Leonard Chan [Mon, 15 Oct 2018 19:59:52 +0000 (19:59 +0000)]
added fix
llvm-svn: 344548
Stella Stamenova [Mon, 15 Oct 2018 19:51:21 +0000 (19:51 +0000)]
[lldbsuite] Make the names of test classes unique
Summary:
If the names are not unique, the tests overwrite each other's results and logs. This also causes failures on platforms where the files are locked for writing.
The names of the class/test pairs *have to* always be unique. The easiest way to achieve that is to name each class differently (usually the same as the file name).
Reviewers: jasonmolenda, asmith
Subscribers: clayborg, nemanjai, kbarton, lldb-commits
Differential Revision: https://reviews.llvm.org/D53297
llvm-svn: 344547
Marshall Clow [Mon, 15 Oct 2018 19:46:03 +0000 (19:46 +0000)]
Mark a couple of test cases as 'C++17-only' pending the resolution of PR#39232
llvm-svn: 344546
Vedant Kumar [Mon, 15 Oct 2018 19:22:20 +0000 (19:22 +0000)]
[CodeExtractor] Erase debug intrinsics in outlined thunks (fix PR22900)
Variable updates within the outlined function are invisible to
debuggers. This could be improved by defining a DISubprogram for the
new function. For the moment, simply erase the debug intrinsics instead.
This fixes verifier failures about function-local metadata being used in
the wrong function, seen while testing the hot/cold splitting pass.
rdar://
45142482
Differential Revision: https://reviews.llvm.org/D53267
llvm-svn: 344545
Sean Fertile [Mon, 15 Oct 2018 19:05:57 +0000 (19:05 +0000)]
[PPC64] Add split - stack support.
This support is slightly different then the X86_64 implementation in that calls
to __morestack don't need to get rewritten to calls to __moresatck_non_split
when a split-stack caller calls a non-split-stack callee. Instead the size of
the stack frame requested by the caller is adjusted prior to the call to
__morestack. The size the stack-frame will be adjusted by is tune-able through a
new --split-stack-adjust-size option.
Differential Revision: https://reviews.llvm.org/D52099
llvm-svn: 344544
Stella Stamenova [Mon, 15 Oct 2018 18:51:28 +0000 (18:51 +0000)]
[lldbsuite] Disable Test128BitsInteger on Windows
Summary: This test is failing on Windows because lldb does not support JIT on Windows.
Reviewers: davide, asmith
Reviewed By: davide
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D53226
llvm-svn: 344543
Lang Hames [Mon, 15 Oct 2018 18:34:36 +0000 (18:34 +0000)]
[ADT] Fix a bug in DenseSet's initializer_list constructor.
Without this fix, DenseSet crashes with an assertion if constructed with an
initializer_list whose length is not a power of two.
llvm-svn: 344542
Sanjay Patel [Mon, 15 Oct 2018 18:05:34 +0000 (18:05 +0000)]
[SelectionDAG] allow FP binops in SimplifyDemandedVectorElts
This is intended to make the backend on par with functionality that was
added to the IR version of SimplifyDemandedVectorElts in:
rL343727
...and the original motivation is that we need to improve demanded-vector-elements
in several ways to avoid problems that would be exposed in D51553.
Differential Revision: https://reviews.llvm.org/D52912
llvm-svn: 344541
Artem Dergachev [Mon, 15 Oct 2018 18:01:34 +0000 (18:01 +0000)]
[analyzer] Add doxygen comments for the new CXXAllocatorCall APIs.
Forgot to squeeze this into r344539.
llvm-svn: 344540
Artem Dergachev [Mon, 15 Oct 2018 17:53:18 +0000 (17:53 +0000)]
[analyzer] Teach CallEvent about C++17 aligned operator new().
In C++17, when class C has large alignment value, a special case of
overload resolution rule kicks in for expression new C that causes the aligned
version of operator new() to be called. The aligned new has two arguments:
size and alignment. However, the new-expression has only one "argument":
the construct-expression for C(). This causes a false positive in
core.CallAndMessage's check for matching number of arguments and number
of parameters.
Update CXXAllocatorCall, which is a CallEvent sub-class for operator new calls
within new-expressions, so that the number of arguments always matched
the number of parameters.
rdar://problem/
44738501
Differential Revision: https://reviews.llvm.org/D52957
llvm-svn: 344539
Artem Dergachev [Mon, 15 Oct 2018 17:47:56 +0000 (17:47 +0000)]
[analyzer] NFC: RetainCountChecker: Don't dump() symbols into program point tags.
We don't need a separate node for every symbol, because whenever the first
symbol leaks, a bug is emitted, the analysis is sinked, and the checker
callback immediately returns due to State variable turning into null,
so we never get to see the second leaking symbol.
Additionally, we are no longer able to break normal analysis while experimenting
with debug dumps.
Differential Revision: https://reviews.llvm.org/D52804
llvm-svn: 344538
Artem Dergachev [Mon, 15 Oct 2018 17:43:23 +0000 (17:43 +0000)]
[python] [tests] Disable python binding tests under LLVM_USE_SANITIZER=Address
They don't work yet.
