Sanjoy Das [Wed, 27 Sep 2017 21:45:21 +0000 (21:45 +0000)]
LoopDeletion: use return value instead of passing in LPMUpdater; NFC
I will use this refactoring in a later patch.
llvm-svn: 314351
Sanjoy Das [Wed, 27 Sep 2017 21:45:19 +0000 (21:45 +0000)]
Rename LoopUnrollStatus to LoopUnrollResult; NFC
A "Result" suffix is more appropriate here
llvm-svn: 314350
Rui Ueyama [Wed, 27 Sep 2017 21:38:02 +0000 (21:38 +0000)]
Fix off-by-one error in TarWriter.
The tar format originally supported up to 99 byte filename. The two
extensions are proposed later: Ustar or PAX.
In the UStar extension, a pathanme is split at a '/' and its "prefix"
and "suffix" are stored in different locations in the tar header. Since
"prefix" can be up to 155 byte, it can represent up to 254 byte
filename (but exact limit depends on the location of '/' character in
a pathname.)
Our TarWriter first attempt to use UStar extension and then fallback to
PAX extension.
But there's a bug in UStar header creation. "Suffix" part must be a NUL-
terminated string, but we didn't handle it correctly. As a result, if
your filename just 100 characters long, the last character was droppped.
This patch fixes the issue.
Differential Revision: https://reviews.llvm.org/D38149
llvm-svn: 314349
Brian Gesiak [Wed, 27 Sep 2017 21:37:33 +0000 (21:37 +0000)]
[CMake] Fix typo: "in-tree" -> "in-source" (NFC)
Summary:
*In-source builds* of LLVM, in which a user invokes `cmake` from within the
LLVM source directory, or invokes `cmake -B/path/to/source/dir/of/llvm`,
are explicitly checked for and disallowed by LLVM's `CMakeLists.txt`.
*In-tree builds*, on the other hand, refer to when the source directories
of projects such as Clang are nested within the `llvm/tools` source
directory. These are not disallowed, and are in fact a common way of
building LLVM and Clang.
Revise the comment to match the logic underneath it: it checks for an
"in-source build", not an "in-tree build".
Reviewers: beanz
Reviewed By: beanz
Subscribers: mgorny
Differential Revision: https://reviews.llvm.org/D38317
llvm-svn: 314348
Francis Ricci [Wed, 27 Sep 2017 21:35:33 +0000 (21:35 +0000)]
Revert "Add support for custom loaders to symbolizer"
This broke the windows buildbots, revert for now.
This reverts commit
24050b5ddef42f6f3306aa94d4a1f42a7893a9a7.
llvm-svn: 314347
Don Hinton [Wed, 27 Sep 2017 21:19:56 +0000 (21:19 +0000)]
Cleanup some problems with LLVM_ENABLE_DUMP in release builds, and
always set LLVM_ENABLE_DUMP=ON for +Asserts builds.
Differential Revision: https://reviews.llvm.org/D38306
llvm-svn: 314346
Rui Ueyama [Wed, 27 Sep 2017 21:19:24 +0000 (21:19 +0000)]
Do not remove a target file in FileOutputBuffer::create().
FileOutputBuffer::create() attempts to remove a target file if the file
is a regular one, which results in an unexpected result in a failure
scenario.
If something goes wrong and the user of FileOutputBuffer decides to not
call commit(), it leaves nothing. An existing file is removed, and no
new file is created.
What we should do is to atomically replace an existing file with a new
file using rename(), so that it wouldn't remove an existing file without
creating a new one.
Differential Revision: https://reviews.llvm.org/D38283
llvm-svn: 314345
Kostya Kortchinsky [Wed, 27 Sep 2017 21:10:25 +0000 (21:10 +0000)]
[scudo] Temporary disabling the valloc test on armhf
Summary:
Weird failure where `errno != ENOMEM` on valloc failure. The returned pointer
is null since it passes the previous assert, so this shouldn't happen.
http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-full/builds/10931
http://lab.llvm.org:8011/builders/clang-cmake-thumbv7-a15-full-sh/builds/2469
Disabling until we figure out what's going on.
Reviewers: alekseyshl
Reviewed By: alekseyshl
Subscribers: aemerson, srhines, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D38324
llvm-svn: 314344
Jason Molenda [Wed, 27 Sep 2017 20:56:32 +0000 (20:56 +0000)]
Update the Objective-C runtime interface code to handle objc objects
whose isa is an index instead of a pointer. Currently, this type
of isa encoding is only used on watchos.
<rdar://problem/
34675497>
llvm-svn: 314343
Francis Ricci [Wed, 27 Sep 2017 20:56:04 +0000 (20:56 +0000)]
Add support for custom loaders to symbolizer
Change-Id: I5594bd6b216deca2c73cf0a7001f9aec1e803c60
llvm-svn: 314342
Jessica Paquette [Wed, 27 Sep 2017 20:47:39 +0000 (20:47 +0000)]
[MachineOutliner] AArch64: Avoid saving + restoring LR if possible
This commit allows the outliner to avoid saving and restoring the link register
on AArch64 when it is dead within an entire class of candidates.
This introduces changes to the way the outliner interfaces with the target.
For example, the target now interfaces with the outliner using a
MachineOutlinerInfo struct rather than by using getOutliningCallOverhead and
getOutliningFrameOverhead.
This also improves several comments on the outliner's cost model.
https://reviews.llvm.org/D36721
llvm-svn: 314341
Jonathan Peyton [Wed, 27 Sep 2017 20:36:27 +0000 (20:36 +0000)]
Remove unnecessary semicolons
Removes semicolons after if {} blocks, function definitions, etc.
