Joachim Protze [Mon, 22 Feb 2021 17:46:28 +0000 (18:46 +0100)]
[sanitizers] Pass CMAKE_C_FLAGS into TSan buildgo script
When compiling with ccache, compiler commands get split into smaller steps
and clang's default -Wunused-command-line-argument complains about unused
include directory arguments. In combination -Werror, compilation aborts.
If CMAKE_C_FLAGS contains -Wno-unused-command-line-argument or
-Wno-error=unused-command-line-argument, the latter flag is passed into the
build script.
This is a re-commit. The previous version was reverted because of failing
tests.
Differential Revision: https://reviews.llvm.org/D96762
Nikita Popov [Mon, 22 Feb 2021 17:46:55 +0000 (18:46 +0100)]
[MemCpyOpt] Fix handling of readnone byval arguments
If the call is readnone, then there may not be any MemoryAccess
associated with the call. Bail out in that case.
This fixes the issue reported at
https://reviews.llvm.org/D94376#2578312.
Nathan James [Mon, 22 Feb 2021 17:43:52 +0000 (17:43 +0000)]
[clang] Tweaked fixit for static assert with no message
If a static assert has a message as the right side of an and condition, suggest a fix it of replacing the '&&' to ','.
`static_assert(cond && "Failed Cond")` -> `static_assert(cond, "Failed cond")`
This use case comes up when lazily replacing asserts with static asserts.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D89065
Nikita Popov [Sun, 21 Feb 2021 15:24:18 +0000 (16:24 +0100)]
[JumpThreading] Clone noalias.scope.decl when threading blocks
When cloning instructions during jump threading, also clone and
adapt any declared scopes. This is primarily important when
threading loop exits, because we'll end up with two dominating
scope declarations in that case (at least after additional loop
rotation). This addresses a loose thread from
https://reviews.llvm.org/rG2556b413a7b8#975012.
Differential Revision: https://reviews.llvm.org/D97154
Fangrui Song [Mon, 22 Feb 2021 17:18:13 +0000 (09:18 -0800)]
Improve diagnostic for ignored GNU 'used' attribute
Differential Revision: https://reviews.llvm.org/D97161
Jez Ng [Mon, 22 Feb 2021 17:06:58 +0000 (12:06 -0500)]
[lld-macho] Clean up comments
Jez Ng [Mon, 22 Feb 2021 17:12:39 +0000 (12:12 -0500)]
[lld-macho] Fix cpuSubtype for non-x86_64 archs
dyld on iOS will complain if the LIB64 bit is set.
Reviewed By: #lld-macho, thakis
Differential Revision: https://reviews.llvm.org/D96565
Florian Hahn [Mon, 22 Feb 2021 10:43:11 +0000 (10:43 +0000)]
[ConstraintElimination] Add initial ICMP_NE test cases.
Tim Keith [Mon, 22 Feb 2021 17:03:19 +0000 (09:03 -0800)]
[flang] Add -J and -module-dir to f18 driver
Add -J to the f18 driver for compatibility with gfortran.
Add -module-dir for compatibility with the new flang driver.
They both set the output directory for .mod files and add the
directory to the search list. -module still only does the former.
Clean up the new driver test to match.
Differential Revision: https://reviews.llvm.org/D97164
Simon Pilgrim [Mon, 22 Feb 2021 17:00:49 +0000 (17:00 +0000)]
[InstSimplify] Cleanup out-of-range shift amount handling.
Use APInt::uge() direct instead of getLimitedValue().
Use KnownBits::getMinValue() to make the bounds check more obvious.
Florian Hahn [Mon, 22 Feb 2021 15:05:45 +0000 (15:05 +0000)]
[LV] Directly use incoming value for single VPBlendRecipes.
VPBlendRecipes with single incoming (value, mask) pair are no-ops. Use
the incoming value directly.
Shilei Tian [Mon, 22 Feb 2021 15:59:55 +0000 (10:59 -0500)]
[Clang][OpenMP] Require CUDA 9.2+ for OpenMP offloading on NVPTX target
In current implementation of `deviceRTLs`, we're using some functions
that are CUDA version dependent (if CUDA_VERSION < 9, it is one; otheriwse, it
is another one). As a result, we have to compile one bitcode library for each
CUDA version supported. A worse problem is forward compatibility. If a new CUDA
version is released, we have to update CMake file as well.
CUDA 9.2 has been released for three years. Instead of using various weird tricks
to make `deviceRTLs` work with different CUDA versions and still have forward
compatibility, we can simply drop support for CUDA 9.1 or lower version. It has at
least two benifits:
- We don't need to generate bitcode libraries for each CUDA version;
- Clang driver doesn't need to search for the bitcode lib based on CUDA version.
We can claim that starting from LLVM 12, OpenMP offloading on NVPTX target requires
CUDA 9.2+.
Reviewed By: jdoerfert, JonChesterfield
Differential Revision: https://reviews.llvm.org/D97003
Nathan James [Mon, 22 Feb 2021 15:25:33 +0000 (15:25 +0000)]
[clang][NFC] Reorder CXXCtorInitializer members
Swapping the order of Init and MemberOrEllipsisLocation removes 8 bytes (20%) of padding on 64bit builds.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D97191
Sanjay Patel [Mon, 22 Feb 2021 14:30:14 +0000 (09:30 -0500)]
[LangRef] fix typo in assume bundle description; NFC
Anastasia Stulova [Mon, 22 Feb 2021 14:22:02 +0000 (14:22 +0000)]
[pp-trace] Fix test for OpenCL pragmas.
After updating clang driver to include standard
OpenCL headers implicitly, the output being checked
in the test does not match because the implicit
header contains other pragmas. The test does not
aim to use the header and therefore it has to be
updated passing '-cl-no-stdinc' command-line flag.
