Rafael Espindola [Tue, 29 Oct 2013 14:25:43 +0000 (14:25 +0000)]
Support names like llvm-ar-3.4 and llvm-ranlib-3.4.
They are used in some packages. For example:
http://packages.ubuntu.com/saucy/i386/llvm-3.4/filelist
This fixes pr17721.
llvm-svn: 193612
Bernard Ogden [Tue, 29 Oct 2013 14:16:09 +0000 (14:16 +0000)]
Test cleanup for v8 instructions
Add some missing tests, factor out a test not specific to v8 into
its own file.
llvm-svn: 193611
Timur Iskhodzhanov [Tue, 29 Oct 2013 14:13:45 +0000 (14:13 +0000)]
Fix an assertion when handling a custom case of virtual inheritance; also reduce code duplication
llvm-svn: 193610
Rafael Espindola [Tue, 29 Oct 2013 13:44:11 +0000 (13:44 +0000)]
Clarify that GlobalVariables definitions must have an initializer.
llvm-svn: 193609
Timur Iskhodzhanov [Tue, 29 Oct 2013 13:29:32 +0000 (13:29 +0000)]
Add a missing lit-test expectation (forgotten to write initially?)
llvm-svn: 193608
Sylvestre Ledru [Tue, 29 Oct 2013 13:01:41 +0000 (13:01 +0000)]
Use LLVMLibsOptions intead of LDFLAGS when setting the libclang soname. This allows 'make LDFLAGS=foo' to be set correctly (and matches the way it is done with libllvm and liblldb)
llvm-svn: 193607
Tim Northover [Tue, 29 Oct 2013 12:32:58 +0000 (12:32 +0000)]
ARM: fix AST for __builtin_arm_strex call
The AST was constructed so that this builtin returned the default BoolTy and
since I'd opted for custom SemaChecking, I should have set it properly at that
point.
This caused an assertion failure when the types didn't match up with what we
generated. This makes it return an IntTy, which is as good as anything.
llvm-svn: 193606
Daniel Jasper [Tue, 29 Oct 2013 12:24:23 +0000 (12:24 +0000)]
clang-format: Fix overloaded operator for global-scoped conversions.
Before:
operator::A();
After:
operator ::A();
llvm-svn: 193605
Timur Iskhodzhanov [Tue, 29 Oct 2013 12:13:22 +0000 (12:13 +0000)]
Quick-fix DebugInfo build on Windows
MSVC can't comprehend
template<typename T, size_t N>
ArrayRef<T> makeArrayRef(const T (&Arr)[N]) {
return ArrayRef<T>(Arr);
}
if Arr is
static const uint8_t sizes[];
declared in a templated and defined a few lines later.
I'll send a proper fix (i.e. get rid of unnecessary templates) for review soon.
llvm-svn: 193604
Sylvestre Ledru [Tue, 29 Oct 2013 11:05:18 +0000 (11:05 +0000)]
Fix a typo in my family name. Tobias: ;)
llvm-svn: 193603
Dmitry Vyukov [Tue, 29 Oct 2013 10:30:39 +0000 (10:30 +0000)]
tsan/asan: support pthread_setname_np to set thread names
llvm-svn: 193602
Chandler Carruth [Tue, 29 Oct 2013 10:27:30 +0000 (10:27 +0000)]
Teach the driver to not try to use 'lib32' multilib spellings on ARM
which doesn't use that multilib. As a consequence, fix Clang's support
for cross compiling environments that were relying on this quirk to
ensure the correct library search path ordering.
This also re-instates the new test cases from Rafael's r193528 for
cross-compiling to ARM on Ubuntu 13.10 without any of the changes to the
existing test cases (they were no longer needed).
This solution was the result of a lot of IRC debugging and trying to
understand *exactly* what quirk was being relied upon. It took some time
for me to figure out that it was the use of 'lib32' is a multilib that
was throwing a wrench in the works.
In case you are thinking that its silly to use a multilib of 'lib' at
all, entertainingly, GCC does so as well (you can see it with the
.../lib/../lib/crt1.o pattern it uses), and the 2-phase sequence of
search paths (multilib followed by non-multilib) has observable (if
dubious) consequences. =/ Yuck.
llvm-svn: 193601
Bernard Ogden [Tue, 29 Oct 2013 09:47:51 +0000 (09:47 +0000)]
ARM: Add -m[no-]crc to dis/enable CRC subtargetfeature from clang
Allow users to disable or enable CRC subtarget feature.
