oota-llvm.git
10 years agoMC: honour IMAGE_SCN_CNT_INITIALIZED_DATA
Saleem Abdulrasool [Wed, 23 Apr 2014 21:29:34 +0000 (21:29 +0000)]
MC: honour IMAGE_SCN_CNT_INITIALIZED_DATA

Emit the flag to indicate to the assembler that a section contains data if there
is pre-populated data present.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207028 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMove the AddressPool from DwarfFile to DwarfDebug.
David Blaikie [Wed, 23 Apr 2014 21:20:10 +0000 (21:20 +0000)]
Move the AddressPool from DwarfFile to DwarfDebug.

There's only ever one address pool, not one per DWARF output file, so
let's just have one.

(similar refactoring of the string pool to come soon)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207026 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoclang-format for my previous commit (I keep forgetting... )
David Blaikie [Wed, 23 Apr 2014 21:20:07 +0000 (21:20 +0000)]
clang-format for my previous commit (I keep forgetting... )

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207025 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse pointer size function where only a pointer is expected
Matt Arsenault [Wed, 23 Apr 2014 21:10:15 +0000 (21:10 +0000)]
Use pointer size function where only a pointer is expected

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207023 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSeparate out the DWARF address pool into its own type/files.
David Blaikie [Wed, 23 Apr 2014 21:04:59 +0000 (21:04 +0000)]
Separate out the DWARF address pool into its own type/files.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207022 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove more default address space argument usage.
Matt Arsenault [Wed, 23 Apr 2014 20:58:57 +0000 (20:58 +0000)]
Remove more default address space argument usage.

These places are inconsequential in practice.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207021 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Fix the information we give to the peephole optimizer for comparison.
Quentin Colombet [Wed, 23 Apr 2014 20:43:38 +0000 (20:43 +0000)]
[ARM64] Fix the information we give to the peephole optimizer for comparison.
ANDS does not use the same encoding scheme as other xxxS instructions (e.g.,
ADDS). Take that into account to avoid wrong peephole optimization.

<rdar://problem/16693089>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207020 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't use default address space arguments in GlobalOpt
Matt Arsenault [Wed, 23 Apr 2014 20:36:10 +0000 (20:36 +0000)]
Don't use default address space arguments in GlobalOpt

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207019 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[python] Fix python bindings tests
Anders Waldenborg [Wed, 23 Apr 2014 20:32:03 +0000 (20:32 +0000)]
[python] Fix python bindings tests

Broke after the changes related to the LLVMGetSymbolFileOffset
removal in r206750

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207018 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Add a test that used to be broken that I forgot to add
Matt Arsenault [Wed, 23 Apr 2014 19:45:05 +0000 (19:45 +0000)]
R600: Add a test that used to be broken that I forgot to add

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207017 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoclang-format r207010
David Blaikie [Wed, 23 Apr 2014 19:44:08 +0000 (19:44 +0000)]
clang-format r207010

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207016 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert test to FileCheck
Matt Arsenault [Wed, 23 Apr 2014 19:32:37 +0000 (19:32 +0000)]
Convert test to FileCheck

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207015 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Fix missing/wrong scheduling model found by code inspection.
Quentin Colombet [Wed, 23 Apr 2014 19:30:26 +0000 (19:30 +0000)]
[X86] Fix missing/wrong scheduling model found by code inspection.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207014 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agollvm-build: Get rid of 'import *'
Anders Waldenborg [Wed, 23 Apr 2014 19:17:42 +0000 (19:17 +0000)]
llvm-build: Get rid of 'import *'

This allows pyflakes catching more errors in the script.

Differential Revision: http://reviews.llvm.org/D3334

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207012 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSplit out DwarfFile from DwarfDebug into its own .h/.cpp files.
David Blaikie [Wed, 23 Apr 2014 18:54:00 +0000 (18:54 +0000)]
Split out DwarfFile from DwarfDebug into its own .h/.cpp files.