This fixes failing bots.
Mikhail Goncharov [Mon, 22 Feb 2021 13:37:03 +0000 (14:37 +0100)]
Revert "Revert "Revert "Implement nullPointerConstant() using a better API."""
This reverts commit
ba1d9546ee389ce1d2e8f353ae777a65f647d508.
Mikhail Goncharov [Mon, 22 Feb 2021 12:30:25 +0000 (13:30 +0100)]
Revert "Revert "Implement nullPointerConstant() using a better API.""
This reverts commit
6984e0d4398592a20055cb12842fc72462ce01a5.
While change by itself seems to be consistent with nullPointerConstant
docs of not matching "int i = 0;" but it's not clear why it's wrong and
9148302a2ae5ac6e5d69ae84042361889247ce64 author just forgot to update
the doc.
Anastasia Stulova [Mon, 22 Feb 2021 11:05:52 +0000 (11:05 +0000)]
[OpenCL] Add builtin declarations by default.
This change enables the builtin function declarations
in clang driver by default using the Tablegen solution
along with the implicit include of 'opencl-c-base.h'
header.
A new flag '-cl-no-stdinc' disabling all default
declarations and header includes is added. If any other
mechanisms were used to include the declarations (e.g.
with -Xclang -finclude-default-header) and the new default
approach is not sufficient the, `-cl-no-stdinc` flag has
to be used with clang to activate the old behavior.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D96515
Simon Pilgrim [Mon, 22 Feb 2021 12:20:06 +0000 (12:20 +0000)]
[InstCombine] Add PR45977 test coverage
Ryan Santhiraraja [Mon, 22 Feb 2021 11:23:31 +0000 (11:23 +0000)]
[AArch64] Adding SHA3 Intrinsics support
This patch adds the following SHA3 Intrinsics:
vsha512hq_u64,
vsha512h2q_u64,
vsha512su0q_u64,
vsha512su1q_u64
veor3q_u8
veor3q_u16
veor3q_u32
veor3q_u64
veor3q_s8
veor3q_s16
veor3q_s32
veor3q_s64
vrax1q_u64
vxarq_u64
vbcaxq_u8
vbcaxq_u16
vbcaxq_u32
vbcaxq_u64
vbcaxq_s8
vbcaxq_s16
vbcaxq_s32
vbcaxq_s64
Note need to include +sha3 and +crypto when building from the front-end
Reviewed By: DavidSpickett
Differential Revision: https://reviews.llvm.org/D96381
Dmitry Preobrazhensky [Mon, 22 Feb 2021 11:59:40 +0000 (14:59 +0300)]
[AMDGPU][MC] Corrected bound_ctrl for compatibility with sp3
Enabled "bound_ctrl:1" and disabled "bound_ctrl:-1" syntax.
Corrected printer to output "bound_ctrl:1" instead of "bound_ctrl:0".
See bug 35397 for detailed issue description.
Differential Revision: https://reviews.llvm.org/D97048
LLVM GN Syncbot [Mon, 22 Feb 2021 11:35:19 +0000 (11:35 +0000)]
[gn build] Port
7dc7f0c2ecc0
Balázs Kéri [Mon, 22 Feb 2021 10:46:36 +0000 (11:46 +0100)]
[clang-tidy] Add new check 'concurrency-thread-canceltype-asynchronous' and alias 'cert-pos47-c'.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D96719
Nashe Mncube [Fri, 19 Feb 2021 11:34:58 +0000 (11:34 +0000)]
[llvm-link] fix IRMover returning wrong modified vector type
Modified scalable vector types weren't correctly returned at link-time.
The previous behaviour was a FixedVectorType was constructed
when expecting a ScalableVectorType. This commit has added a regression
test which re-creates the failure as well as a fix.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D96953
Stephen Kelly [Mon, 22 Feb 2021 11:07:03 +0000 (11:07 +0000)]
Regenerate documentation
LLVM GN Syncbot [Mon, 22 Feb 2021 10:35:33 +0000 (10:35 +0000)]
[gn build] Port
6e3071007b4c
Florian Hahn [Mon, 22 Feb 2021 10:24:57 +0000 (10:24 +0000)]
[VPlan] Skip VPWidenPHIRecipe in VPInterleavedACcessInfo.
Update unit tests that did not expect VPWidenPHIRecipes after
15a74b64dfa9.
Simon Pilgrim [Mon, 22 Feb 2021 10:24:58 +0000 (10:24 +0000)]
[InstCombine] Add smulo NumSignBits test from D97170
Andrzej Warzynski [Wed, 17 Feb 2021 14:13:29 +0000 (14:13 +0000)]
[flang][driver] Add -Xflang and make -test-io a frontend-only flang
This patch adds support for `-Xflang` in `flang-new`. The semantics are
identical to `-Xclang`.
With the addition of `-Xflang`, we can modify `-test-io` to be a
compiler-frontend only flag. This makes more sense, this flag is:
* very frontend specific
* to be used for development and testing only
* not to be exposed to the end user
Originally we added it to the compiler driver, `flang-new`, in order to
facilitate testing. With `-Xflang` this is no longer needed. Tests are
updated accordingly.
Differential Revision: https://reviews.llvm.org/D96864
David Green [Mon, 22 Feb 2021 10:07:53 +0000 (10:07 +0000)]
[ARM] Remove dead lowering code. NFC
Remove the unnecessary code from
21a4faab60c34b8a8c4d09, left over from
a different way of lowering.