Differential Revision: http://llvm-reviews.chandlerc.com/D2037
llvm-svn: 193600
Bernard Ogden [Tue, 29 Oct 2013 09:47:35 +0000 (09:47 +0000)]
ARM: Add subtarget feature for CRC
Adds a subtarget feature for the CRC instructions (optional in v8-A) to the ARM (32-bit) backend.
Differential Revision: http://llvm-reviews.chandlerc.com/D2036
llvm-svn: 193599
Anders Waldenborg [Tue, 29 Oct 2013 09:37:28 +0000 (09:37 +0000)]
Fix misapplied patch in r193597
Sorry Peter Zotov, entirely my fault.
llvm-svn: 193598
Anders Waldenborg [Tue, 29 Oct 2013 09:02:02 +0000 (09:02 +0000)]
llvm-c: Make LLVM{Get,Set}Alignment work on {Load,Store}Inst too
Patch by Peter Zotov
Differential Revision: http://llvm-reviews.chandlerc.com/D1910
llvm-svn: 193597
Chandler Carruth [Tue, 29 Oct 2013 08:57:33 +0000 (08:57 +0000)]
Doh, got this turned around. This is in fact the consistent ordering.
llvm-svn: 193596
Chandler Carruth [Tue, 29 Oct 2013 08:53:03 +0000 (08:53 +0000)]
Clean up the deeply misleading name of the "MultiLibSuffix". This is
actually a MIPS-only hack to shim in random ABI directory suffixes in
numerous places throughout the toolchain's path search. It shouldn't
appear to be anything more general or useful.
llvm-svn: 193595
Alp Toker [Tue, 29 Oct 2013 08:32:41 +0000 (08:32 +0000)]
Use Rewriter::overwriteChangedFiles() directly
This replaces the custom code in RefactoringTool::saveRewrittenFiles() which
lacked atomic file saving and error diagnostics, resolving an old FIXME from
r157331.
Landing this time with the proper return code, plus a very unhelpful comment
cleared up.
Rubber-stamped by Manuel Klimek.
llvm-svn: 193594
Tim Northover [Tue, 29 Oct 2013 08:22:33 +0000 (08:22 +0000)]
AArch64: add 'a' inline asm operand modifier
This is used in the Linux kernel, and effectively just means "print an
address".
llvm-svn: 193593
Alp Toker [Tue, 29 Oct 2013 07:56:03 +0000 (07:56 +0000)]
Revert "Use Rewriter::overwriteChangedFiles() directly"
This wasn't ready for prime time yet, seems to break tools-extra.
This reverts commit r193590.
llvm-svn: 193592
Alp Toker [Tue, 29 Oct 2013 07:47:57 +0000 (07:47 +0000)]
Remove obsolete RewriteBuffer comment
The string buffer in question was replaced with RewriteRope in r43884
back in 2007.
llvm-svn: 193591
Alp Toker [Tue, 29 Oct 2013 07:47:49 +0000 (07:47 +0000)]
Use Rewriter::overwriteChangedFiles() directly
The old code in RefactoringTool::saveRewrittenFiles() lacked
atomic moves and diagnostics and had FIXME to this effect.
Rubber-stamped by Manuel Klimek.
llvm-svn: 193590
Manman Ren [Tue, 29 Oct 2013 05:49:41 +0000 (05:49 +0000)]
Debug Info: instead of calling addToContextOwner which constructs the context
after the DIE creation, we construct the context first.
This touches creation of namespaces and global variables. The purpose is to
handle all DIE creations similarly: constructs the context first, then creates
the DIE and immediately adds the DIE to its parent.
We use createAndAddDIE to wrap around "new DIE(".
llvm-svn: 193589
Shankar Easwaran [Tue, 29 Oct 2013 05:33:52 +0000 (05:33 +0000)]
[FileToMutable] Garbage collect unused private member
llvm-svn: 193588
Alexey Samsonov [Tue, 29 Oct 2013 05:31:25 +0000 (05:31 +0000)]
[Sanitizer] Simplify StackTrace::PrintStack interface: prefer common flags to turn on/off the symbolization
llvm-svn: 193587
Shankar Easwaran [Tue, 29 Oct 2013 05:26:49 +0000 (05:26 +0000)]
[test] Fix a test failure, this test is only applicable in debug mode
llvm-svn: 193586
Shankar Easwaran [Tue, 29 Oct 2013 05:12:14 +0000 (05:12 +0000)]
[PassManager] add ReaderWriter{Native,YAML} to the Driver.
