Luís Marques [Tue, 24 Nov 2020 22:46:38 +0000 (22:46 +0000)]
[RISCV] Set __GCC_HAVE_SYNC_COMPARE_AND_SWAP_x defines
The RISCV target did not set the GCC atomic compare and swap defines,
unlike other targets. This broke builds for things like glib on RISCV.
Patch by Kristof Provost (kprovost)
Differential Revision: https://reviews.llvm.org/D91784
Raphael Isemann [Tue, 24 Nov 2020 22:08:57 +0000 (23:08 +0100)]
[ASTImporter] Make the Import() return value consistent with the map of imported decls when merging ClassTemplateSpecializationDecls
When importing a `ClassTemplateSpecializationDecl` definition into a TU with a matching
`ClassTemplateSpecializationDecl` definition and a more recent forward decl, the ASTImporter
currently will call `MapImported()` for the definitions, but will return the forward declaration
from the `ASTImporter::Import()` call.
This is triggering some assertions in LLDB when we try to fully import some DeclContexts
before we delete the 'From' AST. The returned 'To' Decl before this patch is just the most recent
forward decl but that's not the Decl with the definition to which the ASTImporter will import
the child declarations.
This patch just changes that the ASTImporter returns the definition that the imported Decl was
merged with instead of the found forward declaration.
Reviewed By: martong
Differential Revision: https://reviews.llvm.org/D92016
Luís Marques [Tue, 24 Nov 2020 22:29:56 +0000 (22:29 +0000)]
[RISCV] Add GHC calling convention
This is a special calling convention to be used by the GHC compiler.
Patch by Andreas Schwab (schwab)
Differential Revision: https://reviews.llvm.org/D89788
Luís Marques [Tue, 24 Nov 2020 22:34:20 +0000 (22:34 +0000)]
Revert "[RISCV] Add GHC calling convention"
This reverts commit
f8317bb256be2cd8ed81ebc567f0fa626b645f63 due to lack of
proper attribution.
LLVM GN Syncbot [Tue, 24 Nov 2020 22:27:19 +0000 (22:27 +0000)]
[gn build] Port
8d06a678a5c
Hsiangkai Wang [Fri, 20 Nov 2020 00:52:03 +0000 (08:52 +0800)]
[SelectionDAG] Avoid aliasing analysis if the object size is unknown.
If the size of memory access is unknown, do not use it to analysis. One
example of unknown size memory access is to load/store scalable vector
objects on the stack.
Differential Revision: https://reviews.llvm.org/D91833
cchen [Tue, 24 Nov 2020 22:07:39 +0000 (16:07 -0600)]
[OpenMP50][DOCS] Mark target data non-contiguous as done, NFC.
Adam Czachorowski [Mon, 23 Nov 2020 13:48:57 +0000 (14:48 +0100)]
[clangd] Addusing tweak: find insertion point after definition
When type/function is defined in the middle of the file, previuosly we
would sometimes insert a "using" line before that definition, leading to
a compilation error. With this fix, we pick a point after such
definition in translation unit.
This is not a perfect solution. For example, it still doesn't handle
"using namespace" directives. It is, however, a significant improvement.
Differential Revision: https://reviews.llvm.org/D92053
Luís Marques [Tue, 24 Nov 2020 21:55:06 +0000 (21:55 +0000)]
[RISCV] Add GHC calling convention
This is a special calling convention to be used by the GHC compiler.
Differential Revision: https://reviews.llvm.org/D89788
Sean Silva [Fri, 13 Nov 2020 01:08:56 +0000 (17:08 -0800)]
[mlir] Remove SameOperandsAndResultShape when redundant with ElementwiseMappable
SameOperandsAndResultShape and ElementwiseMappable have similar
verification, but in general neither is strictly redundant with the
other.
Examples:
- SameOperandsAndResultShape allows
`"foo"(%0) : tensor<2xf32> -> tensor<?xf32> but ElementwiseMappable
does not.
- ElementwiseMappable allows
`select %scalar_pred, %true_tensor, %false_tensor` but
SameOperandsAndResultShape does not.
SameOperandsAndResultShape is redundant with ElementwiseMappable when
we can prove that the mixed scalar/non-scalar case cannot happen. In
those situations, `ElementwiseMappable & SameOperandsAndResultShape ==
ElementwiseMappable`:
- Ops with 1 operand: the case of mixed scalar and non-scalar operands
cannot happen since there is only one operand.
- When SameTypeOperands is also present, the mixed scalar/non-scalar
operand case cannot happen.
Differential Revision: https://reviews.llvm.org/D91396
Aart Bik [Mon, 23 Nov 2020 19:01:39 +0000 (11:01 -0800)]
[mlir][sparse] generalize invariant expression handling in sparse compiler
Generalizes invariant handling to anything defined outside the Linalg op
(parameters and SSA computations). Fixes bug that was using parameter number
as tensor number.
Reviewed By: penpornk
Differential Revision: https://reviews.llvm.org/D91985
Sean Fertile [Tue, 24 Nov 2020 15:07:41 +0000 (10:07 -0500)]
[PowerPC] Don't reuse an illegal typed load for int_to_fp conversion.
When the operand to an (s/u)int_to_fp node is an illegally typed load we
cannot reuse the load address since we can not build a proper dependancy
chain. The legalized loads will use a different chain output then the
illegal load. If we reuse the load address then we will build a
conversion node that uses the chain of the illegal load and operations
which modify the memory address in the other dependancy chain can be
scheduled before the floating point load which feeds the conversion.
