Sanjay Patel [Sat, 10 Dec 2016 22:16:29 +0000 (22:16 +0000)]
[InstCombine] add helper for shift-by-shift folds; NFCI
These are currently limited to integer types, but we should
be able to extend to splat vectors and possibly general vectors.
llvm-svn: 289343
Simon Pilgrim [Sat, 10 Dec 2016 22:02:36 +0000 (22:02 +0000)]
[X86][SSE] Add tests for sign extended vXi64 multiplication
llvm-svn: 289342
Simon Pilgrim [Sat, 10 Dec 2016 21:16:45 +0000 (21:16 +0000)]
[X86][SSE] Ensure UNPCK inputs are a consistent value type in LowerHorizontalByteSum
llvm-svn: 289341
Craig Topper [Sat, 10 Dec 2016 21:15:52 +0000 (21:15 +0000)]
[AVX-512] Remove 128/256 masked vpermil instrinsics and autoupgrade to a select around the unmasked avx1 intrinsics.
llvm-svn: 289340
Craig Topper [Sat, 10 Dec 2016 21:15:48 +0000 (21:15 +0000)]
[X86][IR] Move the autoupgrading of store intrinsics out of the main nested if/else chain. This should buy a little more time against the MSVC limit mentioned in PR31034.
The handlers for stores all return at the end of their block so they can be picked off early.
llvm-svn: 289339
Craig Topper [Sat, 10 Dec 2016 20:27:39 +0000 (20:27 +0000)]
[AVX-512] Remove 128/256-bit masked vpermilvar builtins and replace with select and the avx unmasked builtins.
llvm-svn: 289338
Matt Arsenault [Sat, 10 Dec 2016 19:58:00 +0000 (19:58 +0000)]
AMDGPU: Fix asan errors when folding operands
This was failing when trying to fold immediates into operand 1 of a
phi, which only has one statically known operand.
llvm-svn: 289337
Simon Pilgrim [Sat, 10 Dec 2016 19:49:55 +0000 (19:49 +0000)]
[X86][SSE] Move ZeroVector creation into the shuffle pattern case where its actually used.
Also fix the ZeroVector's type - I've no idea how this hasn't caused problems........
llvm-svn: 289336
Craig Topper [Sat, 10 Dec 2016 19:35:39 +0000 (19:35 +0000)]
[AVX-512] Add support for lowering (v2i64 (fp_to_sint (v2f32))) to vcvttps2uqq when AVX512DQ and AVX512VL are available.
llvm-svn: 289335
Craig Topper [Sat, 10 Dec 2016 19:35:36 +0000 (19:35 +0000)]
[X86] Clarify indentation. NFC
llvm-svn: 289334
Craig Topper [Sat, 10 Dec 2016 19:35:33 +0000 (19:35 +0000)]
[X86] Combine LowerFP_TO_SINT and LowerFP_TO_UINT. They only differ by a single boolean flag passed to a helper function. Just check the opcode and create the flag.
llvm-svn: 289333
Sanjay Patel [Sat, 10 Dec 2016 17:40:47 +0000 (17:40 +0000)]
[InstSimplify] improve function name; NFC
llvm-svn: 289332
Simon Atanasyan [Sat, 10 Dec 2016 17:30:09 +0000 (17:30 +0000)]
[mips] Eliminate else-after-return. NFC
llvm-svn: 289331
Rui Ueyama [Sat, 10 Dec 2016 17:23:23 +0000 (17:23 +0000)]
Create a TPI stream only when /debugpdb is given.
Adding type records to TPI stream is too time consuming.
It is reported that linking chrome_child.dll took 5 minutes.
llvm-svn: 289330
Simon Pilgrim [Sat, 10 Dec 2016 17:00:00 +0000 (17:00 +0000)]
[SelectionDAG] Add ability for computeKnownBits to peek through bitcasts from 'large element' scalar/vector to 'small element' vector.
Extension to D27129 which already supported bitcasts from 'small element' vector to 'large element' scalar/vector types.
llvm-svn: 289329
Yaron Keren [Sat, 10 Dec 2016 14:55:14 +0000 (14:55 +0000)]
Simplify parseShowColorsArgs logic, NFC.
llvm-svn: 289328
Simon Pilgrim [Sat, 10 Dec 2016 13:45:08 +0000 (13:45 +0000)]
[X86][XOP] Add permil2ps buildvector combine test
llvm-svn: 289327
Dylan McKay [Sat, 10 Dec 2016 12:08:19 +0000 (12:08 +0000)]
[AVR] Add a stub README file
llvm-svn: 289326
Dylan McKay [Sat, 10 Dec 2016 11:49:07 +0000 (11:49 +0000)]
[AVR] Fix and clean up the inline assembly tests
There was a bug where we would hit an assertion if 'Q' was used as a
constraint.
