Sanjay Patel [Sat, 10 Mar 2018 16:51:28 +0000 (16:51 +0000)]
[InstSimplify] fp_binop X, undef --> NaN
The variable operand could be NaN, so it's always safe to propagate NaN.
llvm-svn: 327212
Sanjay Patel [Sat, 10 Mar 2018 16:39:59 +0000 (16:39 +0000)]
[AMDGPU] fix tests to be independent of FP undef
llvm-svn: 327211
Sanjay Patel [Sat, 10 Mar 2018 16:14:05 +0000 (16:14 +0000)]
[PowerPC] fix tests to be independent of FP undef
llvm-svn: 327210
Matt Arsenault [Sat, 10 Mar 2018 16:05:35 +0000 (16:05 +0000)]
AMDGPU: Fix crash when constant folding with physreg operand
llvm-svn: 327209
Sanjay Patel [Sat, 10 Mar 2018 15:56:25 +0000 (15:56 +0000)]
[ConstantFold] fp_binop AnyConstant, undef --> NaN
With the updated LangRef ( D44216 / rL327138 ) in place, we can proceed with more constant folding.
I'm intentionally taking the conservative path here: no matter what the constant or the FMF, we can
always fold to NaN. This is because the undef operand can be chosen as NaN, and in our simplified
default FP env, nothing else happens - NaN just propagates to the result. If we find some way/need
to propagate undef instead, that can be added subsequently.
The tests show that we always choose the same quiet NaN constant (0x7FF8000000000000 in IR text).
There were suggestions to improve that with a 'NaN' string token or not always print a 64-bit hex
value, but those are independent changes. We might also consider setting/propagating the payload of
NaN constants as an enhancement.
Differential Revision: https://reviews.llvm.org/D44308
llvm-svn: 327208
Florian Hahn [Sat, 10 Mar 2018 14:53:44 +0000 (14:53 +0000)]
[PartialInlining] Use isInlineViable to detect constructs preventing inlining.
Use isInlineViable to prevent inlining of functions with non-inlinable
constructs, in case cost analysis is skipped.
Reviewers: efriedma, sfertile, davide, davidxl
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D42846
llvm-svn: 327207
Akira Hatanaka [Sat, 10 Mar 2018 06:36:08 +0000 (06:36 +0000)]
[ObjC] Allow declaring __weak pointer fields in C structs in ARC.
This patch uses the infrastructure added in r326307 for enabling
non-trivial fields to be declared in C structs to allow __weak fields in
C structs in ARC.
rdar://problem/
33599681
Differential Revision: https://reviews.llvm.org/D44095
llvm-svn: 327206
Craig Topper [Sat, 10 Mar 2018 06:05:13 +0000 (06:05 +0000)]
[X86] Add a missing EVEX instruction to EmitAnyX86InstComments.
The equivalent SSE and VEX instruction are already there.
llvm-svn: 327205
Akira Hatanaka [Sat, 10 Mar 2018 05:55:21 +0000 (05:55 +0000)]
[Driver] Pass Default=false to hasFlag.
I forgot to do this in r326530.
llvm-svn: 327204
Sam Clegg [Sat, 10 Mar 2018 05:34:04 +0000 (05:34 +0000)]
[WebAssembly] Object: Add accessor for wasm symbols
llvm-svn: 327203
Craig Topper [Sat, 10 Mar 2018 05:15:22 +0000 (05:15 +0000)]
[X86] Move the AC_EVEX_2_VEX AsmComments enum to X86InstrInfo.h from X86InstComments.h.
X86InstComments.h is used by tools that only have the MC layer. We shouldn't be importing a file from CodeGen into this.
X86InstrInfo.h isn't a great place, but I couldn't find a better one.
llvm-svn: 327202
George Burgess IV [Sat, 10 Mar 2018 03:34:43 +0000 (03:34 +0000)]
[AliasAnalysis] Shrink AliasResults; NFC
We're persisting AliasResults in some places in MemorySSA, so the size
of these now matters a little bit (well, 8 regular-sized bits, to be
precise).
Do the same for ModRefInfo for consistency.
llvm-svn: 327201
Craig Topper [Sat, 10 Mar 2018 03:12:00 +0000 (03:12 +0000)]
[X86] Rewrite printMasking code in X86InstComments to use TSFlags to determine whether the instruction is masked.
This should have been NFC, but it looks like we were missing PUNPCKLHQDQ/PUNPCKLQDQ instructions in there.
llvm-svn: 327200
George Burgess IV [Sat, 10 Mar 2018 02:51:10 +0000 (02:51 +0000)]
Clean up a temp file on the buildbots.
r327100 made us stop producing vecreduce-propagate-sd-flags.s, but it's
still sticking around on some bots. This makes the bots unhappy.
I'll revert this tomorrow.
llvm-svn: 327199
Rafael Espindola [Sat, 10 Mar 2018 02:42:14 +0000 (02:42 +0000)]
Go back to sometimes assuming intristics are local.
This fixes pr36674.
While it is valid for shouldAssumeDSOLocal to return false anytime,
always returning false for intrinsics is not optimal on i386 and also
hits a bug in the backend.
To use a plt, the caller must first setup ebx to handle the case of
that file being linked into a PIE executable or shared library. In
those cases the generated PLT uses ebx.