Some of these types (DwarfDebug in particular) are quite large to begin
with (and I keep forgetting whether DwarfFile is in DwarfDebug or
DwarfUnit... ) so having a few smaller files seems like goodness.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207010 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoProfileData: Avoid unnecessary copies of CounterData
Justin Bogner [Wed, 23 Apr 2014 18:50:16 +0000 (18:50 +0000)]
ProfileData: Avoid unnecessary copies of CounterData

We're currently copying CounterData from InstrProfWriter into the
OnDiskHashTable, even though we don't need to, and then carelessly
leaking those copies. A const pointer is much better here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207009 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[yaml2obj][ELF] Remove unnecessary space between namespace name and
Simon Atanasyan [Wed, 23 Apr 2014 17:30:29 +0000 (17:30 +0000)]
[yaml2obj][ELF] Remove unnecessary space between namespace name and
colons.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207003 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ASan] Move the shadow range on 32-bit iOS (and iOS Simulator)
Alexander Potapenko [Wed, 23 Apr 2014 17:14:45 +0000 (17:14 +0000)]
[ASan] Move the shadow range on 32-bit iOS (and iOS Simulator)
to 0x40000000-0x60000000 to avoid address space clash with system libraries.
The solution has been proposed by tahabekireren@gmail.com in https://code.google.com/p/address-sanitizer/issues/detail?id=210
This is also known to fix some Chromium iOS tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207002 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove dead code in instcombine.
Matt Arsenault [Wed, 23 Apr 2014 16:48:40 +0000 (16:48 +0000)]
Remove dead code in instcombine.

Don't replace shifts greater than the type with the maximum shift.

This isn't hit anywhere in the tests, and somewhere else is replacing
these with undef.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207000 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoX86AsmParser.cpp: Fix memory leak at replacing movsd to movsl.
NAKAMURA Takumi [Wed, 23 Apr 2014 14:51:35 +0000 (14:51 +0000)]
X86AsmParser.cpp: Fix memory leak at replacing movsd to movsl.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206991 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agocl::ParseCommandLineOptions(): Use StringRef to receive sys::path::filename() instead...
NAKAMURA Takumi [Wed, 23 Apr 2014 14:51:23 +0000 (14:51 +0000)]
cl::ParseCommandLineOptions(): Use StringRef to receive sys::path::filename() instead of std::string.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206990 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMark llvm/test/BugPoint/compile-custom.ll as XFAIL:vg_leak.
NAKAMURA Takumi [Wed, 23 Apr 2014 14:51:12 +0000 (14:51 +0000)]
Mark llvm/test/BugPoint/compile-custom.ll as XFAIL:vg_leak.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206989 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCentralize handling of ELF_Other_ThumbFunc.
Rafael Espindola [Wed, 23 Apr 2014 14:42:32 +0000 (14:42 +0000)]
Centralize handling of ELF_Other_ThumbFunc.
No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206988 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix handling of missing DataLayout in sanitizers.
Evgeniy Stepanov [Wed, 23 Apr 2014 12:51:32 +0000 (12:51 +0000)]
Fix handling of missing DataLayout in sanitizers.

Pass::doInitialization is supposed to return False when it did not
change the program, not when a fatal error occurs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206975 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove AssemblyAnnotationWriter from NamedMDNode::print.
Rafael Espindola [Wed, 23 Apr 2014 12:23:05 +0000 (12:23 +0000)]
Remove AssemblyAnnotationWriter from NamedMDNode::print.

No functionality change, this parameter was always set to nullptr.

Patch by Robert Matusewicz!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206972 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCreate MCTargetOptions.
Evgeniy Stepanov [Wed, 23 Apr 2014 11:16:03 +0000 (11:16 +0000)]
Create MCTargetOptions.

For now it contains a single flag, SanitizeAddress, which enables
AddressSanitizer instrumentation of inline assembly.