Differential Revision: https://reviews.llvm.org/D91265
Alex Zinenko [Mon, 23 Nov 2020 19:29:27 +0000 (20:29 +0100)]
[mlir] Add conversion from SCF parallel loops to OpenMP
Introduce a conversion pass from SCF parallel loops to OpenMP dialect
constructs - parallel region and workshare loop. Loops with reductions are not
supported because the OpenMP dialect cannot model them yet.
The conversion currently targets only one level of parallelism, i.e. only
one top-level `omp.parallel` operation is produced even if there are nested
`scf.parallel` operations that could be mapped to `omp.wsloop`. Nested
parallelism support is left for future work.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D91982
Marek Kurdej [Tue, 24 Nov 2020 20:07:06 +0000 (21:07 +0100)]
[libc++] [P0482] [C++20] Implement missing bits for atomic
Added: ATOMIC_CHAR8_T_LOCK_FREE, atomic<char8_t>, atomic_char8_t.
http://wg21.link/P0482
Reviewed By: ldionne, #libc
Differential Revision: https://reviews.llvm.org/D91706
Teresa Johnson [Tue, 17 Nov 2020 22:44:25 +0000 (14:44 -0800)]
Avoid redundant work when computing vtable vcall visibility
Add a Visited set to avoid repeatedly processing the same base classes
in complex class hierarchies. This cut down the compile time of one
source file from >12min to ~1min.
Differential Revision: https://reviews.llvm.org/D91676
Nicolas Vasilache [Tue, 24 Nov 2020 19:52:24 +0000 (19:52 +0000)]
[mlir] NFC - Refactor and expose a helper printOffsetSizesAndStrides helper function.
Print part of an op of the form:
```
<optional-offset-prefix>`[` offset-list `]`
<optional-size-prefix>`[` size-list `]`
<optional-stride-prefix>[` stride-list `]`
```
Also address some leftover nits.
Differential revision: https://reviews.llvm.org/D92031
Fangrui Song [Tue, 24 Nov 2020 19:59:15 +0000 (11:59 -0800)]
[test] Clean up ppc-features.cpp and improve tests
And add ppc-cpus.cpp for -mcpu= specific tests.
Nicolas Vasilache [Tue, 24 Nov 2020 14:54:06 +0000 (14:54 +0000)]
[mlir] NFC - Refactor and expose a parsing helper for OffsetSizeAndStrideInterface
Parse trailing part of an op of the form:
```
<optional-offset-prefix>`[` offset-list `]`
<optional-size-prefix>`[` size-list `]`
<optional-stride-prefix>[` stride-list `]`
```
Each entry in the offset, size and stride list either resolves to an integer
constant or an operand of index type.
Constants are added to the `result` as named integer array attributes with
name `OffsetSizeAndStrideOpInterface::getStaticOffsetsAttrName()` (resp.
`getStaticSizesAttrName()`, `getStaticStridesAttrName()`).
Append the number of offset, size and stride operands to `segmentSizes`
before adding it to `result` as the named attribute:
`OpTrait::AttrSizedOperandSegments<void>::getOperandSegmentSizeAttr()`.
Offset, size and stride operands resolution occurs after `preResolutionFn`
to give a chance to leading operands to resolve first, after parsing the
types.
```
ParseResult parseOffsetsSizesAndStrides(
OpAsmParser &parser, OperationState &result, ArrayRef<int> segmentSizes,
llvm::function_ref<ParseResult(OpAsmParser &, OperationState &)>
preResolutionFn = nullptr,
llvm::function_ref<ParseResult(OpAsmParser &)> parseOptionalOffsetPrefix =
nullptr,
llvm::function_ref<ParseResult(OpAsmParser &)> parseOptionalSizePrefix =
nullptr,
llvm::function_ref<ParseResult(OpAsmParser &)> parseOptionalStridePrefix =
nullptr);
```
Differential revision: https://reviews.llvm.org/D92030
zhanghb97 [Tue, 24 Nov 2020 18:35:22 +0000 (18:35 +0000)]
[mlir][Python] Sync Python bindings with C API MlirStringRef modification.
MLIR C API use the `MlirStringRef` instead of `const char *` for the string type now. This patch sync the Python bindings with the C API modification.
Differential Revision: https://reviews.llvm.org/D92007
Haojian Wu [Tue, 24 Nov 2020 18:44:42 +0000 (19:44 +0100)]
[clangd] Add more trace spans for rename, NFC.
Sanjay Patel [Tue, 24 Nov 2020 18:33:12 +0000 (13:33 -0500)]
[InstCombine] try difference-of-shifts factorization before negator
We need to preserve wrapping flags to allow better folds.
The cases with geps may be non-intuitive, but that appears to agree with Alive2:
https://alive2.llvm.org/ce/z/JQcqw7
We create 'nsw' ops independent from the original wrapping on the sub.
Sanjay Patel [Tue, 24 Nov 2020 18:22:06 +0000 (13:22 -0500)]
[InstCombine] add tests for sub of muls; NFC
Sanjay Patel [Tue, 24 Nov 2020 18:14:42 +0000 (13:14 -0500)]
[InstCombine] add tests for gep math; NFC
Philip Reames [Tue, 24 Nov 2020 18:44:55 +0000 (10:44 -0800)]
[LoopVec] Add a minor clarifying comment
Arthur Eubanks [Tue, 24 Nov 2020 18:16:22 +0000 (10:16 -0800)]
[test] Pin parallel_deletion_cg_update.ll to legacy PM
This tests legacy PM-specific code.
Florian Hahn [Tue, 24 Nov 2020 18:09:26 +0000 (18:09 +0000)]
[DSE] Precommit test case for PR48279.