Currently we can produce "calll expf@plt" without setting ebx. We
could fix that by correctly setting ebx, but this would produce worse
code for the case where the runtime library is statically linked. It
would also required other tools to handle R_386_PLT32.
llvm-svn: 327198
Nirav Dave [Sat, 10 Mar 2018 02:16:15 +0000 (02:16 +0000)]
Revert: r327172 "Correct load-op-store cycle detection analysis"
r327171 "Improve Dependency analysis when doing multi-node Instruction Selection"
r328170 "[DAG] Enforce stricter NodeId invariant during Instruction selection"
Reverting patch as NodeId invariant change is causing pathological
increases in compile time on PPC
llvm-svn: 327197
Petr Hosek [Sat, 10 Mar 2018 02:00:18 +0000 (02:00 +0000)]
[Fuzzer] When building for Fuchsia, add an explicit libc++ dependency
libFuzzer dependes on C++ library, when building for Fuchsia, we
need to ensure that libFuzzer is only being built after libc++
has been built, so we add an explicity dependency on it.
Differential Revision: https://reviews.llvm.org/D44340
llvm-svn: 327196
Richard Smith [Sat, 10 Mar 2018 01:47:22 +0000 (01:47 +0000)]
Revert r326946. It caused stack overflows by significantly increasing the size of a CallArgList.
llvm-svn: 327195
Petr Hosek [Sat, 10 Mar 2018 01:20:11 +0000 (01:20 +0000)]
[CMake] Copy the generated __config header into build directory
When the generated __config file is being used, it is currently only
copied during installation process. However, that means that the file
that gets copied into LLVM build directory is the vanilla __config file,
and any parts of the build that depend on the just built toolchain like
sanitizers will get that instead of the generated version. To avoid this
issue, we need to copy the generated header into the LLVM build
directory as well.
Differential Revision: https://reviews.llvm.org/D43797
llvm-svn: 327194
Adrian Prantl [Sat, 10 Mar 2018 01:11:25 +0000 (01:11 +0000)]
Rename clang.modules-cache-path to symbols.clang-modules-cache-path
I want to extend the properties on ModuleList to also contain other
more general settings and renaming the settings category to symbols
seems to be the least bad of choices.
llvm-svn: 327193
George Burgess IV [Sat, 10 Mar 2018 01:11:17 +0000 (01:11 +0000)]
[CodeGen] Try to not call a dtor after lifetime.end
If CodeGenFunction::EmitCall is:
- asked to emit a call with an indirectly returned value,
- given an invalid return value slot, and
- told the return value of the function it's calling is unused
then it'll make its own temporary, and add lifetime markers so that the
temporary's lifetime ends immediately after the call.
The early lifetime.end becomes problematic when we need to run a
destructor on the result of the function.
Instead of unconditionally saying that results of all calls are used
here (which would be correct, but would also cause us to never emit
lifetime markers for these temporaries), we just build our own temporary
to pass in when a dtor has to be run.
llvm-svn: 327192
Mike Edwards [Sat, 10 Mar 2018 00:53:05 +0000 (00:53 +0000)]
[libcxx][test] Reverting r327178 and r327190.
Reverting changes made to iter_alloc_deduction.fail.cpp
as my changes seem to be making several Linux bots angry.
llvm-svn: 327191
Mike Edwards [Sat, 10 Mar 2018 00:19:25 +0000 (00:19 +0000)]
[libcxx][test] Marking libcpp-no-deduction-guides unsupported.
This fixes linux bot failures with r327178.
llvm-svn: 327190
Abderrazek Zaafrani [Fri, 9 Mar 2018 23:39:34 +0000 (23:39 +0000)]
[ARM] Add ARMv8.2-A FP16 vector intrinsic
Add the fp16 neon vector intrinsic for ARM as described in the ARM ACLE document.
Reviews in https://reviews.llvm.org/D43650
llvm-svn: 327189
Craig Topper [Fri, 9 Mar 2018 23:36:58 +0000 (23:36 +0000)]
[TwoAddressInstructionPass] Improve tryInstructionCommute of X86 FMA and vpternlog instructions
These instructions have 3 operands that can be commuted. The first commute we find may not be the best. So we should keep searching if we performed an aggressive commute. There may still be an operand that is killed or a physical register constraint that might be better.
Differential Revision: https://reviews.llvm.org/D44324
llvm-svn: 327188
Heejin Ahn [Fri, 9 Mar 2018 23:29:06 +0000 (23:29 +0000)]
[WebAssembly] Add EVT::getEVTString() for except_ref type
Summary: This was missing in D43706 (rL326985).
Reviewers:
Subscribers: jfb, sbc100, jgravelle-google, sunfish, llvm-commits
Differential Revision: https://reviews.llvm.org/D44325
llvm-svn: 327187
Julie Hockett [Fri, 9 Mar 2018 23:26:56 +0000 (23:26 +0000)]
[clang-tidy] Update run-clang-tidy.py with config arg
Updating the run-clang-tidy.py script to allow specification of the
config argument to the clang-tidy invocation.
Differential Revision: https://reviews.llvm.org/D43538
llvm-svn: 327186
Dan Liew [Fri, 9 Mar 2018 23:12:32 +0000 (23:12 +0000)]
[asan] Clean up the REQUIRES of `asan/TestCases/strcat-overlap.cc`
by explicitly listing where we expect this test to work.
Because this test invokes undefined behaviour all sorts of things
can happen (e.g. crash, or on some bots asan manages to catch
something). The test seems to pass okay on Darwin (x86_64/x86_64/i386)
and on Linux (x86_64/i386) so explicitly require one of these platforms
to run the test.
llvm-svn: 327185
Matt Morehouse [Fri, 9 Mar 2018 23:02:22 +0000 (23:02 +0000)]
[clangd-fuzzer] Update ClangdLSPServer constructor call.