Patch by Yuri Gorshenin.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206971 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[yaml2obj][ELF] Add a virtual destructor to the ELFYAML::Section class
Simon Atanasyan [Wed, 23 Apr 2014 11:10:55 +0000 (11:10 +0000)]
[yaml2obj][ELF] Add a virtual destructor to the ELFYAML::Section class
to prevent memory leaks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206969 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LCG] Add the first round of mutation support to the lazy call graph.
Chandler Carruth [Wed, 23 Apr 2014 11:03:03 +0000 (11:03 +0000)]
[LCG] Add the first round of mutation support to the lazy call graph.
This implements the core functionality necessary to remove an edge from
the call graph and correctly update both the basic graph and the SCC
structure. As part of that it has to run a tiny (in number of nodes)
Tarjan-style DFS walk of an SCC being mutated to compute newly formed
SCCs, etc.

This is *very rough* and a WIP. I have a bunch of FIXMEs for code
cleanup that will reduce the boilerplate in this change substantially.
I also have a bunch of simplifications to various parts of both
algorithms that I want to make, but first I'd like to have a more
holistic picture. Ideally, I'd also like more testing. I'll probably add
quite a few more unit tests as I go here to cover the various different
aspects and corner cases of removing edges from the graph.

Still, this is, so far, successfully updating the SCC graph in-place
without disrupting the identity established for the existing SCCs even
when we do challenging things like delete the critical edge that made an
SCC cycle at all and have to reform things as a tree of smaller SCCs.
Getting this to work is really critical for the new pass manager as it
is going to associate significant state with the SCC instance and needs
it to be stable. That is also the motivation behind the return of the
newly formed SCCs. Eventually, I'll wire this all the way up to the
public API so that the pass manager can use it to correctly re-enqueue
newly formed SCCs into a fresh postorder traversal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206968 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Fix formatting.
James Molloy [Wed, 23 Apr 2014 10:50:32 +0000 (10:50 +0000)]
[ARM64] Fix formatting.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206967 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LCG] Implement Tarjan's algorithm correctly this time. We have to walk
Chandler Carruth [Wed, 23 Apr 2014 10:31:17 +0000 (10:31 +0000)]
[LCG] Implement Tarjan's algorithm correctly this time. We have to walk
up the stack finishing the exploration of each entries children before
we're finished in addition to accounting for their low-links. Added
a unittest that really hammers home the need for this with interlocking
cycles that would each appear distinct otherwise and crash or compute
the wrong result. As part of this, nuke a stale fixme and bring the rest
of the implementation still more closely in line with the original
algorithm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206966 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Add a big endian version of the ARM64 target machine, and update all users.
James Molloy [Wed, 23 Apr 2014 10:26:40 +0000 (10:26 +0000)]
[ARM64] Add a big endian version of the ARM64 target machine, and update all users.

This completes the porting of r202024 (cpirker "Add AArch64 big endian Target (aarch64_be)") to ARM64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206965 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFixing typos in commit r206957
Alexey Volkov [Wed, 23 Apr 2014 10:20:31 +0000 (10:20 +0000)]
Fixing typos in commit r206957

Differential Revision: http://reviews.llvm.org/D3451

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206960 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LCG] Add some accessor methods to the SCC to allow iterating over the
Chandler Carruth [Wed, 23 Apr 2014 09:57:18 +0000 (09:57 +0000)]
[LCG] Add some accessor methods to the SCC to allow iterating over the
parents of an SCC, and add a lookup method for finding the SCC for
a given function. These aren't used yet, but will be used shortly in
some unit tests I'm adding and are really part of the broader intended
interface for the analysis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206959 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Silvermont new scheduler model
Alexey Volkov [Wed, 23 Apr 2014 08:57:09 +0000 (08:57 +0000)]
[X86] Silvermont new scheduler model
This model is not final and work is still in progress.
However there are substantial improvements on integer tests mainly because of better RAL with new scheduler.

Differential Revision: http://reviews.llvm.org/D3451

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206957 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LV] Statistics numbers for LoopVectorize introduced: a number of analyzed loops...
Alexander Musman [Wed, 23 Apr 2014 08:40:37 +0000 (08:40 +0000)]
[LV] Statistics numbers for LoopVectorize introduced: a number of analyzed loops & a number of vectorized loops.
Use -stats to see how many loops were analyzed for possible vectorization and how many of them were actually vectorized.
Patch by Zinovy Nis

Differential Revision: http://reviews.llvm.org/D3438

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206956 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LCG] Add a unittest for the LazyCallGraph. I had a weak moment and
Chandler Carruth [Wed, 23 Apr 2014 08:08:49 +0000 (08:08 +0000)]
[LCG] Add a unittest for the LazyCallGraph. I had a weak moment and
resisted this for too long. Just with the basic testing here I was able
to exercise the analysis in more detail and sift out both type signature
bugs in the API and a bug in the DFS numbering. All of these are fixed
here as well.