I was able to apply the large OMPT patch cleanly on top of this one
with no conflicts.
llvm-svn: 314340
Craig Topper [Wed, 27 Sep 2017 20:34:17 +0000 (20:34 +0000)]
Revert r314249 "Recommit r314151 "[X86] Make all the NOREX CodeGenOnly instructions into postRA pseudos like the NOREX version of TEST."""
This caused PR34751
llvm-svn: 314339
Craig Topper [Wed, 27 Sep 2017 20:34:13 +0000 (20:34 +0000)]
Revert r314248 "[X86] Don't emit X86::MOV8rr_NOREX from X86InstrInfo::copyPhysReg."
This contributed to PR34751
llvm-svn: 314338
Simon Pilgrim [Wed, 27 Sep 2017 20:19:53 +0000 (20:19 +0000)]
[X86][SSE] Pull out variable shuffle mask combine logic. NFCI.
Hopefully this will make it easier to vary the combine depth threshold per-target.
llvm-svn: 314337
Sean Callanan [Wed, 27 Sep 2017 19:57:58 +0000 (19:57 +0000)]
Add support for remembering origins to ExternalASTMerger
ExternalASTMerger has hitherto relied on being able to look up
any Decl through its named DeclContext chain. This works for
many cases, but causes problems for function-local structs,
which cannot be looked up in their containing FunctionDecl. An
example case is
void f() {
{ struct S { int a; }; }
{ struct S { bool b; }; }
}
It is not possible to lookup either of the two Ses individually
(or even to provide enough information to disambiguate) after
parsing is over; and there is typically no need to, since they
are invisible to the outside world.
However, ExternalASTMerger needs to be able to complete either
S on demand. This led to an XFAIL on test/Import/local-struct,
which this patch removes. The way the patch works is:
It defines a new data structure, ExternalASTMerger::OriginMap,
which clients are expected to maintain (default-constructing
if the origin does not have an ExternalASTMerger servicing it)
As DeclContexts are imported, if they cannot be looked up by
name they are placed in the OriginMap. This allows
ExternalASTMerger to complete them later if necessary.
As DeclContexts are imported from an origin that already has
its own OriginMap, the origins are forwarded – but only for
those DeclContexts that are actually used. This keeps the
amount of stored data minimal.
The patch also applies several improvements from review:
- Thoroughly documents the interface to ExternalASTMerger;
- Adds optional logging to help track what's going on; and
- Cleans up a bunch of braces and dangling elses.
Differential Revision: https://reviews.llvm.org/D38208
llvm-svn: 314336
Than McIntosh [Wed, 27 Sep 2017 19:34:00 +0000 (19:34 +0000)]
[CodeGen] Emit necessary .note sections for -fsplit-stack
Summary:
According to https://gcc.gnu.org/wiki/SplitStacks, the linker expects a zero-sized .note.GNU-split-stack section if split-stack is used (and also .note.GNU-no-split-stack section if it also contains non-split-stack functions), so it can handle the cases where a split-stack function calls non-split-stack function.
This change adds the sections if needed.
Fixes PR #34670.
Reviewers: thanm, rnk, luqmana
Reviewed By: rnk
Subscribers: llvm-commits
Patch by Cherry Zhang <cherryyz@google.com>
Differential Revision: https://reviews.llvm.org/D38051
llvm-svn: 314335
Artem Belevich [Wed, 27 Sep 2017 19:07:15 +0000 (19:07 +0000)]
[CUDA] Work around conflicting function definitions in CUDA-9 headers.
Differential Revision: https://reviews.llvm.org/D38326
llvm-svn: 314334
Manoj Gupta [Wed, 27 Sep 2017 19:06:47 +0000 (19:06 +0000)]
[builtins] ARM: Revert r314284, r314285 and r314289
Revert r314284, r314285 and r314289 because of a reported
breakage in armv7k watchos builder.
llvm-svn: 314333
Craig Topper [Wed, 27 Sep 2017 18:56:20 +0000 (18:56 +0000)]
[X86] Rewrite the zero vector checks in lowerV2X128VectorShuffle to use the Zeroable APInt
We already have zeroable bits in an APInt. We might as well use that instead of checking for an all zero BUILD_VECTOR.
Differential Revision: https://reviews.llvm.org/D37950
llvm-svn: 314332
Craig Topper [Wed, 27 Sep 2017 18:36:45 +0000 (18:36 +0000)]
[X86] In combineLoopSADPattern, pad result with zeros and use full size add instead of using a smaller add and inserting.
In some cases the result psadbw is smaller than the type of the add that started the match. Currently in these cases we are using a smaller add and inserting the result.
If we instead combine the psadbw with zeros and use the full size add we can take advantage of implicit zeroing we get if we emit a narrower move before the add.
In a future patch, I want to make isel aware that the psadbw itself already zeroed the upper bits and remove the move entirely.
Differential Revision: https://reviews.llvm.org/D37453
llvm-svn: 314331
Jonas Hahnfeld [Wed, 27 Sep 2017 18:12:36 +0000 (18:12 +0000)]
[OpenMP] Fix translation of target args
ToolChain::TranslateArgs() returns nullptr if no changes are performed.
This would currently mean that OpenMPArgs are lost. Patch fixes this
by falling back to simply using OpenMPArgs in that case.
Differential Revision: https://reviews.llvm.org/D38259
llvm-svn: 314330
Jonas Hahnfeld [Wed, 27 Sep 2017 18:12:34 +0000 (18:12 +0000)]
[OpenMP] Fix passing of -m arguments to device toolchain
AuxTriple is not set if host and device share a toolchain. Also,
removing an argument modifies the DAL which needs to be returned
for future use.