Build was broken by r326719.
llvm-svn: 327184
Saleem Abdulrasool [Fri, 9 Mar 2018 23:00:29 +0000 (23:00 +0000)]
test: repair windows build after SVN r327105
Thanks to Nico Weber for pointing out the failure. Add an explicit
target for the test.
llvm-svn: 327183
Rui Ueyama [Fri, 9 Mar 2018 22:59:34 +0000 (22:59 +0000)]
[WebAssembly] Remove a second parameter from toString().
toString(T) is a stringize function for an object of type T. Each type
that has that function defined should know how to stringize itself, and
there should be one string representation of an object. Passing a
"supplemental" argument to toString() breaks that princple. We shouldn't
add a second parameter to that function.
Differential Revision: https://reviews.llvm.org/D44323
llvm-svn: 327182
Peter Collingbourne [Fri, 9 Mar 2018 22:52:22 +0000 (22:52 +0000)]
ADT: Make MapVector::value_type and MapVector::size_type public. NFC.
llvm-svn: 327180
Craig Topper [Fri, 9 Mar 2018 22:24:12 +0000 (22:24 +0000)]
[TargetLowering] Remove redundant term in two ifs in SimplifySetCC. NFC
If we get into the right hand side of the OR, we know that isOpaque is true since the left hand size is just a check that it's false.
llvm-svn: 327179
Mike Edwards [Fri, 9 Mar 2018 22:13:12 +0000 (22:13 +0000)]
XFAIL: libcpp-no-deduction-guides in libcxx/test/std/strings/basic.string/string.cons/iter_alloc_deduction.fail.cpp
Summary: Refactor the previous version method of marking each apple-clang version as UNSUPPORTED and just XFAIL'ing the libcpp-no-deduction-guides instead. This brings this test inline with the same style as iter_alloc_deduction.pass.cpp
Reviewers: EricWF, dexonsmith
Reviewed By: EricWF
Subscribers: EricWF, vsapsai, vsk, cfe-commits
Differential Revision: https://reviews.llvm.org/D44103
llvm-svn: 327178
Fangrui Song [Fri, 9 Mar 2018 22:11:46 +0000 (22:11 +0000)]
[ELF] Add elf64lppc emulation.
Subscribers: emaste, arichardson, llvm-commits
Differential Revision: https://reviews.llvm.org/D44264
llvm-svn: 327177
Ulrich Weigand [Fri, 9 Mar 2018 22:00:10 +0000 (22:00 +0000)]
Revert "[Debug] Retain both sets of debug intrinsics in HoistThenElseCodeToIf"
This reverts commit r327175 as problems in debug info generation were shown.
llvm-svn: 327176
Ulrich Weigand [Fri, 9 Mar 2018 21:37:07 +0000 (21:37 +0000)]
[Debug] Retain both sets of debug intrinsics in HoistThenElseCodeToIf
When hoisting common code from the "then" and "else" branches of a condition
to before the "if", there is no need to require that debug intrinsics match
before moving them (and merging them). Instead, we can simply always keep
all debug intrinsics from both sides of the "if".
This fixes PR36410, which describes a problem where as a result of the attempt
to merge debug locations for two debug intrinsics we end up with an invalid
intrinsic, where the scope indicated in the !dbg location no longer matches
the scope of the variable tracked by the intrinsic.
In addition, this has the benefit that we no longer throw away information
that is actually still valid, helping to generate better debug data.
Reviewed By: vsk
Differential Revision: https://reviews.llvm.org/D44312
llvm-svn: 327175
Stefan Pintilie [Fri, 9 Mar 2018 21:08:35 +0000 (21:08 +0000)]
[Power9] Code Cleaup and adding Comments for Power 9 Scheduler
Did some code cleanup up removing ItinRW that are not needed and resource types
that are no longer used.
Also added more comments to the td files related to the Power 9 sheduler model.
llvm-svn: 327174
Renato Golin [Fri, 9 Mar 2018 21:05:58 +0000 (21:05 +0000)]
[NFC] Consolidate six getPointerOperand() utility functions into one place
There are six separate instances of getPointerOperand() utility.
LoopVectorize.cpp has one of them,
and I don't want to create a 7th one while I'm trying to move
LoopVectorizationLegality into a separate file
(eventual objective is to move it to Analysis tree).
See http://lists.llvm.org/pipermail/llvm-dev/2018-February/120999.html
for llvm-dev discussions
Closes D43323.
Patch by Hideki Saito <hideki.saito@intel.com>.
llvm-svn: 327173
Nirav Dave [Fri, 9 Mar 2018 20:58:07 +0000 (20:58 +0000)]
Correct load-op-store cycle detection analysis
Add missing cycle dependency checks in load-op-store fusion.
Fixes PR36274.
Reviewers: craig.topper, bogner
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D43154
llvm-svn: 327172
Nirav Dave [Fri, 9 Mar 2018 20:57:42 +0000 (20:57 +0000)]
Improve Dependency analysis when doing multi-node Instruction Selection
Relanding after fixing NodeId Invariant.
Cleanup cycle/validity checks in ISel (IsLegalToFold,
HandleMergeInputChains) and X86 (isFusableLoadOpStore). Now do a full
search for cycles / dependencies pruning the search when topological
property of NodeId allows.
As part of this propogate the NodeId-based cutoffs to narrow
hasPreprocessorHelper searches.