Arthur Eubanks [Tue, 24 Nov 2020 18:02:12 +0000 (10:02 -0800)]
[test] Fix rtf_type_checking.ll under NPM
Janek van Oirschot [Tue, 24 Nov 2020 16:54:34 +0000 (16:54 +0000)]
[HardwareLoops] Change order of SCEV expression construction for InitLoopCount.
Putting the +1 before the zero-extend will allow scalar evolution to fold the expression in some cases such as the one shown in PowerPC's `shrink-wrap.ll` test.
Reviewed By: samparker
Differential Revision: https://reviews.llvm.org/D91724
Adam Czachorowski [Wed, 18 Nov 2020 16:43:19 +0000 (17:43 +0100)]
[clangd] AddUsing: Used spelled text instead of type name.
This improves the behavior related to type aliases, as well as cases of
typo correction.
Differential Revision: https://reviews.llvm.org/D91966
Stella Laurenzo [Tue, 24 Nov 2020 17:50:18 +0000 (17:50 +0000)]
[mlir][Python] Fix the last remaining instance of PYTHON_EXECUTABLE.
* Was causing auto-detect of pybind11 to fail on clean configure.
Differential Revision: https://reviews.llvm.org/D92043
Fangrui Song [Tue, 24 Nov 2020 17:51:32 +0000 (09:51 -0800)]
[Driver] Default Generic_GCC aarch64 to -fasynchronous-unwind-tables
In GCC, `aarch64-*-linux` and `aarch64-*-freebsd` made the switch in 2018
(https://gcc.gnu.org/pipermail/gcc-patches/2018-March/495549.html).
In Clang, FreeBSD/Fuchsia/NetBSD/MinGW aarch64 default to -fasynchronous-unwind-tables.
This patch defaults Generic_GCC aarch64 (which affects Linux) to use -fasynchronous-unwind-tables.
Reviewed By: nickdesaulniers
Differential Revision: https://reviews.llvm.org/D91760
Jay Foad [Mon, 23 Nov 2020 14:13:53 +0000 (14:13 +0000)]
[AMDGPU] Add a TRANS bit to TSFlags. NFC.
This is used to mark transcendental instructions that execute on a
separate pipeline from the normal VALU pipeline.
Differential Revision: https://reviews.llvm.org/D92042
Tei Jeong [Tue, 24 Nov 2020 17:34:52 +0000 (09:34 -0800)]
Fix CalibratedQuantizedType's print function to match parser
Reviewed By: liufengdb
Differential Revision: https://reviews.llvm.org/D92034
Teresa Johnson [Thu, 19 Nov 2020 19:10:47 +0000 (11:10 -0800)]
[ThinLTO/WPD] Enable -wholeprogramdevirt-skip in ThinLTO backends
Previously this option could be used to skip devirtualizations of the
given functions in regular LTO and in the ThinLTO indexing step. This
change allows them to be skipped in the backend as well, which is useful
when debugging WPD in a distributed ThinLTO backend.
Differential Revision: https://reviews.llvm.org/D91812
Victor Huang [Tue, 24 Nov 2020 16:48:11 +0000 (10:48 -0600)]
[PowerPC][PCRelative] Add new pseudo instructions for PCRel TLS to fix R2 clobber issue
New pseudo instructions GETtlsADDRPCREL and GETtlsldADDRPCREL are added for properly
setting REGMASK for tls_get_addr function when using PCRelative address.
Differential Revisien: https://reviews.llvm.org/D91420
Reviewed by: bsaleil
Stella Laurenzo [Tue, 24 Nov 2020 16:58:26 +0000 (16:58 +0000)]
[mlir] Add Tosa dialect const folder for tosa.const.
* Was missed in the initial submission and is required for a ConstantLike op.
* Also adds a materializeConstant hook to preserve it.
* Tightens up the argument constraint on tosa.const to match what is actually legal.
Differential Revision: https://reviews.llvm.org/D92040
Louis Dionne [Tue, 24 Nov 2020 17:26:05 +0000 (12:26 -0500)]
[libc++] NFC: Fix confusing indentation in <numeric>
Hubert Tong [Tue, 24 Nov 2020 17:14:03 +0000 (12:14 -0500)]
[NFC][tests] Replace non-portable grep with FileCheck
After commit
2482648a795afbe12774168bbbf70dc14c031267, a GNU grep option
is just passed unconditionally to `grep` in general. This patch fixes
the test for platforms where `grep` is not GNU grep.
Yichao Yu [Tue, 24 Nov 2020 00:11:17 +0000 (19:11 -0500)]
Clear NewGEPBases after finish using them in CodeGenPrep pass
AFAICT all other set/map are correctly cleared in `runOnFunction`.
With assertion enabled this causes a crash when the module is freed and potentially if a later pass delete the instruction (not observed in real world though). Without assertion this can potentially cause confusing result when running on a new Function/Module.
Reviewed By: loladiro
Differential Revision: https://reviews.llvm.org/D84031
Arthur Eubanks [Tue, 24 Nov 2020 16:56:00 +0000 (08:56 -0800)]
[test] Merge parallel_region_merging{,_legacy_pm}.ll
These are identical except for the RUN lines.
Also pin legacy RUN line to legacy PM.
Thomas Preud'homme [Thu, 19 Nov 2020 23:33:23 +0000 (23:33 +0000)]
Add support for STRICT_FSETCC promotion
Add missing handling of STRICT_FSETCC promotion. This prevents assert
failure in llvm::TargetLoweringBase::getTypeToPromoteTo().
Reviewed By: uweigand
Differential Revision: https://reviews.llvm.org/D91962
Mark de Wever [Tue, 24 Nov 2020 13:50:49 +0000 (14:50 +0100)]
[libcxx] Implement P1956 rename low-level bit functions
Implements P1956: On the names of low-level bit manipulation functions.