The unittests will be much more important for the mutation support where
it is necessary to craft minimal mutations and then inspect the state of
the graph. There is just no way to do that with a standard FileCheck
test. However, unittesting these kinds of analyses is really quite easy,
especially as they're designed with the new pass manager where there is
essentially no infrastructure required to rig up the core logic and
exercise it at an API level.

As a minor aside about the DFS numbering bug, the DFS numbering used in
LCG is a bit unusual. Rather than numbering from 0, we number from 1,
and use 0 as the sentinel "unvisited" state. Other implementations often
use '-1' for this, but I find it easier to deal with 0 and it shouldn't
make any real difference provided someone doesn't write silly bugs like
forgetting to actually initialize the DFS numbering. Oops. ;]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206954 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoX86Disassembler - fixed a bug in immediate print
Elena Demikhovsky [Wed, 23 Apr 2014 07:21:04 +0000 (07:21 +0000)]
X86Disassembler - fixed a bug in immediate print

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206953 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIntegrated assbemler, macros: added 'vararg' argument qualifier support.
Stepan Dyatkovskiy [Wed, 23 Apr 2014 06:56:28 +0000 (06:56 +0000)]
Integrated assbemler, macros: added 'vararg' argument qualifier support.
Note, currently we have no 'vararg' support for darwin macros.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206951 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM64] Enable feature predicates for NEON / FP / CRYPTO.
Kevin Qin [Wed, 23 Apr 2014 06:22:48 +0000 (06:22 +0000)]
[ARM64] Enable feature predicates for NEON / FP / CRYPTO.

AArch64 has feature predicates for NEON, FP and CRYPTO instructions.
This allows the compiler to generate code without using FP, NEON
or CRYPTO instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206949 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LCG] Hoist the logic for forming a new SCC from the top of the DFSStack
Chandler Carruth [Wed, 23 Apr 2014 06:09:03 +0000 (06:09 +0000)]
[LCG] Hoist the logic for forming a new SCC from the top of the DFSStack
into a helper function. I plan to re-use it for doing incremental
DFS-based updates to the SCCs when we mutate the call graph.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206948 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[LCG] Switch the Callee sets to be DenseMaps pointing to the index into
Chandler Carruth [Wed, 23 Apr 2014 04:00:17 +0000 (04:00 +0000)]
[LCG] Switch the Callee sets to be DenseMaps pointing to the index into
the Callee list. This is going to be quite important to prevent removal
from going quadratic. No functionality changed at this point, this is
one of the refactoring patches I've broken out of my initial work toward
mutation updates of the call graph.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206938 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix test/CodeGen/arm.ll
Reid Kleckner [Wed, 23 Apr 2014 01:09:29 +0000 (01:09 +0000)]
Fix test/CodeGen/arm.ll

The 'CHECK: add' line was occasionally matching against the filename,
breaking the subsequent CHECK-NOT.  Also use CHECK-LABEL.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206936 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRequisite reformatting for previous commit.
David Blaikie [Tue, 22 Apr 2014 23:09:36 +0000 (23:09 +0000)]
Requisite reformatting for previous commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206927 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPush memory ownership of DwarfUnits into clients of DwarfFile.
David Blaikie [Tue, 22 Apr 2014 22:39:41 +0000 (22:39 +0000)]
Push memory ownership of DwarfUnits into clients of DwarfFile.

This prompted me to push references through most of DwarfDebug. Sorry
for the churn.

Honestly it's a bit silly that we're passing around units all over the
place like that anyway and I think it's mostly due to the DIE attribute
adding utility functions being utilities in DwarfUnit. I should have
another go at moving them out of DwarfUnit...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206925 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] Add a note to docs/README.txt
Sean Silva [Tue, 22 Apr 2014 21:47:53 +0000 (21:47 +0000)]
[docs] Add a note to docs/README.txt

Added note to docs/README.txt on how to check the reachibility of
external links in the documentation.