Reviewers: craig.topper, bogner
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D41293
llvm-svn: 327171
Nirav Dave [Fri, 9 Mar 2018 20:57:15 +0000 (20:57 +0000)]
[DAG] Enforce stricter NodeId invariant during Instruction selection
Instruction Selection makes use of the topological ordering of nodes
by node id (a node's operands have smaller node id than it) when doing
cycle detection. During selection we may violate this property as a
selection of multiple nodes may induce a use dependence (and thus a
node id restriction) between two unrelated nodes. If a selected node
has an unselected successor this may allow us to miss a cycle in
detection an invalid selection.
This patch fixes this by marking all unselected successors of a
selected node have negated node id. We avoid pruning on such negative
ids but still can reconstruct the original id for pruning.
In-tree targets have been updated to replace DAG-level replacements
with ISel-level ones which enforce this property.
This preemptively fixes PR36312 before triggering commit r324359 relands
Reviewers: craig.topper, bogner, jyknight
Subscribers: arsenm, nhaehnle, javed.absar, llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D43198
llvm-svn: 327170
Nirav Dave [Fri, 9 Mar 2018 20:56:51 +0000 (20:56 +0000)]
Make early exit hasPredecessorHelper return true. NFCI.
All uses conservatively assume in early exit case that it will be a
predecessor. Changing default removes checking code in all uses.
llvm-svn: 327169
Peter Collingbourne [Fri, 9 Mar 2018 20:41:56 +0000 (20:41 +0000)]
cfi: Disable simple-pass.cpp on Darwin.
-mretpoline does not work yet on Darwin.
llvm-svn: 327168
Kostya Kortchinsky [Fri, 9 Mar 2018 20:17:12 +0000 (20:17 +0000)]
[sanitizer] Revert rCRT327145
Summary:
It breaks the Chromium toolchain due to:
```
lib/sanitizer_common/sanitizer_allocator_primary32.h:269:34: error: requested alignment is not an integer constant
struct ALIGNED(kCacheLineSize) SizeClassInfo {
```
Reviewers: alekseyshl, thakis
Reviewed By: thakis
Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D44320
llvm-svn: 327167
Peter Collingbourne [Fri, 9 Mar 2018 19:37:28 +0000 (19:37 +0000)]
Fix Clang test case.
llvm-svn: 327166
Evgeniy Stepanov [Fri, 9 Mar 2018 19:35:16 +0000 (19:35 +0000)]
Don't use -pie in relocatable link.
Summary:
Android, in particular, got PIE enabled by default in r316606. It resulted in
relocatable links passing both -r and -pie to the linker, which is not allowed.
Reviewers: srhines
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D44229
llvm-svn: 327165
Rafael Auler [Fri, 9 Mar 2018 19:13:44 +0000 (19:13 +0000)]
[llvm-objdump] Support disassembling by symbol name
Summary:
Add a new option -df to llvm-objdump that takes function names
as arguments and instructs the disassembler to only dump those function
contents. Based on code originally written by Bill Nell.
Reviewers: espindola, JDevlieghere
Differential Revision: https://reviews.llvm.org/D44224
llvm-svn: 327164
Peter Collingbourne [Fri, 9 Mar 2018 19:11:44 +0000 (19:11 +0000)]
Use branch funnels for virtual calls when retpoline mitigation is enabled.
The retpoline mitigation for variant 2 of CVE-2017-5715 inhibits the
branch predictor, and as a result it can lead to a measurable loss of
performance. We can reduce the performance impact of retpolined virtual
calls by replacing them with a special construct known as a branch
funnel, which is an instruction sequence that implements virtual calls
to a set of known targets using a binary tree of direct branches. This
allows the processor to speculately execute valid implementations of the
virtual function without allowing for speculative execution of of calls
to arbitrary addresses.
This patch extends the whole-program devirtualization pass to replace
certain virtual calls with calls to branch funnels, which are
represented using a new llvm.icall.jumptable intrinsic. It also extends
the LowerTypeTests pass to recognize the new intrinsic, generate code
for the branch funnels (x86_64 only for now) and lay out virtual tables
as required for each branch funnel.
The implementation supports full LTO as well as ThinLTO, and extends the
ThinLTO summary format used for whole-program devirtualization to
support branch funnels.
For more details see RFC:
http://lists.llvm.org/pipermail/llvm-dev/2018-January/120672.html
Differential Revision: https://reviews.llvm.org/D42453
llvm-svn: 327163
Aaron Smith [Fri, 9 Mar 2018 18:50:19 +0000 (18:50 +0000)]
[SymbolFilePDB] Keep searching until the file name is found for the pdb compiland
Reviewers: zturner, rnk, lldb-commits
Reviewed By: zturner
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44182
llvm-svn: 327162
Alina Sbirlea [Fri, 9 Mar 2018 18:48:20 +0000 (18:48 +0000)]
Avoid creating a Constant for each value in a ConstantDataSequential.
Summary: We create a ConstantDataSequential (ConstantDataArray or ConstantDataVector) to avoid creating a Constant for each element in an array of constants. But them in AsmPrinter, we do create a ConstantFP for each element in the ConstantDataSequential. This triggers excessive memory use when generating large global FP constants.
Reviewers: bogner, lhames, t.p.northover
Subscribers: jlebar, sanjoy, llvm-commits
Differential Revision: https://reviews.llvm.org/D44277
llvm-svn: 327161
Rafael Espindola [Fri, 9 Mar 2018 18:42:25 +0000 (18:42 +0000)]
Delay creating an alias for @@@.
With this we only create an alias for @@@ once we know if it should
use @ or @@. This avoids last minutes renames and hacks to handle MS
names.