Enable this for the following flavors
a) core
b) gnu
c) darwin
Its disabled for the flavor PECOFF. Convenient markers are added with FIXME
comments in the Driver that would be removed and code removed from each flavor.
llvm-svn: 193585
Rui Ueyama [Tue, 29 Oct 2013 04:54:50 +0000 (04:54 +0000)]
Fix spelling.
llvm-svn: 193584
Nick Lewycky [Tue, 29 Oct 2013 03:54:41 +0000 (03:54 +0000)]
Add missing attributes [[gnu::no_sanitize_address]] and
[[gnu::no_address_safety_analysis]].
llvm-svn: 193583
Nick Lewycky [Tue, 29 Oct 2013 03:33:57 +0000 (03:33 +0000)]
Treat [[gnu::__const]] the same as [[gnu::const]]. GCC's tokenizer can't tell
the difference.
llvm-svn: 193582
Evgeniy Stepanov [Tue, 29 Oct 2013 02:48:49 +0000 (02:48 +0000)]
[msan] Intercept shmat.
llvm-svn: 193581
NAKAMURA Takumi [Tue, 29 Oct 2013 02:46:00 +0000 (02:46 +0000)]
Add llvm/test/Transforms/SLPVectorizer/ARM/lit.local.cfg. Tests there require ARM in targets.
llvm-svn: 193580
Alp Toker [Tue, 29 Oct 2013 02:35:28 +0000 (02:35 +0000)]
Fix "existant" typos
llvm-svn: 193579
Alp Toker [Tue, 29 Oct 2013 02:34:44 +0000 (02:34 +0000)]
Fix "existant" typos
llvm-svn: 193578
Chandler Carruth [Tue, 29 Oct 2013 02:27:56 +0000 (02:27 +0000)]
Cleanup comments and braces. No functionality changed, just trying to
record what is *actually* going on here as the comments existing in the
code are confusing at best, and in places outright misleading.
The API is even more misleading. Yay.
llvm-svn: 193577
Richard Smith [Tue, 29 Oct 2013 01:44:23 +0000 (01:44 +0000)]
Clean up.
llvm-svn: 193576
NAKAMURA Takumi [Tue, 29 Oct 2013 01:43:05 +0000 (01:43 +0000)]
DWARFFormValue.cpp: Appease gcc to give explicit constructors.
error: conversion from `const uint8_t*' to non-scalar type `llvm::ArrayRef<unsigned char>' requested
llvm-svn: 193575
Arnold Schwaighofer [Tue, 29 Oct 2013 01:33:57 +0000 (01:33 +0000)]
ARM cost model: Unaligned vectorized double stores are expensive
Updated a test case that assumed that <2 x double> would vectorize to use
<4 x float>.
radar://
15338229
llvm-svn: 193574
Arnold Schwaighofer [Tue, 29 Oct 2013 01:33:53 +0000 (01:33 +0000)]
ARM cost model: Account for zero cost scalar SROA instructions
By vectorizing a series of srl, or, ... instructions we have obfuscated the
intention so much that the backend does not know how to fold this code away.
radar://
15336950
llvm-svn: 193573
Arnold Schwaighofer [Tue, 29 Oct 2013 01:33:50 +0000 (01:33 +0000)]
SLPVectorizer: Use vector type for vectorized memory operations
No test case, because with the current cost model we don't see a difference.
An upcoming ARM memory cost model change will expose and test this bug.
radar://
15332579
llvm-svn: 193572
Andrew Kaylor [Tue, 29 Oct 2013 01:33:14 +0000 (01:33 +0000)]
Cleaning up comments in lli
llvm-svn: 193571
Andrew Kaylor [Tue, 29 Oct 2013 01:29:56 +0000 (01:29 +0000)]
Adding a workaround for __main linking with remote lli and Cygwin/MinGW
llvm-svn: 193570
Joerg Sonnenberger [Tue, 29 Oct 2013 01:06:17 +0000 (01:06 +0000)]
Move the STT_FILE symbols out of the normal symbol table processing for
ELF. They can overlap with the other symbols, e.g. if a source file
"foo.c" contains a function "foo" with a static variable "c".
llvm-svn: 193569
Manman Ren [Tue, 29 Oct 2013 01:03:01 +0000 (01:03 +0000)]
Debug Info: use createAndAddDIE to wrap around "new DIE" in DwarfDebug.