(Move tests back to offload-openmp.c as they are not related to GPUs.)
Differential Revision: https://reviews.llvm.org/D38258
llvm-svn: 314329
Jonas Hahnfeld [Wed, 27 Sep 2017 18:12:31 +0000 (18:12 +0000)]
[OpenMP] Fix memory leak when translating arguments
Parsing the argument after -Xopenmp-target allocates memory that needs
to be freed. Associate it with the final DerivedArgList after we know
which one will be used.
Differential Revision: https://reviews.llvm.org/D38257
llvm-svn: 314328
Nico Weber [Wed, 27 Sep 2017 17:57:50 +0000 (17:57 +0000)]
clang-format/java: Unbreak genenrics formatting after r299952.
https://reviews.llvm.org/rL299952 merged '>>>' tokens into a single
JavaRightLogicalShift token. This broke formatting of generics nested more than
two deep, e.g. Foo<Bar<Baz>>> because the '>>>' now weren't three '>' for
parseAngle().
Luckily, just deleting JavaRightLogicalShift fixes things without breaking the
test added in r299952, so do that.
https://reviews.llvm.org/D38291
llvm-svn: 314325
Alexey Bataev [Wed, 27 Sep 2017 17:42:49 +0000 (17:42 +0000)]
[SLP] Fix crash on propagate IR flags for undef operands of min/max
reductions.
If both operands of the newly created SelectInst are Undefs the
resulting operation is also Undef, not SelectInst. It may cause crashes
when trying to propagate IR flags because function expects exactly
SelectInst instruction, nothing else.
llvm-svn: 314323
Manoj Gupta [Wed, 27 Sep 2017 17:36:25 +0000 (17:36 +0000)]
[ARM] builtins: Replace abort by assert in clear_cache.
Summary:
__builtion___clear_cache maps to clear_cache function. On Linux,
clear_cache functions makes a syscall and does an abort if syscall fails.
Replace the abort by an assert so that non-debug builds do not abort
if the syscall fails.
Fixes PR34588.
Reviewers: rengolin, compnerd, srhines, peter.smith, joerg
Reviewed By: rengolin
Subscribers: aemerson, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D37788
llvm-svn: 314322
Roman Lebedev [Wed, 27 Sep 2017 17:24:34 +0000 (17:24 +0000)]
[support] mapped_file_region: and fix the windows code too
Followup for r314312 / r314313
Sorry, i really failed to fully grep all the codebase :/
llvm-svn: 314321
Chad Rosier [Wed, 27 Sep 2017 17:16:51 +0000 (17:16 +0000)]
[InstCombine] Gating select arithmetic optimization.
These changes faciliate positive behavior for arithmetic based select
expressions that match its translation criteria, keeping code size gated to
neutral or improved scenarios.
Patch by Michael Berg <michael_c_berg@apple.com>!
Differential Revision: https://reviews.llvm.org/D38263
llvm-svn: 314320
Geoff Berry [Wed, 27 Sep 2017 17:14:10 +0000 (17:14 +0000)]
[AArch64][Falkor] Ignore SP based loads in HW prefetch fixups.
Reviewers: mcrosier
Subscribers: aemerson, rengolin, javed.absar, kristof.beyls
Differential Revision: https://reviews.llvm.org/D38301
llvm-svn: 314319
Alex Shlyapnikov [Wed, 27 Sep 2017 17:10:49 +0000 (17:10 +0000)]
[Sanitizer] Disable compact size class tests on Android
Fixing test failure on Android introduced in D38245. Compact size class
maps defined there are not to be used on Android.
llvm-svn: 314318
Anastasia Stulova [Wed, 27 Sep 2017 17:03:35 +0000 (17:03 +0000)]
[OpenCL] Fixed CL version in failing test.
llvm-svn: 314317
Javed Absar [Wed, 27 Sep 2017 16:39:17 +0000 (16:39 +0000)]
[Misched] : Fix typo in comment. NFC.
llvm-svn: 314316
Sanjay Patel [Wed, 27 Sep 2017 16:32:56 +0000 (16:32 +0000)]
[SLP] fix typos/formatting; NFC
llvm-svn: 314315
Sean Eveson [Wed, 27 Sep 2017 16:20:07 +0000 (16:20 +0000)]
Revert "[llvm-cov] Create directory structure when filtering using -name*= options"
Test failures.
llvm-svn: 314314
Roman Lebedev [Wed, 27 Sep 2017 16:08:33 +0000 (16:08 +0000)]
[Support] mapped_file_region::size() returns size_t
Fixup last commit, found by clang-stage1-cmake-RA-incremental bot.
llvm-svn: 314313
Roman Lebedev [Wed, 27 Sep 2017 15:59:16 +0000 (15:59 +0000)]
[Support] mapped_file_region: store size as size_t
Summary:
Found when testing stage-2 build with D38101.
```
In file included from /build/llvm/lib/Support/Path.cpp:1045:
/build/llvm/lib/Support/Unix/Path.inc:648:14: error: comparison 'uint64_t' (aka 'unsigned long') >
18446744073709551615 is always false [-Werror,-Wtautological-constant-compare]
if (length > std::numeric_limits<size_t>::max()) {
~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
`size_t` is `uint64_t` here, apparently, thus any `uint64_t` value
always fits into `size_t`.
Initial patch was to use some preprocessor logic to
not check if the size is known to fit at compile time.
But Zachary Turner suggested using this approach.