Users may use older versions of libc++ or other standard libraries with the old names. In order to keep compatibility the old functions are kept, but marked as deprecated.
The patch also adds a new config macro `_LIBCPP_DEPRECATED_MSG`. Do you prefer a this is a separate patch?
Reviewed By: ldionne, #libc
Differential Revision: https://reviews.llvm.org/D90551
Nico Weber [Mon, 23 Nov 2020 15:51:56 +0000 (10:51 -0500)]
[lld/mac] Implement basic typo correction for flags
Also use "unknown flag 'flag'" instead of "unknown flag: flag" for
consistency with the other ports.
Differential Revision: https://reviews.llvm.org/D91970
Masoud Ataei [Tue, 24 Nov 2020 16:21:12 +0000 (16:21 +0000)]
[PowerPC] dyn_cast should be dyn_cast_or_null in MASSV pass
It is possible that we have different constants in different slots
of second vector double (float) of pow function. So, in this case
Exp->getSplatValue() will return nullptr. Here, I handle it properly.
Reviewed By: steven.zhang, PowerPC
Differential Revision: https://reviews.llvm.org/D91729
Nico Weber [Tue, 24 Nov 2020 15:59:57 +0000 (10:59 -0500)]
clang: Pass -platform-version to new MachO LLD
New MachO LLD doesn't implement the old -macos_version_min (etc)
flags, but it understands the modern platform_version flag.
So make the clang driver pass that when using new MachO LLD.
Also, while here, don't pass -lto_library to LLD, since it
links in LTO libraries statically (which it can because it's
versioned alongside clang).
Differential Revision: https://reviews.llvm.org/D92037
Sergey Dmitriev [Tue, 24 Nov 2020 15:52:46 +0000 (07:52 -0800)]
[clang-offload-bundler] use std::forward_list for storing temp file names [NFC]
Use a different container that preserves existing elements on modification
for storing temporary file names. Current container can make StringRefs
returned earlier invalid on reallocation.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D92010
Nico Weber [Tue, 24 Nov 2020 16:06:51 +0000 (11:06 -0500)]
lld: Fix darwinnew symlink name added in
e16c0a9a68971
Arthur O'Dwyer [Mon, 23 Nov 2020 17:44:41 +0000 (12:44 -0500)]
[libc++] Remove _LIBCPP_CONSTEXPR_AFTER_CXX17_WITH_IS_CONSTANT_EVALUATED.
Zoe Carver says: "We decided that libc++ only supports C++20 constexpr algorithms
when `is_constant_evaluated` is also supported. Here's a link to the discussion."
https://reviews.llvm.org/D65721#inline-735682
Remove _LIBCPP_HAS_NO_BUILTIN_IS_CONSTANT_EVALUATED from tests, too.
See Louis's
5911e6a8857f146fb5a8f23af1d768aba25e7c3e if needed to fix bots.
I've applied `UNSUPPORTED: clang-8` preemptively to the altered tests;
I don't know for sure that this was needed, because no clang-8 buildbots
are triggered on pull requests.
Sam McCall [Fri, 20 Nov 2020 14:47:55 +0000 (15:47 +0100)]
[clangd] Mention when CXXThis is implicit in exposed AST.
Seeing an implicit this in the AST is pretty confusing I think.
While here, also mention when `this` is const.
Differential Revision: https://reviews.llvm.org/D91868
Nico Weber [Tue, 24 Nov 2020 14:43:22 +0000 (09:43 -0500)]
Fix driver test from
e16c0a9a689719
The test failed silently if lld wasn't built alongside clang.
But the test uses -###, so the "invalid linker name in -fuse-ld=lld"
diag didn't make clang fail, and something else happened to match
"-demangle", so the test passed.
To fix, pass -B to a directory with two empty +x files (which works
on non-Windows), and look for `"-demangle"` instead of just `-demangle`.
Also force linker_version to 0 and pass a darwin triple.
Differential Revision: https://reviews.llvm.org/D92028
Evgeny Leviant [Tue, 24 Nov 2020 15:27:59 +0000 (18:27 +0300)]
[llvm-mca] Fix processing thumb instruction set
Differential revision: https://reviews.llvm.org/D91704
diggerlin [Tue, 24 Nov 2020 15:23:57 +0000 (10:23 -0500)]
[NFC][AIX][XCOFF] change function name from getNumofGPRsSaved to getNumOfGPRsSaved
change function name from getNumofGPRsSaved to getNumOfGPRsSaved for class XCOFFTracebackTable
Reviewers: Jason Liu
Differential Revision: https://reviews.llvm.org/D91882
Evgeny Leviant [Tue, 24 Nov 2020 15:13:10 +0000 (18:13 +0300)]
[MC][ARM] Fix number of operands of tMOVSr
Differential revision: https://reviews.llvm.org/D92029
Nicolas Vasilache [Tue, 24 Nov 2020 14:23:57 +0000 (14:23 +0000)]
[mlir] NFC - Expose an OffsetSizeAndStrideOpInterface
This revision will make it easier to create new ops base on the strided memref abstraction outside of the std dialect.
OffsetSizeAndStrideOpInterface is an interface for ops that allow specifying mixed dynamic and static offsets, sizes and strides variadic operands.
Ops that implement this interface need to expose the following methods:
1. `getArrayAttrRanks` to specify the length of static integer
attributes.
2. `offsets`, `sizes` and `strides` variadic operands.
3. `static_offsets`, resp. `static_sizes` and `static_strides` integer
array attributes.