Patch by Dan Liew!
rdar://problem/
45242886
Differential Revision: https://reviews.llvm.org/D53239
llvm-svn: 344537
Nick Desaulniers [Mon, 15 Oct 2018 17:39:00 +0000 (17:39 +0000)]
Revert 344389 "Revert r344375 "[Driver] check for exit code from SIGPIPE""
Summary:
Add preprocessor guards for UNIX.
This reverts commit r344389.
Reviewers: rnk, majnemer, jfb
Reviewed By: rnk
Subscribers: cfe-commits, pirama, srhines
Differential Revision: https://reviews.llvm.org/D53210
llvm-svn: 344536
Marshall Clow [Mon, 15 Oct 2018 17:33:20 +0000 (17:33 +0000)]
Wrap up the new chrono literals in an #ifdef so that old versions of clang don't complain. I'm looking at you, clang 5.0.1
llvm-svn: 344535
Sanjay Patel [Mon, 15 Oct 2018 16:54:07 +0000 (16:54 +0000)]
[DAGCombiner] allow undef elts in vector fmul matching
llvm-svn: 344534
Sam McCall [Mon, 15 Oct 2018 16:47:45 +0000 (16:47 +0000)]
[clangd] Revert include path change in Dexp. NFC
llvm-svn: 344533
Sanjay Patel [Mon, 15 Oct 2018 16:47:01 +0000 (16:47 +0000)]
[DAGCombiner] refactor folds for fadd (fmul X, -2.0), Y; NFCI
The transform doesn't work if the vector constant has undef elements.
llvm-svn: 344532
Sanjay Patel [Mon, 15 Oct 2018 16:44:00 +0000 (16:44 +0000)]
[AArch64] add tests for fmul x, -2.0 with undef elts; NFC
Also, add tests with commuted operands. There was no coverage for that case.
llvm-svn: 344531
Leonard Chan [Mon, 15 Oct 2018 16:07:02 +0000 (16:07 +0000)]
[Fixed Point Arithmetic] FixedPointCast
This patch is a part of https://reviews.llvm.org/D48456 in an attempt to
split them up. This contains the code for casting between fixed point types
and other fixed point types.
The method for converting between fixed point types is based off the convert()
method in APFixedPoint.
Differential Revision: https://reviews.llvm.org/D50616
llvm-svn: 344530
Marshall Clow [Mon, 15 Oct 2018 16:06:37 +0000 (16:06 +0000)]
Implement the first part of the calendar support for C++20. This is still incomplete; there will be more patches coming. Reviewed as D51762
llvm-svn: 344529
Sanjay Patel [Mon, 15 Oct 2018 15:56:39 +0000 (15:56 +0000)]
[DAGCombiner] allow undef elts in vector fma matching
llvm-svn: 344528
Sanjay Patel [Mon, 15 Oct 2018 15:47:37 +0000 (15:47 +0000)]
[x86] add tests for fma with undef elts; NFC
llvm-svn: 344527
Sean Fertile [Mon, 15 Oct 2018 15:43:00 +0000 (15:43 +0000)]
Revert "[CodeGenCXX] Treat 'this' as noalias in constructors"
This reverts commit https://reviews.llvm.org/rL344150 which causes
MachineOutliner related failures on the ppc64le multistage buildbot.
llvm-svn: 344526
Sanjay Patel [Mon, 15 Oct 2018 15:38:38 +0000 (15:38 +0000)]
[DAGCombiner] allow undef elts in vector fma matching
llvm-svn: 344525
Fedor Sergeev [Mon, 15 Oct 2018 15:36:08 +0000 (15:36 +0000)]
Revert "[NewPM] teach -passes= to emit meaningful error messages"
This reverts r344519 due to failures in pipeline-parsing test.
llvm-svn: 344524
Sanjay Patel [Mon, 15 Oct 2018 15:28:44 +0000 (15:28 +0000)]
[x86] add tests for fma with undef elts; NFC
llvm-svn: 344523
Lang Hames [Mon, 15 Oct 2018 15:26:47 +0000 (15:26 +0000)]
[ADT] Adds equality operators for DenseMap and DenseSet, and an initializer_list
constructor for DenseMap (DenseSet already had an initializer_list constructor).
These changes make it easier to migrate existing code that uses std::map and
std::set (which support initializer_list construction and equality comparison)
to DenseMap and DenseSet.
llvm-svn: 344522
Haojian Wu [Mon, 15 Oct 2018 15:12:40 +0000 (15:12 +0000)]
[clangd] Add createIndex in dexp
Summary:
This would allow easily injecting our internal customization.
Also updates the stale "symbol-collection-file" flag.
Reviewers: sammccall
Subscribers: ilya-biryukov, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53292
llvm-svn: 344521
Sam McCall [Mon, 15 Oct 2018 15:04:03 +0000 (15:04 +0000)]
[clangd] Use SyncAPI in more places in tests. NFC
llvm-svn: 344520
Fedor Sergeev [Mon, 15 Oct 2018 15:00:18 +0000 (15:00 +0000)]
[NewPM] teach -passes= to emit meaningful error messages
Summary:
All the PassBuilder::parse interfaces now return descriptive StringError
instead of a plain bool. It allows to make -passes/aa-pipeline parsing
errors context-specific and thus less confusing.
TODO: ideally we should also make suggestions for misspelled pass names,
but that requires some extensions to PassBuilder.