Balazs Benics [Mon, 22 Feb 2021 10:12:22 +0000 (11:12 +0100)]
[analyzer][CTU] API for CTU macro expansions
Removes `CrossTranslationUnitContext::getImportedFromSourceLocation`
Removes the corresponding unit-test segment.
Introduces the `CrossTranslationUnitContext::getMacroExpansionContextForSourceLocation`
which will return the macro expansion context for an imported TU. Also adds a
few implementation FIXME notes where applicable, since this feature is
not implemented yet. This fact is also noted as Doxygen comments.
Uplifts a few CTU LIT test to match the current **incomplete** behavior.
It is a regression to some extent since now we don't expand any
macros in imported TUs. At least we don't crash anymore.
Note that the introduced function is already covered by LIT tests.
Eg.: Analysis/plist-macros-with-expansion-ctu.c
Reviewed By: balazske, Szelethus
Differential Revision: https://reviews.llvm.org/D94673
Balazs Benics [Mon, 22 Feb 2021 10:12:18 +0000 (11:12 +0100)]
[analyzer] Use the MacroExpansionContext for macro expansions in plists
Removes the obsolete ad-hoc macro expansions during bugreport constructions.
It will skip the macro expansion if the expansion happened in an imported TU.
Also removes the expected plist file, while expanding matching context for
the tests.
Adds a previously crashing `plist-macros-with-expansion.c` testfile.
Temporarily marks `plist-macros-with-expansion-ctu.c ` to `XFAIL`.
Reviewed By: xazax.hun, Szelethus
Differential Revision: https://reviews.llvm.org/D93224
Balazs Benics [Mon, 22 Feb 2021 10:12:14 +0000 (11:12 +0100)]
[analyzer] Create MacroExpansionContext member in AnalysisConsumer
Adds a `MacroExpansionContext` member to the `AnalysisConsumer` class.
Tracks macro expansions only if the `ShouldDisplayMacroExpansions` is set.
Passes a reference down the pipeline letting AnalysisConsumers query macro
expansions during bugreport construction.
Reviewed By: martong, Szelethus
Differential Revision: https://reviews.llvm.org/D93223
Balazs Benics [Mon, 22 Feb 2021 10:11:57 +0000 (11:11 +0100)]
[analyzer] Introduce MacroExpansionContext to libAnalysis
Introduce `MacroExpansionContext` to track what and how macros in a translation
unit expand. This is the first element of the patch-stack in this direction.
The main goal is to substitute the current macro expansion generator in the
`PlistsDiagnostics`, but all the other `DiagnosticsConsumer` could benefit from
this.
`getExpandedText` and `getOriginalText` are the primary functions of this class.
The former can provide you the text that was the result of the macro expansion
chain starting from a `SourceLocation`.
While the latter will tell you **what text** was in the original source code
replaced by the macro expansion chain from that location.
Here is an example:
void bar();
#define retArg(x) x
#define retArgUnclosed retArg(bar()
#define BB CC
#define applyInt BB(int)
#define CC(x) retArgUnclosed
void unbalancedMacros() {
applyInt );
//^~~~~~~~~~^ is the substituted range
// Original text is "applyInt )"
// Expanded text is "bar()"
}
#define expandArgUnclosedCommaExpr(x) (x, bar(), 1
#define f expandArgUnclosedCommaExpr
void unbalancedMacros2() {
int x = f(f(1)) )); // Look at the parenthesis!
// ^~~~~~^ is the substituted range
// Original text is "f(f(1))"
// Expanded text is "((1,bar(),1,bar(),1"
}
Might worth investigating how to provide a reusable component, which could be
used for example by a standalone tool eg. expanding all macros to their
definitions.
I borrowed the main idea from the `PrintPreprocessedOutput.cpp` Frontend
component, providing a `PPCallbacks` instance hooking the preprocessor events.
I'm using that for calculating the source range where tokens will be expanded
to. I'm also using the `Preprocessor`'s `OnToken` callback, via the
`Preprocessor::setTokenWatcher` to reconstruct the expanded text.
Unfortunately, I concatenate the token's string representation without any
whitespaces except if the token is an identifier when I emit an extra space
to produce valid code for `int var` token sequences.
This could be improved later if needed.
Patch-stack:
1) D93222 (this one) Introduces the MacroExpansionContext class and unittests
2) D93223 Create MacroExpansionContext member in AnalysisConsumer and pass
down to the diagnostics consumers
3) D93224 Use the MacroExpansionContext for macro expansions in plists
It replaces the 'old' macro expansion mechanism.
4) D94673 API for CTU macro expansions
You should be able to get a `MacroExpansionContext` for each imported TU.
Right now it will just return `llvm::None` as this is not implemented yet.
5) FIXME: Implement macro expansion tracking for imported TUs as well.
It would also relieve us from bugs like:
- [fixed] D86135
- [confirmed] The `__VA_ARGS__` and other macro nitty-gritty, such as how to
stringify macro parameters, where to put or swallow commas, etc. are not
handled correctly.
- [confirmed] Unbalanced parenthesis are not well handled - resulting in
incorrect expansions or even crashes.
- [confirmed][crashing] https://bugs.llvm.org/show_bug.cgi?id=48358
Reviewed By: martong, Szelethus
Differential Revision: https://reviews.llvm.org/D93222
Florian Hahn [Mon, 22 Feb 2021 09:39:59 +0000 (09:39 +0000)]
[VPlan] Manage pairs of incoming (VPValue, VPBB) in VPWidenPHIRecipe.
This patch extends VPWidenPHIRecipe to manage pairs of incoming
(VPValue, VPBasicBlock) in the VPlan native path. This is made possible
because we now directly manage defined VPValues for recipes.
By keeping both the incoming value and block in the recipe directly,
code-generation in the VPlan native path becomes independent of the
predecessor ordering when fixing up non-induction phis, which currently
can cause crashes in the VPlan native path.
