Zachary Turner [Fri, 18 Nov 2016 20:44:46 +0000 (20:44 +0000)]
Change CreateTarget and dependents to accept StringRef.
llvm-svn: 287376
Geoff Berry [Fri, 18 Nov 2016 20:08:05 +0000 (20:08 +0000)]
[MIRPrinter] XFAIL test for powerpc
This test introduced in r287368 is failing on powerpc for reasons
unrelated to branch probabilities. See PR31062.
llvm-svn: 287375
Mehdi Amini [Fri, 18 Nov 2016 20:02:34 +0000 (20:02 +0000)]
Revert "Add link-time detection of LLVM_ABI_BREAKING_CHECKS mismatch"
This reverts commit r287352, LLDB CI is broken.
llvm-svn: 287374
Eric Fiselier [Fri, 18 Nov 2016 19:53:45 +0000 (19:53 +0000)]
Make LIBCXX_ENABLE_STATIC_ABI_LIBRARY merge libc++.a and libc++abi.a
llvm-svn: 287373
Rui Ueyama [Fri, 18 Nov 2016 19:45:04 +0000 (19:45 +0000)]
Simplify. NFC.
llvm-svn: 287372
Matthias Braun [Fri, 18 Nov 2016 19:43:25 +0000 (19:43 +0000)]
Adapt to llvm NamedRegionTimer changes
We have to specify a name and description for the timers and groups now.
llvm-svn: 287371
Matthias Braun [Fri, 18 Nov 2016 19:43:24 +0000 (19:43 +0000)]
Statistic/Timer: Include timers in PrintStatisticsJSON().
Differential Revision: https://reviews.llvm.org/D25588
llvm-svn: 287370
Matthias Braun [Fri, 18 Nov 2016 19:43:18 +0000 (19:43 +0000)]
Timer: Track name and description.
The previously used "names" are rather descriptions (they use multiple
words and contain spaces), use short programming language identifier
like strings for the "names" which should be used when exporting to
machine parseable formats.
Also removed a unused TimerGroup from Hexxagon.
Differential Revision: https://reviews.llvm.org/D25583
llvm-svn: 287369
Geoff Berry [Fri, 18 Nov 2016 19:37:24 +0000 (19:37 +0000)]
[MIRPrinter] Print raw branch probabilities as expected by MIRParser
Fixes PR28751.
Reviewers: MatzeB, qcolombet
Subscribers: mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D26775
llvm-svn: 287368
Zachary Turner [Fri, 18 Nov 2016 19:23:39 +0000 (19:23 +0000)]
Re-add the StringRef interface changes for Variable.
This concludes the changes I originally tried to make and then
had to back out. This way if anything is still broken, it
should be easier to bisect it back to a more specific changeset.
llvm-svn: 287367
Chris Bieneman [Fri, 18 Nov 2016 19:20:39 +0000 (19:20 +0000)]
[CMake] Fixing variable names that were mistyped
This is a silly bug that I'm sure I caused...
llvm-svn: 287366
Rafael Espindola [Fri, 18 Nov 2016 19:02:15 +0000 (19:02 +0000)]
Simplify handling of SHF_LINK_ORDER.
It seems a lot simpler to just sort the sections and let the
relocations do the rest.
llvm-svn: 287365
Saleem Abdulrasool [Fri, 18 Nov 2016 19:01:53 +0000 (19:01 +0000)]
__cxa_demangle: use default member initialization
Sink the Db initialization into the structure rather than out-of-line at the
declaration size. This just makes it easier to see what initialization is being
performed. NFC.
llvm-svn: 287364
Matt Arsenault [Fri, 18 Nov 2016 18:33:36 +0000 (18:33 +0000)]
AMDGPU: Fix unused variable warning
llvm-svn: 287362
Hans Wennborg [Fri, 18 Nov 2016 18:27:31 +0000 (18:27 +0000)]
Fix test from r287353: don't use /dev/null
llvm-svn: 287360
Saleem Abdulrasool [Fri, 18 Nov 2016 18:21:06 +0000 (18:21 +0000)]
builtins: repair the builtins build with clang 3.8
When the C unwinding personality was corrected to match the ARM EHABI
specification, the unwind header in clang was updated with necessary
declarations. However, when building with an older compiler, we would not have
the necessary declarations. This would result in a build failure. Provide a
supplementary header to ensure that the necessary declarations are present for
the build of the C unwinding personality.
