviewers: theraven, vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6111
llvm-svn: 221518
Daniel Sanders [Fri, 7 Nov 2014 11:10:48 +0000 (11:10 +0000)]
[mips] Move SpecialCallingConv to MipsCCState and use it from tablegen-erated code. NFC
Summary:
In the long run, it should probably become a calling convention in its own
right but for now just move it out of
MipsISelLowering::analyzeCallOperands() so that we can drop this function
in favour of CCState::AnalyzeCallOperands().
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6085
llvm-svn: 221517
Daniel Sanders [Fri, 7 Nov 2014 10:45:16 +0000 (10:45 +0000)]
[mips] Removed IsVarArg from MipsISelLowering::analyzeCallOperands(). NFC.
Summary:
CCState objects already carry this information in their isVarArg() method.
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6084
llvm-svn: 221516
Renato Golin [Fri, 7 Nov 2014 10:30:53 +0000 (10:30 +0000)]
Move ARM failure from FIXME to XFAIL
llvm-svn: 221515
Jay Foad [Fri, 7 Nov 2014 09:08:39 +0000 (09:08 +0000)]
llvm-symbolizer: teach it about PowerPC64 ELF function descriptors
Summary:
Teach llvm-symbolizer about PowerPC64 ELF function descriptors. Symbols in the .opd section point to function descriptors, the first word of which is a pointer to the real function. For the purposes of symbolizing we pretend that the symbol points directly to the function.
This is enough to get decent function names in stack traces for unoptimized binaries, which fixes the sanitizer print-stack-trace test on PowerPC64 Linux.
Reviewers: kcc, willschm, samsonov
Reviewed By: samsonov
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6110
llvm-svn: 221514
David Majnemer [Fri, 7 Nov 2014 08:54:19 +0000 (08:54 +0000)]
SCCP: overdefined calls cannot become constant
We would attempt to fold away a call instruction which had been marked
overdefined. However, it's not valid to transition to constant from
overdefined.
This fixes PR21512.
llvm-svn: 221513
Johannes Doerfert [Fri, 7 Nov 2014 08:31:31 +0000 (08:31 +0000)]
[Refactor][NFC] Generalize the creation of ScopArrayInfo objects.
Differential Revision: http://reviews.llvm.org/D6031
llvm-svn: 221512
David Majnemer [Fri, 7 Nov 2014 07:26:38 +0000 (07:26 +0000)]
CodeGen, Itanium: Properly dllimport RTTI data
We would blindly assume that RTTI data should have the same linkage as
the vtable because we didn't think the RTTI data was external. This
oversight stemmed because we didn't take dllimport into account.
This fixes PR21512.
llvm-svn: 221511
Justin Hibbits [Fri, 7 Nov 2014 04:46:10 +0000 (04:46 +0000)]
Add Position-independent Code model Module API.
Summary:
This makes PIC levels a Module flag attribute, which can be queried by the
backend. The flag is named `PIC Level`, and can have a value of:
0 - Backend-default
1 - Small-model (-fpic)
2 - Large-model (-fPIC)
These match the `-pic-level' command line argument for clang, and the value of the
preprocessor macro `__PIC__'.
Test Plan:
New flags tests specific for the 'PIC Level' module flag.
Tests to be added as part of a future commit for PowerPC, which will use this new API.
Reviewers: rafael, echristo
Reviewed By: rafael, echristo
Subscribers: rafael, llvm-commits
Differential Revision: http://reviews.llvm.org/D5882
llvm-svn: 221510
Rafael Espindola [Fri, 7 Nov 2014 04:24:12 +0000 (04:24 +0000)]
Use llvm::StringRefMemoryObject NFC.
llvm-svn: 221509
Rui Ueyama [Fri, 7 Nov 2014 02:54:52 +0000 (02:54 +0000)]
Fix Mach-O unit tests breakage on Windows
Mach-O normalized file reader assumes that the entire file is aligned
to a large boundary. If the in-memory file is not aligned properly, it will
abort with an assertion failure in read32/read64. This patch forces the
in-memory file for the unit test to be aligned at 64-byte boundary.
I found these tests are failing on Windows, but theoretically they could
fail on other platform.
llvm-svn: 221508
Ahmed Bougacha [Fri, 7 Nov 2014 02:50:00 +0000 (02:50 +0000)]
[AArch64] Keep flags on condition vreg when instantiating a CB branch.