Patch by Dan Liew!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206924 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoChange the prototype for MCContext::FatalError() so it can be called
Kevin Enderby [Tue, 22 Apr 2014 21:42:18 +0000 (21:42 +0000)]
Change the prototype for MCContext::FatalError() so it can be called
from places like MCCodeEmitter() in the MC backend when the
MCContext is const.

I was going to use this in my change for r206669 but Jim convinced
me to use an assert there. But this still is a good tweak.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206923 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse std::unique_ptr to handle ownership of DwarfUnits in DwarfFile.
David Blaikie [Tue, 22 Apr 2014 21:27:37 +0000 (21:27 +0000)]
Use std::unique_ptr to handle ownership of DwarfUnits in DwarfFile.

So Chandler - how about those range algorithms? (would really love a
dereferencing range adapter for this sort of stuff)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206921 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoWhitespace
Rui Ueyama [Tue, 22 Apr 2014 19:52:05 +0000 (19:52 +0000)]
Whitespace

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206919 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoNo need to check condition after grow()
Rui Ueyama [Tue, 22 Apr 2014 19:47:26 +0000 (19:47 +0000)]
No need to check condition after grow()

r206916 was not logically the same as the previous code because the
goto statements did not create loop. This should be the same as the
previous code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206918 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFollow aliases when determining if a symbol is thumb.
Rafael Espindola [Tue, 22 Apr 2014 19:11:07 +0000 (19:11 +0000)]
Follow aliases when determining if a symbol is thumb.

This fixes pr19484.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206917 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoReplace loops using goto with plain while loops
Rui Ueyama [Tue, 22 Apr 2014 19:07:14 +0000 (19:07 +0000)]
Replace loops using goto with plain while loops

Goto statements jumping into previous inner blocks are pretty confusing
to read even though in this case they are valid. No reason to not use
while loops there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206916 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Constant Hoisting] Materialize the constant before the cloned cast instruction.
Juergen Ributzka [Tue, 22 Apr 2014 18:06:58 +0000 (18:06 +0000)]
[Constant Hoisting] Materialize the constant before the cloned cast instruction.

In the case where the constant comes from a cloned cast instruction, the
materialization code has to go before the cloned cast instruction.

This commit fixes the method that finds the materialization insertion point
by making it aware of this case.

This fixes <rdar://problem/15532441>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206913 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Constant Hoisting] Print the instructions in the correct order for debugging. No...
Juergen Ributzka [Tue, 22 Apr 2014 18:06:51 +0000 (18:06 +0000)]
[Constant Hoisting] Print the instructions in the correct order for debugging. No functional change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206912 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix DataLayout::operator==().
Rafael Espindola [Tue, 22 Apr 2014 17:47:03 +0000 (17:47 +0000)]
Fix DataLayout::operator==().

Patch by Maks Naumov!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206911 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix the assembler to print a better relocatable expression error
Kevin Enderby [Tue, 22 Apr 2014 17:27:29 +0000 (17:27 +0000)]
Fix the assembler to print a better relocatable expression error
diagnostic that includes location information.

Currently if one has this assembly:

.quad (0x1234 + (4 * SOME_VALUE))

where SOME_VALUE is undefined ones gets the less than
useful error message with no location information:

% clang -c x.s
clang -cc1as: fatal error: error in backend: expected relocatable expression

With this fix one now gets a more useful error message
with location information:

% clang -c x.s
x.s:5:8: error: expected relocatable expression
 .quad (0x1234 + (4 * SOME_VALUE))
       ^

To do this I plumbed the SMLoc through the MCObjectStreamer
EmitValue() and EmitValueImpl() interfaces so it could be used
when creating the MCFixup.

rdar://12391022

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206906 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSimplify address pool index assignment.
David Blaikie [Tue, 22 Apr 2014 17:21:40 +0000 (17:21 +0000)]
Simplify address pool index assignment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206905 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Emit error instead of unreachable on function call
Matt Arsenault [Tue, 22 Apr 2014 16:42:00 +0000 (16:42 +0000)]
R600: Emit error instead of unreachable on function call