This commit ensures DIEs are constructed within a compile unit and
immediately added to their parents.
Reviewed off-list by Eric.
llvm-svn: 193568
Manman Ren [Tue, 29 Oct 2013 00:58:04 +0000 (00:58 +0000)]
Debug Info: use createAndAddDIE for newly-created Subprogram DIEs.
More patches will be submitted to convert "new DIE(" to use createAddAndDIE in
DwarfCompileUnit.cpp. This will simplify implementation of addDIEEntry where
we have to decide between ref4 and ref_addr, because DIEs that can be shared
across CU will be added to a CU already.
Reviewed off-list by Eric.
llvm-svn: 193567
Manman Ren [Tue, 29 Oct 2013 00:53:03 +0000 (00:53 +0000)]
Debug Info: add a helper function createAndAddDIE.
It wraps around "new DIE(" and handles the bookkeeping part of the newly-created
DIE. It adds the DIE to its parent, and calls insertDIE if necessary. It makes
sure that bookkeeping is done at the earliest time and we should not see
parentless DIEs if all constructions of DIEs go through this helper function.
Later on, we can use an allocator for DIE allocation, and will only need to
change createAndAddDIE instead of modifying all the "new DIE(".
Reviewed off-list by Eric.
llvm-svn: 193566
Rui Ueyama [Tue, 29 Oct 2013 00:33:34 +0000 (00:33 +0000)]
[PECOFF] Fix __ImageBase symbol.
__ImageBase is an absolute symbol whose address is the same as the image base
address. What we did before this patch was to create __ImageBase symbol as a
symbol whose *contents* (not location) is the image base address, which is
clearly wrong.
llvm-svn: 193565
Enrico Granata [Tue, 29 Oct 2013 00:28:35 +0000 (00:28 +0000)]
<rdar://problem/
15144376>
This commit reimplements the TypeImpl class (the class that backs SBType) in terms of a static,dynamic type pair
This is useful for those cases when the dynamic type of an ObjC variable can only be obtained in terms of an "hollow" type with no ivars
In that case, we could either go with the static type (+iVar information) or with the dynamic type (+inheritance chain)
With the new TypeImpl implementation, we try to combine these two sources of information in order to extract as much information as possible
This should improve the functionality of tools that are using the SBType API to do extensive dynamic type inspection
llvm-svn: 193564
Alexey Samsonov [Mon, 28 Oct 2013 23:58:58 +0000 (23:58 +0000)]
Merge DWARFDIE::extractFast and DWARFDIE::extract into one function.
Complicated CU-DIE-specific logic in the latter was never used,
and it makes sense to have safety checks for broken dwarf in the former.
llvm-svn: 193563
Andrew Kaylor [Mon, 28 Oct 2013 23:51:03 +0000 (23:51 +0000)]
Renaming MCJIT .ir files to .ll and moving them to Inputs
llvm-svn: 193562
Alp Toker [Mon, 28 Oct 2013 23:47:09 +0000 (23:47 +0000)]
Switch %clang -cc1 tests to %clang_cc1
llvm-svn: 193561
Alexey Samsonov [Mon, 28 Oct 2013 23:41:49 +0000 (23:41 +0000)]
DWARF parser: Use ArrayRef to represent form sizes and simplify DWARFDIE::extractFast() interface. No functionality change.
llvm-svn: 193560
Alp Toker [Mon, 28 Oct 2013 23:37:49 +0000 (23:37 +0000)]
lit: add missing substitutions for recently added tools
llvm-mcmarkup, obj2yaml and yaml2obj were missing from the substitutions list,
causing the test suite to fail in a sandboxed environment.
llvm-svn: 193559
Alp Toker [Mon, 28 Oct 2013 23:37:45 +0000 (23:37 +0000)]
Quote potential shell expansions found in tests
llvm-svn: 193558
Rui Ueyama [Mon, 28 Oct 2013 23:34:41 +0000 (23:34 +0000)]
Fix comment.
llvm-svn: 193557
Rui Ueyama [Mon, 28 Oct 2013 23:26:33 +0000 (23:26 +0000)]
[PECOFF] Set section name for BSS atoms.
