Simon Pilgrim [Tue, 17 Nov 2020 17:59:05 +0000 (17:59 +0000)]
[X86] segmented-stacks tests - replace X32 check prefix with X86 for non-gnux32 tests. NFC.
Only use X32 for the gnux32 triples in the tests
Simon Pilgrim [Tue, 17 Nov 2020 17:56:09 +0000 (17:56 +0000)]
[X86] select-of-fp-constants.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 17:50:36 +0000 (17:50 +0000)]
[InstCombine] visitAnd - remove unnecessary Value *X, *Y shadow variables. NFCI.
Fixes a number of Wshadow warnings.
Nico Weber [Tue, 17 Nov 2020 17:15:42 +0000 (12:15 -0500)]
lld: Add --color-diagnostic to MachO port, harmonize others
This adds `--[no-]color-diagnostics[=auto,never,always]` to
the MachO port and harmonizes the flag in the other ports:
- Consistently use MetaVarName
- Consistently document the non-eq version as alias of the eq version
- Use B<> in the ports that have it (no-op, shorter)
- Fix oversight in COFF port that made the --no flag have the wrong
prefix
Differential Revision: https://reviews.llvm.org/D91640
Christian Sigg [Tue, 17 Nov 2020 17:00:55 +0000 (18:00 +0100)]
[mlir] Simplify std.alloc lowering to LLVM.
std.alloc only supports memrefs with identity layout, which means we can simplify the lowering to LLVM and compute strides only from (static and dynamic) sizes.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D91549
Stephen Kelly [Tue, 17 Nov 2020 17:48:14 +0000 (17:48 +0000)]
Comment out new test while I figure out what is wrong with it
Simon Pilgrim [Tue, 17 Nov 2020 17:22:36 +0000 (17:22 +0000)]
[InstCombine] visitAnd - use m_SpecificInt instead of m_APInt + comparison. NFCI.
m_SpecificInt has the same 'no undef element' behaviour as m_APInt so no change there, and anyway we have test coverage for undef elements in the fold.
Noticed while fixing a Wshadow warning about shadow Value *X, *Y variables.
Simon Pilgrim [Tue, 17 Nov 2020 16:51:24 +0000 (16:51 +0000)]
[X86] rot16.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 16:50:08 +0000 (16:50 +0000)]
[X86] vec_fabs.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Wei Wang [Fri, 13 Nov 2020 19:17:47 +0000 (11:17 -0800)]
[BPI] Look through bitcasts in calcZeroHeuristic
Constant hoisting may hide the constant value behind bitcast for And's
operand. Track down the constant to make the BFI result consistent
regardless of hoisting.
Differential Revision: https://reviews.llvm.org/D91450
Alessandro Arzilli [Tue, 17 Nov 2020 17:26:20 +0000 (09:26 -0800)]
[debugserver] Add option to propagate SIGSEGV to target process
Adds a command line option that makes debugserver propagate the SIGSEGV
signal to the target process.
Motivation: I'm one of the maintainers of Delve [1] a debugger for Go.
We use debugserver as our backend on macOS and one of the most often
reported bugs is that, on macOS, we don't propagate SIGSEGV back to the
target process [2]. Sometimes some programs will actually cause a
SIGSEGV, by design, and then handle it. Those programs can not be
debugged at all.
Since catching signals isn't very important for a Go debugger I'd much
rather have a command line option in debugserver that causes it to let
SIGSEGV go directly to the target process.
[1] https://github.com/go-delve/delve/
[2] https://github.com/go-delve/delve/issues/852
Differential revision: https://reviews.llvm.org/D89315
Rahul Joshi [Tue, 17 Nov 2020 01:37:19 +0000 (17:37 -0800)]
[NFC] Add unit tests for printing/parsing of variadic operands and results.
Differential Revision: https://reviews.llvm.org/D91557
Sanjay Patel [Tue, 17 Nov 2020 17:03:55 +0000 (12:03 -0500)]
[InstCombine] remove scalar constraint for mask-of-add fold
https://rise4fun.com/Alive/V6fP
Name: add with low mask
Pre: (C1 & (-1 u>> countLeadingZeros(C2))) == 0
%a = add i8 %x, C1
%r = and i8 %a, C2
=>
%r = and i8 %x, C2
Sanjay Patel [Tue, 17 Nov 2020 17:02:27 +0000 (12:02 -0500)]
[InstCombine] add vector test for mask of add; NFC
Sanjay Patel [Tue, 17 Nov 2020 16:46:34 +0000 (11:46 -0500)]
[InstCombine] relax constraints on mask-of-add
There are 2 changes:
1. Remove the unnecessary one-use check.
2. Remove the unnecessary power-of-2 check.
https://rise4fun.com/Alive/V6fP
Name: add with low mask
Pre: (C1 & (-1 u>> countLeadingZeros(C2))) == 0
%a = add i8 %x, C1
%r = and i8 %a, C2
=>
%r = and i8 %x, C2
Sanjay Patel [Tue, 17 Nov 2020 16:28:33 +0000 (11:28 -0500)]
[InstCombine] add tests for masked add; NFC
ergawy [Tue, 17 Nov 2020 16:45:32 +0000 (11:45 -0500)]
[MLIR][SPIRV] Rename `spv._address_of` to `spv.mlir.addressof`
This commit does the renaming mentioned in the title in order to bring
`spv` dialect closer to the MLIR naming conventions.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D91609
Fangrui Song [Tue, 17 Nov 2020 17:11:19 +0000 (09:11 -0800)]
[ELF] --gc-sections: collect unused .gcc_except_table in section groups and associated text sections
`try ... catch` in an inline function produces `.gcc_except_table.*` in a COMDAT
group with GCC or newer Clang (since D83655). For --gc-sections, currently we
scan `.eh_frame` pieces and mark liveness of such a `.gcc_except_table.*` and
then the associated `.text.*` (if a member in a section group is retained, the
others should be retained as well).
Essentially all `.text.*` and `.gcc_except_table.*` compiled from inline
functions with `try ... catch` cannot be discarded by the imprecise
--gc-sections. Compared with the state before D83655, the output
`.gcc_except_table` is smaller (non-prevailing copies in COMDAT groups can now
be discarded) but `.text` may be larger, i.e. size regression.
This patch teaches the .eh_frame piece scanning code to not mark
`.gcc_except_table` in a section group, thus allow unused `.text.*` and
`.gcc_except_table.*` in a section group to be discarded.
Note, non-group `.gcc_except_table` can still not be discarded. That is the status quo.
Reviewed By: grimar, echristo
Differential Revision: https://reviews.llvm.org/D91579
Fangrui Song [Tue, 17 Nov 2020 17:07:38 +0000 (09:07 -0800)]
[ARC] Correct ARCInstPrinter::getMnemonic after D90039
Nikita Popov [Thu, 12 Nov 2020 20:51:32 +0000 (21:51 +0100)]
[BasicAA] Make alias GEP positive offset handling symmetric
aliasGEP() currently implements some special handling for the case
where all variable offsets are positive, in which case the constant
offset can be taken as the minimal offset. However, it does not
perform the same handling for the all-negative case. This means that
the alias-analysis result between two GEPs is asymmetric:
If GEP1 - GEP2 is all-positive, then GEP2 - GEP1 is all-negative,
and the first will result in NoAlias, while the second will result
in MayAlias.
Apart from producing sub-optimal results for one order, this also
violates our caching assumption. In particular, if BatchAA is used,
the cached result depends on the order of the GEPs in the first query.
This results in an inconsistency in BatchAA and AA results, which
is how I noticed this issue in the first place.
Differential Revision: https://reviews.llvm.org/D91383
Alex Zinenko [Tue, 17 Nov 2020 13:17:22 +0000 (14:17 +0100)]
[mlir] Add a _get_default_loc_context utility to Python bindings
This utility function is helpful for dialect-specific builders that need
to access the context through location, and the location itself may be
either provided as an argument or expected to be recovered from the
implicit location stack.
Reviewed By: stellaraccident
Differential Revision: https://reviews.llvm.org/D91623
Stephen Kelly [Thu, 5 Nov 2020 18:42:21 +0000 (18:42 +0000)]
[AST] Update matchers to be traverse-aware
Don't match Stmt or Decl nodes not spelled in the source when using
TK_IgnoreUnlessSpelledInSource. This prevents accidental modification
of source code at incorrect locations.
Differential Revision: https://reviews.llvm.org/D90984
Stephen Kelly [Wed, 4 Nov 2020 22:57:42 +0000 (22:57 +0000)]
[AST] Ignore implicit nodes in IgnoreUnlessSpelledInSource mode
Update the ASTNodeTraverser to dump only nodes spelled in source. There
are only a few which need to be handled, but Decl nodes for which
isImplicit() is true are handled together.
Update the RAV instances used in ASTMatchFinder to ignore the nodes too.
As with handling of template instantiations, it is necessary to allow
the RAV to process the implicit nodes because they need to be visitable
before the first traverse() matcher is encountered. An exception to
this is in the MatchChildASTVisitor, because we sometimes wish to make a
node matchable but make its children not-matchable. This is the case
for defaulted CXXMethodDecls for example.
Extend TransformerTests to illustrate the kinds of problems that can
arise when performing source code rewriting due to matching implicit
nodes.