Note that this is NOT an ABI break. It merely is a compile time failure due to
the constants not being present. The constants here are reproduced
equivalently. This header should permit building with clang[<3.9] as well as
gcc.
Addresses PR31035!
llvm-svn: 287359
Adam Nemet [Fri, 18 Nov 2016 18:17:36 +0000 (18:17 +0000)]
LTO support for -fsave-optimization-record on Darwin
I guess this would have to be added for each linker.
Differential Revision: https://reviews.llvm.org/D26833
llvm-svn: 287358
Adam Nemet [Fri, 18 Nov 2016 18:17:33 +0000 (18:17 +0000)]
Fix a comment for -fsave-optimization-record
Differential Revision: https://reviews.llvm.org/D26807
llvm-svn: 287357
Adam Nemet [Fri, 18 Nov 2016 18:06:28 +0000 (18:06 +0000)]
[LTO] Add option to generate optimization records
It is used to drive this from the clang driver via -mllvm.
Same option name is used as in opt.
Differential Revision: https://reviews.llvm.org/D26832
llvm-svn: 287356
Eugene Zelenko [Fri, 18 Nov 2016 18:00:19 +0000 (18:00 +0000)]
[DebugInfo] Fix some Clang-tidy modernize-use-default, modernize-use-equal-delete and Include What You Use warnings; other minor fixes (NFC).
Per Zachary Turner and Mehdi Amini suggestion to make only post-commit reviews.
llvm-svn: 287355
Zachary Turner [Fri, 18 Nov 2016 17:55:04 +0000 (17:55 +0000)]
Resubmit "Remove an output-parameter from Variable function".
The scanning algorithm had a few little subtleties that I
overlooked, but this patch should fix everything.
I still haven't changed the function to take a StringRef since
that has some trickle down effect and is mostly mechanical,
I just wanted to get the tricky part as isolated as possible.
llvm-svn: 287354
Hans Wennborg [Fri, 18 Nov 2016 17:33:05 +0000 (17:33 +0000)]
IRMover: Avoid accidentally mapping types from the destination module (PR30799)
During Module linking, it's possible for SrcM->getIdentifiedStructTypes();
to return types that are actually defined in the destination module
(DstM). Depending on how the bitcode file was read,
getIdentifiedStructTypes() might do a walk over all values, including
metadata nodes, looking for types. In my case, a debug info metadata
node was shared between the two modules, and it referred to a type
defined in the destination module (see test case).
Differential Revision: https://reviews.llvm.org/D26212
llvm-svn: 287353
Mehdi Amini [Fri, 18 Nov 2016 17:28:10 +0000 (17:28 +0000)]
Add link-time detection of LLVM_ABI_BREAKING_CHECKS mismatch
Summary:
LLVM will define a symbol, either EnableABIBreakingChecks or
DisableABIBreakingChecks depending on the configuration setting for
LLVM_ABI_BREAKING_CHECKS.
The llvm-config.h header will add weak references to these symbols in
every clients that includes this header. This should ensure that
a mismatch triggers a link failure (or a load time failure for DSO).
On MSVC, the pragma "detect_mismatch" is used instead.
Reviewers: rnk, jroelofs
Subscribers: llvm-commits, mgorny
Differential Revision: https://reviews.llvm.org/D26841
llvm-svn: 287352
Ehsan Amiri [Fri, 18 Nov 2016 16:24:27 +0000 (16:24 +0000)]
[PPC] limit line width to 80 characters
NFC. Forgot to fix this in the original commit.
llvm-svn: 287350
Simon Dardis [Fri, 18 Nov 2016 16:17:44 +0000 (16:17 +0000)]
[mips][msa] Implement f16 support
The MIPS MSA ASE provides instructions to convert to and from half precision
floating point. This patch teaches the MIPS backend to treat f16 as a legal
type and how to promote such values to f32 for the usual set of operations.
As a result of this, the fexup[lr].w intrinsics no longer crash LLVM during
type legalization.