This change has no effect now, but will be needed to emit BSS atoms
to .bss section.
llvm-svn: 193556
Alexey Samsonov [Mon, 28 Oct 2013 23:15:15 +0000 (23:15 +0000)]
DWARF parser: since DWARF4, DW_AT_high_pc may be a constant representing function size
llvm-svn: 193555
Rafael Espindola [Mon, 28 Oct 2013 23:14:34 +0000 (23:14 +0000)]
Revert "Support cross linking to arm on ubuntu."
This reverts commit r193528 (and fixes).
It broke Chandler's setup.
llvm-svn: 193554
Alexey Samsonov [Mon, 28 Oct 2013 23:01:48 +0000 (23:01 +0000)]
DebugInfo: Introduce the notion of "form classes"
Summary:
Use DWARF4 table of form classes to fetch attributes from DIE
in a more consistent way. This shouldn't change the functionality and
serves as a refactoring for upcoming change: DW_AT_high_pc has different
semantics depending on its form class.
Reviewers: dblaikie, echristo
Reviewed By: echristo
CC: echristo, llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1961
llvm-svn: 193553
Alp Toker [Mon, 28 Oct 2013 22:51:25 +0000 (22:51 +0000)]
Fix the lli --extra-module value_desc
llvm-svn: 193552
Nick Kledzik [Mon, 28 Oct 2013 22:48:33 +0000 (22:48 +0000)]
[MachO] pointer align linker generated (non)lazy pointers
llvm-svn: 193551
Rui Ueyama [Mon, 28 Oct 2013 22:42:54 +0000 (22:42 +0000)]
Add a few tests for StringRef::{start,end}with.
llvm-svn: 193550
Richard Smith [Mon, 28 Oct 2013 22:18:19 +0000 (22:18 +0000)]
Allow a new syntax in a module requires-declaration:
requires ! feature
The purpose of this is to allow (for instance) the module map for /usr/include
to exclude <tgmath.h> and <complex.h> when building in C++ (these headers are
instead provided by the C++ standard library in this case, and the glibc C
<tgmath.h> header would otherwise try to include <complex.h>, resulting in a
module cycle).
llvm-svn: 193549
Rafael Espindola [Mon, 28 Oct 2013 22:17:19 +0000 (22:17 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193548
Rafael Espindola [Mon, 28 Oct 2013 22:11:47 +0000 (22:11 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193547
Rafael Espindola [Mon, 28 Oct 2013 22:05:05 +0000 (22:05 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193546
Richard Smith [Mon, 28 Oct 2013 22:04:30 +0000 (22:04 +0000)]
Fix a parser crash when there are #pragmas in a context which requires a single
statement (after a case label, if, etc). Patch by Olivier Goffart!
llvm-svn: 193545
Andrew Kaylor [Mon, 28 Oct 2013 21:58:15 +0000 (21:58 +0000)]
Standardizing lli's extra module command line option
llvm-svn: 193544
Bill Wendling [Mon, 28 Oct 2013 21:43:54 +0000 (21:43 +0000)]
Remove stray '_'.
llvm-svn: 193543
Bill Wendling [Mon, 28 Oct 2013 21:43:11 +0000 (21:43 +0000)]
Use the correct reference. Spotted by Sean Silva.
llvm-svn: 193542
Bill Wendling [Mon, 28 Oct 2013 21:22:23 +0000 (21:22 +0000)]
Remove 2.4 from the list of supported Python versions.
llvm-svn: 193541
Akira Hatanaka [Mon, 28 Oct 2013 21:21:36 +0000 (21:21 +0000)]
[mips] Simplify LowerFormalArguments using getRegClassFor.
No functionality change.
llvm-svn: 193540
Rafael Espindola [Mon, 28 Oct 2013 21:12:15 +0000 (21:12 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193539
Rafael Espindola [Mon, 28 Oct 2013 21:06:12 +0000 (21:06 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193538
Rafael Espindola [Mon, 28 Oct 2013 20:59:41 +0000 (20:59 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193537
Rafael Espindola [Mon, 28 Oct 2013 20:54:33 +0000 (20:54 +0000)]
Convert another llc -filetype=obj test.
llvm-svn: 193536
Lang Hames [Mon, 28 Oct 2013 20:51:11 +0000 (20:51 +0000)]
Return early from getUnconditionalBranchTargetOpValue if the branch target is
an MCExpr, in order to avoid writing an encoded zero value in the immediate
field.