Reviewed By: philip.pfaffe, chandlerc
Differential Revision: https://reviews.llvm.org/D53246
llvm-svn: 344519
Sid Manning [Mon, 15 Oct 2018 14:44:52 +0000 (14:44 +0000)]
[ELF][HEXAGON] Let input determine e_flag.
Differential Revision: https://reviews.llvm.org/D53204
llvm-svn: 344518
Aleksandar Beserminji [Mon, 15 Oct 2018 14:39:12 +0000 (14:39 +0000)]
[mips][micromips] Fix overlaping FDEs error
When compiling static executable for micromips, CFI symbols
are incorrectly labeled as MICROMIPS, which cause
".eh_frame_hdr refers to overlapping FDEs." error.
This patch does not label CFI symbols as MICROMIPS, and FDEs do not
overlap anymore. This patch also exposes another bug, which is fixed
here: https://reviews.llvm.org/D52985
Differential Revision: https://reviews.llvm.org/D52987
llvm-svn: 344516
Aleksandar Beserminji [Mon, 15 Oct 2018 14:36:48 +0000 (14:36 +0000)]
[mips][micromips] Revert "Fix overlaping FDEs error"
This reverts r344511.
llvm-svn: 344515
George Rimar [Mon, 15 Oct 2018 14:21:43 +0000 (14:21 +0000)]
[LLD][ELF] - Check options before processing the -v/-version options.
This is https://bugs.llvm.org/show_bug.cgi?id=39289.
Currently both gold and bfd report errors about invalid options values
even with -v/-versions. But LLD does not.
This makes complicated to check the options available when LLD is used.
Patch makes LLD behavior to be consistent with GNU linkers.
Differential revision: https://reviews.llvm.org/D53278
llvm-svn: 344514
Sam McCall [Mon, 15 Oct 2018 13:34:10 +0000 (13:34 +0000)]
[clangd] Minimal implementation of automatic static index (not enabled).
Summary:
See tinyurl.com/clangd-automatic-index for design and goals.
Lots of limitations to keep this patch smallish, TODOs everywhere:
- no serialization to disk
- no changes to dynamic index, which now has a much simpler job
- no partitioning of symbols by file to avoid duplication of header symbols
- no reindexing of edited files
- only a single worker thread
- compilation database is slurped synchronously (doesn't scale)
- uses memindex, rebuilds after every file (should be dex, periodically)
It's not hooked up to ClangdServer/ClangdLSPServer yet: the layering
isn't clear (it should really be in ClangdServer, but ClangdLSPServer
has all the CDB interactions).
Reviewers: ioeric
Subscribers: mgorny, ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, jfb, cfe-commits
Differential Revision: https://reviews.llvm.org/D53032
llvm-svn: 344513
Simon Pilgrim [Mon, 15 Oct 2018 13:20:41 +0000 (13:20 +0000)]
[ARM][NEON] Improve vector popcnt lowering with PADDL (PR39281)
As I suggested on PR39281, this patch uses PADDL pairwise addition to widen from the vXi8 CTPOP result to the target vector type.
This is a blocker for moving more x86 code to generic vector CTPOP expansion (P32655 + D53258) - ARM's vXi64 CTPOP currently expands, which would generate a vXi64 MUL but ARM's custom lowering expands the general MUL case and vectors aren't well handled in LegalizeDAG - improving the CTPOP lowering was a lot easier than fixing the MUL lowering for this one case......
Differential Revision: https://reviews.llvm.org/D53257
llvm-svn: 344512
Aleksandar Beserminji [Mon, 15 Oct 2018 12:59:17 +0000 (12:59 +0000)]
[mips][micromips] Fix overlaping FDEs error
When compiling static executable for micromips, CFI symbols
are incorrectly labeled as MICROMIPS, which cause
".eh_frame_hdr refers to overlapping FDEs." error.
This patch does not label CFI symbols as MICROMIPS, and FDEs do not
overlap anymore. This patch also exposes another bug, which is fixed
here: https://reviews.llvm.org/D52985
Differential Revision: https://reviews.llvm.org/D52987
llvm-svn: 344511
Haojian Wu [Mon, 15 Oct 2018 12:39:45 +0000 (12:39 +0000)]
[clangd] Remove an unused include header, NFC.
llvm-svn: 344510
Eric Liu [Mon, 15 Oct 2018 12:37:23 +0000 (12:37 +0000)]
[CodeComplete] Make sure keyword 'template' is added even when code pattern is disabled.
Reviewers: sammccall, hokein
Subscribers: arphaman, cfe-commits
Differential Revision: https://reviews.llvm.org/D53284
llvm-svn: 344509
Haojian Wu [Mon, 15 Oct 2018 12:32:49 +0000 (12:32 +0000)]
[clangd] dump xrefs information in dexp tool.
Reviewers: sammccall
Subscribers: ilya-biryukov, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53019
llvm-svn: 344508
Haojian Wu [Mon, 15 Oct 2018 11:46:26 +0000 (11:46 +0000)]
[clangd] Fix some references missing in dynamic index.
Summary:
Previously, SymbolCollector postfilters all references at the end to
find all references of interesting symbols.
It was incorrect when indxing main AST where we don't see locations
of symbol declarations and definitions in the main AST (as those are in
preamble AST).