Reviewers: zoran.jovanvoic, vkalintiris
Differential Revision: https://reviews.llvm.org/D26398
llvm-svn: 287349
Simon Pilgrim [Fri, 18 Nov 2016 15:31:36 +0000 (15:31 +0000)]
[X86][AVX512] Split AVX512F/AVX512VL tests to demonstrate missed int2fp opportunities without AVX512VL
llvm-svn: 287348
Johannes Doerfert [Fri, 18 Nov 2016 14:37:08 +0000 (14:37 +0000)]
[NFC] Adjust naming scheme of statistic variables
Suggested-by: Tobias Grosser <tobias@grosser.es>
llvm-svn: 287347
Eugene Leviant [Fri, 18 Nov 2016 14:35:03 +0000 (14:35 +0000)]
[ELF] Convert PltSection to input section
Differential revision: https://reviews.llvm.org/D26842
llvm-svn: 287346
Tom Stellard [Fri, 18 Nov 2016 14:14:35 +0000 (14:14 +0000)]
GlobalISel: Fix unconditional fallback with global isel abort is disabled
Reviewers: t.p.northover, ab, qcolombet
Subscribers: mehdi_amini, vkalintiris, wdng, dberris, llvm-commits, rovka
Differential Revision: https://reviews.llvm.org/D26765
llvm-svn: 287344
Joey Gouly [Fri, 18 Nov 2016 14:10:54 +0000 (14:10 +0000)]
[OpenCL] Introduce ReadPipeType and WritePipeType.
This allows Sema to diagnose passing a read_only pipe to a
write_only pipe argument.
llvm-svn: 287343
Tom Stellard [Fri, 18 Nov 2016 13:53:34 +0000 (13:53 +0000)]
AMDGPU/SI: Remove zero_extend patterns for i16 ops selected to 32-bit insts
Summary:
The 32-bit instructions don't zero the high 16-bits like the 16-bit
instructions do.
Reviewers: arsenm
Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, llvm-commits, tony-tye
Differential Revision: https://reviews.llvm.org/D26828
llvm-svn: 287342
Florian Hahn [Fri, 18 Nov 2016 13:12:07 +0000 (13:12 +0000)]
[simplifycfg][loop-simplify] Preserve loop metadata in 2 transformations.
insertUniqueBackedgeBlock in lib/Transforms/Utils/LoopSimplify.cpp now
propagates existing llvm.loop metadata to newly the added backedge.
llvm::TryToSimplifyUncondBranchFromEmptyBlock in lib/Transforms/Utils/Local.cpp
now propagates existing llvm.loop metadata to the branch instructions in the
predecessor blocks of the empty block that is removed.
Differential Revision: https://reviews.llvm.org/D26495
llvm-svn: 287341
Simon Pilgrim [Fri, 18 Nov 2016 12:16:18 +0000 (12:16 +0000)]
Cleanup function with clang-format. NFCI.
llvm-svn: 287340
Nicolai Haehnle [Fri, 18 Nov 2016 11:55:52 +0000 (11:55 +0000)]
AMDGPU: Fix legalization of MUBUF instructions in shaders
Summary:
The addr64-based legalization is incorrect for MUBUF instructions with idxen
set as well as for BUFFER_LOAD/STORE_FORMAT_* instructions. This affects
e.g. shaders that access buffer textures.
Since we never actually need the addr64-legalization in shaders, this patch
takes the easy route and keys off the calling convention. If this ever
affects (non-OpenGL) compute, the type of legalization needs to be chosen
based on some TSFlag.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98664
Reviewers: arsenm, tstellarAMD
Subscribers: kzhuravl, wdng, yaxunl, tony-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D26747
llvm-svn: 287339
Simon Pilgrim [Fri, 18 Nov 2016 11:53:36 +0000 (11:53 +0000)]
Fix spelling mistakes in MIPS target comments. NFC.