The invariants of this interface are:
1. `static_offsets`, `static_sizes` and `static_strides` have length
exactly `getArrayAttrRanks()`[0] (resp. [1], [2]).
2. `offsets`, `sizes` and `strides` have each length at most
`getArrayAttrRanks()`[0] (resp. [1], [2]).
3. if an entry of `static_offsets` (resp. `static_sizes`,
`static_strides`) is equal to a special sentinel value, namely
`ShapedType::kDynamicStrideOrOffset` (resp. `ShapedType::kDynamicSize`,
`ShapedType::kDynamicStrideOrOffset`), then the corresponding entry is
a dynamic offset (resp. size, stride).
4. a variadic `offset` (resp. `sizes`, `strides`) operand must be present
for each dynamic offset (resp. size, stride).
This interface is useful to factor out common behavior and provide support
for carrying or injecting static behavior through the use of the static
attributes.
Differential Revision: https://reviews.llvm.org/D92011
Nico Weber [Fri, 20 Nov 2020 18:57:44 +0000 (13:57 -0500)]
clang+lld: Improve clang+ld.darwinnew.lld interaction, pass -demangle
This patch:
- adds an ld64.lld.darwinnew symlink for lld, to go with
f2710d4b576,
so that `clang -fuse-ld=lld.darwinnew` can be used to test new
Mach-O lld while it's in bring-up. (The expectation is that we'll
remove this again once new Mach-O lld is the defauld and only Mach-O
lld.)
- lets the clang driver know if the linker is lld (currently
only triggered if `-fuse-ld=lld` or `-fuse-ld=lld.darwinnew` is
passed). Currently only used for the next point, but could be used
to implement other features that need close coordination between
compiler and linker, e.g. having a diag for calling `clang++` instead
of `clang` when link errors are caused by a missing C++ stdlib.
- lets the clang driver pass `-demangle` to Mach-O lld (both old and
new), in addition to ld64
- implements -demangle for new Mach-O lld
- changes demangleItanium() to accept _Z, __Z, ___Z, ____Z prefixes
(and updates one test added in D68014). Mach-O has an extra
underscore for symbols, and the three (or, on Mach-O, four)
underscores are used for block names.
Differential Revision: https://reviews.llvm.org/D91884
Ayal Zaks [Tue, 24 Nov 2020 10:15:18 +0000 (12:15 +0200)]
[LV] Keep Primary Induction alive when folding tail by masking
Fix PR47390.
The primary induction should be considered alive when folding tail by masking,
because it will be used by said masking; even when it may otherwise appear
useless: feeding only its own 'bump', which is correctly considered dead, and
as the 'bump' of another induction variable, which may wrongfully want to
consider its bump = the primary induction, dead.
Differential Revision: https://reviews.llvm.org/D92017
Yaxun (Sam) Liu [Fri, 23 Oct 2020 20:24:48 +0000 (16:24 -0400)]
[HIP] Fix regressions due to fp contract change
Recently HIP toolchain made a change to use clang instead of opt/llc to do compilation
(https://reviews.llvm.org/D81861). The intention is to make HIP toolchain canonical like
other toolchains.
However, this change introduced an unintentional change regarding backend fp fuse
option, which caused regressions in some HIP applications.
Basically before the change, HIP toolchain used clang to generate bitcode, then use
opt/llc to optimize bitcode and generate ISA. As such, the amdgpu backend takes
the default fp fuse mode which is 'Standard'. This mode respect contract flag of
fmul/fadd instructions and do not fuse fmul/fadd instructions without contract flag.
However, after the change, HIP toolchain now use clang to generate IR, do optimization,
and generate ISA as one process. Now amdgpu backend fp fuse option is determined
by -ffp-contract option, which is 'fast' by default. And this -ffp-contract=fast language option
is translated to 'Fast' fp fuse option in backend. Suddenly backend starts to fuse fmul/fadd
instructions without contract flag.
This causes wrong result for some device library functions, e.g. tan(-1e20), which should
return 0.8446, now returns -0.933. What is worse is that since backend with 'Fast' fp fuse
option does not respect contract flag, there is no way to use #pragma clang fp contract
directive to enforce fp contract requirements.
This patch fixes the regression by introducing a new value 'fast-honor-pragmas' for -ffp-contract
and use it for HIP by default. 'fast-honor-pragmas' is equivalent to 'fast' in frontend but
let the backend to use 'Standard' fp fuse option. 'fast-honor-pragmas' is useful since 'Fast'
fp fuse option in backend does not honor contract flag, it is of little use to HIP
applications since all code with #pragma STDC FP_CONTRACT or any IR from a
source compiled with -ffp-contract=on is broken.
Differential Revision: https://reviews.llvm.org/D90174
Hans Wennborg [Tue, 24 Nov 2020 13:08:42 +0000 (14:08 +0100)]
[docs] Try to make this bullet list in ThinLTO.rst actually be a bullet list
Ulysse Beaugnon [Tue, 24 Nov 2020 12:57:18 +0000 (13:57 +0100)]
[MLIR] Fix TableGen generator for attribute interfaces.
Use the correct interface base type name when generating attribute interfaces
with TabeGen.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D92023
Fraser Cormack [Fri, 20 Nov 2020 17:10:24 +0000 (17:10 +0000)]
[RISCV] Combine GREVI sequences
This combine step performs the following type of transformation:
rev.p a0, a0 # grevi a0, a0, 0b01
rev2.n a0, a0 # grevi a0, a0, 0b10
-->
rev.n a0, a0 # grevi a0, a0, 0b11
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D91877
Kadir Cetinkaya [Fri, 20 Nov 2020 12:02:18 +0000 (13:02 +0100)]
[clangd] Fix shared-lib builds
Differential Revision: https://reviews.llvm.org/D91859
Simon Pilgrim [Tue, 24 Nov 2020 11:38:30 +0000 (11:38 +0000)]
[IR] Constant::getAggregateElement - early-out for ScalableVectorType
We can't call getNumElements() for ScalableVectorType types - just bail for now, although ConstantAggregateZero/UndefValue could return a reasonable value.