This fixes PR45958.
Reviewed By: sguggill
Differential Revision: https://reviews.llvm.org/D96773
David Green [Mon, 22 Feb 2021 09:29:47 +0000 (09:29 +0000)]
[ARM] Move double vector insert patterns using vins to DAG combine
This removes the existing patterns for inserting two lanes into an
f16/i16 vector register using VINS, instead using a DAG combine to
pattern match the same code sequences. The tablegen patterns were
already on the large side (foreach LANE = [0, 2, 4, 6]) and were not
handling all the cases they could. Moving that to a DAG combine, whilst
not less code, allows us to better control and expand the selection of
VINSs. Additionally this allows us to remove the AddedComplexity on
VCVTT.
The extra trick that this has learned in the process is to move two
adjacent lanes using a single f32 vmov, allowing some extra
inefficiencies to be removed.
Differenial Revision: https://reviews.llvm.org/D96876
Andy Wingo [Fri, 12 Feb 2021 10:22:13 +0000 (11:22 +0100)]
[WebAssembly] call_indirect issues table number relocs
If the reference-types feature is enabled, call_indirect will explicitly
reference its corresponding function table via `TABLE_NUMBER`
relocations against a table symbol.
Also, as before, address-taken functions can also cause the function
table to be created, only with reference-types they additionally cause a
symbol table entry to be emitted.
We abuse the used-in-reloc flag on symbols to indicate which tables
should end up in the symbol table. We do this because unfortunately
older wasm-ld will carp if it see a table symbol.
Differential Revision: https://reviews.llvm.org/D90948
Kadir Cetinkaya [Thu, 18 Feb 2021 12:48:43 +0000 (13:48 +0100)]
[clang][CodeComplete] Ensure there are no crashes when completing with ParenListExprs as LHS
Differential Revision: https://reviews.llvm.org/D96950
Jan Svoboda [Mon, 22 Feb 2021 08:54:16 +0000 (09:54 +0100)]
[clang][cli] Pass '-Wspir-compat' to cc1 from driver
This patch moves the creation of the '-Wspir-compat' argument from cc1 to the driver.
Without this change, generating command line arguments from `CompilerInvocation` cannot be done reliably: there's no way to distinguish whether '-Wspir-compat' was passed to cc1 on the command line (should be generated), or if it was created within `CompilerInvocation::CreateFromArgs` (should not be generated).
This is also in line with how other '-W' flags are handled.
(This was introduced in D21567.)
Reviewed By: Anastasia
Differential Revision: https://reviews.llvm.org/D97041
Jan Svoboda [Mon, 22 Feb 2021 08:24:12 +0000 (09:24 +0100)]
[clang][cli] Stop creating '-Wno-stdlibcxx-not-found' in cc1
This patch stops creating the '-Wno-stdlibcxx-not-found' argument in `CompilerInvocation::CreateFromArgs`.
The code was added in
2e7ab55e657f (a follow-up to D48297). However, D61963 removes relevant tests and starts explicitly passing '-Wno-stdlibcxx-not-found' to the driver. I think it's fair to assume this is a dead code.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D97042
Tres Popp [Thu, 18 Feb 2021 13:25:57 +0000 (14:25 +0100)]
[mlir] Mark std.subview as NoSideEffect
Differential Revision: https://reviews.llvm.org/D96951
Djordje Todorovic [Mon, 22 Feb 2021 07:20:37 +0000 (23:20 -0800)]
[NFC][llvm-dwarfdump] Don't calculate unnecessary stats
Small optimization of the code -- No need to calculate any stats
for NULL nodes, and also no need to call the collectStatsForDie()
if it is the CU itself.
Differential Revision: https://reviews.llvm.org/D96871
Petr Hosek [Mon, 22 Feb 2021 07:47:18 +0000 (23:47 -0800)]
[InstrProfiling] Fix instrprof-gc-sections.c test
After D97110 __llvm_prof_cnts has the nobits type so it's empty.
Kern Handa [Sun, 21 Feb 2021 09:13:48 +0000 (01:13 -0800)]
[mlir] Export CUDA and Vulkan runtime wrappers on Windows
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D97140
Amara Emerson [Mon, 22 Feb 2021 05:16:20 +0000 (21:16 -0800)]
[AArch64][GlobalISel] Fix <16 x s8> G_DUP regbankselect to assign source to gpr.
We can only select this type if the source is on GPR, not FPR.
Kazu Hirata [Mon, 22 Feb 2021 03:58:07 +0000 (19:58 -0800)]
[CodeGen] Use range-based for loops (NFC)
Kazu Hirata [Mon, 22 Feb 2021 03:58:05 +0000 (19:58 -0800)]
[llvm] Fix header guards (NFC)
Identified with llvm-header-guard.
Kazu Hirata [Mon, 22 Feb 2021 03:58:03 +0000 (19:58 -0800)]
[Analysis] Use ListSeparator (NFC)
Nico Weber [Mon, 22 Feb 2021 03:13:59 +0000 (22:13 -0500)]
Revert "[sanitizers] Pass CMAKE_C_FLAGS into TSan buildgo script"
This reverts commit
ac6c13bfc49f2d67a77144c839ecf49e48cb994c.
Breaks building with PGO, see https://reviews.llvm.org/D96762#2574009
Jacques Pienaar [Sat, 30 Jan 2021 16:55:55 +0000 (08:55 -0800)]
[mlir] Add simple jupyter kernel
Simple jupyter kernel using mlir-opt and reproducer to run passes.