I also removed hardcoded register names to prefer regexes so the tests
don't break when the register allocator changes.
llvm-svn: 289325
Dylan McKay [Sat, 10 Dec 2016 11:23:16 +0000 (11:23 +0000)]
[AVR] Explicitly set the target in all CodeGen tests
This seems to have caused failures on the buildbot.
llvm-svn: 289324
Dylan McKay [Sat, 10 Dec 2016 11:18:37 +0000 (11:18 +0000)]
[AVR] Fix an inline asm assertion which would always trigger
It looks like some time in the past, constraint codes were changed from
chars being passed around to enums.
llvm-svn: 289323
Dylan McKay [Sat, 10 Dec 2016 10:51:55 +0000 (10:51 +0000)]
[AVR] Use the register scavenger when expanding 'LDDW' instructions
Summary: This gets rid of the hardcoded 'r0' that was used previously.
Reviewers: asl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D27567
llvm-svn: 289322
Dylan McKay [Sat, 10 Dec 2016 10:16:13 +0000 (10:16 +0000)]
[AVR] Support stores to undefined pointers
This would previously trigger an assertion error in AVRISelDAGToDAG.
llvm-svn: 289321
Ismail Donmez [Sat, 10 Dec 2016 09:29:47 +0000 (09:29 +0000)]
Fix shared library build
llvm-svn: 289320
Chandler Carruth [Sat, 10 Dec 2016 07:59:59 +0000 (07:59 +0000)]
[PM] Further broaden this test's regex as both the CGSCC and Function
inner AM proxies are now being rendered differently.
llvm-svn: 289319
Chandler Carruth [Sat, 10 Dec 2016 07:46:51 +0000 (07:46 +0000)]
[PM] Try to support the new spelling of one of the proxy names that are
showing up on the build bots.
llvm-svn: 289318
Chandler Carruth [Sat, 10 Dec 2016 06:34:44 +0000 (06:34 +0000)]
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit.
Summary:
This never really got implemented, and was very hard to test before
a lot of the refactoring changes to make things more robust. But now we
can test it thoroughly and cleanly, especially at the CGSCC level.
The core idea is that when an inner analysis manager proxy receives the
invalidation event for the outer IR unit, it needs to walk the inner IR
units and propagate it to the inner analysis manager for each of those
units. For example, each function in the SCC needs to get an
invalidation event when the SCC gets one.
The function / module interaction is somewhat boring here. This really
becomes interesting in the face of analysis-backed IR units. This patch
effectively handles all of the CGSCC layer's needs -- both invalidating
SCC analysis and invalidating function analysis when an SCC gets
invalidated.
However, this second aspect doesn't really handle the
LoopAnalysisManager well at this point. That one will need some change
of design in order to fully integrate, because unlike the call graph,
the entire function behind a LoopAnalysis's results can vanish out from
under us, and we won't even have a cached API to access. I'd like to try
to separate solving the loop problems into a subsequent patch though in
order to keep this more focused so I've adapted them to the API and
updated the tests that immediately fail, but I've not added the level of
testing and validation at that layer that I have at the CGSCC layer.
An important aspect of this change is that the proxy for the
FunctionAnalysisManager at the SCC pass layer doesn't work like the
other proxies for an inner IR unit as it doesn't directly manage the
FunctionAnalysisManager and invalidation or clearing of it. This would
create an ever worsening problem of dual ownership of this
responsibility, split between the module-level FAM proxy and this
SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy
to work in terms of the module-level proxy and defer to it to handle
much of the updates. It only does SCC-specific invalidation. This will
become more important in subsequent patches that support more complex
invalidaiton scenarios.
Reviewers: jlebar
Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits
Differential Revision: https://reviews.llvm.org/D27197
llvm-svn: 289317
Craig Topper [Sat, 10 Dec 2016 06:02:48 +0000 (06:02 +0000)]
[X86] Use X86ISD::CVTTP2SI and X86ISD::CVTTP2UI for lowering 128-bit cvttps2qq and cvttps2uqq intrinsics since there is a mismatch between number of input and output elements.
Ideally ISD::FP_TO_SINT and ISD::FP_TO_UINT would only be used for cases with the same number of input and output elements.
Similar things have already been done for other convert intrinsics.
llvm-svn: 289316
Dylan McKay [Sat, 10 Dec 2016 05:48:48 +0000 (05:48 +0000)]
[AVR] Fix a bunch of incorrect assertion messages
These should've been checking whether the immediate is a 6-bit unsigned
integer.
If the immediate was '63', this would cause an assertion error which
shouldn't have occurred.
llvm-svn: 289315
Kostya Serebryany [Sat, 10 Dec 2016 02:48:42 +0000 (02:48 +0000)]
[libFuzzer] test cleanup (3)
llvm-svn: 289314
Kostya Serebryany [Sat, 10 Dec 2016 02:47:00 +0000 (02:47 +0000)]
[libFuzzer] test cleanup (2)
llvm-svn: 289313
Kostya Serebryany [Sat, 10 Dec 2016 02:45:56 +0000 (02:45 +0000)]
[libFuzzer] test cleanup
llvm-svn: 289312
Kostya Serebryany [Sat, 10 Dec 2016 02:26:23 +0000 (02:26 +0000)]
[libFuzzer] switch all libFuzzer tests to use -fsanitize-coverage=trace-pc-guard. Support for the previosly used instrumentation will be removed in the following changes
llvm-svn: 289311
Kostya Serebryany [Sat, 10 Dec 2016 01:19:35 +0000 (01:19 +0000)]
[libFuzzer] use __sanitizer_get_module_and_offset_for_pc to get the module name while printing the coverage
llvm-svn: 289310
Devin Coughlin [Sat, 10 Dec 2016 01:16:09 +0000 (01:16 +0000)]
[analyzer] Improve VirtualCallChecker diagnostics and move into optin package.