The fix is to do earily check during collecting references.
Reviewers: sammccall
Subscribers: ilya-biryukov, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53273
llvm-svn: 344507
Nicolai Haehnle [Mon, 15 Oct 2018 11:37:04 +0000 (11:37 +0000)]
AMDGPU: Test showing a scalar buffer load deficiency
Change-Id: I5b64a565f22a8482aa0712488d85e45163ac3d12
llvm-svn: 344506
Fedor Sergeev [Mon, 15 Oct 2018 10:46:35 +0000 (10:46 +0000)]
[NewPM] implement SCC printing for -print-before-all/-print-after-all
Removing deficiency of initial implementation of -print-before-all/-after-all
- it was effectively skipping IR printing for all the SCC passes.
Now LazyCallGraph:SCC gets its IR printed.
Reviewed By: skatkov
Differential Revision: https://reviews.llvm.org/D53270
llvm-svn: 344505
Chandler Carruth [Mon, 15 Oct 2018 10:42:50 +0000 (10:42 +0000)]
[TI removal] Make `getTerminator()` return a generic `Instruction`.
This removes the primary remaining API producing `TerminatorInst` which
will reduce the rate at which code is introduced trying to use it and
generally make it much easier to remove the remaining APIs across the
codebase.
Also clean up some of the stragglers that the previous mechanical update
of variables missed.
Users of LLVM and out-of-tree code generally will need to update any
explicit variable types to handle this. Replacing `TerminatorInst` with
`Instruction` (or `auto`) almost always works. Most of these edits were
made in prior commits using the perl one-liner:
```
perl -i -ple 's/TerminatorInst(\b.* = .*getTerminator\(\))/Instruction\1/g'
```
This also my break some rare use cases where people overload for both
`Instruction` and `TerminatorInst`, but these should be easily fixed by
removing the `TerminatorInst` overload.
llvm-svn: 344504
Chandler Carruth [Mon, 15 Oct 2018 10:10:54 +0000 (10:10 +0000)]
[TI removal] Rework `InstVisitor` to support visiting instructions that
are terminators without relying on the specific `TerminatorInst` type.
This required cleaning up two users of `InstVisitor`s usage of
`TerminatorInst` as well.
llvm-svn: 344503
Chandler Carruth [Mon, 15 Oct 2018 10:04:59 +0000 (10:04 +0000)]
[TI removal] Make variables declared as `TerminatorInst` and initialized
by `getTerminator()` calls instead be declared as `Instruction`.
This is the biggest remaining chunk of the usage of `getTerminator()`
that insists on the narrow type and so is an easy batch of updates.
Several files saw more extensive updates where this would cascade to
requiring API updates within the file to use `Instruction` instead of
`TerminatorInst`. All of these were trivial in nature (pervasively using
`Instruction` instead just worked).
llvm-svn: 344502
Chandler Carruth [Mon, 15 Oct 2018 10:00:15 +0000 (10:00 +0000)]
[TI removal] Remove `TerminatorInst` from GVN.h and GVN.cpp.
This is the last interesting usage in all of LLVM's headers. The
remaining usages in headers are the core typesystem bits (Core.h,
instruction types, and InstVisitor) and as the return of
`BasicBlock::getTerminator`. The latter is the big remaining API point
that I'll remove after mass updates to user code.
llvm-svn: 344501
Chandler Carruth [Mon, 15 Oct 2018 09:47:26 +0000 (09:47 +0000)]
[TI removal] Remove `TerminatorInst` from SparsePropagation.h and
related code.
This is simple as we just need to replace the type and move to the
concept of visiting a "terminator" rather than a specific instruction
subclass.
llvm-svn: 344500
Chandler Carruth [Mon, 15 Oct 2018 09:34:31 +0000 (09:34 +0000)]
[TI removal] Remove a dead forward declaration of TerminatorInst. NFC.
llvm-svn: 344499
Chandler Carruth [Mon, 15 Oct 2018 09:34:05 +0000 (09:34 +0000)]
[TI removal] Remove `TerminatorInst` from BasicBlockUtils.h
This requires updating a number of .cpp files to adapt to the new API.
I've just systematically updated all uses of `TerminatorInst` within
these files te `Instruction` so thta I won't have to touch them again in
the future.
llvm-svn: 344498
Chandler Carruth [Mon, 15 Oct 2018 09:33:40 +0000 (09:33 +0000)]
[TI removal] Just use Instruction in the CFG printer code. NFC.
llvm-svn: 344497
Guillaume Chatelet [Mon, 15 Oct 2018 09:21:21 +0000 (09:21 +0000)]
[llvm-exegesis] Fix missing std::move.
llvm-svn: 344496
Chandler Carruth [Mon, 15 Oct 2018 09:17:38 +0000 (09:17 +0000)]
[TI removal] Remove a unnecessary use of `TerminatorInst` from an IR
header. NFC.
Part of the removal of `TerminatorInst` from the type hierarchy.
llvm-svn: 344495
Chandler Carruth [Mon, 15 Oct 2018 09:17:09 +0000 (09:17 +0000)]
[TI removal] Remove TerminatorInst as an input parameter from all public
LLVM APIs. There weren't very many.