This change accounts for handling nodes not spelled in source when using
direct matching of nodes, and when using the has() and hasDescendant()
matchers. Other matchers such as
cxxRecordDecl(hasMethod(cxxMethodDecl())) still succeed for
compiler-generated methods for example after this change. Updating the
implementations of hasMethod() and other matchers is for a follow-up
patch.
Differential Revision: https://reviews.llvm.org/D90982
Stephen Kelly [Tue, 17 Nov 2020 15:52:46 +0000 (15:52 +0000)]
Move ASTMatchFinder definition so it can be accessed from other functions
Stephen Kelly [Tue, 17 Nov 2020 15:42:03 +0000 (15:42 +0000)]
Rename API to not be constrained to template instantiations
A follow-up commit will use this method to ignore implcit AST nodes.
Stephen Kelly [Tue, 17 Nov 2020 15:38:34 +0000 (15:38 +0000)]
Invert accessor for checking traversal mode
For now, there is a traversal mode which ignores implicit casts and
parenthesis. Invert the check in order to not obviously break that.
Simon Pilgrim [Tue, 17 Nov 2020 16:27:32 +0000 (16:27 +0000)]
[PPC] Fix dead store value clang static analyzer warning. NFCI.
Simplify the SplatBits 2-byte -> 4-byte 'splat'.
Christian Sigg [Tue, 17 Nov 2020 08:31:47 +0000 (09:31 +0100)]
[mlir] Remove unused ConvertToLLVMPattern::linearizeSubscripts().
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D91594
Florian Hahn [Tue, 17 Nov 2020 09:46:22 +0000 (09:46 +0000)]
[VPlan] Add VPDef class.
This patch introduces a new VPDef class, which can be used to
manage VPValues defined by recipes/VPInstructions.
The idea here is to mirror VPUser for values defined by a recipe. A
VPDef can produce either zero (e.g. a store recipe), one (most recipes)
or multiple (VPInterleaveRecipe) result VPValues.
To traverse the def-use chain from a VPDef to its users, one has to
traverse the users of all values defined by a VPDef.
VPValues now contain a pointer to their corresponding VPDef, if one
exists. To traverse the def-use chain upwards from a VPValue, we first
need to check if the VPValue is defined by a VPDef. If it does not have
a VPDef, this means we have a VPValue that is not directly defined
iniside the plan and we are done.
If we have a VPDef, it is defined inside the region by a recipe, which
is a VPUser, and the upwards def-use chain traversal continues by
traversing all its operands.
Note that we need to add an additional field to to VPVAlue to link them
to their defs. The space increase is going to be offset by being able to
remove the SubclassID field in future patches.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D90558
Simon Pilgrim [Tue, 17 Nov 2020 16:05:03 +0000 (16:05 +0000)]
[IR] ShuffleVectorInst::isIdentityWithPadding - bail on non-fixed-type vector shuffles.
Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=27416
Simon Pilgrim [Tue, 17 Nov 2020 14:26:53 +0000 (14:26 +0000)]
[X86] nontemporal.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 14:16:44 +0000 (14:16 +0000)]
[X86] subvector-broadcast.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Peyton, Jonathan L [Tue, 17 Nov 2020 16:14:36 +0000 (10:14 -0600)]
[OpenMP] NFC: remove tabs in message catalog file
Andy Wingo [Tue, 17 Nov 2020 16:03:50 +0000 (08:03 -0800)]
[WebAssembly] Fix parsing of linking section for named global imports
Differential Revision: https://reviews.llvm.org/D91635
Peyton, Jonathan L [Thu, 5 Nov 2020 18:34:39 +0000 (12:34 -0600)]
[OpenMP][stats] reset serial state when re-entering serial region
Differential Revision: https://reviews.llvm.org/D90867
Matt Arsenault [Tue, 17 Nov 2020 15:47:43 +0000 (10:47 -0500)]
Linker: Fix linking of byref types
This wasn't properly remapping the type like with the other
attributes, so this would end up hitting a verifier error after
linking different modules using byref.
Jay Foad [Mon, 16 Nov 2020 16:08:25 +0000 (16:08 +0000)]
[AMDGPU] Precommit more vccz workaround tests
Anton Afanasyev [Tue, 17 Nov 2020 15:46:31 +0000 (18:46 +0300)]
[SLPVectorizer] Fix assert
Ben Shi [Tue, 17 Nov 2020 15:28:41 +0000 (23:28 +0800)]
[clang][AVR] Improve avr-ld command line options
Reviewed By: dylanmckay
Differential Revision: https://reviews.llvm.org/D88410
Alexey Bataev [Mon, 16 Nov 2020 17:43:53 +0000 (09:43 -0800)]
[OPENMP]Fix PR48076: mapping of data member pointer.