Useful for local experimentation & generating examples. The export to
markdown from here is not immediately useful nor did I define a
CodeMirror synax to make the HTML output prettier. It only supports one
level of history (e.g., `_`) as I was mostly using with expanding a
pipeline one pass at a time and so was all I needed.
I placed this in utils directory next to editor & debugger utils.
Differential Revision: https://reviews.llvm.org/D95742
Petr Hosek [Sat, 13 Jul 2019 21:02:07 +0000 (14:02 -0700)]
[InstrProfiling] Use ELF section groups for counters, data and values
__start_/__stop_ references retain C identifier name sections such as
__llvm_prf_*. Putting these into a section group disables this logic.
The ELF section group semantics ensures that group members are retained
or discarded as a unit. When a function symbol is discarded, this allows
allows linker to discard counters, data and values associated with that
function symbol as well.
Note that `noduplicates` COMDAT is lowered to zero-flag section group in
ELF. We only set this for functions that aren't already in a COMDAT and
for those that don't have available_externally linkage since we already
use regular COMDAT groups for those.
Differential Revision: https://reviews.llvm.org/D96757
Craig Topper [Sun, 21 Feb 2021 22:48:27 +0000 (14:48 -0800)]
[KnownBits][RISCV] Improve known bits for srem.
The result must be less than or equal to the LHS side, so any
leading zeros in the left hand side must also exist in the result.
This is stronger than the previous behavior where we only considered
the sign bit being 0.
The affected test case used the sign bit being known 0 to change
a sign extend to a zero extend pre type legalization. After type
legalization the types were promoted to i64, but we no longer
knew bit 31 was zero. This shifts are are the equivalent of an
AND with 0xffffffff or zext_inreg X, i32. This patch allows us to
see that bit 31 is zero and remove the shifts.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D97124
Stella Laurenzo [Sun, 21 Feb 2021 06:46:27 +0000 (22:46 -0800)]
Implement simple type polymorphism for linalg named ops.
* It was decided that this was the end of the line for the existing custom tc parser/generator, and this is the first step to replacing it with a declarative format that maps well to mathy source languages.
* One such source language is implemented here: https://github.com/stellaraccident/mlir-linalgpy/blob/main/samples/mm.py
* In fact, this is the exact source of the declarative `polymorphic_matmul` in this change.
* I am working separately to clean this python implementation up and add it to MLIR (probably as `mlir.tools.linalg_opgen` or equiv). The scope of the python side is greater than just generating named ops: the ops are callable and directly emit `linalg.generic` ops fully dynamically, and this is intended to be a feature for frontends like npcomp to define custom linear algebra ops at runtime.
* There is more work required to handle full type polymorphism, especially with respect to integer formulations, since they require more specificity wrt types.
* Followups to this change will bring the new generator to feature parity with the current one and delete the current. Roughly, this involves adding support for interface declarations and attribute symbol bindings.
Differential Revision: https://reviews.llvm.org/D97135
Simon Pilgrim [Sun, 21 Feb 2021 21:51:27 +0000 (21:51 +0000)]
[X86] Add vector support to sub(C1, xor(X, C2)) -> add(xor(X, ~C2), C1+1) fold.
Simon Pilgrim [Sun, 21 Feb 2021 21:40:32 +0000 (21:40 +0000)]
[X86] Replace explicit constant handling in sub(C1, xor(X, C2)) -> add(xor(X, ~C2), C1+1) fold. NFCI.
NFC cleanup before adding vector support - rely on the SelectionDAG to handle everything for us.
Simon Pilgrim [Sun, 21 Feb 2021 21:25:26 +0000 (21:25 +0000)]
[X86] Regenerate sub.ll test
Simon Pilgrim [Sun, 21 Feb 2021 21:19:39 +0000 (21:19 +0000)]
[X86] Add 'sub C1, (xor X, C1) -> add (xor X, ~C2), C1+1' tests
This is also in sub.ll but that's for a specific i686 pattern - this adds x86_64 and vector tests
Simon Pilgrim [Sun, 21 Feb 2021 21:10:08 +0000 (21:10 +0000)]
[X86] Add common CHECK check-prefix to sub combine tests
António Afonso [Sun, 21 Feb 2021 21:08:06 +0000 (13:08 -0800)]
Revert "[lldb-vscode] Emit the breakpoint changed event on location resolved"
This reverts commit
1f21d488bd79a06c9cf405cc5db985fcd71c4f70.
Michał Górny [Sat, 20 Feb 2021 09:31:42 +0000 (10:31 +0100)]
[lldb] [docs] Update platform support status
Update supported features on FreeBSD, and supported platform list
on FreeBSD, Linux and NetBSD.
Differential Revision: https://reviews.llvm.org/D97114
Martin Storsjö [Wed, 17 Feb 2021 06:56:07 +0000 (08:56 +0200)]
[LLDB] [docs] Update the list of supported architectures on Windows
Differential Revision: https://reviews.llvm.org/D96840
Pavel Labath [Thu, 4 Feb 2021 19:53:15 +0000 (20:53 +0100)]
Reapply "[lldb/test] Automatically find debug servers to test"
This reapplies
7df4eaaa93/D96202, which was reverted due to issues on
windows. These were caused by problems in the computation of the liblldb
directory, which was fixed by D96779.
The original commit message was:
Our test configuration logic assumes that the tests can be run either
with debugserver or with lldb-server. This is not entirely correct,
since lldb server has two "personalities" (platform server and debug
server) and debugserver is only a replacement for the latter.
A consequence of this is that it's not possible to test the platform
behavior of lldb-server on macos, as it is not possible to get a hold of
the lldb-server binary.
One solution to that would be to duplicate the server configuration
logic to be able to specify both executables. However, that seems
excessively redundant.