The VirtualCallChecker is in alpha because its interprocedural diagnostics
represent the call path textually in the diagnostic message rather than with a
path sensitive diagnostic.
This patch turns off the AST-based interprocedural analysis in the checker so
that no call path is needed and improves with diagnostic text. With these
changes, the checker is ready to be moved into the optin package.
Ultimately the right fix is to rewrite this checker to be path sensitive -- but
there is still value in enabling the checker for intraprocedural analysis only
The interprocedural mode can be re-enabled with an -analyzer-config flag.
Differential Revision: https://reviews.llvm.org/D26768
llvm-svn: 289309
Vedant Kumar [Sat, 10 Dec 2016 00:54:13 +0000 (00:54 +0000)]
[clang-format] Another attempt at python 3 compatibility
The entries in vim.current.buffer appear to be decoded strings, which
means that python3 won't allow invoking 'decode' on them. Keep the old
behavior when running under python2, but skip the error-inducing decode
step with python3..
llvm-svn: 289308
Matt Arsenault [Sat, 10 Dec 2016 00:52:50 +0000 (00:52 +0000)]
AMDGPU: Fix AMDGPUPromoteAlloca breaking addrspacecasts
The users of the addrspacecast were having their types incorrectly
changed, producing invalid bitcasts between address spaces.
llvm-svn: 289307
Matt Arsenault [Sat, 10 Dec 2016 00:39:12 +0000 (00:39 +0000)]
AMDGPU: Fix handling of 16-bit immediates
Since 32-bit instructions with 32-bit input immediate behavior
are used to materialize 16-bit constants in 32-bit registers
for 16-bit instructions, determining the legality based
on the size is incorrect. Change operands to have the size
specified in the type.
Also adds a workaround for a disassembler bug that
produces an immediate MCOperand for an operand that
is supposed to be OPERAND_REGISTER.
The assembler appears to accept out of bounds immediates and
truncates them, but this seems to be an issue for 32-bit
already.
llvm-svn: 289306
Rui Ueyama [Sat, 10 Dec 2016 00:34:06 +0000 (00:34 +0000)]
Rename find(SymbolVersion) to avoid overloading with find(StringRef).
llvm-svn: 289293
Matt Arsenault [Sat, 10 Dec 2016 00:29:55 +0000 (00:29 +0000)]
AMDGPU: Fix vintrp disassembly
llvm-svn: 289292
Matt Arsenault [Sat, 10 Dec 2016 00:23:12 +0000 (00:23 +0000)]
AMDGPU: Change vintrp printing to better match sc
Some of the immediates need to be printed differently
eventually.
llvm-svn: 289291
Vedant Kumar [Fri, 9 Dec 2016 23:48:18 +0000 (23:48 +0000)]
[ubsan] Treat ObjC's BOOL as if its range is always {0, 1}
On some Apple platforms, the ObjC BOOL type is defined as a signed char.
When performing instrumentation for -fsanitize=bool, we'd like to treat
the range of BOOL like it's always {0, 1}. While we can't change clang's
IRGen for char-backed BOOL's due to ABI compatibility concerns, we can
teach ubsan to catch potential abuses of this type.
rdar://problem/
29502773
Differential Revision: https://reviews.llvm.org/D27607
llvm-svn: 289290
Peter Collingbourne [Fri, 9 Dec 2016 23:34:49 +0000 (23:34 +0000)]
COFF: Fix test failure on Windows.
llvm-svn: 289289
Paul Robinson [Fri, 9 Dec 2016 23:08:17 +0000 (23:08 +0000)]
Bigger-hammer REQUIRES to fix Windows bot.
llvm-svn: 289288
Artem Belevich [Fri, 9 Dec 2016 22:59:17 +0000 (22:59 +0000)]
[CUDA,Driver] Added --no-cuda-gpu-arch= option.
This allows us to negate preceding --cuda-gpu-arch=X.
This comes handy when user needs to override default
flags set for them by the build system.
Differential Revision: https://reviews.llvm.org/D27631
llvm-svn: 289287
Richard Smith [Fri, 9 Dec 2016 22:56:20 +0000 (22:56 +0000)]
[c++17] P0490R0, NB comment FI 20: allow direct-initialization of decomposition declarations.
llvm-svn: 289286
Simon Pilgrim [Fri, 9 Dec 2016 22:45:21 +0000 (22:45 +0000)]
Fix unused variable warnings. NFCI.
llvm-svn: 289285
Rafael Espindola [Fri, 9 Dec 2016 22:40:49 +0000 (22:40 +0000)]
Fix a bogus warning.