If the data member pointer is mapped, the compiler tries to optimize the
mapping of such data by discarding explicit mapping flags and trying to
emit combined data instead. In some cases, this optimization is not
quite correctly implemented and it leads to a program crash at the
runtime. Instead, if the data member is mapped, just emit it as is and
do not emit combined mapping flags for it.
Differential Revision: https://reviews.llvm.org/D91552
Florian Hahn [Tue, 17 Nov 2020 14:24:37 +0000 (14:24 +0000)]
[Matrix] Add inline assembly test case.
This patch adds a new test cases which uses a matrix value as memory
inline assembly argument. Currently the pointer element type does not
match the vector type.
Anton Afanasyev [Fri, 30 Oct 2020 07:22:50 +0000 (10:22 +0300)]
[SLP] Make SLPVectorizer to use `llvm.masked.gather` intrinsic
For the scattered operands of load instructions it makes sense
to use gathering load intrinsic, which can lower to native instruction
for X86/AVX512 and ARM/SVE. This also enables building
vectorization tree with entries containing scattered operands.
The next step is to add scattered store.
Fixes PR47629 and PR47623
Differential Revision: https://reviews.llvm.org/D90445
Andy Wingo [Tue, 17 Nov 2020 15:09:30 +0000 (07:09 -0800)]
[WebAssembly][MC] Remove useless overrides in MCWasmStreamer
Differential Revision: https://reviews.llvm.org/D91604
Joe Ellis [Tue, 17 Nov 2020 14:34:31 +0000 (14:34 +0000)]
[AArch64][SVE] Support implicit lax vector conversions for SVE types
Lax vector conversions was behaving incorrectly for implicit casts
between scalable and fixed-length vector types. For example, this:
#include <arm_sve.h>
#define N __ARM_FEATURE_SVE_BITS
#define FIXED_ATTR __attribute__((arm_sve_vector_bits(N)))
typedef svfloat32_t fixed_float32_t FIXED_ATTR;
void allowed_depending() {
fixed_float32_t fs32;
svfloat64_t s64;
fs32 = s64;
}
... would fail because the vectors have differing lane sizes. This patch
implements the correct behaviour for
-flax-vector-conversions={none,all,integer}. Specifically:
- -flax-vector-conversions=none prevents all lax vector conversions
between scalable and fixed-sized vectors.
- -flax-vector-conversions=integer allows lax vector conversions between
scalable and fixed-size vectors whose element types are integers.
- -flax-vector-conversions=all allows all lax vector conversions between
scalable and fixed-size vectors (including those with floating point
element types).
The implicit conversions are implemented as bitcasts.
Reviewed By: fpetrogalli
Differential Revision: https://reviews.llvm.org/D91067
Paul C. Anagnostopoulos [Sat, 14 Nov 2020 18:06:58 +0000 (13:06 -0500)]
[TableGen] Clean up Target .td include files
Differential Revision: https://reviews.llvm.org/D91483
Benjamin Kramer [Tue, 17 Nov 2020 14:42:45 +0000 (15:42 +0100)]
Andrzej Warzynski [Tue, 17 Nov 2020 14:32:39 +0000 (14:32 +0000)]
[flang][nfc] Add missing dependency in CMake
This missing dependency has been causing the Flang buildbots
(with BUILD_SHARED_LIBS set to ON) to fail:
* http://lab.llvm.org:8011/#/builders/66/builds/542
* http://lab.llvm.org:8011/#/builders/33/builds/764
This missing dependency was exposed by this change:
* https://reviews.llvm.org/D91461
This change is fine - the root cause of the failing builds is the
missing dependency.
Florian Hahn [Wed, 5 Aug 2020 14:24:54 +0000 (15:24 +0100)]
[ConstraintElimination] Add support for And.
When processing conditional branches, if the condition is an AND of 2 compares
and the true successor only has the current block as predecessor, queue both
conditions for the true successor.
Kadir Cetinkaya [Tue, 17 Nov 2020 09:28:52 +0000 (10:28 +0100)]
[clang-tidy] Make clang-format and include-order-check coherent
LLVM style puts both gtest and gmock to the end of the include list.
But llvm-include-order-check was only moving gtest headers to the end, resulting
in a false tidy-warning.
Differential Revision: https://reviews.llvm.org/D91602
Erich Keane [Mon, 16 Nov 2020 23:26:22 +0000 (15:26 -0800)]
Update 'note-candiate' functions to skip lambda-conversion-op-overloads
In the wake of https://reviews.llvm.org/D89559, we discovered that a
couple of tests (the ones modified below to have additional triple
versions) would fail on Win32, for 1 of two reasons. We seem to not
have a win32 buildbot anymore, so the triple is to make sure this
doesn't get broken in the future.