Reviewers: Bigcheese, rafael, zturner, mehdi_amini
Reviewed by (via email): zturner
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38132
llvm-svn: 314312
Alex Shlyapnikov [Wed, 27 Sep 2017 15:38:05 +0000 (15:38 +0000)]
[Sanitizers] Allocator: new "release memory to OS" implementation
Summary:
The current implementation of the allocator returning freed memory
back to OS (controlled by allocator_release_to_os_interval_ms flag)
requires sorting of the free chunks list, which has two major issues,
first, when free list grows to millions of chunks, sorting, even the
fastest one, is just too slow, and second, sorting chunks in place
is unacceptable for Scudo allocator as it makes allocations more
predictable and less secure.
The proposed approach is linear in complexity (altough requires quite
a bit more temporary memory). The idea is to count the number of free
chunks on each memory page and release pages containing free chunks
only. It requires one iteration over the free list of chunks and one
iteration over the array of page counters. The obvious disadvantage
is the allocation of the array of the counters, but even in the worst
case we support (4T allocator space, 64 buckets, 16 bytes bucket size,
full free list, which leads to 2 bytes per page counter and ~17M page
counters), requires just about 34Mb of the intermediate buffer (comparing
to ~64Gb of actually allocated chunks) and usually it stays under 100K
and released after each use. It is expected to be a relatively rare event,
releasing memory back to OS, keeping the buffer between those runs
and added complexity of the bookkeeping seems unnesessary here (it can
always be improved later, though, never say never).
The most interesting problem here is how to calculate the number of chunks
falling into each memory page in the bucket. Skipping all the details,
there are three cases when the number of chunks per page is constant:
1) P >= C, P % C == 0 --> N = P / C
2) C > P , C % P == 0 --> N = 1
3) C <= P, P % C != 0 && C % (P % C) == 0 --> N = P / C + 1
where P is page size, C is chunk size and N is the number of chunks per
page and the rest of the cases, where the number of chunks per page is
calculated on the go, during the page counter array iteration.
Among the rest, there are still cases where N can be deduced from the
page index, but they require not that much less calculations per page
than the current "brute force" way and 2/3 of the buckets fall into
the first three categories anyway, so, for the sake of simplicity,
it was decided to stick to those two variations. It can always be
refined and improved later, should we see that brute force way slows
us down unacceptably.
Reviewers: eugenis, cryptoad, dvyukov
Subscribers: kubamracek, mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D38245
llvm-svn: 314311
Sean Eveson [Wed, 27 Sep 2017 15:37:40 +0000 (15:37 +0000)]
[llvm-cov] Create directory structure when filtering using -name*= options
Before this change using any of the -name*= command line options with an output
directory would result in a single file (functions.txt/functions.html)
containing the coverage for those specific functions. Now you get the same
directory structure as when not using any -name*= options.
Differential Revision: https://reviews.llvm.org/D38280
llvm-svn: 314310
Marc-Andre Laperle [Wed, 27 Sep 2017 15:31:17 +0000 (15:31 +0000)]
[clangd] Handle InitializeParams and store rootUri
Summary:
The root Uri is the workspace location and will be useful in the context of
indexing. We could also add more things to InitializeParams in order to
configure Clangd for C/C++ sepecific extensions.
Reviewers: ilya-biryukov, bkramer, krasimir, Nebiroth
Reviewed By: ilya-biryukov
Subscribers: ilya-biryukov
Tags: #clang-tools-extra
Differential Revision: https://reviews.llvm.org/D38093
llvm-svn: 314309
Sanjay Patel [Wed, 27 Sep 2017 14:54:16 +0000 (14:54 +0000)]
[SimplifyCFG] add a struct to house optional folds (PR34603)
This was intended to be no-functional-change, but it's not - there's a test diff.
So I thought I should stop here and post it as-is to see if this looks like what was expected
based on the discussion in PR34603:
https://bugs.llvm.org/show_bug.cgi?id=34603
Notes:
1. The test improvement occurs because the existing 'LateSimplifyCFG' marker is not carried
through the recursive calls to 'SimplifyCFG()->SimplifyCFGOpt().run()->SimplifyCFG()'.
The parameter isn't passed down, so we pick up the default value from the function signature
after the first level. I assumed that was a bug, so I've passed 'Options' down in all of the
'SimplifyCFG' calls.
2. I split 'LateSimplifyCFG' into 2 bits: ConvertSwitchToLookupTable and KeepCanonicalLoops.
This would theoretically allow us to differentiate the transforms controlled by those params
independently.
3. We could stash the optional AssumptionCache pointer and 'LoopHeaders' pointer in the struct too.
I just stopped here to minimize the diffs.
4. Similarly, I stopped short of messing with the pass manager layer. I have another question that
could wait for the follow-up: why is the new pass manager creating the pass with LateSimplifyCFG
set to true no matter where in the pipeline it's creating SimplifyCFG passes?
// Create an early function pass manager to cleanup the output of the
// frontend.
EarlyFPM.addPass(SimplifyCFGPass());
-->
/// \brief Construct a pass with the default thresholds
/// and switch optimizations.
SimplifyCFGPass::SimplifyCFGPass()
: BonusInstThreshold(UserBonusInstThreshold),
LateSimplifyCFG(true) {} <-- switches get converted to lookup tables and loops may not be in canonical form
If this is unintended, then it's possible that the current behavior of dropping the 'LateSimplifyCFG'
setting via recursion was masking this bug.
Differential Revision: https://reviews.llvm.org/D38138
llvm-svn: 314308
Haicheng Wu [Wed, 27 Sep 2017 14:44:56 +0000 (14:44 +0000)]
[InlineCost] add visitSelectInst()
InlineCost can understand Select IR now. This patch finds free Select IRs and
continue the propagation of SimplifiedValues, ConstantOffsetPtrs, and
SROAArgValues.