Identified by Pedro Giffuni in PR27636.
llvm-svn: 287338
Eric Fiselier [Fri, 18 Nov 2016 11:29:05 +0000 (11:29 +0000)]
Make merge_archives.py executable
llvm-svn: 287337
Eric Fiselier [Fri, 18 Nov 2016 11:26:14 +0000 (11:26 +0000)]
Add merge_archives.py utility
llvm-svn: 287336
Simon Pilgrim [Fri, 18 Nov 2016 11:18:28 +0000 (11:18 +0000)]
Wdocumentation fix
llvm-svn: 287335
Ehsan Amiri [Fri, 18 Nov 2016 11:05:55 +0000 (11:05 +0000)]
[Power9] Add patterns for vnegd, vnegw
Exploit new instructions by adding patterns to .td file.
https://reviews.llvm.org/D26551
llvm-svn: 287334
Simon Pilgrim [Fri, 18 Nov 2016 11:04:02 +0000 (11:04 +0000)]
Fix spelling mistakes in AMDGPU target comments. NFC.
Identified by Pedro Giffuni in PR27636.
llvm-svn: 287333
Simon Pilgrim [Fri, 18 Nov 2016 11:00:36 +0000 (11:00 +0000)]
[X86][AVX2] Add v8i32->v8i64 mul test (PR30845)
llvm-svn: 287332
Simon Pilgrim [Fri, 18 Nov 2016 10:52:12 +0000 (10:52 +0000)]
Fix typo in comment. NFC.
Identified by Pedro Giffuni in PR27636.
llvm-svn: 287331
Haojian Wu [Fri, 18 Nov 2016 10:51:16 +0000 (10:51 +0000)]
[clang-move] Fix not moving using-decls in global namespace in old.cc
Reviewers: ioeric
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D26844
llvm-svn: 287330
Ehsan Amiri [Fri, 18 Nov 2016 10:41:44 +0000 (10:41 +0000)]
[PPC][DAGCombine] Convert SETCC to subtract when the result is zero extended
When we see a SETCC whose only users are zero extend operations, we can replace
it with a subtraction. This results in doing all calculations in GPRs and
avoids CR use.
Currently we do this only for ULT, ULE, UGT and UGE condition codes. There are
ways that this can be extended. For example for signed condition codes. In that
case we will be introducing additional sign extend instructions, so more careful
profitability analysis may be required.
Another direction to extend this is for equal, not equal conditions. Also when
users of SETCC are any_ext or sign_ext, we might be able to do something
similar.
llvm-svn: 287329
Amaury Sechet [Fri, 18 Nov 2016 10:11:02 +0000 (10:11 +0000)]
Fix go binding to adapt the new attribute API
https://reviews.llvm.org/D26339
llvm-svn: 287328
Eric Fiselier [Fri, 18 Nov 2016 09:54:49 +0000 (09:54 +0000)]
Rename TU names to not conflict with libc++.
In order to easily merge libc++ and libc++abi static archives it's important
that none of the source files share the same name.
(See http://stackoverflow.com/questions/3821916/how-to-merge-two-ar-static-libraries-into-one)
This patch renames source files which share a name with libc++ sources.
llvm-svn: 287327
Eugene Leviant [Fri, 18 Nov 2016 09:06:47 +0000 (09:06 +0000)]
[ELF] Convert HashTableSection to input section
Differential revision: https://reviews.llvm.org/D26834
llvm-svn: 287326
Rui Ueyama [Fri, 18 Nov 2016 07:03:56 +0000 (07:03 +0000)]
Update comment.
llvm-svn: 287325
Rui Ueyama [Fri, 18 Nov 2016 06:49:09 +0000 (06:49 +0000)]
Omit empty parameter list.
llvm-svn: 287324
Rui Ueyama [Fri, 18 Nov 2016 06:49:07 +0000 (06:49 +0000)]
Use consume() instead of peek() and skip().
llvm-svn: 287323
Eugene Leviant [Fri, 18 Nov 2016 06:44:18 +0000 (06:44 +0000)]
[ELF] Convert GnuHashTableSection to input section
Differential revision: https://reviews.llvm.org/D26792
llvm-svn: 287322
Eric Fiselier [Fri, 18 Nov 2016 06:42:17 +0000 (06:42 +0000)]
Remove _LIBCPP_HAS_NO_DEFAULTED_FUNCTIONS
libc++ no longer supports C++11 compilers that don't implement `= default`.
This patch removes all instances of the feature test macro
_LIBCPP_HAS_NO_DEFAULTED_FUNCTIONS as well as the potentially dead code it hides.
llvm-svn: 287321
Zachary Turner [Fri, 18 Nov 2016 06:34:45 +0000 (06:34 +0000)]
Revert "Remove an out param from ValueObject::GetValueForExpressionPath."