We first decide that the symbol is global, than that it should have
version foo. Since it was already not the default version, we were
producing a bogus warning.
llvm-svn: 289284
Stephan T. Lavavej [Fri, 9 Dec 2016 22:35:53 +0000 (22:35 +0000)]
[libcxx] [test] Fix string_view tests.
test/std/strings/string.view/string.view.ops/compare.pointer_size.pass.cpp
Passing -1 to size_t triggers signed/unsigned mismatch warnings because it's
a value-modifying conversion. Add static_cast<size_t> to soothe the compiler.
(This file refers to size_t unqualified.)
test/std/strings/string.view/string.view.ops/substr.pass.cpp
Add <algorithm> for std::min() and <stdexcept> for std::out_of_range.
N4618 21.4.2.4 [string.view.access]/1: "Requires: pos < size()."
/4: "[ Note: Unlike basic_string::operator[], basic_string_view::operator[](size())
has undefined behavior instead of returning charT(). -end note ]"
Fixes D27633.
llvm-svn: 289283
Eugene Zelenko [Fri, 9 Dec 2016 22:06:55 +0000 (22:06 +0000)]
[AMDGPU, PowerPC, TableGen] Fix some Clang-tidy modernize and Include What You Use warnings; other minor fixes (NFC).
llvm-svn: 289282
Paul Robinson [Fri, 9 Dec 2016 21:59:00 +0000 (21:59 +0000)]
Speculative REQUIRES to fix Windows bot.
llvm-svn: 289281
Peter Collingbourne [Fri, 9 Dec 2016 21:55:24 +0000 (21:55 +0000)]
COFF: New symbol table design.
This ports the ELF linker's symbol table design, introduced in r268178,
to the COFF linker.
Differential Revision: http://reviews.llvm.org/D21166
llvm-svn: 289280
Simon Pilgrim [Fri, 9 Dec 2016 21:53:12 +0000 (21:53 +0000)]
[X86] Regenerate test
llvm-svn: 289279
Meador Inge [Fri, 9 Dec 2016 21:51:37 +0000 (21:51 +0000)]
[ELF] Fix test case thinko from r289152
It was pointed out in a post-commit review that the tests
were structured oddly. Fixed thusly.
llvm-svn: 289278
Rafael Espindola [Fri, 9 Dec 2016 21:46:39 +0000 (21:46 +0000)]
Add support for R_X86_64_GOT64.
llvm-svn: 289277
Graydon Hoare [Fri, 9 Dec 2016 21:45:49 +0000 (21:45 +0000)]
[modules] Add optional out-param to ASTReader::ReadAST for imported submodules.
Summary:
The Swift frontend is acquiring the ability to load non-module PCH files containing
bridging definitions from C/ObjC. As part of this work, it needs to know which submodules
were imported by a PCH in order to wrap them in local Swift modules. This information
is collected by ASTReader::ReadAST in a local vector, but is currently kept private.
The change here is just to make the type of the vector elements public, and provide
an optional out-parameter to the ReadAST method to provide the vector's contents to
a caller after a successful read.
Reviewers: manmanren, rsmith, doug.gregor
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D27580
llvm-svn: 289276
Reid Kleckner [Fri, 9 Dec 2016 21:10:43 +0000 (21:10 +0000)]
Remove special error recovery for ::(id)
The code pattern used to implement the token rewriting hack doesn't
interact well with token caching in the pre-processor. As a result,
clang would crash on 'int f(::(id));' while doing a tenative parse of
the contents of the outer parentheses. The original code from PR11852
still doesn't crash the compiler.
This error recovery also often does the wrong thing with member function
pointers. The test case from the original PR doesn't recover the right
way either:
void S::(*pf)() = S::f; // should be 'void (S::*pf)()'
Instead we were recovering as 'void S::*pf()', which is still wrong.
If we still think that users mistakenly parenthesize identifiers in
nested name specifiers, we should change clang to intentionally parse
that form with an error, rather than doing a token rewrite.
Fixes PR26623, but I think there will be many more bugs like this around
token rewriting in the parser.
Reviewers: rsmith, rtrieu
Differential Revision: https://reviews.llvm.org/D25882
llvm-svn: 289273
Matt Arsenault [Fri, 9 Dec 2016 21:10:41 +0000 (21:10 +0000)]
AMDGPU: Cleanup checks in sext_inreg test
llvm-svn: 289272
Hans Wennborg [Fri, 9 Dec 2016 20:54:44 +0000 (20:54 +0000)]
COFF: Support both /lldmap and /lldmap:<file>
The former option bases the filename on the output name, e.g. if the
link output is a.exe, the map will be written to a.map. This matches the
behaviour of link.exe's /MAP option and is useful for creating a map
file of each executable when building a large project.
Differential Revision: https://reviews.llvm.org/D27595
llvm-svn: 289271
Peter Collingbourne [Fri, 9 Dec 2016 20:51:33 +0000 (20:51 +0000)]
COFF: Move a call to toString() out of the hot path.
Profiling revealed that we were spending 5% of our time linking
chrome_child.dll just in this call to toString().
Differential Revision: https://reviews.llvm.org/D27628
llvm-svn: 289270
Adrian Prantl [Fri, 9 Dec 2016 20:43:40 +0000 (20:43 +0000)]
Fix LLVM's use of DW_OP_bit_piece in DWARF expressions.