Differential Revision: https://reviews.llvm.org/D37198
llvm-svn: 314307
Gadi Haber [Wed, 27 Sep 2017 14:44:15 +0000 (14:44 +0000)]
[X86][SKX][KNL] Updated regression tests to use -mattr instead of -mcpu flag.NFC.
NFC.
Updated 8 regression tests to use -mattr instead of -mcpu flag as follows:
-mcpu=knl --> -mattr=+avx512f
-mcpu=skx --> -mattr=+avx512f,+avx512bw,+avx512vl,+avx512dq
The updates are as part of the preparation of a large commit to add all instruction scheduling for the SKX target.
Reviewers: delena, zvi, RKSimon
Differential Revision: https://reviews.llvm.org/D38222
Change-Id: I2381c9b5bb75ecacfca017243c22d054f6eddd14
llvm-svn: 314306
Zvi Rackover [Wed, 27 Sep 2017 14:38:05 +0000 (14:38 +0000)]
X86 Tests: Unsigned saturation subtraction tests. NFC.
Summary:
Adding tests for D37534.
Commit on behalf of julia.koval@intel.com
Reviewers: n.bozhenov, zvi, spatel, DavidKreitzer
Reviewed By: zvi
Differential Revision: https://reviews.llvm.org/D37510
llvm-svn: 314305
Anastasia Stulova [Wed, 27 Sep 2017 14:37:00 +0000 (14:37 +0000)]
[OpenCL] Handle address space conversion while setting type alignment.
Added missing addrspacecast case in alignment computation
logic of pointer type emission in IR generation.
Differential Revision: https://reviews.llvm.org/D37804
llvm-svn: 314304
Gheorghe-Teodor Bercea [Wed, 27 Sep 2017 14:31:08 +0000 (14:31 +0000)]
[OpenMP] Add an additional test for D34888
Summary: Test for checking if the mapping is performed correctly. This is a test initially included in Patch https://reviews.llvm.org/D29905
Reviewers: Hahnfeld, carlo.bertolli, caomhin, ABataev
Reviewed By: Hahnfeld
Subscribers: tra, cfe-commits
Differential Revision: https://reviews.llvm.org/D38040
llvm-svn: 314303
Coby Tayree [Wed, 27 Sep 2017 13:02:44 +0000 (13:02 +0000)]
revert rL314300
accidently added only tests w/o the respective changes..
llvm-svn: 314302
Krzysztof Parzyszek [Wed, 27 Sep 2017 12:48:48 +0000 (12:48 +0000)]
Typo: const MCSchedModel SchedModel -> const MCSchedModel &SchedModel
llvm-svn: 314301
Coby Tayree [Wed, 27 Sep 2017 12:36:54 +0000 (12:36 +0000)]
[X86][MS-InlineAsm] Extended support for variables / identifiers on memory / immediate expressions
Allow the proper recognition of Enum values and global variables inside ms inline-asm memory / immediate expressions, as they require some additional overhead and treated incorrect if doesn't early recognized.
supersedes D33277, D35775
Corrsponds with D37412, D37413
llvm-svn: 314300
Mikael Holmen [Wed, 27 Sep 2017 11:27:50 +0000 (11:27 +0000)]
[RegAllocGreedy] Fix spelling error, "inteference" -> "interference", NFC
llvm-svn: 314299
Artem Dergachev [Wed, 27 Sep 2017 10:59:06 +0000 (10:59 +0000)]
[analyzer] Fix an outdated comment in a test. NFC.
llvm-svn: 314298
Hiroshi Inoue [Wed, 27 Sep 2017 10:33:02 +0000 (10:33 +0000)]
[PowerPC] eliminate unconditional branch to the next instruction
This patch makes analyzeBranch eliminate unconditional branch to the next instruction.
After basic blocks are re-organized by optimizers, such as machine block placement, a BB may end with an unconditional branch to the next (fallthrough) BB. This patch removes such redundant branch instruction.
Differential Revision: https://reviews.llvm.org/D37730
llvm-svn: 314297
Javed Absar [Wed, 27 Sep 2017 10:31:58 +0000 (10:31 +0000)]
[Misched]: Remove double call getMicroOpFactor.NFC.
Reviewed by: @MatzeB
Differential Revision: https://reviews.llvm.org/D38176
llvm-svn: 314296
Coby Tayree [Wed, 27 Sep 2017 10:29:29 +0000 (10:29 +0000)]
[X86][AsmParser] fix PR32035
Differential Revision: https://reviews.llvm.org/D37473
llvm-svn: 314295
Jonas Devlieghere [Wed, 27 Sep 2017 10:23:34 +0000 (10:23 +0000)]
[test] Don't verify .debug_line offsets in bitcode tests.
The exact values of the .debug_line offsets should not be hard-coded in
the checks for bitcode tests.
Fixes: http://bb.pgr.jp/builders/test-llvm-i686-linux-RA/builds/543
llvm-svn: 314294
Simon Pilgrim [Wed, 27 Sep 2017 10:10:17 +0000 (10:10 +0000)]
[X86][AVX] Improve (i4 bitcast (v4i1 x)) handling for 256-bit vector compare results.
As commented on D37849 and rL313547, AVX1 targets were missing a chance to use vmovmskpd for v4f64/v4i64 results for bool vector bitcasts
llvm-svn: 314293
Simon Pilgrim [Wed, 27 Sep 2017 10:03:17 +0000 (10:03 +0000)]
Use const where possible. NFCI.
llvm-svn: 314292
Jonas Devlieghere [Wed, 27 Sep 2017 10:00:27 +0000 (10:00 +0000)]
[dwarfdump] Fix printing of .debug_line offset.