This reverts commit r287315, as it introduces a bug that breaks
many things.
llvm-svn: 287320
Rui Ueyama [Fri, 18 Nov 2016 06:30:09 +0000 (06:30 +0000)]
Split ScriptParser::readVersionDeclaration.
readVersionDeclaration was to read anonymous version definition and
named version definition. Splitting it into two functions should
improve readability as the two cases are different enough.
I also changed a few helper functions to return values instead of
mutating given references.
llvm-svn: 287319
Rui Ueyama [Fri, 18 Nov 2016 06:30:08 +0000 (06:30 +0000)]
Use consistent variable name.
llvm-svn: 287318
Ekaterina Romanova [Fri, 18 Nov 2016 06:26:01 +0000 (06:26 +0000)]
Add doxygen comments for lzcntintrin.h's intrinsics.
The doxygen comments are automatically generated based on Sony's intrinsics document.
I got an OK from Eric Christopher to commit doxygen comments without prior code
review upstream. This patch was internally reviewed by Charles Li.
llvm-svn: 287317
Craig Topper [Fri, 18 Nov 2016 06:04:33 +0000 (06:04 +0000)]
[InstCombine][AVX-512] Teach InstCombineCalls how to handle the intrinsics for variable shift with 16-bit elements.
This is a straightforward extension of the existing support for 32/64-bit element types. Just needed to add the additional instrinsics to the switches.
llvm-svn: 287316
Zachary Turner [Fri, 18 Nov 2016 05:45:41 +0000 (05:45 +0000)]
Remove an out param from ValueObject::GetValueForExpressionPath.
This argument was only used in one place in the codebase, and
it was in a non-critical log statement and can be easily
substituted for an equally meaningful field instead. The
payoff of computing this value is not worth the added
complexity.
llvm-svn: 287315
Rui Ueyama [Fri, 18 Nov 2016 05:05:43 +0000 (05:05 +0000)]
Simplify MergeOutputSection.
MergeOutputSection class was a bit hard to use because it provdes
a series of finalize functions that have to be called in a right way
at a right time. It also intereacted with MergeInputSection, and the
logic was somewhat entangled between the two classes.
This patch simplifies it by providing only one finalize function.
Now, all you have to do is to call MergeOutputSection::finalize
when you have added all sections to the output section. Then, it
internally merges strings and initliazes StringPiece objects.
I think this is much easier to understand.
This patch also adds comments.
llvm-svn: 287314
Craig Topper [Fri, 18 Nov 2016 05:04:51 +0000 (05:04 +0000)]
[AVX-512] Replace masked 16-bit element variable shift builtins with new unmasked versions and selects.
llvm-svn: 287313
Craig Topper [Fri, 18 Nov 2016 05:04:44 +0000 (05:04 +0000)]
[AVX-512] Replace masked 16-bit element variable shift intrinsics with new unmasked versions and selects.
The same thing was done to 32-bit and 64-bit element sizes previously.
This will allow us to support these shuffls in InstCombineCalls along with the other variable shift intrinsics.
llvm-svn: 287312
Matt Arsenault [Fri, 18 Nov 2016 04:42:59 +0000 (04:42 +0000)]
AMDGPU: Move redundant setting of inst properties
llvm-svn: 287311
Matt Arsenault [Fri, 18 Nov 2016 04:42:57 +0000 (04:42 +0000)]
AMDGPU: Fix crash on illegal type for inlineasm
There are still crashes on non-MVT types in other
places.
llvm-svn: 287310
Shoaib Meenai [Fri, 18 Nov 2016 04:31:09 +0000 (04:31 +0000)]
[libc++] Fix preprocessor guard for overload declaration
Fix a typo in the conditional. Caught by going through list of removed
symbols when building with hidden visibility.
Differential Revision: https://reviews.llvm.org/D26825
llvm-svn: 287309
Zachary Turner [Fri, 18 Nov 2016 04:30:47 +0000 (04:30 +0000)]
Delete more dead code in ValueObject.
Apparently these two enormous functions were dead. Which is
good, since one was largely a copy of another function with
only a few minor tweaks.
llvm-svn: 287308
Zachary Turner [Fri, 18 Nov 2016 03:51:19 +0000 (03:51 +0000)]
Remove some dead code in ValueObject.