LLVM's use of DW_OP_bit_piece is incorrect and a based on a
misunderstanding of the wording in the DWARF specification. The offset
argument of DW_OP_bit_piece refers to the offset into the location
that is on the top of the DWARF expression stack, and not an offset
into the source variable. This has since also been clarified in the
DWARF specification.
This patch fixes all uses of DW_OP_bit_piece to emit the correct
offset and simplifies the DwarfExpression class to semi-automaticaly
emit empty DW_OP_pieces to adjust the offset of the source variable,
thus simplifying the code using DwarfExpression.
While this is an incompatible bugfix, in practice I don't expect this
to be much of a problem since LLVM's old interpretation and the
correct interpretation of DW_OP_bit_piece differ only when there are
gaps in the fragmented locations of the described variables or if
individual fragments are smaller than a byte. LLDB at least won't
interpret locations with gaps in them because is has no way to present
undefined bits in a variable, and there is a high probability that an
old-form expression will be malformed when interpreted correctly,
because the DW_OP_bit_piece offset will be outside of the location at
the top of the stack.
As a nice side-effect, this patch enables us to use a more efficient
encoding for subregisters: In order to express a sub-register at a
non-zero offset we now use a DW_OP_bit_piece instead of shifting the
value into place manually.
This patch also adds missing test coverage for code paths that weren't
exercised before.
<rdar://problem/
29335809>
Differential Revision: https://reviews.llvm.org/D27550
llvm-svn: 289266
Matthias Braun [Fri, 9 Dec 2016 20:16:12 +0000 (20:16 +0000)]
Add README describing the intention of test/CodeGen/MIR
llvm-svn: 289265
Stephan T. Lavavej [Fri, 9 Dec 2016 19:53:08 +0000 (19:53 +0000)]
[libcxx] [test] Add LIBCPP_ASSERT_NOEXCEPT/LIBCPP_ASSERT_NOT_NOEXCEPT, remove an unused variable.
test/support/test_macros.h
For convenience/greppability, add macros for libcxx-specific static_asserts about noexceptness.
(Moving the definitions of ASSERT_NOEXCEPT/ASSERT_NOT_NOEXCEPT isn't technically necessary
because they're macros, but I think it's better style to define stuff before using it.)
test/std/utilities/tuple/tuple.tuple/tuple.apply/apply.pass.cpp
There was a completely unused `TrackedCallable obj;`.
apply() isn't depicted with conditional noexcept in C++17.
test/std/utilities/tuple/tuple.tuple/tuple.apply/make_from_tuple.pass.cpp
Now that we have LIBCPP_ASSERT_NOEXCEPT, use it.
Fixes D27622.
llvm-svn: 289264
Marek Olsak [Fri, 9 Dec 2016 19:49:58 +0000 (19:49 +0000)]
AMDGPU/SI: Remove XNACK feature from CI
Summary: CI doesn't have XNACK.
Reviewers: tstellarAMD
Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27175
llvm-svn: 289263
Marek Olsak [Fri, 9 Dec 2016 19:49:54 +0000 (19:49 +0000)]
AMDGPU/SI: Don't reserve XNACK when it's disabled
Summary:
This frees 2 additional scalar registers.
These are results from all of my 3 patches combined:
Polaris:
Spilled SGPRs: 2231 -> 1517 (-32.00 %)
Tonga:
Spilled SGPRs: 3829 -> 2608 (-31.89 %)
Spilled VGPRs: 100 -> 84 (-16.00 %)
Tonga even spills SGPRs via VGPRs to scratch. That's a compute shader
limited to 64 VGPRs.
Reviewers: tstellarAMD
Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27151
llvm-svn: 289262
Marek Olsak [Fri, 9 Dec 2016 19:49:48 +0000 (19:49 +0000)]
AMDGPU/SI: Don't reserve FLAT_SCR on non-HSA targets & without stack objects
Summary: This frees 2 scalar registers.
Reviewers: tstellarAMD
Subscribers: qcolombet, arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27150
llvm-svn: 289261
Marek Olsak [Fri, 9 Dec 2016 19:49:40 +0000 (19:49 +0000)]
AMDGPU/SI: Allow using SGPRs 96-101 on VI
Summary:
There is no point in setting SGPRS=104, because VI allocates SGPRs
in multiples of 16, so 104 -> 112. That enables us to use all 102 SGPRs
for general purposes.
Reviewers: tstellarAMD
Subscribers: qcolombet, arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27149
llvm-svn: 289260
Reid Kleckner [Fri, 9 Dec 2016 19:47:58 +0000 (19:47 +0000)]
Improve error message when referencing a non-tag type with a tag
Other compilers accept invalid code here that we reject, and we need a
better error message to try to convince users that the code is really
incorrect. Consider:
class Foo {
typedef MyIterHelper<Foo> iterator;
friend class iterator;
};
Previously our wording was "elaborated type refers to a typedef".
"elaborated type" isn't widely known terminology, so the new diagnostic
says "typedef 'iterator' cannot be referenced with class specifier".