We still have the instruction visitor, and APIs with TerminatorInst as
a return type or an output parameter.
llvm-svn: 344494
Guillaume Chatelet [Mon, 15 Oct 2018 09:09:19 +0000 (09:09 +0000)]
[llvm-exegesis][NFC] Return many CodeTemplates instead of one.
Summary: This is part one of the change where I simply changed the signature of the functions. More work need to be done to actually produce more than one CodeTemplate per instruction.
Reviewers: courbet
Subscribers: tschuett, llvm-commits
Differential Revision: https://reviews.llvm.org/D53209
llvm-svn: 344493
Bjorn Pettersson [Mon, 15 Oct 2018 08:36:03 +0000 (08:36 +0000)]
[TwoAddressInstructionPass] Replace subregister uses when processing tied operands
Summary:
TwoAddressInstruction pass typically rewrites
%1:short = foo %0.sub_lo:long
as
%1:short = COPY %0.sub_lo:long
%1:short = foo %1:short
when having tied operands.
If there are extra un-tied operands that uses the same reg and
subreg, such as the second and third inputs to fie here:
%1:short = fie %0.sub_lo:long, %0.sub_hi:long, %0.sub_lo:long
then there was a bug which replaced the register %0 also for
the un-tied operand, but without changing the subregister indices.
So we used to get:
%1:short = COPY %0.sub_lo:long
%1:short = fie %1, %1.sub_hi:short, %1.sub_lo:short
With this fix we instead get:
%1:short = COPY %0.sub_lo:long
%1:short = fie %1, %0.sub_hi:long, %1
Reviewers: arsenm, JesperAntonsson, kparzysz, MatzeB
Reviewed By: MatzeB
Subscribers: bjope, kparzysz, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D36224
llvm-svn: 344492
Craig Topper [Mon, 15 Oct 2018 05:31:24 +0000 (05:31 +0000)]
[X86] Autogenerate checks. NFC
llvm-svn: 344490
Lang Hames [Mon, 15 Oct 2018 05:07:54 +0000 (05:07 +0000)]
[ORC] Simplify naming for JITDylib definition generators.
Renames:
JITDylib's setFallbackDefinitionGenerator method to setGenerator.
DynamicLibraryFallbackGenerator class to DynamicLibrarySearchGenerator.
ReexportsFallbackDefinitionGenerator to ReexportsGenerator.
llvm-svn: 344489
Dean Michael Berris [Mon, 15 Oct 2018 02:57:06 +0000 (02:57 +0000)]
[XRay][compiler-rt] FDR Mode Controller
Summary:
This change implements a controller for abstracting away the details of
what happens when tracing with FDR mode. This controller type allows us
to test in isolation the various cases where we're encountering function
entry, exit, and other kinds of events we are handling when FDR mode is
enabled.
This change introduces a number of testing facilities we've needed to
better support expressing the conditions we need for the unit tests. We
leave some TODOs for moving those utilities into the LLVM project,
sitting in the `Testing` library, to make matching conditions on XRay
`Trace` instances through googlemock more manageable and declarative.
We don't wire in the controller right away, to allow us to incrementally
update the implementation(s) as we increase testing coverage of the
controller type. There's a need to re-think the way we're managing
buffers in a multi-threaded environment, which is more invasive than
this implementation.
This step in the process allows us to encode our assumptions in the
implementation of the controller, and then evolve the buffer queue
implementation to support generational buffer management to ensure we
can continue to support the cases we're already supporting with the
controller.
Reviewers: mboerger, eizan
Subscribers: mgorny, llvm-commits, jfb
Differential Revision: https://reviews.llvm.org/D52588
llvm-svn: 344488
Craig Topper [Mon, 15 Oct 2018 01:51:58 +0000 (01:51 +0000)]
[X86] Move promotion of vector and/or/xor from legalization to DAG combine
Summary:
I've noticed that the bitcasts we introduce for these make computeKnownBits and computeNumSignBits not work well in LegalizeVectorOps. LegalizeVectorOps legalizes bottom up while LegalizeDAG legalizes top down. The bottom up strategy for LegalizeVectorOps means operands are legalized before their uses. So we promote and/or/xor before we legalize the operands that use them making computeKnownBits/computeNumSignBits in places like LowerTruncate suboptimal. I looked at changing LegalizeVectorOps to be top down as well, but that was more disruptive and caused some regressions. I also looked at just moving promotion of binops to LegalizeDAG, but that had a few issues one around matching AND,ANDN,OR into VSELECT because I had to create ANDN as vXi64, but the other nodes hadn't legalized yet, I didn't look too hard at fixing that.
This patch seems to produce better results overall than my other attempts. We now form broadcasts of constants better in some cases. For at least some of them the AND was being introduced in LegalizeDAG, promoted to vXi64, and the BUILD_VECTOR was also legalized there. I think we got bad ordering of that. Now the promotion is out of the legalizer so we handle this better.
In the longer term I think we really should evaluate whether we should be doing this promotion at all. It's really there to reduce isel pattern count, but I'm wondering if we'd be better served just eating the pattern cost or doing C++ based isel for vector and/or/xor in X86ISelDAGToDAG. The masked and/or/xor will definitely be difficult in patterns if a bitcast gets between the vselect and the and/or/xor node. That becomes a lot of permutations to cover.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D53107
llvm-svn: 344487
Craig Topper [Mon, 15 Oct 2018 01:51:53 +0000 (01:51 +0000)]
[X86] Add 128 MOVDDUP to the constant pool printing in X86AsmPrinter::EmitInstruction.