Reversing a CB* instruction used to drop the flags on the condition. On the
included testcase, this lead to a read from an undefined vreg.
Using addOperand keeps the flags, here <undef>.
Differential Revision: http://reviews.llvm.org/D6159
llvm-svn: 221507
Rui Ueyama [Fri, 7 Nov 2014 01:51:00 +0000 (01:51 +0000)]
Disable a Mach-O test on Windows that depends on Unix path separator.
llvm-svn: 221506
Rui Ueyama [Fri, 7 Nov 2014 01:45:32 +0000 (01:45 +0000)]
Add missing -target triple to ELF tests.
llvm-svn: 221505
Rui Ueyama [Fri, 7 Nov 2014 01:30:49 +0000 (01:30 +0000)]
XFAIL elf/demangle.test on Windows.
Because LLD on that platform can't demangle symbols.
llvm-svn: 221504
Rafael Espindola [Fri, 7 Nov 2014 01:09:51 +0000 (01:09 +0000)]
Use a StringRefMemoryObject. NFC.
llvm-svn: 221503
Rafael Espindola [Fri, 7 Nov 2014 00:52:15 +0000 (00:52 +0000)]
Use StringRefMemoryObject. NFC.
llvm-svn: 221502
David Majnemer [Fri, 7 Nov 2014 00:31:14 +0000 (00:31 +0000)]
LoopVectorize: Don't assume pointees are sized
A pointer's pointee might not be sized: the pointee could be a function.
Report this as IK_NoInduction when calculating isInductionVariable.
This fixes PR21508.
llvm-svn: 221501
Rui Ueyama [Thu, 6 Nov 2014 23:50:48 +0000 (23:50 +0000)]
[PECOFF] Improve subsystem inference
If /subsystem option is not given, the linker needs to infer the
subsystem based on the entry point symbol. If it fails to infer
that, the linker should error out on it.
LLD was almost correct, but it would fail to infer the subsystem
if the entry point is specified with /entry. This is because the
subsystem inference was coupled with the entry point function
searching (if no entry point name is specified, the linker needs
to find the right entry name).
This patch makes the subsystem inference an independent pass to
fix the issue. Now, as long as an entry point function is defined,
LLD can infer the subsystem no matter how it resolved the entry
point.
I don't think scanning all the defined symbols is fast, although
it shouldn't be that slow. The file class there does not provide
any easy way to find an atom by name, so this is what we can do
at this moment. I'd like to revisit this later to make it more
efficient.
llvm-svn: 221499
David Majnemer [Thu, 6 Nov 2014 23:23:30 +0000 (23:23 +0000)]
InstCombine: Rely on cmpxchg's return code when it's strong
Comparing the result of a cmpxchg instruction can be replaced with an
extractvalue of the cmpxchg success indicator.
llvm-svn: 221498
Rafael Espindola [Thu, 6 Nov 2014 23:16:57 +0000 (23:16 +0000)]
Remove unused variable. NFC.
llvm-svn: 221497
Greg Clayton [Thu, 6 Nov 2014 22:59:28 +0000 (22:59 +0000)]
Fixes so tests compile and run remotely.
Fixes include:
- dont set or change LDFLAGS, but set LD_EXTRAS instead
- fix compilation errors for iOS based builds with objective C code
- fix test cases to create classes instead of relying on classes from AppKit
- rename things where it makes sense
llvm-svn: 221496
Rafael Espindola [Thu, 6 Nov 2014 22:57:10 +0000 (22:57 +0000)]
Don't repeat names in comments. NFC.
llvm-svn: 221495
Greg Clayton [Thu, 6 Nov 2014 22:56:56 +0000 (22:56 +0000)]
Changed program to go along with the previous commit to allow this test to be run remotely via lldb-platform.
llvm-svn: 221494
Greg Clayton [Thu, 6 Nov 2014 22:56:15 +0000 (22:56 +0000)]
Fix the test to not have to run the binary separately first to get the golden output, use the process STDOUT instead.