Reviewers: rsmith
Differential Revision: https://reviews.llvm.org/D25216
llvm-svn: 289259
Richard Smith [Fri, 9 Dec 2016 19:35:45 +0000 (19:35 +0000)]
Add tests for a couple more DRs.
llvm-svn: 289258
Reid Kleckner [Fri, 9 Dec 2016 19:20:28 +0000 (19:20 +0000)]
Remove /Zc:sizedDealloc- from the MSVC build
According to the connect bug
(https://connect.microsoft.com/VisualStudio/feedback/details/1351894),
this was only necessary with pre-release versions of MSVC 2015.
Fixes PR23513
llvm-svn: 289257
Paul Robinson [Fri, 9 Dec 2016 19:15:32 +0000 (19:15 +0000)]
[DWARF] Suppress .loc directives from CFI instructions
Like DBG_VALUE, these emit nothing to the .text section, and sometimes
have no source location specified. Just ignore them.
Differential Revision: http://reviews.llvm.org/D27492
llvm-svn: 289256
Richard Smith [Fri, 9 Dec 2016 19:11:50 +0000 (19:11 +0000)]
cxx_dr_status: update to latest issue list and add a couple more tests.
llvm-svn: 289255
Matthias Braun [Fri, 9 Dec 2016 19:08:15 +0000 (19:08 +0000)]
Move .mir tests to appropriate directories
test/CodeGen/MIR should contain tests that intent to test the MIR
printing or parsing. Tests that test something else should be in
test/CodeGen/TargetName even when they are written in .mir.
As a rule of thumb, only tests using "llc -run-pass none" should be in
test/CodeGen/MIR.
llvm-svn: 289254
Renato Golin [Fri, 9 Dec 2016 19:02:04 +0000 (19:02 +0000)]
[DFSAN] Another unstable test in AArch64 breaking bots unnecessarily
llvm-svn: 289253
Yaxun Liu [Fri, 9 Dec 2016 19:01:11 +0000 (19:01 +0000)]
Add support for non-zero null pointer for C and OpenCL
In amdgcn target, null pointers in global, constant, and generic address space take value 0 but null pointers in private and local address space take value -1. Currently LLVM assumes all null pointers take value 0, which results in incorrectly translated IR. To workaround this issue, instead of emit null pointers in local and private address space, a null pointer in generic address space is emitted and casted to local and private address space.
Tentative definition of global variables with non-zero initializer will have weak linkage instead of common linkage since common linkage requires zero initializer and does not have explicit section to hold the non-zero value.
Virtual member functions getNullPointer and performAddrSpaceCast are added to TargetCodeGenInfo which by default returns ConstantPointerNull and emitting addrspacecast instruction. A virtual member function getNullPointerValue is added to TargetInfo which by default returns 0. Each target can override these virtual functions to get target specific null pointer and the null pointer value for specific address space, and perform specific translations for addrspacecast.
Wrapper functions getNullPointer is added to CodegenModule and getTargetNullPointerValue is added to ASTContext to facilitate getting the target specific null pointers and their values.
This change has no effect on other targets except amdgcn target. Other targets can provide support of non-zero null pointer in a similar way.
This change only provides support for non-zero null pointer for C and OpenCL. Supporting for other languages will be added later incrementally.
Differential Revision: https://reviews.llvm.org/D26196
llvm-svn: 289252
Rui Ueyama [Fri, 9 Dec 2016 18:49:37 +0000 (18:49 +0000)]
Remove `REQUIRES: shell` hack to workaround an echo issue.
These tests are disabled on Windows, but they seem to work just fine now,
so I'll enable them.
llvm-svn: 289251
Richard Smith [Fri, 9 Dec 2016 18:49:13 +0000 (18:49 +0000)]
DR1295 and cleanup for P0135R1: Make our initialization code more directly
mirror the description in the standard. Per DR1295, this means that binding a
const / rvalue reference to a bit-field no longer "binds directly", and per
P0135R1, this means that we materialize a temporary in reference binding
after adjusting cv-qualifiers and before performing a derived-to-base cast.
In C++11 onwards, this should have fixed the last case where we would
materialize a temporary of the wrong type (with a subobject adjustment inside
the MaterializeTemporaryExpr instead of outside), but we still have to deal
with that possibility in C++98, unless we want to start using xvalues to
represent materialized temporaries there too.
llvm-svn: 289250
Ekaterina Romanova [Fri, 9 Dec 2016 18:35:50 +0000 (18:35 +0000)]
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers.
Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font.
In the past, \c command was used, unfortunately it applied to only one word.
<c> .. </c> has the same meaning, but applies to all words in between the tags.
llvm-svn: 289249
Matt Arsenault [Fri, 9 Dec 2016 17:57:43 +0000 (17:57 +0000)]
AMDGPU: Fix isTypeDesirableForOp for i16
This should do nothing for targets without i16.
llvm-svn: 289235
Reid Kleckner [Fri, 9 Dec 2016 17:56:04 +0000 (17:56 +0000)]
Relax IRgen check in decl-in-prototype.c to match signext on PPC64
llvm-svn: 289234
Greg Clayton [Fri, 9 Dec 2016 17:54:59 +0000 (17:54 +0000)]
Fix i386 being able to show member variables correctly by not returning empty objective C types from the runtime.
We don't parse ObjC v1 types from the runtime metadata like we do for ObjC v2, but doing so by creating empty types was ruining the i386 v1 debugging experience.