Fixes crash shown in OSS-Fuzz #25272 https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=25272
Eugene Zhulenev [Tue, 24 Nov 2020 11:17:33 +0000 (03:17 -0800)]
[mlir] AsyncRuntime: fix concurrency bugs + fix exports in methods definitions
1. Move ThreadPool ownership to the runtime, and wait for the async tasks completion in the destructor.
2. Remove MLIR_ASYNCRUNTIME_EXPORT from method definitions because they are unnecessary in .cpp files, as only function declarations need to be exported, not their definitions.
3. Fix concurrency bugs in group emplace and potential use-after-free in token emplace.
Tested internally 10k runs in `async.mlir` and `async-group.mlir`.
Fixed: https://bugs.llvm.org/show_bug.cgi?id=48267
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D91988
Max Kazantsev [Tue, 24 Nov 2020 10:56:59 +0000 (17:56 +0700)]
Revert "[NFC][SCEV] Generalize monotonicity check for full and limited iteration space"
This reverts commit
2734a9ebf4a31df0131acdfc739395a5e692c342.
This patch appeared to not be a NFC. It introduced an execution path where
monotonicity check on limited space started relying in existing nsw/nuw
flags, which is illegal. The motivating test will follow-up.
Evgeny Leviant [Tue, 24 Nov 2020 10:52:32 +0000 (13:52 +0300)]
[SchedModels] Improve diagnostics. NFC
Alexander Belyaev [Tue, 24 Nov 2020 10:09:33 +0000 (11:09 +0100)]
[mlir][linalg] Add bufferization pattern for `linalg.indexed_generic`.
Differential Revision: https://reviews.llvm.org/D92014
AndreyChurbanov [Tue, 24 Nov 2020 10:08:24 +0000 (13:08 +0300)]
[OpenMP] fix asm code for for arm64 (AARCH64) for Darwin/macOS
Adjusted external reference for Darwin/AARCH64 link compatibility.
Made size directive conditional only if __ELF__ defined.
Patch by Michael_Pique <mpique@icloud.com>
Differential Revision: https://reviews.llvm.org/D88252
Marek Kurdej [Tue, 24 Nov 2020 10:06:08 +0000 (11:06 +0100)]
[libc++] [www] Fix HTML. NFC.
Needed for a future automatic update to RST.
Evgeny Leviant [Tue, 24 Nov 2020 09:44:10 +0000 (12:44 +0300)]
[SchedModels] Limit set of predicates seen by mutuallyExclusive
Patch limits set of predicates seen by mutuallyExclusive to ones which belong
to current processor model. This needs to be done, because same predicate can
be used by multiple processor models which can make mutuallyExclusive over
optimistic.
Kai Luo [Tue, 24 Nov 2020 08:17:46 +0000 (08:17 +0000)]
[DAGCombine][PowerPC] Convert negated abs to trivial arithmetic ops
This patch converts `0 - abs(x)` to `Y = sra (X, size(X)-1); sub (Y, xor (X, Y))` for better codegen.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D91120
Georgii Rymar [Mon, 23 Nov 2020 13:18:00 +0000 (16:18 +0300)]
[llvm-readelf/obj] - Refine the implementation of `printGNUVersionSectionProlog`
This:
1) Changes its signature.
2) Refines the name of local variable (`SymTabName`->`LinkedSecName`,
because SHT_GNU_verneed/SHT_GNU_verdef are linked with the string table, not with the symbol table).
3) Stops using the `unwrapOrError` inside.
Differential revision: https://reviews.llvm.org/D91964
Georgii Rymar [Tue, 17 Nov 2020 13:15:45 +0000 (16:15 +0300)]
[llvm-readelf/obj] - Improve the error reporting in printStackSize().
This stops using `RelocationRef` API in the `printStackSize` method
and starts using the "regular" API that is used in almost all other places
in ELFDumper.cpp.
This is not only makes the code to be more consistent, but helps to diagnose
issues better, because the `ELFObjectFile` API, which is used
currently to implement stack sized dumping sometimes has a behavior
that just doesn't work well for broken inputs.
E.g see how it gets the `symbol_end` iterator. It will just not work
well for a case when the `sh_size` is broken.
```
template <class ELFT>
basic_symbol_iterator ELFObjectFile<ELFT>::symbol_end() const {
...
DataRefImpl Sym = toDRI(SymTab, SymTab->sh_size / sizeof(Elf_Sym));
return basic_symbol_iterator(SymbolRef(Sym, this));
}
```
Differential revision: https://reviews.llvm.org/D91624
Evgeny Leviant [Tue, 24 Nov 2020 08:43:51 +0000 (11:43 +0300)]
Add support for branch forms of ALU instructions to Cortex-A57 model
Patch fixes scheduling of ALU instructions which modify pc register. Patch
also fixes computation of mutually exclusive predicates for sequences of
variants to be properly expanded
Differential revision: https://reviews.llvm.org/D91266
Georgii Rymar [Mon, 23 Nov 2020 10:40:01 +0000 (13:40 +0300)]
[[lvm-readelf/obj] - Remove `tryGetSectionName` helper.
D91867 introduced the `tryGetSectionName` helper.