This helps this test be able to be run remotely.
llvm-svn: 221493
Greg Clayton [Thu, 6 Nov 2014 22:55:09 +0000 (22:55 +0000)]
Make sure stderr is filtered out in case xcodebuild or xcrun print errors when getting the SDK path and use xcrun to find the SDK path.
llvm-svn: 221492
Simon Atanasyan [Thu, 6 Nov 2014 22:46:24 +0000 (22:46 +0000)]
[ELF][yaml2obj] Handle additional MIPS specific st_other field flags
The ELF symbol `st_other` field might contain additional flags besides
visibility ones. This patch implements support for some MIPS specific
flags.
llvm-svn: 221491
Rafael Espindola [Thu, 6 Nov 2014 22:39:16 +0000 (22:39 +0000)]
Factor out call to push_back. NFC.
llvm-svn: 221490
Simon Pilgrim [Thu, 6 Nov 2014 22:15:41 +0000 (22:15 +0000)]
[X86][SSE] Vector integer/float conversion memory folding (cvttps2dq / cvttpd2dq)
Fixed an issue with the (v)cvttps2dq and (v)cvttpd2dq instructions being incorrectly put in the 2 source operand folding tables instead of the 1 source operand and added the missing SSE/AVX versions.
Also added missing (v)cvtps2dq and (v)cvtpd2dq instructions to the folding tables.
Differential Revision: http://reviews.llvm.org/D6001
llvm-svn: 221489
Ahmed Bougacha [Thu, 6 Nov 2014 22:04:15 +0000 (22:04 +0000)]
[X86] Add VFMADDSUB cases for the 213->231 custom inserter.
Also add tests for vfmadd/vfmsub.
llvm-svn: 221488
Ahmed Bougacha [Thu, 6 Nov 2014 21:58:11 +0000 (21:58 +0000)]
[X86] Add missing FMA3 VFMADDSUB in the emitter.
Also reuse the fma4 intrinsic test to cover fma3 instructions too.
llvm-svn: 221487
Enrico Granata [Thu, 6 Nov 2014 21:55:30 +0000 (21:55 +0000)]
This is a large, but clearical, commit that enables the C++ formatters to take on the additional TypeSummaryOptions argument. It is still not used for anything, but it is now there. Adding support for this extra argument to Python formatters will follow suit
llvm-svn: 221486
David Majnemer [Thu, 6 Nov 2014 21:46:55 +0000 (21:46 +0000)]
Object, COFF: Don't consider AuxFunctionDefinition for getSymbolSize
mingw lies about the size of a function's AuxFunctionDefinition. Ignore
the field and rely on our heuristic to determine the symbol's size.
llvm-svn: 221485
Ahmed Bougacha [Thu, 6 Nov 2014 21:46:23 +0000 (21:46 +0000)]
[X86] Split FMA4 RM tests into a separate file. NFC.
While there, remove useless comments.
llvm-svn: 221484
Tobias Grosser [Thu, 6 Nov 2014 21:25:54 +0000 (21:25 +0000)]
Fix formatting
llvm-svn: 221483
Enrico Granata [Thu, 6 Nov 2014 21:23:20 +0000 (21:23 +0000)]
Introduce the notion of "type summary options" as flags that can be passed down to individual summary formatters to alter their behavior in a formatter-dependent way
Two flags are introduced:
- preferred display language (as in, ObjC vs. C++)
- summary capping (as in, should a limit be put to the amount of data retrieved)
The meaning - if any - of these options is for individual formatters to establish
The topic of a subsequent commit will be to actually wire these through to individual data formatters
llvm-svn: 221482
Frederic Riss [Thu, 6 Nov 2014 21:12:06 +0000 (21:12 +0000)]
Return a DIDescriptor from CGDebugInfo::getDeclarationOrDefinition...
... instead of a DIScope that might wrap something that's not actually
a DIScope (eg. a DIGlobalVariable);
llvm-svn: 221481
Rafael Espindola [Thu, 6 Nov 2014 20:01:34 +0000 (20:01 +0000)]
Base check on the section name, not the variable name.
The variable is private, so the name should not be relied on. Also, the
linker uses the sections, so asan should too when trying to avoid causing
the linker problems.
llvm-svn: 221480
Tobias Grosser [Thu, 6 Nov 2014 19:35:21 +0000 (19:35 +0000)]
Explicitly annotate loops we want to run thread-parallel
We introduces a new flag -polly-parallel and use it to annotate the for-nodes in
the isl ast that we want to execute thread parallel (e.g., using OpenMP). We
previously already emmitted openmp annotations, but we did this for various
kinds of parallel loops, including some which we can not run in parallel.