This only handles the ELF writer. LTO still has issues with @@@
aliases.
llvm-svn: 327160
Simon Pilgrim [Fri, 9 Mar 2018 18:37:56 +0000 (18:37 +0000)]
[X86][AVX] createVariablePermute - fix v2i64/v2f64 VPERMILPD index creation.
The input indices vector will put the index in bit0, but VPERMILPD actually selects off bit1 - so we need to scale accordingly.
llvm-svn: 327159
Nicolai Haehnle [Fri, 9 Mar 2018 18:32:04 +0000 (18:32 +0000)]
TableGen: Remove space at EOL in TGLexer.{h,cpp}
Change-Id: Ica5f39470174e85f173d3b6db95789033f75ce17
llvm-svn: 327158
Simon Pilgrim [Fri, 9 Mar 2018 18:08:08 +0000 (18:08 +0000)]
[X86][SSE] createVariablePermute - move source vector canonicalization to top of function. NFCI.
This is to make it easier to return early from the switch statement with custom lowering.
llvm-svn: 327157
Fangrui Song [Fri, 9 Mar 2018 18:03:22 +0000 (18:03 +0000)]
[ELF] Convert {read,write}*be to endianness-aware read/write.
Subscribers: emaste, nemanjai, arichardson, kbarton, llvm-commits
Differential Revision: https://reviews.llvm.org/D44227
llvm-svn: 327156
Renato Golin [Fri, 9 Mar 2018 18:02:36 +0000 (18:02 +0000)]
[LV] Adding test for r327109
llvm-svn: 327155
Peter Collingbourne [Fri, 9 Mar 2018 17:54:43 +0000 (17:54 +0000)]
ELF: Do not create multiple thunks for the same virtual address.
This avoids creating multiple thunks for symbols with aliases or which
belong to ICF'd sections. This patch reduces the size of Chromium for
Android by 260KB (0.8% of .text).
Differential Revision: https://reviews.llvm.org/D44284
llvm-svn: 327154
Farhana Aleen [Fri, 9 Mar 2018 17:41:39 +0000 (17:41 +0000)]
[AMDGPU] Supported ds_read_b128 generation; Widened vector length for local address-space.
Summary: Starting from GCN 2nd generation, ISA supports ds_read_b128 on top of ds_read_b64.
This patch supports ds_read_b128 instruction pattern and generation of this instruction.
In the vectorizer, this patch also widen the vector length so that vectorizer generates
128 bit loads for local address-space which gets translated to ds_read_b128.
Since the performance benefit is not clear; compiler generates ds_read_b128 under -amdgpu-ds128.
Author: FarhanaAleen
Reviewed By: rampitec, arsenm
Subscribers: llvm-commits, AMDGPU
Differential Revision: https://reviews.llvm.org/D44210
llvm-svn: 327153
Aditya Nandakumar [Fri, 9 Mar 2018 17:31:51 +0000 (17:31 +0000)]
[GISel]: Add helpers for easy building G_FCONSTANT along with matchers
Added helpers to build G_FCONSTANT, along with matching ConstantFP and
unit tests for the same.
Sample usage.
auto MIB = Builder.buildFConstant(s32, 0.5); // Build IEEESingle
For Matching the above
const ConstantFP* Tmp;
mi_match(DstReg, MRI, m_GFCst(Tmp));
https://reviews.llvm.org/D44128
reviewed by: volkan
llvm-svn: 327152
Nicholas Wilson [Fri, 9 Mar 2018 17:06:38 +0000 (17:06 +0000)]
[WebAssembly] Handle weak undefined functions with a synthetic stub
This error case is described in Linking.md. The operand for call requires
generation of a synthetic stub.
Differential Revision: https://reviews.llvm.org/D44028
llvm-svn: 327151
Chad Rosier [Fri, 9 Mar 2018 16:43:46 +0000 (16:43 +0000)]
[JumpThreading] Don't restrict cast-traversal to i1
In r263618, JumpThreading learned to look trough simple cast instructions, but
only if the source of those cast instructions was a phi/cmp i1 (in an effort to
limit compile time effects). I think this condition is too restrictive. For
switches with limited value range, InstCombine will readily introduce an extra
trunc instruction to a smaller integer type (e.g. from i8 to i2), leaving us in
the somewhat perverse situation that jump-threading would work before running
instcombine, but not after. Since instcombine produces this pattern, I think we
need to consider it canonical and support it in JumpThreading. In general,
for limiting recursion, I think the existing restriction to phi and cmp nodes
should be sufficient to avoid looking through unprofitable chains of
instructions.
Patch by Keno Fischer!
Differential Revision: https://reviews.llvm.org/D42262
llvm-svn: 327150
Nicholas Wilson [Fri, 9 Mar 2018 16:43:05 +0000 (16:43 +0000)]
[WebAssembly] Refactor order of creation for SyntheticFunction
Previously we created __wasm_call_ctors with null InputFunction, and
added the InputFunction later. Now we create the SyntheticFunction with
null body, and set the body later.
Differential Revision: https://reviews.llvm.org/D44206
llvm-svn: 327149
Adrian Prantl [Fri, 9 Mar 2018 16:42:05 +0000 (16:42 +0000)]
Move generic test to the Generic directory
llvm-svn: 327148
Sanjay Patel [Fri, 9 Mar 2018 16:33:34 +0000 (16:33 +0000)]
[AMDGPU] fix test to be independent of FP undef
llvm-svn: 327147
Nicholas Wilson [Fri, 9 Mar 2018 16:30:44 +0000 (16:30 +0000)]
[WebAssembly] Disallow weak undefined globals in the object format
This implements https://github.com/WebAssembly/tool-conventions/pull/47
Differential Revision: https://reviews.llvm.org/D44201
llvm-svn: 327146
Kostya Kortchinsky [Fri, 9 Mar 2018 16:18:38 +0000 (16:18 +0000)]
[sanitizer] Align & pad the allocator structures to the cacheline size
Summary:
Both `SizeClassInfo` structures for the 32-bit primary & `RegionInfo`
structures for the 64-bit primary can be used by different threads, and as such
they should be aligned & padded to the cacheline size to avoid false sharing.