A well-behaved lldb should be able to find the debug server on its own,
and testing lldb with a different (lldb-|debug)server does not seem very
useful (even in the out-of-tree debugserver setup, we copy the server
into the build tree to make it appear "real").
Therefore, this patch deletes the configuration altogether and changes
the low-level server retrieval functions to be able to both lldb-server
and debugserver paths. They do this by consulting the "support
executable" directory of the lldb under test.
Differential Revision: https://reviews.llvm.org/D96202
Craig Topper [Sun, 21 Feb 2021 19:13:34 +0000 (11:13 -0800)]
[SelectionDAG][RISCV] Teach ComputeNumSignBits to handle SREM.
This also removes a pattern from RISCV that is no longer needed
since the sexti32 on the LHS of the srem in the pattern implies
the result is sign extended so the sign_extend_inreg should be
removed in DAG combine now.
Reviewed By: luismarques, RKSimon
Differential Revision: https://reviews.llvm.org/D97133
Simon Pilgrim [Sun, 21 Feb 2021 18:40:23 +0000 (18:40 +0000)]
[X86][AVX] canonicalizeLaneShuffleWithRepeatedOps - remove unnecessary BITCASTs.
In conjunction with the 'vperm2x128(bitcast(x),bitcast(y),c) -> bitcast(vperm2x128(x,y,c))' fold in combineTargetShuffle, this should remove any unnecessary bitcasts around vperm2x128 lane shuffles.
António Afonso [Sun, 21 Feb 2021 18:38:25 +0000 (10:38 -0800)]
Revert "Make sure the interpreter module was loaded before making checks against it"
This reverts commit
a83a825e9902b54b315870e9ed85723525208f09.
madhur13490 [Sun, 21 Feb 2021 18:02:44 +0000 (18:02 +0000)]
[NFC] Remove redundant word in comment
Differential Revision: https://reviews.llvm.org/D97157
António Afonso [Fri, 19 Feb 2021 17:43:38 +0000 (09:43 -0800)]
[lldb-vscode] Emit the breakpoint changed event on location resolved
VSCode was not being informed whenever a location had been resolved (after being initated as non-resolved), so even though it was actually resolved, the IDE would show a hollow dot (instead of a red dot) because it didn't know about the change.
Differential Revision: https://reviews.llvm.org/D96680
Nikita Popov [Sun, 21 Feb 2021 16:49:49 +0000 (17:49 +0100)]
[Loads] Add optimized FindAvailableLoadedValue() overload (NFCI)
FindAvailableLoadedValue() accepts an iterator by reference. If no
available value is found, then the iterator will either be left
at a clobbering instruction or the beginning of the basic block.
This allows using FindAvailableLoadedValue() across multiple blocks.
If this functionality is not needed, as is the case in InstCombine,
then we can use a much more efficient implementation: First try
to find an available value, and only perform clobber checks if
we actually found one. As this function only looks at a very small
number of instructions (6 by default) and usually doesn't find an
available value, this saves many expensive alias analysis queries.
Sanjay Patel [Sun, 21 Feb 2021 17:33:53 +0000 (12:33 -0500)]
[IR] restrict vector reduction intrinsic types
The arguments in all cases should be vectors of exactly one of integer or FP.
All of the tests currently pass the verifier because we check for any vector
type regardless of the type of reduction.
This obviously can't work if we mix up integer and FP, and based on current
LangRef text it was not intended to work for pointers either.
The pointer case from https://llvm.org/PR49215 is what led me here. That
example was avoided with
5b250a27ec.
Differential Revision: https://reviews.llvm.org/D96904
António Afonso [Fri, 19 Feb 2021 04:31:55 +0000 (20:31 -0800)]
Make sure the interpreter module was loaded before making checks against it
This issue was introduced in https://reviews.llvm.org/D92187.
The guard I'm changing were is supposed to act when linux is loading the linker for the second time (due to differences in paths like symlinks).
This is done by checking `module_sp != m_interpreter_module.lock()` however this will be true when `m_interpreter_module` wasn't initialized, making linux unload the linker module (the most visible result here is that lldb will stop getting notified about new modules loaded by the process, because it can't set the rendezvous breakpoint again after the stepping over it once).
The `m_interpreter_module` is not getting initialize when it goes through this path: https://github.com/llvm/llvm-project/blob/
dbfdb139f75470a9abc78e7c9faf743fdd963c2d/lldb/source/Plugins/DynamicLoader/POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp#L332, which happens when lldb was able to read the address from the dynamic section of the executable.
What I'm not sure about though, is if when we go through this path if we still load the linker twice on linux. If that's the case then it means we need to somehow set the m_interpreter_module instead of the fix I provide here. I've only tested this on Android.
Differential Revision: https://reviews.llvm.org/D96637
Nikita Popov [Sun, 21 Feb 2021 17:22:34 +0000 (18:22 +0100)]
[Loads] Extract helper frunction for available load/store (NFC)
This contains the logic for extracting an available load/store
from a given instruction, to be reused in a following patch.
Kristina Bessonova [Tue, 2 Feb 2021 10:13:43 +0000 (12:13 +0200)]
[ThinLTO] Fix import of multiply defined global variables
Currently, if there is a module that contains a strong definition of
a global variable and a module that has both a weak definition for
the same global and a reference to it, it may result in an undefined symbol error
while linking with ThinLTO.
It happens because:
* the strong definition become internal because it is read-only and can be imported;
* the weak definition gets replaced by a declaration because it's non-prevailing;
* the strong definition failed to be imported because the destination module
already contains another definition of the global yet this def is non-prevailing.
The patch adds a check to computeImportForReferencedGlobals() that allows
considering a global variable for being imported even if the module contains
a definition of it in the case this def has an interposable linkage type.