First, two of the three 'note-candidate' functions weren't appropriately
skipping the remaining conversion functions.
Second, in 1 situation (note surrogate candidates) we actually print the
type of the conversion operator. The two tests that ran into that
needed updating to make sure it printed the proper one in the win32
case.
Sander de Smalen [Thu, 12 Nov 2020 12:33:36 +0000 (12:33 +0000)]
Reland [LoopVectorizer] NFCI: Calculate register usage based on TLI.getTypeLegalizationCost.
This relands https://reviews.llvm.org/D91059 and reverts commit
30fded75b48bcbc034120154a57a00c7f3d07e06.
GetRegUsage now returns 0 when Ty is not a valid vector element type.
Kazushi (Jam) Marukawa [Tue, 17 Nov 2020 13:38:49 +0000 (22:38 +0900)]
[VE] Implement JumpTable
Implement JumpTable to make BRIND work on VE. Update an existing
br_jt regression test also.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91582
Stephan Herhut [Tue, 17 Nov 2020 12:59:26 +0000 (13:59 +0100)]
[mlir][std] Canonicalize extract_element(tensor_cast).
Canonicalize extract_element(tensor_cast(v)) to just extract_element(v).
Differential Revision: https://reviews.llvm.org/D91621
Stephan Herhut [Tue, 17 Nov 2020 12:39:08 +0000 (13:39 +0100)]
[mlir][std] Fold dim(dynamic_tensor_from_elements, %cst)
The shape of the result of a dynamic_tensor_from_elements is defined via its
result type and operands. We already fold dim operations when they reference
one of the statically sized dimensions. Now, also fold dim on the dynamically
sized dimensions by picking the corresponding operand.
Differential Revision: https://reviews.llvm.org/D91616
Stephan Herhut [Tue, 17 Nov 2020 12:19:57 +0000 (13:19 +0100)]
[mlir][linalg] Allow tensor_to_memref in dependence analysis
This enables the use of fusion on buffers in partially lowered
programs.
Differential Revision: https://reviews.llvm.org/D91613
Kazushi (Jam) Marukawa [Tue, 17 Nov 2020 10:28:49 +0000 (19:28 +0900)]
[VE] Correct getMnemonic
https://reviews.llvm.org/D90039 breaks VE backend. So, fix it.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91619
Alex Zinenko [Sun, 15 Nov 2020 16:49:37 +0000 (17:49 +0100)]
[mlir] Allow for using interface class name in ODS interface definitions
It may be necessary for interface methods to process or return variables with
the interface class type, in particular for attribute and type interfaces that
can return modified attributes and types that implement the same interface.
However, the code generated by ODS in this case would not compile because the
signature (and the body if provided) appear in the definition of the Model
class and before the interface class, which derives from the Model. Change the ODS
interface method generator to emit only method declarations in the Model class
itself, and emit method definitions after the interface class. Mark as "inline"
since their definitions are still emitted in the header and are no longer
implicitly inline. Add a forward declaration of the interface class before the
Concept+Model classes to make the class name usable in declarations.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D91499
Alex Zinenko [Tue, 17 Nov 2020 13:15:11 +0000 (14:15 +0100)]
[mlir] Fix Python tests after "module_terminator" migrated to ODS
The "module_terminator" op now has a custom syntax and therefore is
printed without quotes. Adapt Python tests to check for this syntax.
Nathan James [Tue, 17 Nov 2020 13:02:58 +0000 (13:02 +0000)]
[clang][NFC] Use SmallString instead of SmallVector<char
Simplifies code in some places and is more explicit about what is being used.
No additional includes were added here so no impact on compile time.
Simon Pilgrim [Tue, 17 Nov 2020 12:39:08 +0000 (12:39 +0000)]
[X86] vector-sext.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 12:37:59 +0000 (12:37 +0000)]
[X86] insert-into-constant-vector.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 12:35:29 +0000 (12:35 +0000)]
[X86] avx-intel-ocl.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 12:34:10 +0000 (12:34 +0000)]
[X86] 2011-10-19-widen_vselect.ll - replace X32 check prefix with X86. NFC.
We typically use X32 for gnux32 triples
Simon Pilgrim [Tue, 17 Nov 2020 12:22:33 +0000 (12:22 +0000)]
[clang][driver] Fix definition/declaration argument name mismatches. NFCI.
AddCXXStdlibLibArgs args were using the names for the clang equivalent methods.
Silences cppcheck warnings.
Muhammad Omair Javaid [Tue, 17 Nov 2020 12:15:38 +0000 (17:15 +0500)]
[LLDB] Fix SVE reginfo for sequential offset in g packet
This moves in the direction of our effort to synchronize register descriptions
between LLDB and GDB xml description. We want to able to send registers in a
way that their offset fields can be re-constructed based on register sizes
in the increasing order of register number.