The former was padded but the array was not aligned, the latter was not padded
but we lucked up as the size of the structure was 192 bytes, and aligned by
the properties of `mmap`.
I plan on adding a couple of fields to the `RegionInfo`, and some highly
threaded tests pointed out that without proper padding & alignment, performance
was getting a hit - and it is going away with proper padding.
This patch makes sure that we are properly padded & aligned for both. I used
a template to avoid padding if the size is already a multiple of the cacheline
size. There might be a better way to do this, I am open to suggestions.
Reviewers: alekseyshl, dvyukov
Reviewed By: alekseyshl
Subscribers: kubamracek, delcypher, #sanitizers, llvm-commits
Differential Revision: https://reviews.llvm.org/D44261
llvm-svn: 327145
Sanjay Patel [Fri, 9 Mar 2018 16:14:02 +0000 (16:14 +0000)]
[InstSimplify] fix FP infinite hex constant values in tests; NFC
Really should improve this...
llvm-svn: 327144
Stefan Pintilie [Fri, 9 Mar 2018 16:08:55 +0000 (16:08 +0000)]
Revert "[PowerPC] LSR tunings for PowerPC"
Revert the rest of the LST tune commit.
It seems that the LSR tune commit breaks internal tests.
Reverting the commit.
llvm-svn: 327143
Stefan Pintilie [Fri, 9 Mar 2018 16:08:48 +0000 (16:08 +0000)]
Revert "[PowerPC] Move test to correct location."
Revert part of the LSR tune commit.
llvm-svn: 327142
Simon Pilgrim [Fri, 9 Mar 2018 15:50:09 +0000 (15:50 +0000)]
Tidyup comment that was destroyed by clang-format. NFCI.
llvm-svn: 327141
Simon Pilgrim [Fri, 9 Mar 2018 15:48:56 +0000 (15:48 +0000)]
[X86][SSE] createVariablePermute - move index vector canonicalization to top of function. NFCI.
This is to make it easier to return early from the switch statement with custom lowering.
llvm-svn: 327140
Tim Northover [Fri, 9 Mar 2018 15:44:59 +0000 (15:44 +0000)]
Try to fix Windows bot by forcing "rm".
llvm-svn: 327139
Sanjay Patel [Fri, 9 Mar 2018 15:27:48 +0000 (15:27 +0000)]
[LangRef] make it clear that FP instructions do not have side effects
Also, fix the undef vs. UB example to use 'sdiv' because that can trigger div-by-zero UB.
The existing text for the constrained intrinsics says:
"By default, LLVM optimization passes assume that the rounding mode is round-to-nearest
and that floating point exceptions will not be monitored. Constrained FP intrinsics are
used to support non-default rounding modes and accurately preserve exception behavior
without compromising LLVM’s ability to optimize FP code when the default behavior is
used."
...so the additional text with the normal FP opcodes should make the different modes
clear.
Differential Revision: https://reviews.llvm.org/D44216
llvm-svn: 327138
Jonas Devlieghere [Fri, 9 Mar 2018 15:22:42 +0000 (15:22 +0000)]
[dsymutil] Unify error handling and add color
We improved the handling of errors and warnings in dwarfdump's verifier
in rL314498. This patch does the same thing for dsymutil.
Differential revision: https://reviews.llvm.org/D44052
llvm-svn: 327137
Alexey Bataev [Fri, 9 Mar 2018 15:20:30 +0000 (15:20 +0000)]
[OPENMP] Fix the address of the original variable in task reductions.
If initialization of the task reductions requires pointer to original
variable, which is stored in the threadprivate storage, we used the
address of this pointer instead.
llvm-svn: 327136
Hans Wennborg [Fri, 9 Mar 2018 14:46:44 +0000 (14:46 +0000)]
CMake: Make libxml2 show up in --system-libs (PR36660)
lib/WindowsManifest/CMakeLists.txt adds it to LLVM_SYSTEM_LIBS on that
target, but it was never getting picked up in
tools/llvm-config/CMakeLists.txt.
Differential Revision: https://reviews.llvm.org/D44302
llvm-svn: 327135
Ilya Biryukov [Fri, 9 Mar 2018 14:43:29 +0000 (14:43 +0000)]
[SemaOverload] Fixed crash on code completion
Summary:
The relevant failing assertion message is:
../tools/clang/lib/Sema/SemaInit.cpp:8411: PerformCopyInitialization(): Assertion `InitE && "No initialization expression?"' failed.
See the added test case for a repro.
Reviewers: bkramer, sammccall, ioeric, hokein
Reviewed By: sammccall
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D44300
llvm-svn: 327134
Pavel Labath [Fri, 9 Mar 2018 14:32:16 +0000 (14:32 +0000)]
Make TestCompletion work on windows
The test I added in r327110 is failing on windows because of "import
pexpect". However, this import is no longer necessary as these tests
don't use pexpect anymore.