Note that currently the check is based only on the linkage type
(and this seems to be enough at the moment), but it might be worth to account
the information whether the def is prevailing or not.
Reviewed By: tejohnson
Differential Revision: https://reviews.llvm.org/D95943
Simon Pilgrim [Sun, 21 Feb 2021 15:26:54 +0000 (15:26 +0000)]
[DAG] Match USUBSAT patterns through zext/trunc
This patch handles usubsat patterns hidden through zext/trunc and uses the getTruncatedUSUBSAT helper to determine if the USUBSAT can be correctly performed in the truncated form:
zext(x) >= y ? x - trunc(y) : 0 --> usubsat(x,trunc(umin(y,SatLimit)))
zext(x) > y ? x - trunc(y) : 0 --> usubsat(x,trunc(umin(y,SatLimit)))
Based on original examples:
void foo(unsigned short *p, int max, int n) {
int i;
unsigned m;
for (i = 0; i < n; i++) {
m = *--p;
*p = (unsigned short)(m >= max ? m-max : 0);
}
}
Differential Revision: https://reviews.llvm.org/D25987
Simon Pilgrim [Sun, 21 Feb 2021 14:50:43 +0000 (14:50 +0000)]
[X86][AVX] Fold concat(extract_subvector(v0,c0), extract_subvector(v1,c1)) -> vperm2x128
Fixes regression exposed by removing bitcasts across logic-ops in D96206.
Differential Revision: https://reviews.llvm.org/D96206
Simon Pilgrim [Sun, 21 Feb 2021 14:40:54 +0000 (14:40 +0000)]
[X86] Fold bitcast(logic(bitcast(X), Y)) --> logic'(X, bitcast(Y)) for int-int bitcasts
Extend the existing combine that handles bitcasting for fp-logic ops to also help remove logic ops across bitcasts to/from the same integer types.
This helps improve AVX512 predicate handling for D/Q logic ops and also allows DAGCombine's scalarizeExtractedBinop to remove some annoying gpr->simd->gpr transfers.
The concat_vectors regression in pr40891.ll will be addressed in a followup commit on this patch.
Differential Revision: https://reviews.llvm.org/D96206
Craig Topper [Sun, 21 Feb 2021 08:06:51 +0000 (00:06 -0800)]
[RISCV] Add test cases for add/sub/mul overflow intrinsics. NFC
Largely copied from AArch64/arm64-xaluo.ll
Petr Hosek [Tue, 16 Feb 2021 07:41:16 +0000 (23:41 -0800)]
[lld][ELF] __start_/__stop_ refs don't retain C-ident named group sections
The special root semantics for identifier-named sections is meant
specifically for the metadata sections. In the context of group
semantics, where group members are always retained or discarded as a
unit, it's natural not to have this semantics apply to a section in a
group, otherwise we would never discard the group defeating the purpose
of using the group in the first place.
This change modifies the GC behavior so that __start_/__stop_ references
don't retain C identifier named sections in section groups which allows
for these groups to be collected. This matches the behavior of BFD ld.
The only kind of existing case that might break is interdependent
metadata sections that are all in a group together, but that group
doesn't contain any other sections referenced by anything except
implicit inclusion in a `__start_` and/or `__stop_`-referenced
identifier-named section, but such cases should be unlikely.
Differential Revision: https://reviews.llvm.org/D96753
Kazu Hirata [Sun, 21 Feb 2021 05:46:02 +0000 (21:46 -0800)]
[CodeGen] Use range-based for loops (NFC)
Kazu Hirata [Sun, 21 Feb 2021 05:46:00 +0000 (21:46 -0800)]
[TableGen] Use ListSeparator (NFC)
Jianzhou Zhao [Sun, 21 Feb 2021 03:29:56 +0000 (03:29 +0000)]
[dfsan] Comment out unused methods by D97087 temporarily
Brad Smith [Sun, 21 Feb 2021 01:43:16 +0000 (20:43 -0500)]
[clang][Driver][OpenBSD] libcxx also requires pthread
Dave Lee [Fri, 19 Feb 2021 19:26:28 +0000 (11:26 -0800)]
[lldb] Refine ThreadPlan::ShouldAutoContinue
Adjust `ShouldAutoContinue` to be available to any thread plan previous to the plan that
explains a stop, not limited to the parent to the plan that explains the stop.
Before this change, `Thread::ShouldStop` did the following:
1. find the plan that explains the stop
2. if it's not a master plan, continue processing previous (aka parent) plans
3. first, call `ShouldAutoContinue` on the immediate parent of the explaining plan
4. then loop over previous plans, calling `ShouldStop` and `MischiefManaged`
Of note, the iteration in step 4 does not call `ShouldAutoContinue`, so again only the
plan just prior to the explaining plan is given the opportunity to override whether to
continue or stop.
This commit changes the loop call `ShouldAutoContinue`, giving each plan the opportunity
to override `ShouldStop` of previous plans.
Why? This allows a plan to do the following:
1. mark itself done and be popped off the stack
2. allow parent plans to finish their work, and to also be popped off the stack
3. and finally, have the thread continue, not stop
This is useful for stepping into async functions. A plan will would step far enough
enough to set a breakpoint on the async target, and then use `ShouldAutoContinue` to
unwind the necessary stepping, and then have the calling thread continue.
Differential Revision: https://reviews.llvm.org/D97076
Jacques Pienaar [Sat, 20 Feb 2021 23:54:52 +0000 (15:54 -0800)]
Update test error string post pass registration change
Jacques Pienaar [Sat, 20 Feb 2021 23:42:02 +0000 (15:42 -0800)]
[mlir] Register the print-op-graph pass using ODS
Move over to ODS & use pass options.