We use this instruction to broadcast a single 64-bit value to a v2i64/v2f64 vector.
llvm-svn: 344486
Craig Topper [Mon, 15 Oct 2018 01:51:50 +0000 (01:51 +0000)]
[X86] Autogenerate complete checks. NFC
llvm-svn: 344485
Simon Pilgrim [Sun, 14 Oct 2018 20:14:33 +0000 (20:14 +0000)]
[InstCombine] Add PR27343 test cases
llvm-svn: 344484
Ayal Zaks [Sun, 14 Oct 2018 17:53:02 +0000 (17:53 +0000)]
[LV] Fix comments reported when not vectorizing single iteration loops; NFC
Landing this as a separate part of https://reviews.llvm.org/D50480, being a
seemingly unrelated change ([LV] Vectorizing loops of arbitrary trip count
without remainder under opt for size).
llvm-svn: 344483
Sid Manning [Sun, 14 Oct 2018 17:51:36 +0000 (17:51 +0000)]
[Hexagon] Update tests account for non-hardcoded linker name.
Tests should not assume the linker's name, CLANG_DEFAULT_LINKER could
change it.
Differential Revision: https://reviews.llvm.org/D53219
llvm-svn: 344482
Simon Pilgrim [Sun, 14 Oct 2018 17:34:20 +0000 (17:34 +0000)]
[X86][AVX] Enable lowerVectorShuffleAsLanePermuteAndPermute v16i16/v32i8 shuffle lowering
Extends D53148 from v4f64 now that we have test coverage for v16i16/v32i8 shuffles.
llvm-svn: 344481
Jonathan Metzman [Sun, 14 Oct 2018 17:07:40 +0000 (17:07 +0000)]
[libfuzzer][Windows] Silence linker warning in unittest
Summary:
Silence warning when linking unittest binary by not passing
-lstdc++ to the linker since it is ignored.
Reviewers: morehouse
Reviewed By: morehouse
Subscribers: mgorny
Differential Revision: https://reviews.llvm.org/D53225
llvm-svn: 344480
Simon Pilgrim [Sun, 14 Oct 2018 16:49:04 +0000 (16:49 +0000)]
[ARM] Regenerate cttz tests
Improve codegen view as part of PR32655
llvm-svn: 344479
Lang Hames [Sun, 14 Oct 2018 16:09:59 +0000 (16:09 +0000)]
[ORC] Remove XXLayer::add methods that default to using the main JITDylib.
They're not currently used and may complicate upcoming changes to add's
signature and behavior.
llvm-svn: 344478
Simon Pilgrim [Sun, 14 Oct 2018 15:56:28 +0000 (15:56 +0000)]
[LegalizeDAG] Don't bother with final MUL+SRL stage for byte CTPOP.
The final stage of CTPOP expansion (v = (v * 0x01010101...) >> (Len - 8)) is completely pointless for the byte (Len = 8) case as it reduces to (v = (v * 0x01...) >> 0), but annoyingly this doesn't always get optimized away.
Found while investigating generic vector CTPOP expansion (PR32655).
llvm-svn: 344477
Sanjay Patel [Sun, 14 Oct 2018 15:25:06 +0000 (15:25 +0000)]
[InstCombine] combine a shuffle and an extract subvector shuffle
This is part of the missing IR-level folding noted in D52912.
This should be ok as a canonicalization because the new shuffle mask can't
be any more complicated than the existing shuffle mask. If there's some
target where the shorter vector shuffle is not legal, it should just end up
expanding to something like the pair of shuffles that we're starting with here.
Differential Revision: https://reviews.llvm.org/D53037
llvm-svn: 344476
Dorit Nuzman [Sun, 14 Oct 2018 08:50:06 +0000 (08:50 +0000)]
recommit 344472 after fixing build failure on ARM and PPC.
llvm-svn: 344475
Vadim Chugunov [Sun, 14 Oct 2018 07:24:56 +0000 (07:24 +0000)]
Fix double import of _lldb module.
Fix llvm.org/pr39054:
- Register _lldb as a built-in module during initialization of script interpreter,
- Reverse the order of imports in __init__.py: first try to import by absolute name, which will find the built-in module in the context of lldb (and other hosts that embed liblldb), then try relative import, in case the module is being imported from Python interpreter.
This works for SWIG>=3.0.11; before that, SWIG did not support custom module import code.
Differential revision: https://reviews.llvm.org/D52404
llvm-svn: 344474
Dorit Nuzman [Sun, 14 Oct 2018 07:21:20 +0000 (07:21 +0000)]
revert 344472 due to failures.
llvm-svn: 344473
Dorit Nuzman [Sun, 14 Oct 2018 07:06:16 +0000 (07:06 +0000)]
[IAI,LV] Add support for vectorizing predicated strided accesses using masked
interleave-group
The vectorizer currently does not attempt to create interleave-groups that
contain predicated loads/stores; predicated strided accesses can currently be
vectorized only using masked gather/scatter or scalarization. This patch makes
predicated loads/stores candidates for forming interleave-groups during the
Loop-Vectorizer's analysis, and adds the proper support for masked-interleave-
groups to the Loop-Vectorizer's planning and transformation stages. The patch
also extends the TTI API to allow querying the cost of masked interleave groups
(which each target can control); Targets that support masked vector loads/
stores may choose to enable this feature and allow vectorizing predicated
strided loads/stores using masked wide loads/stores and shuffles.