With this patch we now have three annotations:
1) #pragma known-parallel [reduction]
2) #pragma omp for
3) #pragma simd
meaning:
1) loop has no loop carried dependences
2) loop will be executed thread-parallel
3) loop can possibly be vectorized
This patch introduces 1) and reduces the use of 2) to only the cases where we
will actually generate thread parallel code.
It is in preparation of openmp code generation in our isl backend.
Legacy:
- We also have a command line option -enable-polly-openmp. This option controls
the OpenMP code generation in CLooG. It will become an alias of
-polly-parallel after the CLooG code generation has been dropped.
http://reviews.llvm.org/D6142
llvm-svn: 221479
Nick Kledzik [Thu, 6 Nov 2014 19:33:59 +0000 (19:33 +0000)]
remove unneeded { }
llvm-svn: 221478
Nick Kledzik [Thu, 6 Nov 2014 19:33:57 +0000 (19:33 +0000)]
[mach-o] Add support for -force_load option
The darwin linker has two ways to force all members of an archive to be loaded.
The -all_load option applies to all static libraries. The -force_load takes
a path to a library and just that library's members are force loaded.
llvm-svn: 221477
Sean Callanan [Thu, 6 Nov 2014 19:26:10 +0000 (19:26 +0000)]
Handle types from the runtime that conform to
protocols.
<rdar://problem/
18883778>
llvm-svn: 221476
Lang Hames [Thu, 6 Nov 2014 19:24:18 +0000 (19:24 +0000)]
[RegAlloc] Remove reference to the trivial spiller in test case.
This test case was never actually testing the trivial spiller: the -spiller
option has not been hooked up for a while now.
llvm-svn: 221475
Lang Hames [Thu, 6 Nov 2014 19:12:38 +0000 (19:12 +0000)]
[RegAlloc] Kill off the trivial spiller - nobody is using it any more.
llvm-svn: 221474
Kevin Enderby [Thu, 6 Nov 2014 19:12:28 +0000 (19:12 +0000)]
[mach-o] Update test case to work with latest llvm-objdump output for ARM thumb
code. Same basic change that was done in r218429 for ARM code.
Where the ARM thumb symbolizer in llvm-objdump’s Mach-O disassembler is now
plumbed in with r221470 from the llvm trunk.
llvm-svn: 221473
Michael Liao [Thu, 6 Nov 2014 19:05:57 +0000 (19:05 +0000)]
Indentation fixes
llvm-svn: 221472
Frederic Riss [Thu, 6 Nov 2014 19:00:47 +0000 (19:00 +0000)]
Try to appease MSVC buildbots after r221466.
llvm-svn: 221471
Kevin Enderby [Thu, 6 Nov 2014 19:00:13 +0000 (19:00 +0000)]
Plumb in the ARM thumb symbolizer in llvm-objdump’s Mach-O disassembler and
add the code and test cases for 32-bit ARM symbolizer.
Also fixed the printing of data in code as it was not using the table correctly
and needed to fix one of the test cases too.
This will break lld’s test/mach-o/arm-interworking-movw.yaml till the tweak
for that is made. Which I’ll be committing immediately after this commit.
llvm-svn: 221470
Alexey Samsonov [Thu, 6 Nov 2014 18:43:45 +0000 (18:43 +0000)]
[Sanitizer] Introduce "stack_trace_format" runtime flag.
This flag can be used to specify the format of stack frames - user
can now provide a string with placeholders, which should be printed
for each stack frame with placeholders replaced with actual data.
For example "%p" will be replaced by PC, "%s" will be replaced by
the source file name etc.
"DEFAULT" value enforces default stack trace format currently used in
all the sanitizers except TSan.
This change also implements __sanitizer_print_stack_trace interface
function in TSan.
llvm-svn: 221469
Greg Clayton [Thu, 6 Nov 2014 18:39:39 +0000 (18:39 +0000)]
Pick better floating point numbers (ones that can be exactly represented) in floating point instead of something that can't to avoid test suite failures on different devices and architectures.
<rdar://problem/
18826900>
llvm-svn: 221468
Shawn Best [Thu, 6 Nov 2014 17:52:15 +0000 (17:52 +0000)]
fixed minor code indenting reviews.llvm.org/D6127
llvm-svn: 221467
Frederic Riss [Thu, 6 Nov 2014 17:46:55 +0000 (17:46 +0000)]
Change DIBuilder::createImportedDeclaration from taking a DIScope to a DIDescriptor.