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206904 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Reorganize SIInstructions.td
Tom Stellard [Tue, 22 Apr 2014 16:33:57 +0000 (16:33 +0000)]
R600/SI: Reorganize SIInstructions.td

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206902 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAVX-512: store and truncstore for i1 values
Elena Demikhovsky [Tue, 22 Apr 2014 14:13:10 +0000 (14:13 +0000)]
AVX-512: store and truncstore for i1 values

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206897 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove DOS CRLF.
NAKAMURA Takumi [Tue, 22 Apr 2014 13:35:50 +0000 (13:35 +0000)]
Remove DOS CRLF.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206894 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64/ARM64: more testing from AArch64 to ARM64
Tim Northover [Tue, 22 Apr 2014 12:45:47 +0000 (12:45 +0000)]
AArch64/ARM64: more testing from AArch64 to ARM64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206889 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64/ARM64: make use of ANDS and BICS instructions for comparisons.
Tim Northover [Tue, 22 Apr 2014 12:45:42 +0000 (12:45 +0000)]
AArch64/ARM64: make use of ANDS and BICS instructions for comparisons.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206888 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64/ARM64: add extra testing from AArch64 to ARM64
Tim Northover [Tue, 22 Apr 2014 12:45:32 +0000 (12:45 +0000)]
AArch64/ARM64: add extra testing from AArch64 to ARM64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206887 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Require HasBMI2 for the new BZHI tablegen patterns.
Lang Hames [Tue, 22 Apr 2014 12:04:53 +0000 (12:04 +0000)]
[X86] Require HasBMI2 for the new BZHI tablegen patterns.

Evidently tablegen doesn't infer this from the HasBMI2 predicate on the BZHI
instructions. This should fix the recent bot failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206885 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AVX512] Implemented integer conversions up/down with masking.
Robert Khasanov [Tue, 22 Apr 2014 11:36:19 +0000 (11:36 +0000)]
[AVX512] Implemented integer conversions up/down with masking.
Added encoding tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206884 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[asan] Support outline instrumentation for wide types and delete dead code, patch...
Kostya Serebryany [Tue, 22 Apr 2014 11:19:45 +0000 (11:19 +0000)]
[asan] Support outline instrumentation for wide types and delete dead code, patch by Yuri Gribov

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206883 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Remove Tablegen def of X86bzhi SDNode: It's not needed as of r206879.
Lang Hames [Tue, 22 Apr 2014 10:50:46 +0000 (10:50 +0000)]
[X86] Remove Tablegen def of X86bzhi SDNode: It's not needed as of r206879.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206880 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Use tablegen instead of DAG combines to match BZHI instructions, as
Lang Hames [Tue, 22 Apr 2014 10:41:56 +0000 (10:41 +0000)]
[X86] Use tablegen instead of DAG combines to match BZHI instructions, as
suggested by Ben Kramer in review of r206738.

Thanks again Ben!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206879 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] Clang-format MipsAsmParser.
Matheus Almeida [Tue, 22 Apr 2014 10:15:54 +0000 (10:15 +0000)]
[mips] Clang-format MipsAsmParser.

No functional changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206878 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64/ARM64: enable various AArch64 tests on ARM64.
Tim Northover [Tue, 22 Apr 2014 10:10:26 +0000 (10:10 +0000)]
AArch64/ARM64: enable various AArch64 tests on ARM64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206877 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64/ARM64: add patterns for scalar_to_vector/extract pairs
Tim Northover [Tue, 22 Apr 2014 10:10:18 +0000 (10:10 +0000)]
AArch64/ARM64: add patterns for scalar_to_vector/extract pairs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206876 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64/ARM64: mark fmul intrinsic as commutative.
Tim Northover [Tue, 22 Apr 2014 10:10:14 +0000 (10:10 +0000)]
AArch64/ARM64: mark fmul intrinsic as commutative.