Originally I converted this entire function and all dependents
to use StringRef, but there were some test failures that
were tricky to track down, as this is a complicated function.
So I'm starting over, this time in smaller increments.
llvm-svn: 287307
Peter Collingbourne [Fri, 18 Nov 2016 03:20:36 +0000 (03:20 +0000)]
Object: Simplify; remove unnecessary use of unique_ptr.
llvm-svn: 287305
Matthias Braun [Fri, 18 Nov 2016 02:40:40 +0000 (02:40 +0000)]
MachineOperand: Add dump() method
llvm-svn: 287302
Eric Fiselier [Fri, 18 Nov 2016 02:36:58 +0000 (02:36 +0000)]
update Apples ABI list to ignore non-std symbols
llvm-svn: 287301
Alexei Starovoitov [Fri, 18 Nov 2016 02:32:35 +0000 (02:32 +0000)]
convert bpf assembler to look like kernel verifier output
since bpf instruction set was introduced people learned to
read and understand kernel verifier output whereas llvm asm
output stayed obscure and unknown. Convert llvm to emit
assembler text similar to kernel to avoid this discrepancy
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
llvm-svn: 287300
Craig Topper [Fri, 18 Nov 2016 02:28:50 +0000 (02:28 +0000)]
[Docs][TableGen] Remove reference to tablegen supporting octal integers. It doesn't and hasn't for at least 9 years.
llvm-svn: 287299
Craig Topper [Fri, 18 Nov 2016 02:25:34 +0000 (02:25 +0000)]
[AVX-512] Support FCOPYSIGN for v16f32 and v8f64
Summary:
This extends FCOPYSIGN support to 512-bit vectors.
I've also added tests to show what the 128-bit and 256-bit cases look like with broadcast loads.
Reviewers: delena, zvi, RKSimon, spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D26791
llvm-svn: 287298
Davide Italiano [Fri, 18 Nov 2016 02:23:48 +0000 (02:23 +0000)]
[ELF] Rename an historical leftover, `Chunk` is now `InputSection`.
llvm-svn: 287297
Davide Italiano [Fri, 18 Nov 2016 02:18:04 +0000 (02:18 +0000)]
[ELF] Use std::for_each() and hoist common code in a lambda.
llvm-svn: 287296
Ekaterina Romanova [Fri, 18 Nov 2016 01:42:01 +0000 (01:42 +0000)]
Add doxygen comments to fxsrintrin.h's intrinsics.
The doxygen comments are automatically generated based on Sony's intrinsics document.
I got an OK from Eric Christopher to commit doxygen comments without prior code
review upstream. This patch was internally reviewed by Paul Robinson and Charles Li.
llvm-svn: 287295
Eric Fiselier [Fri, 18 Nov 2016 01:40:20 +0000 (01:40 +0000)]
Change sym_check to filter non-stdlib symbols.
Currently sym_check almost all names found in the binary, including those
which are defined in other libraries. This makes our ABI lists harder to maintain.
This patch adds a --only-stdlib-symbols option to sym_check which removes
all symbols which aren't possibly provided by libc++. It also re-generates
the linux ABI list after making this change.
llvm-svn: 287294
Yichao Yu [Fri, 18 Nov 2016 01:25:49 +0000 (01:25 +0000)]
Add an option to disable libedit
Summary: This should provide the function similar to `--disable-libedit` with the autotools build system, which seems to be missing from the commit (r200595) that adds this.
Reviewers: pcc, beanz
Subscribers: mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D26550
llvm-svn: 287293
Justin Lebar [Fri, 18 Nov 2016 01:11:32 +0000 (01:11 +0000)]
[CUDA] Attempt to fix test failures in cuda-macos-includes.cu.
Run clang -cc1 -E instead of -S, in an attempt to make this test work
cross-platform.
llvm-svn: 287292
John McCall [Fri, 18 Nov 2016 01:08:24 +0000 (01:08 +0000)]
Forward ns_consumed delegate arguments with a move.