When getUnconditionalBranchTargetOpValue is called with an MCExpr target, we
don't know what the final immediate field value should be. We shouldn't
explicitly set the immediate field to an encoded zero value as zero is encoded
with a non-zero bit pattern. This leads to bits being set that pollute the
final immediate value. The nature of the encoding is such that the polluted
bits only affect very large immediate values, explaining why this hasn't
caused problems earlier.
Fixes <rdar://problem/
15155975>.
llvm-svn: 193535
Rafael Espindola [Mon, 28 Oct 2013 20:40:20 +0000 (20:40 +0000)]
Convert a llc -filetype=obj test into a llvm-mc test.
llvm-svn: 193534
Rafael Espindola [Mon, 28 Oct 2013 19:40:25 +0000 (19:40 +0000)]
More windows path handling.
llvm-svn: 193533
Rafael Espindola [Mon, 28 Oct 2013 19:19:29 +0000 (19:19 +0000)]
Relax test to pass with windows paths.
llvm-svn: 193532
Jim Ingham [Mon, 28 Oct 2013 19:00:42 +0000 (19:00 +0000)]
Change the default handling for SIGALRM and SIGCHLD to not notify.
<rdar://problem/
15208799>
llvm-svn: 193530
Evgeniy Stepanov [Mon, 28 Oct 2013 18:53:37 +0000 (18:53 +0000)]
[msan] Fix a typo and enable poison_in_free flag.
llvm-svn: 193529
Rafael Espindola [Mon, 28 Oct 2013 18:52:30 +0000 (18:52 +0000)]
Support cross linking to arm on ubuntu.
With this patch we correctly determine that ubuntu's ARM tree is not biarch
and use "lib" istead of "lib32".
Without this patch the search inside the arm tree for the crt files was failing
and we would end up trying to use the i686 ones in lib32.
llvm-svn: 193528
Ahmed Bougacha [Mon, 28 Oct 2013 18:19:04 +0000 (18:19 +0000)]
TableGen: remove unused variable.
llvm-svn: 193527
Ahmed Bougacha [Mon, 28 Oct 2013 18:07:21 +0000 (18:07 +0000)]
TableGen: Refactor DAG patterns to enable parsing one pattern at a time.
llvm-svn: 193526
Ahmed Bougacha [Mon, 28 Oct 2013 18:07:17 +0000 (18:07 +0000)]
TableGen: Refactor AsmWriterEmitter to keep AsmWriterInsts.
These used to be referenced by the CGI->AWI map (in AsmWriterEmitter), but
stored in a vector local to EmitPrintInstruction. Move the vector to
AsmWriterEmitter too.
llvm-svn: 193525
Logan Chien [Mon, 28 Oct 2013 17:51:12 +0000 (17:51 +0000)]
[arm] Implement eabi_attribute, cpu, and fpu directives.
This commit allows the ARM integrated assembler to parse
and assemble the code with .eabi_attribute, .cpu, and
.fpu directives.
To implement the feature, this commit moves the code from
AttrEmitter to ARMTargetStreamers, and several new test
cases related to cortex-m4, cortex-r5, and cortex-a15 are
added.
Besides, this commit also change the Subtarget->isFPOnlySP()
to Subtarget->hasD16() to match the usage of .fpu directive.
This commit changes the test cases:
* Several .eabi_attribute directives in
2010-09-29-mc-asm-header-test.ll are removed because the .fpu
directive already cover the functionality.