In context to Arm64 SVE, FPCR and FPSR are same registers in FPU regset and
SVE regset. Previously FPSR/FPCR offset was set at the end of SVE data
because Linux ptrace data placed FPCR and FPSR at the end of SVE register set.
Considering interoperability with other stubs like QEMU and that g packets
should generate register data in increasing order of register numbers. We
have to move FPCR/FPSR offset up to its original location according to
register numbering scheme of ARM64 registers with SVE registers included.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D90741
Luke Drummond [Wed, 4 Nov 2020 12:10:25 +0000 (12:10 +0000)]
Escape command line arguments in backtraces
A common routine is to have the compiler crash, and attempt to rerun the
cc1 command-line by copying and pasting the arguments printed by
`llvm::Support::PrettyStackProgram::print`. However, these arguments are
not quoted or escaped which means they must be manually edited before
working correctly. This patch ensures that shell-unfriendly characters
are C-escaped, and arguments with spaces are double-quoted reducing the
frustration of running cc1 inside a debugger.
As the quoting is C, this is "best effort for most shells", but should
be fine for at least bash, zsh, csh, and cmd.exe.
Reviewed by: jhenderson
Differential Revision: https://reviews.llvm.org/D90759
Muhammad Omair Javaid [Tue, 17 Nov 2020 12:11:42 +0000 (17:11 +0500)]
[LLDB] Update SVE Z reg info to remove invalidate regs
In our recent discussion we are aiming to make LLDB registers exchange minimum
possible information in qRegisterInfo or XMl register descriptions.
For SVE registers, Z registers are catagorized as primary registers and should
not have any infomration about any pseudo registers. All pseudo registers
should have the information on which primary register they belong to.
This patch removes invalidate_regs list from Z registers and will mitigate its
impact on SVE resize patch in a follow up update.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D91057
Florian Hahn [Tue, 17 Nov 2020 11:53:01 +0000 (11:53 +0000)]
[AsmPrinter] Use getMnemonic for instruction-mix remark.
This patch uses the new `getMnemonic` helper from D90039
to display mnemonics instead of the internal opcodes.
The main motivation behind using the mnemonics is that they
are more user-friendly and more directly related to the assembly
the users will be presented.
Reviewed By: paquette
Differential Revision: https://reviews.llvm.org/D90040
Sven van Haastregt [Tue, 17 Nov 2020 12:07:40 +0000 (12:07 +0000)]
[OpenCL] Stop opencl-c-base.h leaking extension enabling
opencl-c.h disables all extensions at its end, but opencl-c-base.h
does not, and that causes any inclusion of only opencl-c-base.h to
leave some extensions (such as cl_khr_fp16) enabled. This affects the
-fdeclare-opencl-builtins option for example.
This violates the OpenCL Extension Specification which specifies that
"The initial state of the compiler is as if the directive #pragma
OPENCL EXTENSION all : disable was issued".
Fix by disabling all extensions at the end of opencl-c-base.h and
enable extensions inside opencl.h which relied on opencl-c-base.h
enabling the cl_khr_fp16/64 extensions.
Differential Revision: https://reviews.llvm.org/D91429
Serge Guelton [Tue, 17 Nov 2020 11:15:02 +0000 (06:15 -0500)]
[build] Fix dependencies of LLVMExegesis
Alex Zinenko [Mon, 16 Nov 2020 15:17:03 +0000 (16:17 +0100)]
[mlir] Add basic support for attributes in ODS-generated Python bindings
In ODS, attributes of an operation can be provided as a part of the "arguments"
field, together with operands. Such attributes are accepted by the op builder
and have accessors generated.
Implement similar functionality for ODS-generated op-specific Python bindings:
the `__init__` method now accepts arguments together with operands, in the same
order as in the ODS `arguments` field; the instance properties are introduced
to OpView classes to access the attributes.
This initial implementation accepts and returns instances of the corresponding
attribute class, and not the underlying values since the mapping scheme of the
value types between C++, C and Python is not yet clear. Default-valued
attributes are not supported as that would require Python to be able to parse
C++ literals.