Fixes 32-bit buildbots:
http://bb.pgr.jp/builders/test-llvm-i686-linux-RA/builds/542
http://lab.llvm.org:8011/builders/clang-cmake-thumbv7-a15/builds/11533
http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15/builds/11494
llvm-svn: 314291
Artem Dergachev [Wed, 27 Sep 2017 09:50:45 +0000 (09:50 +0000)]
[analyzer] Match more patterns in bugreporter::getDerefExpr() API.
This function can now track null pointer through simple pointer arithmetic,
such as '*&*(p + 2)' => 'p' and so on, displaying intermediate diagnostic pieces
for the user to understand where the null pointer is coming from.
Differential Revision: https://reviews.llvm.org/D37025
llvm-svn: 314290
Manoj Gupta [Wed, 27 Sep 2017 09:49:32 +0000 (09:49 +0000)]
[builtins] fix build error on non-ARM for r314285.
llvm-svn: 314289
Jonas Devlieghere [Wed, 27 Sep 2017 09:33:45 +0000 (09:33 +0000)]
[dwarfdump] Add support for -debug-line=OFFSET
This patch adds support for passing an offset to -debug-line.
Differential revision: https://reviews.llvm.org/D38240
llvm-svn: 314288
Artem Dergachev [Wed, 27 Sep 2017 09:33:37 +0000 (09:33 +0000)]
[analyzer] Fix and refactor bugreporter::getDerefExpr() API.
This API is used by checkers (and other entities) in order to track where does
a value originate from, by jumping from an expression value of which is equal
to that value to the expression from which this value has "appeared". For
example, it may be an lvalue from which the rvalue was loaded, or a function
call from which the dereferenced pointer was returned.
The function now avoids incorrectly unwrapping implicit lvalue-to-rvalue casts,
which caused crashes and incorrect intermediate diagnostic pieces. It also no
longer relies on how the expression is written when guessing what it means.
Fixes pr34373 and pr34731.
rdar://problem/
33594502
Differential Revision: https://reviews.llvm.org/D37023
llvm-svn: 314287
Jonas Devlieghere [Wed, 27 Sep 2017 09:33:36 +0000 (09:33 +0000)]
[dwarfdump] Add support for -debug-loc=OFFSET
This patch adds support for passing an offset to -debug-loc.
Differential revision: https://reviews.llvm.org/D38237
llvm-svn: 314286
Manoj Gupta [Wed, 27 Sep 2017 09:33:14 +0000 (09:33 +0000)]
[Builtins] ARM: Fix assembling files in thumb mode.
Summary:
clang does not assemble files in thumb mode unless .thumb declaration
is present. Add .thumb/.arm decl to _FUNCTION macros to ensure that
files are assembled correctly.
Fixes PR 34715.
Reviewers: compnerd, peter.smith, srhines, weimingz, rengolin
Reviewed By: compnerd
Subscribers: aemerson, javed.absar, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D38227
llvm-svn: 314285
Manoj Gupta [Wed, 27 Sep 2017 09:29:57 +0000 (09:29 +0000)]
[Builtins] ARM: Fix msr assembly instruction use for Thumb2.
Summary:
MSR instruction in Thumb2 does not support immediate operand.
Fix this by moving the condition for V7-M to Thumb2 since V7-M support
Thumb2 only. With this change, aeabi_cfcmp.s and aeabi_cdcmp.S files can
be assembled in Thumb2 mode. (This is split out from the review D38227).
Reviewers: compnerd, peter.smith, srhines, weimingz, rengolin, kristof.beyls
Reviewed By: compnerd
Subscribers: aemerson, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D38268
llvm-svn: 314284
George Rimar [Wed, 27 Sep 2017 09:14:59 +0000 (09:14 +0000)]
[ELF] - Detemplate of HashTableSection<ELFT>
Detemplation of one more synthetic section.
Differential revision: https://reviews.llvm.org/D38241
llvm-svn: 314283
George Rimar [Wed, 27 Sep 2017 09:08:53 +0000 (09:08 +0000)]
[ELF] - Speedup -r and --emit-relocs
This is "Bug 34688 - lld much slower than bfd when linking the linux kernel"
Inside copyRelocations() we have O(N*M) algorithm, where N - amount of
relocations and M - amount of symbols in symbol table. It isincredibly slow
for linking linux kernel.
Patch creates local search tables to speedup.
With this fix link time goes for me from 12.95s to 0.55s what is almost 23x
faster. (used release LLD).
Differential revision: https://reviews.llvm.org/D38129
llvm-svn: 314282
Sean Eveson [Wed, 27 Sep 2017 08:32:36 +0000 (08:32 +0000)]
[llvm-cov] Improve const-correctness of filters. NFC.
llvm-svn: 314281
Sam Parker [Wed, 27 Sep 2017 08:30:45 +0000 (08:30 +0000)]
[ARM] isTruncateFree fix
I implemented isTruncateFree in rL313533, this patch fixes the logic
to match my comment, as the previous logic was too general. Now the
only truncates that are free are i64 -> i32.
Differential Revision: https://reviews.llvm.org/D38234
llvm-svn: 314280
Marek Kurdej [Wed, 27 Sep 2017 07:51:51 +0000 (07:51 +0000)]
[clang-format] Fix FixNamespaceComments when BraceWrapping AfterNamespace is true.
Summary:
NamespaceEndCommentsFixer did not fix namespace comments when the brace opening the namespace was not on the same line as the "namespace" keyword.
It occurs in Allman, GNU and Linux styles and whenever BraceWrapping.AfterNamespace is true.