Imported declarations can be DIGlobalVariables which aren't a DIScope. Today
clang (unknowingly I believe) shoehorns these into a DIScope and it all works
just because we never access the fields.
llvm-svn: 221466
Colin LeMahieu [Thu, 6 Nov 2014 17:05:51 +0000 (17:05 +0000)]
[Hexagon] Adding basic Hexagon ELF object emitter.
llvm-svn: 221465
Eli Bendersky [Thu, 6 Nov 2014 17:05:49 +0000 (17:05 +0000)]
Clean up NVPTXLowerStructArgs.cpp. NFC
* Remove unnecessary const_casts and C-style casts
* Simplify attribute access code
* Simplify ArrayRef creation
* 80-col and clang-format
llvm-svn: 221464
Daniel Sanders [Thu, 6 Nov 2014 16:48:57 +0000 (16:48 +0000)]
[mips] Removed IsSoftFloat from MipsISelLowering::analyzeCallOperands(). NFC
Summary:
It isn't used anymore.
Depends on D6081
Reviewers: vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6083
llvm-svn: 221463
Chad Rosier [Thu, 6 Nov 2014 16:46:37 +0000 (16:46 +0000)]
[Reassociate] Don't reassociate when mixing regular and fast-math FP
instructions. Inlining might cause such cases and it's not valid to
reassociate floating-point instructions without the unsafe algebra flag.
Patch by Mehdi Amini <mehdi_amini@apple.com>!
llvm-svn: 221462
Daniel Sanders [Thu, 6 Nov 2014 16:36:30 +0000 (16:36 +0000)]
[mips] Removed MipsISelLowering::analyzeFormalArguments() in favour of CCState::AnalyzeFormalArguments()
Summary:
As with returns, we must be able to identify f128 arguments despite them
being lowered away. We do this with a pre-analyze step that builds a
vector and then we use this vector from the tablegen-erated code.
Reviewers: vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6081
llvm-svn: 221461
Eric Fiselier [Thu, 6 Nov 2014 15:26:20 +0000 (15:26 +0000)]
Add support for UNSUPPORTED tag to lit
llvm-svn: 221460
Rafael Espindola [Thu, 6 Nov 2014 15:05:51 +0000 (15:05 +0000)]
Use FileCheck in a few tests.
llvm-svn: 221459
Charlie Turner [Thu, 6 Nov 2014 14:59:30 +0000 (14:59 +0000)]
Remove references to the cortex-a9-mp CPU.
This CPU definition is redundant. The Cortex-A9 is defined as
supporting multiprocessing extensions. Remove references to this CPU.
This CPU was recently removed from LLVM. See http://reviews.llvm.org/D6057
Change-Id: I62ae7cc656fcae54fbaefc4b6976e77e694a8678
llvm-svn: 221458
Will Schmidt [Thu, 6 Nov 2014 14:58:06 +0000 (14:58 +0000)]
Use GET_CURRENT_FRAME() to calculate the memory layout for power. This works
for both PPC64 Big and Little endian modes, so also eliminates the need for
the BIG_ENDIAN/LITTLE_ENDIAN #ifdeffery.
By trial and error, it also looks like the kPPC64_ShadowOffset64 value is
valid using (1ULL << 41) for both BE and LE, so that #if/#elif/#endif block
has also been simplified.
Differential Revision: http://reviews.llvm.org/D6044
llvm-svn: 221457
Rafael Espindola [Thu, 6 Nov 2014 14:39:49 +0000 (14:39 +0000)]
Compute the correct jump table entries on 32 bit windows.
On 32 bit windows we use label differences and .set does not suppress
rolocations, a combination that was not used before r220256.
This fixes PR21497.
llvm-svn: 221456
Andrea Di Biagio [Thu, 6 Nov 2014 14:36:45 +0000 (14:36 +0000)]
[X86] When commuting SSE immediate blend, make sure that the new blend mask is a valid imm8.
Example:
define <4 x i32> @test(<4 x i32> %a, <4 x i32> %b) {
%shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 4, i32 5, i32 6, i32 3>
ret <4 x i32> %shuffle
}
Before llc (-mattr=+sse4.1), produced the following assembly instruction:
pblendw $
4294967103, %xmm1, %xmm0
After
pblendw $63, %xmm1, %xmm0
llvm-svn: 221455