Since attributes in ODS are tightely related to the actual C++ type system,
provide a separate Tablegen file with the mapping between ODS storage type for
attributes (typically, the underlying C++ attribute class), and the
corresponding class name. So far, this might look unnecessary since all names
match exactly, but this is not necessarily the cases for non-standard,
out-of-tree attributes, which may also be placed in non-default namespaces or
Python modules. This also allows out-of-tree users to generate Python bindings
without having to modify the bindings generator itself. Storage type was
preferred over the Tablegen "def" of the attribute class because ODS
essentially encodes attribute _constraints_ rather than classes, e.g. there may
be many Tablegen "def"s in the ODS that correspond to the same attribute type
with additional constraints
The presence of the explicit mapping requires the change in the .td file
structure: instead of just calling the bindings generator directly on the main
ODS file of the dialect, it becomes necessary to create a new file that
includes the main ODS file of the dialect and provides the mapping for
attribute types. Arguably, this approach offers better separability of the
Python bindings in the build system as the main dialect no longer needs to know
that it is being processed by the bindings generator.
Reviewed By: stellaraccident
Differential Revision: https://reviews.llvm.org/D91542
Haojian Wu [Tue, 17 Nov 2020 09:26:44 +0000 (10:26 +0100)]
[clang-tidy] Fix an abseil-redundant-strcat-calls crash on 0-parameter StrCat().
Differential Revision: https://reviews.llvm.org/D91601
Kazushi (Jam) Marukawa [Sat, 14 Nov 2020 20:55:02 +0000 (05:55 +0900)]
[VE] Add vbrd intrinsic instructions
Add vbrd intrinsic instructions and a regression test.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D91569
Ben Shi [Tue, 17 Nov 2020 09:51:58 +0000 (17:51 +0800)]
[AVR] Optimize the 16-bit NEGW pseudo instruction
Reviewed By: dylanmckay
Differential Revision: https://reviews.llvm.org/D88658
Florian Hahn [Tue, 17 Nov 2020 09:38:17 +0000 (09:38 +0000)]
[AsmWriter] Factor out mnemonic generation to accessible getMnemonic.
This patch factors out the part of printInstruction that gets the
mnemonic string for a given MCInst. This is intended to be used
subsequently for the instruction-mix remarks to display the final
mnemonic (D90040).
Unfortunately making `getMnemonic` available to the AsmPrinter
seems to require making it virtual. Not sure if there's a way around
that with the current layering of the AsmPrinters.
Reviewed By: Paul-C-Anagnostopoulos
Differential Revision: https://reviews.llvm.org/D90039
serge-sans-paille [Fri, 13 Nov 2020 21:45:55 +0000 (22:45 +0100)]
[build] normalize components dependencies
Use LINK_COMPONENTS instead of explicit target_link_libraries for components.
This avoids redundancy and potential inconsistencies.
Differential Revision: https://reviews.llvm.org/D91461
Kai Luo [Tue, 17 Nov 2020 09:28:56 +0000 (09:28 +0000)]
[PowerPC] Add negated abs test using llvm.abs intrinsic. NFC.
Haojian Wu [Tue, 17 Nov 2020 09:15:29 +0000 (10:15 +0100)]
[clang-tidy] Verify the fixes in abseil-redundant-strcat-calls test, NFC
River Riddle [Tue, 17 Nov 2020 08:38:10 +0000 (00:38 -0800)]
[mlir][NFC] Remove references to Module.h and Function.h
These includes have been deprecated in favor of BuiltinDialect.h, which contains the definitions of ModuleOp and FuncOp.
Differential Revision: https://reviews.llvm.org/D91572
River Riddle [Tue, 17 Nov 2020 08:37:14 +0000 (00:37 -0800)]
[mlir][IR] Use tablegen for the BuiltinDialect and operations
This has been a long standing TODO, and cleans up a bit of IR/. This will also make it easier to move FuncOp out of IR/ at some point in the future. For now, Module.h and Function.h just forward BuiltinDialect.h. These files will be removed in a followup.
Differential Revision: https://reviews.llvm.org/D91571
Serge Pavlov [Tue, 17 Nov 2020 05:46:13 +0000 (12:46 +0700)]
[Driver] Quote executable in reports generated by -fproc-stat-report
The option -fproc-stat-report=<file> makes driver to generate child
process resource comsumption report. In the report executable name was
not quoted and it made parsing the report more difficult. With this
change the executable name is surrounded by quotation marks.
Yevgeny Rouban [Tue, 17 Nov 2020 07:24:50 +0000 (14:24 +0700)]
[JumpThreading] Fix branch probabilities in DuplicateCondBranchOnPHIIntoPred()
When instructions are cloned from block BB to PredBB in the method
DuplicateCondBranchOnPHIIntoPred() number of successors of PredBB
changes from 1 to number of successors of BB. So we have to copy
branch probabilities from BB to PredBB.
Reviewed By: Kazu Hirata
Differential Revision: https://reviews.llvm.org/D90841
Max Kazantsev [Tue, 17 Nov 2020 07:36:51 +0000 (14:36 +0700)]
[Test] Add one more elaborate test with LCSSA users
Jonas Devlieghere [Tue, 17 Nov 2020 07:26:35 +0000 (23:26 -0800)]
[crashlog] Improve patch-crashlog.py script
Compute the real addresses and offsets for the json crashlog test.