Before:
```lang=cpp
namespace a
{
void f();
void g();
}
```
After:
```lang=cpp
namespace a
{
void f();
void g();
} // namespace a
```
Reviewers: krasimir
Reviewed By: krasimir
Subscribers: klimek, cfe-commits
Differential Revision: https://reviews.llvm.org/D37904
llvm-svn: 314279
Martin Pelikan [Wed, 27 Sep 2017 07:30:48 +0000 (07:30 +0000)]
[XRay] initialize all members of YAMLXRayRecord for -Wmissing-field-initializers
llvm-svn: 314278
Martin Storsjo [Wed, 27 Sep 2017 06:08:23 +0000 (06:08 +0000)]
[X86] Fix SJLJ struct offsets for x86_64
This is necessary, but not sufficient, for having working SJLJ exception
handling on x86_64.
Differential Revision: https://reviews.llvm.org/D38254
llvm-svn: 314277
Martin Storsjo [Wed, 27 Sep 2017 06:08:16 +0000 (06:08 +0000)]
[X86] Remove erroneous callsite offsetting in SJLJ landing pads
The callsite value is already stored indexed from 0 in
the _Unwind_Context struct. When accessed via the functions
_Unwind_GetIP and _Unwind_SetIP, the value is indexed from 1,
but those functions handle the offseting. When reading directly
from the struct here, we shouldn't subtract 1.
This matches the code generated by the ARM target, where SJLJ
exception handling is used by default on iOS.
This makes clang-built object files for 32 bit x86 mingw work when
linked with libgcc/libstdc++.
Differential Revision: https://reviews.llvm.org/D38251
llvm-svn: 314276
Martin Storsjo [Wed, 27 Sep 2017 06:08:04 +0000 (06:08 +0000)]
[X86] Correct byte offsets and data types in a comment. NFC.
This matches the types of the struct members defined in
lib/CodeGen/SjLjEHPrepare.cpp, and the definition of this struct in libgcc.
Differential Revision: https://reviews.llvm.org/D38248
llvm-svn: 314275
Craig Topper [Wed, 27 Sep 2017 06:04:55 +0000 (06:04 +0000)]
[X86] Use extract128BitVector in LowerMULH so we can extract from constant build vectors.
llvm-svn: 314274
Daniel Berlin [Wed, 27 Sep 2017 05:35:19 +0000 (05:35 +0000)]
MemorySSAUpdater: Only add phis to insertedphis if we actually inserted them, not if we just found existing ones
llvm-svn: 314273
Craig Topper [Wed, 27 Sep 2017 05:17:15 +0000 (05:17 +0000)]
[X86] Add avx512bw command lines to the 256-bit vector idiv tests.
Some of the operations are being sign extended to 512 bits with avx512bw.
llvm-svn: 314272
Craig Topper [Wed, 27 Sep 2017 05:17:14 +0000 (05:17 +0000)]
[SelectionDAG] Make NewSDValueDbgMsg print target specific nodes correctly by passing in the SelectionDAG.
llvm-svn: 314271
Martin Pelikan [Wed, 27 Sep 2017 05:10:31 +0000 (05:10 +0000)]
[XRay] fix the -Werror build by handling all enum cases in switches
Followup to D32840.
llvm-svn: 314270
Martin Pelikan [Wed, 27 Sep 2017 04:48:03 +0000 (04:48 +0000)]
[XRay] convert FDR arg1 log entries
Summary:
A new FDR metadata record will support logging a function call argument;
appending multiple metadata records will represent a sequence of arguments
meaning that "holes" are not representable by the buffer format. Each
call argument is currently a 64-bit value (useful for "this" pointers and
synchronization objects).
If present, we put this argument to the function call "entry" record it
belongs to, and alter its type to notify the user of its presence.
Reviewers: dberris
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D32840
llvm-svn: 314269
Erich Keane [Wed, 27 Sep 2017 03:23:02 +0000 (03:23 +0000)]
Add test forgotten in r314262.
llvm-svn: 314268
Erich Keane [Wed, 27 Sep 2017 03:20:13 +0000 (03:20 +0000)]
Fix capitalization of a bunch of parameters in SemaDeclAttr [NFC]
llvm-svn: 314267
Hongbin Zheng [Wed, 27 Sep 2017 03:11:46 +0000 (03:11 +0000)]
[SimplifyIndVar] Constant fold IV users
This patch tries to transform cases like:
for (unsigned i = 0; i < N; i += 2) {
bool c0 = (i & 0x1) == 0;
bool c1 = ((i + 1) & 0x1) == 1;
}
To
for (unsigned i = 0; i < N; i += 2) {
bool c0 = true;
bool c1 = true;
}
This commit also update test/Transforms/IndVarSimplify/replace-srem-by-urem.ll to prevent constant folding.
Differential Revision: https://reviews.llvm.org/D38272
llvm-svn: 314266
Jason Molenda [Wed, 27 Sep 2017 02:49:18 +0000 (02:49 +0000)]
Update ABIMacOSX_arm::PrepareTrivialCall to correctly align the
stack pointer for apple's armv7 ABI. When in a frameless function
or in a prologue/epilogue where sp wasn't properly aligned, we could
try to make function calls with an unaligned sp; the expression
would crash.
llvm-svn: 314265
Chih-Hung Hsieh [Wed, 27 Sep 2017 00:58:45 +0000 (00:58 +0000)]
[clang-format] Adjust space around &/&& of structured bindings
Keep space before or after the &/&& tokens, but not both. For example,
auto [x,y] = a;
auto &[xr, yr] = a; // LLVM style
auto& [xr, yr] = a; // google style
Differential Revision:https://reviews.llvm.org/D35743
llvm-svn: 314264
Jake Ehrlich [Wed, 27 Sep 2017 00:44:00 +0000 (00:44 +0000)]
Reland: [llvm-objcopy] Add support for dynamic relocations
This change adds support for dynamic relocations (allocated
SHT_REL/SHT_RELA sections with a dynamic symbol table as their link).