Reviewers: Ayal, hsaito, dcaballe, fhahn, javed.absar
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D53011
llvm-svn: 344472
Craig Topper [Sun, 14 Oct 2018 04:01:40 +0000 (04:01 +0000)]
[X86] Fix bad indentation. NFC
llvm-svn: 344471
Craig Topper [Sun, 14 Oct 2018 03:36:27 +0000 (03:36 +0000)]
[X86] Type legalize v2f32 stores by widening to v4f32, casting to v2f64, extracting f64 and storing.
Summary: This is similar to what D52528 did for loads. It should match what generic type legalization does in 64-bit mode where it uses a v2i64 cast and an i64 store.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D53173
llvm-svn: 344470
Hans Wennborg [Sat, 13 Oct 2018 22:22:03 +0000 (22:22 +0000)]
Try harder to fix test/Driver/cl-showfilenames.c
Follow-up to r344462.
llvm-svn: 344469
Benjamin Kramer [Sat, 13 Oct 2018 22:18:22 +0000 (22:18 +0000)]
Move some helpers from the global namespace into anonymous ones.
llvm-svn: 344468
Lang Hames [Sat, 13 Oct 2018 21:53:40 +0000 (21:53 +0000)]
[ORC] During lookup, do not match against hidden symbols in other JITDylibs.
This adds two arguments to the main ExecutionSession::lookup method:
MatchNonExportedInJD, and MatchNonExported. These control whether and where
hidden symbols should be matched when searching a list of JITDylibs.
A similar effect could have been achieved by filtering search results, but
this would have involved materializing symbol definitions (since materialization
is triggered on lookup) only to throw the results away, among other issues.
llvm-svn: 344467
Simon Pilgrim [Sat, 13 Oct 2018 21:50:15 +0000 (21:50 +0000)]
[AARCH64] Regenerate popcnt tests
Improve codegen view as part of PR32655
llvm-svn: 344466
Simon Pilgrim [Sat, 13 Oct 2018 21:32:49 +0000 (21:32 +0000)]
[ARM] Regenerate popcnt tests
Improve codegen view as part of PR32655
llvm-svn: 344465
Enrico Steffinlongo [Sat, 13 Oct 2018 19:45:48 +0000 (19:45 +0000)]
[analyzer] Improved cmake configuration for Z3
Summary:
Enhanced support for Z3 in the cmake configuration of clang; now it is possible to specify any arbitrary Z3 install prefix (CLANG_ANALYZER_Z3_PREFIX) to cmake with lib (or bin) and include folders. Before the patch only in cmake default locations
were searched (https://cmake.org/cmake/help/v3.4/command/find_path.html).
Specifying any CLANG_ANALYZER_Z3_PREFIX will force also CLANG_ANALYZER_BUILD_Z3 to ON.
Removed also Z3 4.5 version requirement since it was not checked, and now Clang works with Z3 4.7
Reviewers: NoQ, george.karpenkov, mikhail.ramalho
Reviewed By: george.karpenkov
Subscribers: rnkovacs, NoQ, esteffin, george.karpenkov, delcypher, ddcc, mgorny, xazax.hun, szepet, a.sidorin, Szelethus
Tags: #clang
Differential Revision: https://reviews.llvm.org/D50818
llvm-svn: 344464
Enrico Steffinlongo [Sat, 13 Oct 2018 19:42:10 +0000 (19:42 +0000)]
[analyzer] Small SMT API improvement
Summary: Removed const qualifier from reset method of SMTSolver and Z3Solver objects.
Reviewers: mikhail.ramalho, george.karpenkov, NoQ, ddcc
Reviewed By: NoQ
Subscribers: xazax.hun, szepet, a.sidorin, Szelethus
Differential Revision: https://reviews.llvm.org/D52031
llvm-svn: 344463
Hans Wennborg [Sat, 13 Oct 2018 19:13:14 +0000 (19:13 +0000)]
Re-commit r344234 "clang-cl: Add /showFilenames option (PR31957)"
The test was failing on e.g. PPC which can't target Windows. Fix by
requiring X86 target in the test. Also, make sure the output goes to a
temporary directory, since CWD may not be writable.
llvm-svn: 344462
Simon Pilgrim [Sat, 13 Oct 2018 18:40:48 +0000 (18:40 +0000)]
Pull out repeated variables from SelectionDAGLegalize::ExpandBitCount.
The CTPOP case has been changed from VT.getSizeInBits to VT.getScalarSizeInBits - but this fits in with future work for vector support (PR32655) and doesn't affect any current (scalar) uses.
llvm-svn: 344461
Craig Topper [Sat, 13 Oct 2018 17:47:20 +0000 (17:47 +0000)]
[LegalizeTypes] Prevent an assertion from PromoteIntRes_BSWAP and PromoteIntRes_BITREVERSE if the shift amount is too large for the VT returned by getShiftAmountTy
Summary:
getShiftAmountTy for X86 returns MVT::i8. If a BSWAP or BITREVERSE is created that requires promotion and the difference between the original VT and the promoted VT is more than 255 then we won't able to create the constant.