In fact, it seems that all TestCompletion tests are passing on windows
after this, so I enable all of them.
llvm-svn: 327133
Sebastian Pop [Fri, 9 Mar 2018 14:29:21 +0000 (14:29 +0000)]
[x86][aarch64] ask the backend whether it has a vector blend instruction
The code to match and produce more x86 vector blends was enabled for all
architectures even though the transform may pessimize the code for other
architectures that do not provide a vector blend instruction.
Added an aarch64 testcase to check that a VZIP instruction is generated instead
of byte movs.
Differential Revision: https://reviews.llvm.org/D44118
llvm-svn: 327132
Haojian Wu [Fri, 9 Mar 2018 14:16:46 +0000 (14:16 +0000)]
[clangd] Fix failing lit test.
This test is missed in r327129.
llvm-svn: 327131
Eric Liu [Fri, 9 Mar 2018 14:06:43 +0000 (14:06 +0000)]
[clangd-vscode] Add package-lock.json to .gitignore
llvm-svn: 327130
Haojian Wu [Fri, 9 Mar 2018 14:00:34 +0000 (14:00 +0000)]
[clangd] Use identifier range as the definition range.
Summary: This also matches the range in symbol index.
Reviewers: sammccall
Subscribers: klimek, ilya-biryukov, jkorous-apple, ioeric, cfe-commits
Differential Revision: https://reviews.llvm.org/D44247
llvm-svn: 327129
Andrea Di Biagio [Fri, 9 Mar 2018 13:52:03 +0000 (13:52 +0000)]
[llvm-mca] Move the logic that prints the summary into its own view. NFCI
llvm-svn: 327128
Sam McCall [Fri, 9 Mar 2018 13:25:29 +0000 (13:25 +0000)]
[clangd] Don't index template specializations.
Summary:
These have different USRs than the underlying entity, but are not typically
interesting in their own right and can be numerous (e.g. generated traits).
Reviewers: ioeric
Subscribers: klimek, ilya-biryukov, jkorous-apple, cfe-commits
Differential Revision: https://reviews.llvm.org/D44298
llvm-svn: 327127
David Zarzycki [Fri, 9 Mar 2018 13:09:36 +0000 (13:09 +0000)]
[CMake] Properly conditionalize installation of lld libraries
In what appears to be a copy-and-paste error, lld currently only
installs libraries if the lld tools are configured to build.
Instead, lld should allow the libraries to be installed even if the lld
tools are not being built. Additionally, if users want to only install
the tools and not the libraries, the LLVM way of doing that is by
checking for LLVM_INSTALL_TOOLCHAIN_ONLY.
This fixes PR35960.
llvm-svn: 327126
Andrea Di Biagio [Fri, 9 Mar 2018 12:50:42 +0000 (12:50 +0000)]
[llvm-mca] Run clang-format on the source code. NFC
llvm-svn: 327125
Nico Weber [Fri, 9 Mar 2018 12:41:04 +0000 (12:41 +0000)]
[lld-link] Add support for /ignore:4037.
Fixes PR36657.
https://reviews.llvm.org/D44286
llvm-svn: 327124
Pavel Labath [Fri, 9 Mar 2018 12:30:09 +0000 (12:30 +0000)]
[elf] Remove one copy of the section merging code
Summary:
Besides being superfluous, this double merging was actually wrong and
causing some sections to be added twice. The reason for that was that
the code assumes section IDs are unique in the section list, but this is
only true if all sections in the list come from the same object file.
Reviewers: fjricci, jankratochvil
Subscribers: emaste, lldb-commits, arichardson
Differential Revision: https://reviews.llvm.org/D44157
llvm-svn: 327123
Dan Liew [Fri, 9 Mar 2018 12:26:29 +0000 (12:26 +0000)]
[asan] Try to unbreak arm build bots by only running the test on
x86_64 and x86_64H (and not in the iOS simulator) where it seems
to work. We need to properly fix the test when time permits.
llvm-svn: 327122
Nicolai Haehnle [Fri, 9 Mar 2018 12:24:42 +0000 (12:24 +0000)]
TableGen: Add a defset statement
Allows capturing a list of concrete instantiated defs.
This can be combined with foreach to create parallel sets of def
instantiations with less repetition in the source. This purpose is
largely also served by multiclasses, but in some cases multiclasses
can't be used.
The motivating example for this change is having a large set of
intrinsics, which are generated from the IntrinsicsBackend.td file
included by Intrinsics.td, and a corresponding set of instruction
selection patterns, which are generated via the backend's .td files.
Multiclasses cannot be used to eliminate the redundancy in this case,
because a multiclass cannot span both LLVM's common .td files and
the backend .td files at the same time.
Change-Id: I879e35042dceea542a5e6776fad23c5e0e69e76b
Differential revision: https://reviews.llvm.org/D44109
llvm-svn: 327121
Nicolai Haehnle [Fri, 9 Mar 2018 12:24:30 +0000 (12:24 +0000)]
TableGen: Allow arbitrary list values as ranges of foreach
The changes to FieldInit are required to make field references (Def.field)
work inside a ForeachDeclaration: previously, Def.field wasn't resolved
immediately when Def was already a fully resolved DefInit.
Change-Id: I9875baec2fc5aac8c2b249e45b9cf18c65ae699b
llvm-svn: 327120
Nicolai Haehnle [Fri, 9 Mar 2018 12:24:20 +0000 (12:24 +0000)]
TableGen: Remove unused ParseForeachMode
Use the default ParseValueMode instead of ParseForeachMode when
parsing the rule
ForeachDeclaration ::= ID '=' '[' ValueList ']'
because the only difference between the two is how an open brace '{'
is handled at the end. In the context of foreach, the 'in' keyword
will appear after the ForeachDeclaration, so this special handling
of '{' is not required.