I had to reland this because of a I wasn't initilizing some pointers.
llvm-svn: 314263
Erich Keane [Tue, 26 Sep 2017 23:42:34 +0000 (23:42 +0000)]
Emit section information for extern variables.
Currently, if _attribute_((section())) is used for extern variables,
section information is not emitted in generated IR when the variables are used.
This is expected since sections are not generated for external linkage objects.
However NiosII requires this information as it uses special GP-relative accesses
for any objects that use attribute section (.sdata). GCC keeps this attribute in
middle-end.
This change emits the section information for all targets.
Patch By: Elizabeth Andrews
Differential Revision:https://reviews.llvm.org/D36487
llvm-svn: 314262
Jan Vesely [Tue, 26 Sep 2017 23:15:54 +0000 (23:15 +0000)]
Add missing HAVE_LLVM define to fix build with latest llvm
Broken since r314111
V2: pointed out by Jan Vesely
- Use format() instead of % formating
Patch-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 314261
Stephan T. Lavavej [Tue, 26 Sep 2017 23:08:43 +0000 (23:08 +0000)]
[libcxx] [test] Fix unused local typedef warnings.
llvm-svn: 314259
Stephan T. Lavavej [Tue, 26 Sep 2017 23:08:41 +0000 (23:08 +0000)]
[libcxx] [test] Strip trailing whitespace.
llvm-svn: 314258
Stephan T. Lavavej [Tue, 26 Sep 2017 23:08:39 +0000 (23:08 +0000)]
[libcxx] [test] Silence warning C4324 for MSVC.
This warning "structure was padded due to alignment specifier" says
that the compiler is going to do exactly what you asked it to do.
It's triggered by the tests for over-aligned dynamic memory allocation.
llvm-svn: 314257
James Y Knight [Tue, 26 Sep 2017 22:44:01 +0000 (22:44 +0000)]
Initialize the RelocationSectionBase::Section member.
In r314227, it wasn't always, and would thus contain random garbage.
llvm-svn: 314256
Manoj Gupta [Tue, 26 Sep 2017 22:37:08 +0000 (22:37 +0000)]
[Builtins] Use 4 byte alignment for __aeabi_memclr.
Summary:
Align __aeabi_memclr to 4 bytes. All other ARM functions are already aligned to
4-bytes in compiler-rt.
(Split off from review D38227)
Reviewers: compnerd, peter.smith, srhines, weimingz, rengolin, kristof.beyls
Reviewed By: compnerd
Subscribers: aemerson, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D38271
llvm-svn: 314255
Jakub Kuderski [Tue, 26 Sep 2017 21:56:55 +0000 (21:56 +0000)]
[Dominators] Invalidate DFS numbers upon edge deletions
This patch makes DeleteEdge correctly invalidate DFS numbers in the
incremental updater. This should fix PR34466 and related bugs.
llvm-svn: 314254
Sanjoy Das [Tue, 26 Sep 2017 21:54:27 +0000 (21:54 +0000)]
[BypassSlowDivision] Improve our handling of divisions by constants
Summary:
Don't bail out on constant divisors for divisions that can be narrowed without
introducing control flow . This gives us a 32 bit multiply instead of an
emulated 64 bit multiply in the generated PTX assembly.
Reviewers: jlebar
Subscribers: jholewinski, mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D38265
llvm-svn: 314253
Geoff Berry [Tue, 26 Sep 2017 21:40:46 +0000 (21:40 +0000)]
[AArch64][Falkor] Fix bug in falkor prefetcher fix pass.
Summary:
In rare cases, loads that don't get prefetched that were marked as
strided loads could cause a crash if they occurred in a loop with other
colliding loads.
Reviewers: mcrosier
Subscribers: aemerson, rengolin, javed.absar, kristof.beyls
Differential Revision: https://reviews.llvm.org/D38261
llvm-svn: 314252
Geoff Berry [Tue, 26 Sep 2017 21:40:41 +0000 (21:40 +0000)]
[AArch64][Falkor] Fix correctness bug in falkor prefetcher fix pass and correct some opcode tag computations.
Summary:
This addresses a correctness bug for LD[1234]*_POST opcodes that have
the prefetcher fix applied to them: the base register was not being
written back from the temp after being incremented, so it would appear
to never be incremented.
Also, fix some opcode tag computations based on some updated HW details
to get better tag avoidance and thus better prefetcher performance.
Reviewers: mcrosier
Subscribers: aemerson, rengolin, javed.absar, kristof.beyls
Differential Revision: https://reviews.llvm.org/D38256
llvm-svn: 314251
Craig Topper [Tue, 26 Sep 2017 21:35:11 +0000 (21:35 +0000)]
[X86] Fix register class name in a comment. NFC
llvm-svn: 314250
Craig Topper [Tue, 26 Sep 2017 21:35:09 +0000 (21:35 +0000)]
Recommit r314151 "[X86] Make all the NOREX CodeGenOnly instructions into postRA pseudos like the NOREX version of TEST.""
The late MOV8rr_NOREX that caused the crash has been removed.
llvm-svn: 314249
Craig Topper [Tue, 26 Sep 2017 21:35:06 +0000 (21:35 +0000)]
[X86] Don't emit X86::MOV8rr_NOREX from X86InstrInfo::copyPhysReg.
This hook is called after register allocation with two physical registers. We don't need a separate instruction at that time to force register class constraints. I left in the assert though. We also have a fatal error in X86MCCodeEmitter if we ever encode an H-reg and a REX prefix.
llvm-svn: 314248