This patch adds a check to replace the result from getShiftAmountTy to MVT::i32 if the difference won't fit. This should get legalized later when the shift is ultimately expanded since its clearly an illegal type that we're only promoting to make it a power of 2 bit width. Alternatively we could base the decision completely on the largest shift amount the promoted VT could use.
Vectors should be immune here because getShiftAmountTy always returns the incoming VT for vectors. Only the scalar shift amount can be changed by the targets.
Reviewers: eli.friedman, RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D53232
llvm-svn: 344460
Thomas Lively [Sat, 13 Oct 2018 16:58:03 +0000 (16:58 +0000)]
[WebAssembly][NFC] Fix signed/unsigned comparison warning
llvm-svn: 344459
Sanjay Patel [Sat, 13 Oct 2018 16:15:37 +0000 (16:15 +0000)]
[InstCombine] fix complexity canonicalization with fake unary vector ops
This is a preliminary step to avoid regressions when we add
an actual 'fneg' instruction to IR. See D52934 and D53205.
llvm-svn: 344458
Simon Pilgrim [Sat, 13 Oct 2018 16:11:15 +0000 (16:11 +0000)]
[X86][SSE] Remove most of vector CTTZ custom lowering and use LegalizeDAG instead.
There is one remnant - AVX1 custom splitting of 256-bit vectors - which is due to a regression where the X86ISD::ANDNP is still performed as a YMM.
I've also tightened the CTLZ or CTPOP lowering in SelectionDAGLegalize::ExpandBitCount to require a legal CTLZ - it doesn't affect existing users and fixes an issue with AVX512 codegen.
llvm-svn: 344457
Sanjay Patel [Sat, 13 Oct 2018 16:02:47 +0000 (16:02 +0000)]
[InstCombine] add tests for operand complexity canonicalization; NFC
The tests with undef vector elements demonstrate a hole in
the current pattern matching.
llvm-svn: 344456
David Bolvansky [Sat, 13 Oct 2018 15:26:13 +0000 (15:26 +0000)]
[NFC] Fixed duplicated test file
llvm-svn: 344455
David Bolvansky [Sat, 13 Oct 2018 15:21:55 +0000 (15:21 +0000)]
[InstCombine] Fixed crash with aliased functions
Summary: Fixes PR39177
Reviewers: spatel, jbuening
Reviewed By: jbuening
Subscribers: jbuening, llvm-commits
Differential Revision: https://reviews.llvm.org/D53129
llvm-svn: 344454
Simon Pilgrim [Sat, 13 Oct 2018 15:16:55 +0000 (15:16 +0000)]
[X86][SSE] Begin removing vector CTTZ custom lowering and use LegalizeDAG instead.
Adds CTTZ vector legalization support and begins the removal of the X86/SSE custom lowering.
llvm-svn: 344453
Simon Pilgrim [Sat, 13 Oct 2018 14:45:44 +0000 (14:45 +0000)]
[X86][SSE] combineIncDecVector - use isConstantSplat
Use isConstantSplat instead of ISD::isConstantSplatVector to let us us peek through to illegal types (in this case for i686 targets to recognise i64 constants)
llvm-svn: 344452
Simon Pilgrim [Sat, 13 Oct 2018 14:28:40 +0000 (14:28 +0000)]
[X86] Pull out target constant splat helper function. NFCI.
The code in LowerScalarImmediateShift is just a more powerful version of ISD::isConstantSplatVector.
llvm-svn: 344451
Simon Pilgrim [Sat, 13 Oct 2018 13:33:32 +0000 (13:33 +0000)]
Pull out repeated getOperand(). NFCI.
llvm-svn: 344450
Simon Pilgrim [Sat, 13 Oct 2018 13:30:10 +0000 (13:30 +0000)]
Remove unused variable. NFCI.
llvm-svn: 344449
Simon Pilgrim [Sat, 13 Oct 2018 13:05:19 +0000 (13:05 +0000)]
[X86][SSE] Improve CTTZ lowering when CTLZ is legal
If we have better CTLZ support than CTPOP, then use cttz(x) = width - ctlz(~x & (x - 1)) - and remove the CTTZ_ZERO_UNDEF handling as it no longer gives better codegen.
Similar to rL344447, this is also closer to LegalizeDAG's approach
llvm-svn: 344448
Simon Pilgrim [Sat, 13 Oct 2018 12:12:06 +0000 (12:12 +0000)]
[X86][SSE] Change CTTZ vector lowering to cttz(x) = ctpop(~x & (x - 1))
This patch changes the vector CTTZ lowering from:
cttz(x) = ctpop((x & -x) - 1)
to:
cttz(x) = ctpop(~x & (x - 1))
Not only does this make better use of the PANDN instruction, but it also matches the LegalizeDAG method which should allow us to remove the x86 specific code at some point in the future (we need to fix some issues with the bitcasted logic ops and CTPOP lowering first).
Differential Revision: https://reviews.llvm.org/D53214
llvm-svn: 344447