Change-Id: I4d86bb73bab9ec26752e1273e5213df77cf28d1d
llvm-svn: 327119
Nicolai Haehnle [Fri, 9 Mar 2018 12:24:14 +0000 (12:24 +0000)]
TableGen: More helpful error messages
Change-Id: Ic78afd0cd765fdb4cf1b7ecfb6bba22653ce6d29
llvm-svn: 327118
Nicolai Haehnle [Fri, 9 Mar 2018 12:24:06 +0000 (12:24 +0000)]
TableGen: add !isa operation
Change-Id: Iddb724c3ae706d82933a2d82c91d07e0e36b30e3
Differential revision: https://reviews.llvm.org/D44105
llvm-svn: 327117
Pavel Labath [Fri, 9 Mar 2018 11:58:59 +0000 (11:58 +0000)]
[DebugInfo/AccelTable] Fix inconsistency in getDIEOffset implementations
Summary:
Even though the getDIEOffset offset function was common for the two
accelerator table implementations, it was doing two different things:
for the Apple tables, it was returning the die offset relative to the
start of the section, whereas for DWARF v5 tables, it was relative to
the start of the CU.
I resolve this by renaming the function to getDIESectionOffset to make
it obvious what the function returns, and change the DWARF
implementation to return the section offset. I also keep the CU-relative
accessor, but only in the DWARF implementation (there is no way to get
this information for the Apple tables). This was not caught by existing
tests because the hand-written inputs also erroneously used section
offsets instead of CU-relative ones.
While looking at this, I noticed that the Apple implementation was not
fully correct either -- the header contains a DIEOffsetBase field, which
should be added to offsets encoded with the DW_FORM_ref*** family, but
this was not being used. This went unnoticed because all current writers
set this field to zero anyway. I fix this as well and add a hand-written
test which demonstrates the issue.
Reviewers: JDevlieghere, dblaikie
Subscribers: aprantl, llvm-commits
Differential Revision: https://reviews.llvm.org/D44202
llvm-svn: 327116
Sam McCall [Fri, 9 Mar 2018 11:47:37 +0000 (11:47 +0000)]
[clang-tidy] fix header guard
llvm-svn: 327115
Haojian Wu [Fri, 9 Mar 2018 10:47:14 +0000 (10:47 +0000)]
[clang-tidy] Add check: replace string::find(...) == 0 with absl::StartsWith
Patch by Niko Weh!
Reviewers: hokein
Subscribers: klimek, cfe-commits, ioeric, ilya-biryukov, ahedberg
Differential Revision: https://reviews.llvm.org/D43847
llvm-svn: 327111
Pavel Labath [Fri, 9 Mar 2018 10:39:40 +0000 (10:39 +0000)]
Move option parsing out of the Args class
Summary:
The args class is used in plenty of places (a lot of them in the lower lldb
layers) for representing a list of arguments, and most of these places don't
care about option parsing. Moving the option parsing out of the class removes
the largest external dependency (there are a couple more, but these are in
static functions), and brings us closer to being able to move it to the
Utility module).
The new home for these functions is the Options class, which was already used
as an argument to the parse calls, so this just inverts the dependency between
the two.
The functions are themselves are mainly just copied -- the biggest functional
change I've made to them is to avoid modifying the input Args argument (getopt
likes to permute the argument vector), as it was weird to have another class
reorder the entries in Args class. So now the functions don't modify the input
arguments, and (for those where it makes sense) return a new Args vector
instead. I've also made the addition of a "fake arg0" (required for getopt
compatibility) an implementation detail rather than a part of interface.
While doing that I noticed that ParseForCompletion function was recording the
option indexes in the shuffled vector, but then the consumer was looking up the
entries in the unshuffled one. This manifested itself as us not being able to
complete "watchpoint set variable foo --" (because getopt would move "foo" to
the end). Surprisingly all other completions (e.g. "watchpoint set variable foo
--w") were not affected by this. However, I couldn't find a comprehensive test
for command argument completion, so I consolidated the existing tests and added
a bunch of new ones.
Reviewers: davide, jingham, zturner
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D43837
llvm-svn: 327110
Renato Golin [Fri, 9 Mar 2018 10:31:31 +0000 (10:31 +0000)]
[LV] Fix vectorizer's isUniform() abuse triggers assert in SCEV
Fixes PR36311.
See more detailed analysis in
https://bugs.llvm.org/show_bug.cgi?id=36311.
isUniform() information is recomputed after LV started transforming the
underlying IR and that triggered an assert in SCEV.
From vectorizer's architectural perspective, such information, while
still useful in vector code gen, should not be recomputed after the
start of transforming the LLVM IR. Instead, we should collect and cache
such information during the analysis phase of LV and use the cached info
during code gen.
From the symptom perspective, this assert as it stands right now is not
very useful. Legality already rejected loops that would trigger the
assert. As such, commenting out the assert is NFC from vectorizer's
functionality perspective. On top of that, just above the assertion, we
check for unit-strided load/store or
gather scatter. Addresses can't be uniform below that check.
From vectorization theory point of view, we don't have to reject all
cases of stores to uniform addresses. Eventually, we should support
safe/profitable cases.
This patch resolves the issue by removing the useless assertion that is
invoking LAA's isUniform() that requires up-to-date DomTree ---- once
vector code gen starts modifying CFG, we don't have an up-to-date
DomTree.
Patch by Hideki Saito <hideki.saito@intel.com>.
llvm-svn: 327109