This gives DAG patterns matching indexed patterns where either side is an
indexed vector.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206875 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: disable emission of __XYZvfp in soft-float environment.
Tim Northover [Tue, 22 Apr 2014 10:10:09 +0000 (10:10 +0000)]
ARM: disable emission of __XYZvfp in soft-float environment.

The point of these calls is to allow Thumb-1 code to make use of the VFP unit
to perform its operations. This is not desirable with -msoft-float, since most
of the reasons you'd want that apply equally to the runtime library.

rdar://problem/13766161

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206874 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix an infinite loop bug in DAG Combine about keeping transfering between ANY_EXTEND...
Hao Liu [Tue, 22 Apr 2014 09:57:06 +0000 (09:57 +0000)]
Fix an infinite loop bug in DAG Combine about keeping transfering between ANY_EXTEND and SIGN_EXTEND.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206873 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Don't use BZHI for short masks (>=32 bits). Thanks to Ben Kramer for the
Lang Hames [Tue, 22 Apr 2014 07:40:34 +0000 (07:40 +0000)]
[X86] Don't use BZHI for short masks (>=32 bits). Thanks to Ben Kramer for the
review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206869 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert "Use value semantics to manage DbgVariables rather than dynamic allocation...
David Blaikie [Tue, 22 Apr 2014 05:41:06 +0000 (05:41 +0000)]
Revert "Use value semantics to manage DbgVariables rather than dynamic allocation/pointers."

This reverts commit r206780.

This commit was regressing gdb.opt/inline-locals.exp in the GDB 7.5 test
suite. Reverting until I can fix the issue.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206867 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse unique_ptr to manage ParsedBinariesAndObjects in LLVMSymbolizer
David Blaikie [Tue, 22 Apr 2014 05:26:14 +0000 (05:26 +0000)]
Use unique_ptr to manage ParsedBinariesAndObjects in LLVMSymbolizer

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206866 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Change how vector truncating stores are packed.
Matt Arsenault [Tue, 22 Apr 2014 04:11:14 +0000 (04:11 +0000)]
R600: Change how vector truncating stores are packed.

Don't introduce new operations on an illegal sub 32-bit type.
Do the operations on a 32-bit value, and then use a truncating store.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206864 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Make sign_extend_inreg legal.
Matt Arsenault [Tue, 22 Apr 2014 03:49:30 +0000 (03:49 +0000)]
R600: Make sign_extend_inreg legal.

Don't know why I didn't just do this in the first place.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206862 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AArch64] Enable global merge pass.
Jiangning Liu [Tue, 22 Apr 2014 03:33:26 +0000 (03:33 +0000)]
[AArch64] Enable global merge pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206861 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Skip irreducible backedges inside functions
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:53 +0000 (03:31 +0000)]
blockfreq: Skip irreducible backedges inside functions

The branch that skips irreducible backedges was only active when
propagating mass at the top-level.  In particular, when propagating mass
through a loop recognized by `LoopInfo` with irreducible control flow
inside, irreducible backedges would not be skipped.

Not sure where that idea came from, but the result was that mass was
lost until after loop exit.  Added a testcase that covers this case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206860 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Rename PackagedLoops => Loops
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:50 +0000 (03:31 +0000)]
blockfreq: Rename PackagedLoops => Loops

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206859 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Use a pointer for ContainingLoop too
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:44 +0000 (03:31 +0000)]
blockfreq: Use a pointer for ContainingLoop too

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206858 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Use pointers to loops instead of an index
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:37 +0000 (03:31 +0000)]
blockfreq: Use pointers to loops instead of an index

Store pointers directly to loops inside the nodes.  This could have been
done without changing the type stored in `std::vector<>`.  However,
rather than computing the number of loops before constructing them
(which `LoopInfo` doesn't provide directly), I've switched to a
`vector<unique_ptr<LoopData>>`.