<rdar://problem/
24093343>
llvm-svn: 289233
Simon Pilgrim [Fri, 9 Dec 2016 17:53:11 +0000 (17:53 +0000)]
[SelectionDAG] Add knownbits support for EXTRACT_VECTOR_ELT opcodes (REAPPLIED)
Reapplied with fix for PR31323 - X86 SSE2 vXi16 multiplies for illegal types were creating CONCAT_VECTORS nodes with vector inputs that might not total the number of elements in the result type.
llvm-svn: 289232
Matt Arsenault [Fri, 9 Dec 2016 17:49:14 +0000 (17:49 +0000)]
AMDGPU: Fix i128 mul
llvm-svn: 289231
Matt Arsenault [Fri, 9 Dec 2016 17:49:11 +0000 (17:49 +0000)]
AMDGPU: Allow TBA, TMA, TTMP* registers with SMEM instructions
Fixes assembler regressions.
llvm-svn: 289230
Matt Arsenault [Fri, 9 Dec 2016 17:49:08 +0000 (17:49 +0000)]
AMDGPU: Clean up instruction bits
Sort the instruction bits by type and make sure there is one
for each format.
Also cleanup namespaces.
llvm-svn: 289229
Nico Weber [Fri, 9 Dec 2016 17:32:52 +0000 (17:32 +0000)]
Don't assert when redefining a built-in macro in a PCH, PR29119
PCH files store the macro history for a given macro, and the whole history list
for one identifier is given to the Preprocessor at once via
Preprocessor::setLoadedMacroDirective(). This contained an assert that no macro
history exists yet for that identifier. That's usually true, but it's not true
for builtin macros, which are created in Preprocessor() before flags and pchs
are processed. Luckily, ASTWriter stops writing macro history lists at builtins
(see shouldIgnoreMacro() in ASTWriter.cpp), so the head of the history list was
missing for builtin macros. So make the assert weaker, and splice the history
list to the existing single define for builtins.
https://reviews.llvm.org/D27545
llvm-svn: 289228
Sean Fertile [Fri, 9 Dec 2016 17:21:42 +0000 (17:21 +0000)]
[PPC] Add intrinsics for vector extract word and vector insert word.
Revision: https://reviews.llvm.org/D26547
llvm-svn: 289227
Nirav Dave [Fri, 9 Dec 2016 17:18:24 +0000 (17:18 +0000)]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled."
This reverts commit r289221 which appears to be triggering an assertion
llvm-svn: 289226
Reid Kleckner [Fri, 9 Dec 2016 17:14:05 +0000 (17:14 +0000)]
Store decls in prototypes on the declarator instead of in the AST
This saves two pointers from FunctionDecl that were being used for some
rare and questionable C-only functionality. The DeclsInPrototypeScope
ArrayRef was added in r151712 in order to parse this kind of C code:
enum e {x, y};
int f(enum {y, x} n) {
return x; // should return 1, not 0
}
The challenge is that we parse 'int f(enum {y, x} n)' it its own
function prototype scope that gets popped before we build the
FunctionDecl for 'f'. The original change was doing two questionable
things:
1. Saving all tag decls introduced in prototype scope on a TU-global
Sema variable. This is problematic when you have cases like this, where
'x' and 'y' shouldn't be visible in 'f':
void f(void (*fp)(enum { x, y } e)) { /* no x */ }
This patch fixes that, so now 'f' can't see 'x', which is consistent
with GCC.
2. Storing the decls in FunctionDecl in ActOnFunctionDeclarator so that
they could be used in ActOnStartOfFunctionDef. This is just an
inefficient way to move information around. The AST lives forever, but
the list of non-parameter decls in prototype scope is short lived.
Moving these things to the Declarator solves both of these issues.
Reviewers: rsmith
Subscribers: jmolloy, cfe-commits
Differential Revision: https://reviews.llvm.org/D27279
llvm-svn: 289225
Rafael Espindola [Fri, 9 Dec 2016 16:44:05 +0000 (16:44 +0000)]
Fix parsing when one extern follows another.
llvm-svn: 289224
Greg Clayton [Fri, 9 Dec 2016 16:25:13 +0000 (16:25 +0000)]
Fix buildbots that are failing due to this test by adding all expected fails that TestMultipleDebuggers.py has.
llvm-svn: 289223
Greg Clayton [Fri, 9 Dec 2016 16:22:10 +0000 (16:22 +0000)]
Rename multiple target test so it is unique.
llvm-svn: 289222
Nirav Dave [Fri, 9 Dec 2016 16:15:12 +0000 (16:15 +0000)]
In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled.
Retrying after fixing overly aggressive load-store forwarding optimization.
Simplify Consecutive Merge Store Candidate Search
Now that address aliasing is much less conservative, push through
simplified store merging search which only checks for parallel stores
through the chain subgraph. This is cleaner as the separation of
non-interfering loads/stores from the store-merging logic.
Whem merging stores, search up the chain through a single load, and
finds all possible stores by looking down from through a load and a
TokenFactor to all stores visited. This improves the quality of the
output SelectionDAG and generally the output CodeGen (with some
exceptions).