But we have `getPrintableSectionName` member with the similar
behavior which we can reuse. This patch does it.
Differential revision: https://reviews.llvm.org/D91954
Nathan Ridge [Tue, 24 Nov 2020 08:17:44 +0000 (03:17 -0500)]
[clangd] Sort results of incomingCalls request by container name
Differential Revision: https://reviews.llvm.org/D92009
Alex Zinenko [Mon, 23 Nov 2020 13:19:07 +0000 (14:19 +0100)]
[mlir] move lib/Bindings/Python/Attributes.td to include/mlir/Bindings/Python
This file is intended to be included by other files, including
out-of-tree dialects, and makes more sense in `include` than in `lib`.
Depends On D91652
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D91961
Haowei Wu [Tue, 13 Oct 2020 03:01:52 +0000 (20:01 -0700)]
[llvm-elfabi] Emit ELF .dynsym, .dynamic sections
This change makes llvm-elfabi tool to emit .dynsym and .dynamic
sections.
Differential Revision: https://reviews.llvm.org/D89432
Alex Zinenko [Tue, 17 Nov 2020 17:28:16 +0000 (18:28 +0100)]
[mlir] Make attributes mutable in Python bindings
Attributes represent additional data about an operation and are intended to be
modifiable during the lifetime of the operation. In the dialect-specific Python
bindings, attributes are exposed as properties on the operation class. Allow
for assigning values to these properties. Also support creating new and
deleting existing attributes through the generic "attributes" property of an
operation. Any validity checking must be performed by the op verifier after the
mutation, similarly to C++. Operations are not invalidated in the process: no
dangling pointers can be created as all attributes are owned by the context and
will remain live even if they are not used in any operation.
Introduce a Python Test dialect by analogy with the Test dialect and to avoid
polluting the latter with Python-specific constructs. Use this dialect to
implement a test for the attribute access and mutation API.
Reviewed By: stellaraccident, mehdi_amini
Differential Revision: https://reviews.llvm.org/D91652
Martin Storsjö [Tue, 17 Nov 2020 14:05:29 +0000 (16:05 +0200)]
[LLD] [COFF] Allow wrapping dllimported functions
GNU ld doesn't seem to do this though, but it looks like a reasonable
use case, is easy to implement, and was requested in
https://bugs.llvm.org/show_bug.cgi?id=47384.
Differential Revision: https://reviews.llvm.org/D91689
Pavel Labath [Tue, 24 Nov 2020 08:00:42 +0000 (09:00 +0100)]
Revert "[DebugInfo] Refactor code for emitting DWARF expressions for FP constants"
The commit introduced a crash when emitting (debug info for) complex
floats (pr48277).
Arthur Eubanks [Tue, 24 Nov 2020 07:46:08 +0000 (23:46 -0800)]
[test] Fix gvn-inline-iteration.ll under NPM
Amara Emerson [Tue, 24 Nov 2020 06:57:37 +0000 (22:57 -0800)]
[AArch64][GlobalISel] Add pre-isel lowering to convert p0 G_DUPs to use s64.
This uses the same reasoning as other similar conversions just before selection,
without it we miss out on selection because the importer considers s64 and p0
distinct types.
Amara Emerson [Tue, 24 Nov 2020 06:49:27 +0000 (22:49 -0800)]
[AArch64][GlobalISel] Make <2 x p0> of G_SHUFFLE_VECTOR legal.
Gabriel Hjort Åkerlund [Tue, 24 Nov 2020 06:12:54 +0000 (07:12 +0100)]
[GlobalISel][TableGen] Fix seg fault for zero instruction
Tablegen seg faulted when parsing a Pat where the destination part has
no output (zero instruction), due to a register class lookup using
nullptr.
Reviewed By: Paul-C-Anagnostopoulos
Differential Revision: https://reviews.llvm.org/D90829
Kai Luo [Tue, 24 Nov 2020 06:32:38 +0000 (06:32 +0000)]
[PowerPC] Pre-commit more tests for `select` codegen. NFC.
Arthur Eubanks [Thu, 15 Oct 2020 00:56:38 +0000 (17:56 -0700)]
Reland [CGSCC] Detect devirtualization in more cases
The devirtualization wrapper misses cases where if it wraps a pass
manager, an individual pass may devirtualize an indirect call created by
a previous pass. For example, inlining may create a new indirect call
which is devirtualized by instcombine. Currently the devirtualization
wrapper will not see that because it only checks cgscc edges at the very
beginning and end of the pass (manager) it wraps.
This fixes some tests testing this exact behavior in the legacy PM.
Instead of checking WeakTrackingVHs for CallBases at the very beginning
and end of the pass it wraps, check every time
updateCGAndAnalysisManagerForPass() is called.
check-llvm and check-clang with -abort-on-max-devirt-iterations-reached
on by default doesn't show any failures outside of tests specifically
testing it so it doesn't needlessly rerun passes more than necessary.
(The NPM -O2/3 pipeline run the inliner/function simplification pipeline
under a devirtualization repeater pass up to 4 times by default).
http://llvm-compile-time-tracker.com/?config=O3&stat=instructions&remote=aeubanks
shows that 7zip has ~1% compile time regression. I looked at it and saw
that there indeed was devirtualization happening that was not previously
caught, so now it reruns the CGSCC pipeline on some SCCs, which is WAI.
The initial land assumed CallBase WeakTrackingVHs would always be
CallBases, but they can be RAUW'd with undef.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D89587
Arthur Eubanks [Tue, 20 Oct 2020 17:57:02 +0000 (10:57 -0700)]
[FunctionAttrs][NPM] Fix handling of convergent
The legacy pass didn't properly detect indirect calls.