StartFunction enters a release cleanup for ns_consumed arguments in
ARC, so we need to balance that somehow. We could teach StartFunction
that it's emitting a delegating function, so that the cleanup is
unnecessary, but that would be invasive and somewhat fraught. We could
balance the consumed argument with an extra retain, but clearing the
original variable should be easier to optimize and avoid some extra work
at -O0. And there shouldn't be any difference as long as nothing else
uses the argument, which should always be true for the places we emit
delegate arguments.
Fixes PR 27887.
llvm-svn: 287291
Justin Lebar [Fri, 18 Nov 2016 00:42:00 +0000 (00:42 +0000)]
[CUDA] Update docs to indicate that MacOS is now supported.
llvm-svn: 287290
Justin Lebar [Fri, 18 Nov 2016 00:41:40 +0000 (00:41 +0000)]
[CUDA] Update docs; CUDA 8.0 is supported as of a while ago.
llvm-svn: 287289
Justin Lebar [Fri, 18 Nov 2016 00:41:35 +0000 (00:41 +0000)]
[CUDA] Wrapper header changes necessary to support MacOS.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D26780
llvm-svn: 287288
Justin Lebar [Fri, 18 Nov 2016 00:41:31 +0000 (00:41 +0000)]
[CUDA] Use the right section and constant names for fatbins when compiling for macos.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D26777
llvm-svn: 287287
Justin Lebar [Fri, 18 Nov 2016 00:41:27 +0000 (00:41 +0000)]
[CUDA] Initialize our header search using the host triple.
Summary:
This used to work because system headers are found in a (somewhat)
predictable set of locations on Linux. But this is not the case on
MacOS; without this change, we don't look in the right places for our
headers when doing device-side compilation on Mac.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D26776
llvm-svn: 287286
Justin Lebar [Fri, 18 Nov 2016 00:41:22 +0000 (00:41 +0000)]
[CUDA] Driver changes to support CUDA compilation on MacOS.
Summary:
Compiling CUDA device code requires us to know the host toolchain,
because CUDA device-side compiles pull in e.g. host headers.
When we only supported Linux compilation, this worked because
CudaToolChain, which is responsible for device-side CUDA compilation,
inherited from the Linux toolchain. But in order to support MacOS,
CudaToolChain needs to take a HostToolChain pointer.
Because a CUDA toolchain now requires a host TC, we no longer will
create a CUDA toolchain from Driver::getToolChain -- you have to go
through CreateOffloadingDeviceToolChains. I am *pretty* sure this is
correct, and that previously any attempt to create a CUDA toolchain
through getToolChain() would eventually have resulted in us throwing
"error: unsupported use of NVPTX for host compilation".
In any case hacking getToolChain to create a CUDA+host toolchain would
be wrong, because a Driver can be reused for multiple compilations,
potentially with different host TCs, and getToolChain will cache the
result, causing us to potentially use a stale host TC.
So that's the main change in this patch.
In addition, we have to pull CudaInstallationDetector out of Generic_GCC
and into a top-level class. It's now used by the Generic_GCC and MachO
toolchains.
Reviewers: tra
Subscribers: rryan, hfinkel, sfantao
Differential Revision: https://reviews.llvm.org/D26774
llvm-svn: 287285
Rafael Espindola [Fri, 18 Nov 2016 00:11:12 +0000 (00:11 +0000)]
Add missing REQUIRES.
llvm-svn: 287284
Saleem Abdulrasool [Thu, 17 Nov 2016 23:53:35 +0000 (23:53 +0000)]
EHABI: mark some functions as exported
These are part of the EHABI specification and are exported to be available to
users. Mark them as `_LIBUNWIND_EXPORT` like the rest of the unwind interfaces.
llvm-svn: 287283
Zachary Turner [Thu, 17 Nov 2016 23:47:31 +0000 (23:47 +0000)]
Resubmit "Change RegisterValue getters / setters to use StringRef."
This resubmits r287279 with a fix for the original issue, which
was a trivial typo.
llvm-svn: 287282
Zachary Turner [Thu, 17 Nov 2016 23:32:26 +0000 (23:32 +0000)]
Revert "Change RegisterValue getters / setters to use StringRef."
This reverts commit r287279, which breaks some register
tests on Linux.
llvm-svn: 287281
Rafael Espindola [Thu, 17 Nov 2016 23:16:39 +0000 (23:16 +0000)]
Always compute sh_link for SHF_LINK_ORDER sections.