Nathan James [Sat, 20 Feb 2021 23:35:58 +0000 (23:35 +0000)]
[NFC] Refactor PreferMemberInitializerCheck
Martin Storsjö [Fri, 19 Feb 2021 22:10:56 +0000 (00:10 +0200)]
[libcxx] [test] Call create_directory_symlink when linking directories
This makes the symlinks work properly on windows.
A similar round of cleanup was done in
c41bda7f5fc26e4602a029646991d6a5c59cb365, but these tests were
added after that.
Differential Revision: https://reviews.llvm.org/D97089
Martin Storsjö [Fri, 6 Nov 2020 09:57:47 +0000 (11:57 +0200)]
[libcxx] Make path::format a non-class enum
The spec doesn't declare it as an enum class, and being declared
as an enum class breaks referring to the values as e.g.
path::auto_format.
Differential Revision: https://reviews.llvm.org/D97084
Petr Hosek [Sat, 20 Feb 2021 05:41:45 +0000 (21:41 -0800)]
[InstrProfiling] Use nobits as __llvm_prf_cnts section type in ELF
This can reduce the binary size because counters will no longer occupy
space in the binary, instead they will be allocated by dynamic linker.
Differential Revision: https://reviews.llvm.org/D97110
Stephen Kelly [Mon, 28 Dec 2020 01:54:04 +0000 (01:54 +0000)]
[clang-tidy] Simplify throw keyword missing check
Extend test to verify that it does not match in template instantiations.
Differential Revision: https://reviews.llvm.org/D96132
Stephen Kelly [Sat, 26 Dec 2020 21:40:51 +0000 (21:40 +0000)]
[clang-tidy] Simplify function complexity check
Update test to note use of lambda instead of the invisible operator().
Differential Revision: https://reviews.llvm.org/D96131
Craig Topper [Sat, 20 Feb 2021 21:48:18 +0000 (13:48 -0800)]
[RISCV] Add another test case showing failure to use remw when the RHS has been zero extended from less than i32. NFC
Stephen Kelly [Tue, 29 Dec 2020 13:18:57 +0000 (13:18 +0000)]
[clang-itdy] Simplify virtual near-miss check
Diagnose the problem in templates in the context of the template
declaration instead of in the context of all of the (possibly very many)
template instantiations.
Differential Revision: https://reviews.llvm.org/D96224
Nikita Popov [Fri, 1 Jan 2021 20:57:34 +0000 (21:57 +0100)]
[ConstantRange] Handle wrapping ranges in min/max (PR48643)
When one of the inputs is a wrapping range, intersect with the
union of the two inputs. The union of the two inputs corresponds
to the result we would get if we treated the min/max as a simple
select.
This fixes PR48643.
Sanjay Patel [Sat, 20 Feb 2021 14:02:39 +0000 (09:02 -0500)]
[InstCombine] fold fdiv with exp/exp2 divisor (PR49147)
Follow-up to:
D96648 /
b40fde062
...for the special-case base calls.
From the earlier commit:
This is unusual in the general (non-reciprocal) case because we need
an extra instruction, but that should be better for general FP
reassociation and codegen. We conservatively check for "arcp" FMF
here as we do with existing fdiv folds, but it is not strictly
necessary to have that.
Sanjay Patel [Fri, 19 Feb 2021 21:36:02 +0000 (16:36 -0500)]
[InstCombine] add tests for fdiv of exp/exp2; NFC
Nikita Popov [Sat, 20 Feb 2021 20:42:26 +0000 (21:42 +0100)]
[ConstantRange] Handle wrapping range in binaryNot()
We don't need any special handling for wrapping ranges (or empty
ranges for that matter). The sub() call will already compute a
correct and precise range.
We only need to adjust the test expectation: We're now computing
an optimal result, rather than an unsigned envelope.
AndreyChurbanov [Sat, 20 Feb 2021 20:26:35 +0000 (23:26 +0300)]
[OpenMP] libomp: cleanup some resource leaks
Close mutexattr and condattr local objects to eliminate resource leaks.
Differential Revision: https://reviews.llvm.org/D96892
Craig Topper [Sat, 20 Feb 2021 19:51:41 +0000 (11:51 -0800)]
[RISCV] Add an additional remw test to rv64m-exhaustive-w-insts.ll. NFC
This adds the IR for this C code
int32_t foo(uint16_t x, int16_t y) {
x %= y;
return x;
}
Note the dividend is unsigned and the divisor is signed. C type
promotion rules will extend them and use a 32-bit srem and the
function returns a 32-bit result.
We fail to use remw for this case. The zero extended input has
enough sign bits, but we won't consider (i64 AssertZext X, i16) in
the sexti32 isel pattern.
We also end up with a extra shifts to zero upper bits on the result.
computeKnownBits knew the result was positive before type legalization
and allowed the SIGN_EXTEND to become ZERO_EXTEND. But after promoting
to i64 we no longer know that bit 31 (and all bits above it) should
be 0.
Shilei Tian [Sat, 20 Feb 2021 20:13:22 +0000 (15:13 -0500)]
[Clang][OpenMP] Update driver test case for OpenMP offload to use sm_35
`sm_35` is the minimum requirement for OpenMP offloading on NVPTX device.
Current driver test case is using `sm_20`. D97003 is going to switch the minimum
CUDA version to 9.2, which only supports `sm_30+`. This patch makes step for the
change.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D97120
Stephen Kelly [Tue, 29 Dec 2020 13:42:20 +0000 (13:42 +0000)]
[clang-tidy] Simplify braced init check
The normalization of matchers means that this now works in all language
modes.
Differential Revision: https://reviews.llvm.org/D96135