We can still remove the convergent attribute when there are indirect
calls. The LangRef says:
> When it appears on a call/invoke, the convergent attribute indicates
that we should treat the call as though we’re calling a convergent
function. This is particularly useful on indirect calls; without this we
may treat such calls as though the target is non-convergent.
So don't skip handling of convergent when there are unknown calls.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D89826
Philip Reames [Tue, 24 Nov 2020 05:06:46 +0000 (21:06 -0800)]
[LoopVec] Reuse a lambda [NFC]
Minor code refactor to improve readability.
Fangrui Song [Tue, 24 Nov 2020 03:46:52 +0000 (19:46 -0800)]
[CMake] Unify LLVM_LINKER_IS_GOLD -Wl,--gc-sections setting with GNU ld and LLD
John Paul Adrian Glaubitz [Tue, 24 Nov 2020 03:25:36 +0000 (19:25 -0800)]
[Driver] Enable getOSLibDir() lib32 workaround for SPARC on Linux
This fixes the Builtins-sparc-linux testsuite failures on Linux
SPARC which occur because clang cannot find the 32-bit runtime
libraries when -m32 is passed on the command line. The same
workaround is already being used on X86 and PPC.
Also, switch the CHECK-DEBIAN-SPARC tests to use debian_multiarch_tree
as both sparc and sparc64 are using the MultiArch mechanism on modern Debian
systems the same way as x86_64, powerpc64el and others. Thus, switch the
CHECK-DEBIAN-SPARC32 and CHECK-DEBIAN-SPARC64 tests to use the files from
the debian_multiarch_tree directory for the header and linker path tests.
Finally, rename CHECK-DEBIAN-SPARC32 to CHECK-DEBIAN-SPARC to match the naming
scheme of the Debian MultiArch checks for the other Debian architectures.
Reviewed By: MaskRay, phosek
Differential Revision: https://reviews.llvm.org/D90524
Brandon Bergren [Tue, 24 Nov 2020 03:07:21 +0000 (19:07 -0800)]
[libunwind] Multiple preprocessor fixes on PowerPC*
* Remove misnamed `PPC64_HAS_VMX` in preference of directly checking `defined(__VSX__)`.
libunwind was using "VMX" to mean "VSX". "VMX" is just another name for Altivec, while "VSX" is the vector-scalar extensions first used in POWER7. Exposing a "PPC64_HAS_VMX" define was misleading and incorrect.
* Add `defined(__ALTIVEC__)` guards around vector register operations to fix non-altivec CPUS such as the e5500.
When compiling for certain Book-E processors such as the e5500, we want to skip vector save/restore, as the Altivec registers are illegal on non-Altivec implementations.
* Add `!defined(__NO_FPRS__)` guards around traditional floating-point save/restore.
When compiling for powerpcspe, we cannot access floating point registers, as there aren't any. (The SPE on e500v2 is a 64-bit extension of the GPRs, and it doesn't have the normal floating-point registers at all.)
This fixes building for powerpcspe, although no actual handling for SPE save/restore is written yet.
Reviewed By: MaskRay, #libunwind, compnerd
Differential Revision: https://reviews.llvm.org/D91906
Nico Weber [Tue, 24 Nov 2020 02:28:51 +0000 (21:28 -0500)]
[gn build] modify hotfix in
17df195f705cef76a9 to work with all enabled targets
llvm-config output with the gn build is just good enough to make
tests pass, but llvm-config wants all .a files it knows about to
actually exist. So let it know about fewer .a files that don't
exist if not all targets are enabled.
Muhammad Omair Javaid [Tue, 24 Nov 2020 02:05:02 +0000 (07:05 +0500)]
[LLDB] Fix typo in RegisterContextPOSIXProcessMonitor_arm64
This patch fixes a minor typo in RegisterContextPOSIXProcessMonitor_arm64
constructor where memset target was wrongly specified as m_fpr instead of
m_gpr_arm64.
Jonas Devlieghere [Mon, 23 Nov 2020 23:10:46 +0000 (15:10 -0800)]
[lldb] Add platform select to TestProcessConnect.py
Extend TestProcessConnect to cover the scenario fixed by
6c0cd5676e0a0feaf836e0399023a6e21224467b. This replaces
command-process-connect.test which would fail if port 4321
was open.
LLVM GN Syncbot [Tue, 24 Nov 2020 01:44:50 +0000 (01:44 +0000)]
[gn build] Port
3e6e6a2db67
Nathan Ridge [Mon, 23 Nov 2020 02:47:07 +0000 (21:47 -0500)]
[clangd] Use WorkScheduler.run() in ClangdServer::resolveTypeHierarchy()
Differential Revision: https://reviews.llvm.org/D91941
Nathan Ridge [Mon, 16 Nov 2020 03:45:17 +0000 (22:45 -0500)]
[clangd] Call hierarchy (ClangdLSPServer layer)
Differential Revision: https://reviews.llvm.org/D91124
Nathan Ridge [Mon, 16 Nov 2020 02:36:46 +0000 (21:36 -0500)]
[clangd] Call hierarchy (ClangdServer layer)
Differential Revision: https://reviews.llvm.org/D91123
Nathan Ridge [Mon, 16 Nov 2020 03:59:10 +0000 (22:59 -0500)]
[clangd] Call hierarchy (XRefs layer, incoming calls)
Support for outgoing calls is left for a future change.
Differential Revision: https://reviews.llvm.org/D91122
Richard Smith [Tue, 24 Nov 2020 01:20:19 +0000 (17:20 -0800)]
Fix signed integer overflow bug that's causing test failures with UBSan.