Since the output has a section table too, it is meaningful to compute
the sh_link. In a more practical note, the binutils' strip crashes if
sh_link is not set for SHT_ARM_EXIDX.
llvm-svn: 287280
Zachary Turner [Thu, 17 Nov 2016 23:05:28 +0000 (23:05 +0000)]
Change RegisterValue getters / setters to use StringRef.
In the process, found some functions that were duplicates of
existing StringRef member functions. So deleted those functions
and used the StringRef functions instead.
llvm-svn: 287279
Ekaterina Romanova [Thu, 17 Nov 2016 23:02:00 +0000 (23:02 +0000)]
Minor changes in x86 intrinsics headers; NFC
I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better.
1. avxintrin.h – Moved the opening bracket on a separate line for several
intrinsics (for consistency with the rest of the intrinsics).
2. emmintrin.h - Moved the doxygen comment next to the body of the function;
- Added braces after extern "C" even though there is only
one declaration each time
3. xmmintrin.h - Moved the doxygen comment next to the body of the function;
- Added intrinsic prototypes for a couple of macro definitions
into the doxygen comment;
- Added braces after extern "C" even though there is only one
declaration each time
4. ammintrin.h – Removed extra line between the doxygen comment and the body
of the functions (for consistency with the rest of the files).
Desk reviewed by Paul Robinson.
llvm-svn: 287278
Davide Italiano [Thu, 17 Nov 2016 22:59:13 +0000 (22:59 +0000)]
[lli] Prefer `exit(1)` to `return 1` for consistency.
llvm-svn: 287277
Davide Italiano [Thu, 17 Nov 2016 22:58:13 +0000 (22:58 +0000)]
[lli] Factor out error handling. NFCI.
llvm-svn: 287276
Bruno Cardoso Lopes [Thu, 17 Nov 2016 22:45:31 +0000 (22:45 +0000)]
[Preprocessor] Support for '-dI' flag
Re-introduce r285411.
Implement the -dI as supported by GCC: Output ‘#include’ directives in addition
to the result of preprocessing.
This change aims to add this option, pass it through to the preprocessor via
the options class, and when inclusions occur we output some information (+ test
cases).
Patch by Steve O'Brien!
Differential Revision: https://reviews.llvm.org/D26089
llvm-svn: 287275
Sam McCall [Thu, 17 Nov 2016 22:29:31 +0000 (22:29 +0000)]
Fix step-over when SymbolContext.function is missing and symbol is present.
Summary:
Fix step-over when SymbolContext.function is missing and symbol is present.
With targets from our build configuration,
ThreadPlanStepOverRange::IsEquivalentContext fails to fire for relevant frames,
leading to ShouldStop() returning true prematurely.
The frame's SymbolContext, and m_addr_context have:
- comp_unit set and matching
- function = nullptr
- symbol set and matching (but this is never checked)
My naive guess is that the context should be equivalent in this case :-)
Reviewers: jingham
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D26804
llvm-svn: 287274
Dylan McKay [Thu, 17 Nov 2016 22:26:09 +0000 (22:26 +0000)]
[ReleaseNotes] Mention the completion of the upstreaming of the AVR backend
llvm-svn: 287273
Johannes Doerfert [Thu, 17 Nov 2016 22:25:17 +0000 (22:25 +0000)]
Probably overwritten loads should not be considered hoistable
Do not assume a load to be hoistable/invariant if the pointer is used by
another instruction in the SCoP that might write to memory and that is
always executed.
llvm-svn: 287272
Johannes Doerfert [Thu, 17 Nov 2016 22:16:35 +0000 (22:16 +0000)]
[NFC] Add flag to disable error block assumptions
The declaration as an "error block" is currently aggressive and not very
smart. This patch allows to disable error blocks completely. This might
be useful to prevent SCoP expansion to a point where the assumed context
becomes infeasible, thus the SCoP has to be discarded.
llvm-svn: 287271
Johannes Doerfert [Thu, 17 Nov 2016 22:11:56 +0000 (22:11 +0000)]
[FIX] Do not try to hoist memory intrinsic
Since we do not necessarily treat memory intrinsics as non-affine
anymore, we have to check for them explicitly before we try to hoist an
access.
llvm-svn: 287270