Additional Minor Changes:
1. Finishes removing unused AliasLoad code
2. Unifies the the chain aggregation in the merged stores across
code paths
3. Re-add the Store node to the worklist after calling
SimplifyDemandedBits.
4. Increase GatherAllAliasesMaxDepth from 6 to 18. That number is
arbitrary, but seemed sufficient to not cause regressions in
tests.
This finishes the change Matt Arsenault started in r246307 and
jyknight's original patch.
Many tests required some changes as memory operations are now
reorderable. Some tests relying on the order were changed to use
volatile memory operations
Noteworthy tests:
CodeGen/AArch64/argument-blocks.ll -
It's not entirely clear what the test_varargs_stackalign test is
supposed to be asserting, but the new code looks right.
CodeGen/AArch64/arm64-memset-inline.lli -
CodeGen/AArch64/arm64-stur.ll -
CodeGen/ARM/memset-inline.ll -
The backend now generates *worse* code due to store merging
succeeding, as we do do a 16-byte constant-zero store efficiently.
CodeGen/AArch64/merge-store.ll -
Improved, but there still seems to be an extraneous vector insert
from an element to itself?
CodeGen/PowerPC/ppc64-align-long-double.ll -
Worse code emitted in this case, due to the improved store->load
forwarding.
CodeGen/X86/dag-merge-fast-accesses.ll -
CodeGen/X86/MergeConsecutiveStores.ll -
CodeGen/X86/stores-merging.ll -
CodeGen/Mips/load-store-left-right.ll -
Restored correct merging of non-aligned stores
CodeGen/AMDGPU/promote-alloca-stored-pointer-value.ll -
Improved. Correctly merges buffer_store_dword calls
CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll -
Improved. Sidesteps loading a stored value and
merges two stores
CodeGen/X86/pr18023.ll -
This test has been removed, as it was asserting incorrect
behavior. Non-volatile stores *CAN* be moved past volatile loads,
and now are.
CodeGen/X86/vector-idiv.ll -
CodeGen/X86/vector-lzcnt-128.ll -
It's basically impossible to tell what these tests are actually
testing. But, looks like the code got better due to the memory
operations being recognized as non-aliasing.
CodeGen/X86/win32-eh.ll -
Both loads of the securitycookie are now merged.
Reviewers: arsenm, hfinkel, tstellarAMD, jyknight, nhaehnle
Subscribers: wdng, nhaehnle, nemanjai, arsenm, weimingz, niravd, RKSimon, aemerson, qcolombet, dsanders, resistor, tstellarAMD, t.p.northover, spatel
Differential Revision: https://reviews.llvm.org/D14834
llvm-svn: 289221
Simon Pilgrim [Fri, 9 Dec 2016 16:01:50 +0000 (16:01 +0000)]
Use SelectionDAG.getSplatBuildVector helper. NFCI.
llvm-svn: 289220
Tom Stellard [Fri, 9 Dec 2016 15:57:15 +0000 (15:57 +0000)]
AMDGPU/SI: Don't mark VINTRP instructions as mayLoad
Summary:
These instructions technically do read from memory, but the memory
is considered to be out of bounds for normal load/store instructions.
shader-db stats:
SGPRS: 1416075 -> 1413323 (-0.19 %)
VGPRS: 867413 -> 863935 (-0.40 %)
Spilled SGPRs: 1409 -> 1354 (-3.90 %)
Spilled VGPRs: 63 -> 63 (0.00 %)
Private memory VGPRs: 880 -> 880 (0.00 %)
Scratch size: 2648 -> 2632 (-0.60 %) dwords per thread
Code Size:
37889052 ->
37897340 (0.02 %) bytes
LDS: 2147 -> 2147 (0.00 %) blocks
Max Waves: 279243 -> 280369 (0.40 %)
Wait states: 0 -> 0 (0.00 %)
Reviewers: nhaehnle, mareko, arsenm
Subscribers: kzhuravl, wdng, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27593
llvm-svn: 289219
Simon Pilgrim [Fri, 9 Dec 2016 15:23:41 +0000 (15:23 +0000)]
[SelectionDAG] Use SelectionDAG.getBuildVector helper. NFCI.
Makes interception of BUILD_VECTOR creation easier for debugging.
llvm-svn: 289218
Rafael Espindola [Fri, 9 Dec 2016 15:08:40 +0000 (15:08 +0000)]
Don't crash on an extra symbol in a version script.
llvm-svn: 289217
Sanjoy Das [Fri, 9 Dec 2016 14:42:14 +0000 (14:42 +0000)]
[SCEVExpander] Remove \brief, reflow comments; NFC
llvm-svn: 289216
Sanjoy Das [Fri, 9 Dec 2016 14:42:11 +0000 (14:42 +0000)]
[SCEVExpander] Use llvm data structures; NFC
llvm-svn: 289215
Simon Pilgrim [Fri, 9 Dec 2016 14:27:52 +0000 (14:27 +0000)]
[SelectionDAG] Add additional checks to CONCAT_VECTORS creation
Part of the work for PR31323 - add extra asserts checking that the input vectors are of consistent type and result in the correct number of vector elements.
llvm-svn: 289214