Max Kazantsev [Tue, 17 Nov 2020 07:00:37 +0000 (14:00 +0700)]
[NFC] Collect ext users into vector instead of finding them twice
Ta-Wei Tu [Tue, 17 Nov 2020 06:40:44 +0000 (14:40 +0800)]
[NPM] Add remaining test of -slsr
Add RUN lines for NPM of -slsr tests that require -separate-const-offset-from-gep
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D91588
Ta-Wei Tu [Tue, 17 Nov 2020 06:39:33 +0000 (14:39 +0800)]
[NFC] Add const qualifier to isEmpty()
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D91587
Kazu Hirata [Tue, 17 Nov 2020 06:09:14 +0000 (22:09 -0800)]
[Transforms] Use pred_empty (NFC)
Max Kazantsev [Tue, 17 Nov 2020 05:52:08 +0000 (12:52 +0700)]
[Test] One more widening opportunity test
Kazu Hirata [Tue, 17 Nov 2020 05:45:21 +0000 (21:45 -0800)]
[SanitizerCoverage] Use [&] for lambdas (NFC)
Esme-Yi [Tue, 17 Nov 2020 04:49:19 +0000 (04:49 +0000)]
[NFC][PowerPC] Add testcase of constant-i64.
Arthur Eubanks [Mon, 16 Nov 2020 19:47:50 +0000 (11:47 -0800)]
[Debugify] Skip debugifying on special/immutable passes
With a function pass manager, it would insert debuginfo metadata before
getting to function passes while processing the pass manager, causing
debugify to skip while running the function passes.
Skip special passes + verifier + printing passes. Compared to the legacy
implementation of -debugify-each, this additionally skips verifier
passes. Probably no need to update the legacy version since it will be
obsolete soon.
This fixes 2 instcombine tests using -debugify-each under NPM.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D91558
Mehdi Amini [Tue, 17 Nov 2020 04:08:35 +0000 (04:08 +0000)]
Fix python bindings tests after change in visibility requirement for symbol declarations
Fangrui Song [Tue, 17 Nov 2020 04:02:25 +0000 (20:02 -0800)]
ELFAsmParser: Remove non-SHF_ALLOC or non-executable sections' line info/address ranges contribution for -g
I filed the issue https://sourceware.org/bugzilla/show_bug.cgi?id=26850 ,
which was acknowledged and fixed in GNU binutils 2.36
This patch adds the similar behavior to MC.
Reviewed By: #debug-info, dblaikie
Differential Revision: https://reviews.llvm.org/D91505
Tei Jeong [Tue, 17 Nov 2020 03:56:04 +0000 (03:56 +0000)]
Add const qualifier to Type's utility functions
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D91491
Michael Kruse [Tue, 17 Nov 2020 03:26:37 +0000 (21:26 -0600)]
[Polly] Add override qualifiers.
Silence compiler warnings that suggest to add override qualifiers.
Wang, Pengfei [Tue, 17 Nov 2020 02:17:07 +0000 (10:17 +0800)]
[CodeGen][X86] Remove some unused check-prefixes and regenerate tests.
Lang Hames [Tue, 17 Nov 2020 03:10:53 +0000 (14:10 +1100)]
[ORC] Include config.h in RegisterEHFrames.cpp.
RegisterEHFrames.cpp needs access to the HAVE_REGISTER_FRAME /
HAVE_DEREGISTER_FRAME defines.
rdar://
71458921
Peter Steinfeld [Mon, 16 Nov 2020 20:06:44 +0000 (12:06 -0800)]
[flang] Duplicate names for ac-implied-do variables erroneously cause errors
According to section 19.4, paragraph 5, the scope of an ac-implied-do variable
is the enclosing ac-implied-do. But we were not creating new scopes upon
entry to an ac-implied-do. This was causing error messages to be erroneously
emitted.
I fixed, the code, added a test to array-constr-values.f90, added the test
folding15.f90 and corrected the test symbol05.f90.
Differential Revision: https://reviews.llvm.org/D91560
Yaxun (Sam) Liu [Mon, 16 Nov 2020 15:58:48 +0000 (10:58 -0500)]
[AMDGPU] Add option -munsafe-fp-atomics
Add an option -munsafe-fp-atomics for AMDGPU target.
When enabled, clang adds function attribute "amdgpu-unsafe-fp-atomics"
to any functions for amdgpu target. This allows amdgpu backend to use
unsafe fp atomic instructions in these functions.
Differential Revision: https://reviews.llvm.org/D91546