This adds some heap overhead, but the number of loops is typically
small.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206857 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Implement clear() explicitly
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:34 +0000 (03:31 +0000)]
blockfreq: Implement clear() explicitly

This was implicitly with copy assignment before, which fails to actually
clear `std::vector<>`'s heap storage.  Move assignment would work, but
since MSVC can't imply those anyway, explicitly `clear()`-ing members
makes more sense.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206856 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Rename PackagedLoopData => LoopData
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:31 +0000 (03:31 +0000)]
blockfreq: Rename PackagedLoopData => LoopData

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206855 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Move PackagedLoopData above WorkingData
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:25 +0000 (03:31 +0000)]
blockfreq: Move PackagedLoopData above WorkingData

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206854 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoblockfreq: Remove "dead" comment
Duncan P. N. Exon Smith [Tue, 22 Apr 2014 03:31:23 +0000 (03:31 +0000)]
blockfreq: Remove "dead" comment

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206853 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 03:10:36 +0000 (03:10 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all of the header #include lines, tools edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206848 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 03:07:47 +0000 (03:07 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all of the header #include lines, lib/Support edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206847 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 03:06:00 +0000 (03:06 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all of the header #include lines, TableGen edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206846 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 03:04:17 +0000 (03:04 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all the header #include lines. This updates most of the
miscellaneous other lib/... directories. A few left though.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206845 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 02:55:47 +0000 (02:55 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all of the header #include lines, lib/Transforms/...
edition.

This one is tricky for two reasons. We again have a couple of passes
that define something else before the includes as well. I've sunk their
name macros with the DEBUG_TYPE.

Also, InstCombine contains headers that need DEBUG_TYPE, so now those
headers #define and #undef DEBUG_TYPE around their code, leaving them
well formed modular headers. Fixing these headers was a large motivation
for all of these changes, as "leaky" macros of this form are hard on the
modules implementation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206844 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 02:48:03 +0000 (02:48 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all the header #include lines, lib/Analysis/...
edition.

This one has a bit extra as there were *other* #define's before #include
lines in addition to DEBUG_TYPE. I've sunk all of them as a block.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206843 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 02:41:26 +0000 (02:41 +0000)]
[Modules] Fix potential ODR violations by sinking the DEBUG_TYPE
definition below all of the header #include lines, lib/Target/...
edition.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206842 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[cleanup] Fix two headers where we included a standard library header
Chandler Carruth [Tue, 22 Apr 2014 02:28:45 +0000 (02:28 +0000)]
[cleanup] Fix two headers where we included a standard library header
after including the generated code from tablegen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206841 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[cleanup] Fix another place where we were including the tablegen'ed code
Chandler Carruth [Tue, 22 Apr 2014 02:25:17 +0000 (02:25 +0000)]
[cleanup] Fix another place where we were including the tablegen'ed code
of a '.inc' file before including actual headers. In this case we had
both duplicated a header's include and were including a standard header.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206840 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoThis reverts r206828 until David has time to figure out that is going on.
Quentin Colombet [Tue, 22 Apr 2014 02:17:11 +0000 (02:17 +0000)]
This reverts r206828 until David has time to figure out that is going on.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206839 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[cleanup] Lift using directives, DEBUG_TYPE definitions, and even some
Chandler Carruth [Tue, 22 Apr 2014 02:03:14 +0000 (02:03 +0000)]
[cleanup] Lift using directives, DEBUG_TYPE definitions, and even some
system headers above the includes of generated '.inc' files that
actually contain code. In a few targets this was already done pretty
consistently, but it wasn't done *really* consistently anywhere. It is
strictly cleaner IMO and necessary in a bunch of places where the
DEBUG_TYPE is referenced from the generated code. Consistency with the
necessary places trumps. Hopefully the build bots are OK with the
movement of intrin.h...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206838 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Modules] Remove potential ODR violations by sinking the DEBUG_TYPE
Chandler Carruth [Tue, 22 Apr 2014 02:02:50 +0000 (02:02 +0000)]
[Modules] Remove potential ODR violations by sinking the DEBUG_TYPE
define below all header includes in the lib/CodeGen/... tree. While the
current modules implementation doesn't check for this kind of ODR
violation yet, it is likely to grow support for it in the future. It
also removes one layer of macro pollution across all the included
headers.

Other sub-trees will follow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206837 91177308-0d34-0410-b5e6-96231b3b80d8