* In the Cortex-A15 test case, the value for
Tag_Advanced_SIMD_arch has be changed from 1 to 2,
which is more precise.
llvm-svn: 193524
Nuno Lopes [Mon, 28 Oct 2013 16:52:38 +0000 (16:52 +0000)]
simplify ConstantRange::getSetSize()
llvm-svn: 193523
Alexey Samsonov [Mon, 28 Oct 2013 16:31:39 +0000 (16:31 +0000)]
[Sanitizer] Revert r193501 and properly fix r193448
llvm-svn: 193522
Richard Sandiford [Mon, 28 Oct 2013 13:53:37 +0000 (13:53 +0000)]
[SystemZ] Set usaAA to true
useAA significantly improves the handling of vector code that has TBAA
information attached. It also helps other cases, as shown by the testsuite
changes here. The only real downside I've seen is that it interferes with
MergeConsecutiveStores. The problem is that that optimization works top
down, starting at the first store in the chain, and looks for cases where
the chain result is only used by a single related store. These related
stores don't alias, so useAA will have rewritten all the later stores to
use a different chain input (typically the same one as the first store).
I think the advantages outweigh the disadvantages though, so for now I've
just disabled alias analysis for the unaligned-01.ll test.
llvm-svn: 193521
Dmitry Vyukov [Mon, 28 Oct 2013 13:05:32 +0000 (13:05 +0000)]
asan/msan: separate different report blocks with new lines
this makes the reports consistent with tsan, and much more readable.
llvm-svn: 193520
Dmitry Vyukov [Mon, 28 Oct 2013 12:29:32 +0000 (12:29 +0000)]
tsan: start the background thread with signals blocked, otherwise it can steal users signals
llvm-svn: 193519
Richard Sandiford [Mon, 28 Oct 2013 12:00:00 +0000 (12:00 +0000)]
[DAGCombiner] Respect volatility when checking for aliases
Making useAA() default to true for SystemZ showed that the combiner alias
analysis wasn't handling volatile accesses. This hit many of the SystemZ
tests, but I arbitrarily picked one for the purpose of this patch.
llvm-svn: 193518
Richard Sandiford [Mon, 28 Oct 2013 11:17:59 +0000 (11:17 +0000)]
Keep TBAA info when rewriting SelectionDAG loads and stores
Most SelectionDAG code drops the TBAA info when creating a new form of a
load and store (e.g. during legalization, or when converting a plain
load to an extending one). This patch tries to catch all cases where
the TBAA information can legitimately be carried over.
The patch adds alternative forms of getLoad() and getExtLoad() that take
a MachineMemOperand instead of individual fields. (The corresponding
getTruncStore() already exists.) The idea is to use the MachineMemOperand
forms when all fields are carried over (size, pointer info, isVolatile,
isNonTemporal, alignment and TBAA info). If some adjustment is being
made, e.g. to narrow the load, then we still pass the individual fields
but also pass the TBAA info.
llvm-svn: 193517
Alp Toker [Mon, 28 Oct 2013 10:26:13 +0000 (10:26 +0000)]
lit: multiprocessing platform fix attempt
The error raised by Python varies by platform(!), so let's just catch any
exception and fall back.
Thanks to Sylvestre Ledru for noticing this on a Debian / Python 2.7 system
running code coverage.
llvm-svn: 193516
Alexander Potapenko [Mon, 28 Oct 2013 08:21:35 +0000 (08:21 +0000)]
[libsanitizer] configure+make: install include/sanitizer into Clang
When building Clang with compiler-rt, copy the contents of compiler-rt/include/sanitizer into lib/clang/3.4/include/sanitizer
llvm-svn: 193515
Benjamin Kramer [Mon, 28 Oct 2013 07:30:06 +0000 (07:30 +0000)]
SCEV: Make the final add of an inbounds GEP nuw if we know that the index is positive.
We can't do this for the general case as saying a GEP with a negative index
doesn't have unsigned wrap isn't valid for negative indices.
%gep = getelementptr inbounds i32* %p, i64 -1
But an inbounds GEP cannot run past the end of address space. So we check for
the very common case of a positive index and make GEPs derived from that NUW.
Together with Andy's recent non-unit stride work this lets us analyze loops
like
void foo3(int *a, int *b) {
for (; a < b; a++) {}
}
PR12375, PR12376.
Differential Revision: http://llvm-reviews.chandlerc.com/D2033
llvm-svn: 193514
NAKAMURA Takumi [Mon, 28 Oct 2013 04:14:33 +0000 (04:14 +0000)]
StaticAnalyzer/Checkers/DynamicTypePropagation.cpp: Fix in comments. 80-col and an utf8 char.
llvm-svn: 193513
NAKAMURA Takumi [Mon, 28 Oct 2013 04:07:38 +0000 (04:07 +0000)]
Prune utf8 chars in comments.
llvm-svn: 193512