Justin Hibbits [Mon, 6 Apr 2020 15:08:02 +0000 (10:08 -0500)]
PowerPC: Don't hoist float multiply + add to fused operation on SPE
SPE doesn't have a fmadd instruction, so don't bother hoisting a
multiply and add sequence to this, as it'd become just a library call.
Hoisting happens too late for the CTR usability test to veto using the
CTR in a loop, and results in an assert "Invalid PPC CTR loop!".
Mohammed Nurul Hoque [Wed, 10 Aug 2022 14:59:23 +0000 (10:59 -0400)]
[ConstantFolding] Eliminate atan and atan2 calls
From the opengroup specifications, atan2 may fail if the result
underflows and atan may fail if the argument is subnormal, but
we assume that does not happen and eliminate the calls if we
can constant fold the result at compile-time.
Differential Revision: https://reviews.llvm.org/D127964
Louis Dionne [Mon, 8 Aug 2022 15:53:34 +0000 (11:53 -0400)]
[libc++] Make __libcpp_verbose_abort [[noreturn]]
This will allow using it in functions that are [[noreturn]] themselves.
Differential Revision: https://reviews.llvm.org/D131408
Simon Pilgrim [Wed, 10 Aug 2022 14:25:04 +0000 (15:25 +0100)]
Revert rGa772f775a2ba401e95a0bbe73deb6300f1dc12c0 "[clang-tidy] Support C++14 in bugprone-signal-handler."
This was breaking a number of buildbots: https://lab.llvm.org/buildbot/#/builders/139/builds/26335
David Truby [Wed, 10 Aug 2022 14:02:55 +0000 (15:02 +0100)]
[clang] Correct documentation for NEON and SVE operator support
Previously the language extension documentation didn't mention SVE and
was incomplete in listing the C/C++ operators supported on NEON. This
corrects the documentation to be in line with the implementation.
Simon Pilgrim [Wed, 10 Aug 2022 13:35:50 +0000 (14:35 +0100)]
[X86] Use DAG.getFreeze() to create freeze node. NFC.
Sam Estep [Wed, 10 Aug 2022 14:01:18 +0000 (14:01 +0000)]
[clang][dataflow] Analyze constructor bodies
This patch adds the ability to context-sensitively analyze constructor bodies, by changing `pushCall` to allow both `CallExpr` and `CXXConstructExpr`, and extracting the main context-sensitive logic out of `VisitCallExpr` into a new `transferInlineCall` method which is now also called at the end of `VisitCXXConstructExpr`.
Reviewed By: ymandel, sgatev, xazax.hun
Differential Revision: https://reviews.llvm.org/D131438
Alexander Belyaev [Wed, 10 Aug 2022 13:26:49 +0000 (15:26 +0200)]
[mlir] Extract offsets-sizes-strides computation from `makeTiledShape(s)`.
This change separates computation of the actual parameters of the subset and
the materialization of subview/extract_slice. That way the users can still use
Linalg tiling logic even if they use different operations to materialize the
subsets.
Differential Revision: https://reviews.llvm.org/D131053
Pavel Kosov [Wed, 10 Aug 2022 13:34:52 +0000 (16:34 +0300)]
[LLDB] Add basic floating point ops to IR interpreter
Patch adds support for fadd, fsub, fdiv, fmul and fcmp to IR interpreter.
~~~
OS Laboratory. Huawei RRI. Saint-Petersburg
Reviewed By: clayborg
Differential Revision: https://reviews.llvm.org/D126359
Jake Egan [Wed, 10 Aug 2022 13:31:04 +0000 (09:31 -0400)]
[AIX][tests] XFAIL for system-aix instead
The Clang folding for floating-point sometimes calls out to the host.
Alex Zinenko [Wed, 10 Aug 2022 13:11:26 +0000 (15:11 +0200)]
[mlir] Generate C++ doc comments for interfaces
When emitting the declarations for interface methods defined in ODS,
also emit their descriptions as C++ comments. This makes the
documentation accessible to C++ tooling such as IDEs that offers better
usability than reading it form the .td or the website.
Reviewed By: jpienaar
Differential Revision: https://reviews.llvm.org/D130478
Edd Barrett [Wed, 10 Aug 2022 12:13:28 +0000 (13:13 +0100)]
Migrate llvm.experimental.patchpoint() to ptr.
This intrinsic used a typed pointer for a call target operand. This
change updates the operand to be an opaque pointer and updates all
pointers in all test files that use the intrinsic.
Differential revision: https://reviews.llvm.org/D131261
Simon Pilgrim [Wed, 10 Aug 2022 12:12:04 +0000 (13:12 +0100)]
[DAG] visitFREEZE - generalize freeze(op()) -> op(freeze()) to any number of operands
canCreateUndefOrPoison currently only handles unary ops, but we intend to change that soon - this more closely matches the pushFreezeToPreventPoisonFromPropagating behaviour where the freeze is pushed up to a single operand value, as long as all others are guaranteed not to be poison/undef.
However, pushFreezeToPreventPoisonFromPropagating would freeze all uses of the value - whilst this variant requires the frozen value to be only used in the op - we can look at generalize multiple uses later if the need arises.
Jacques Pienaar [Wed, 10 Aug 2022 12:08:24 +0000 (05:08 -0700)]
[mlir][shape] Update meet to handle all size & shape types
Also tighten up return type inference & compatibility functions.
Differential Revision: https://reviews.llvm.org/D130866
Wei Yi Tee [Wed, 3 Aug 2022 14:48:49 +0000 (14:48 +0000)]
[clang][dataflow] Store DeclContext of block being analysed in Environment if available.
Differential Revision: https://reviews.llvm.org/D131065
David Truby [Tue, 2 Aug 2022 11:54:05 +0000 (11:54 +0000)]
[clang][AArch64][SVE] Add unary +/- operators for SVE types
This patch enables the unary promotion and negation operators on
SVE types.
Differential Revision: https://reviews.llvm.org/D130984
Simon Pilgrim [Wed, 10 Aug 2022 10:27:08 +0000 (11:27 +0100)]
[DAG] canCreateUndefOrPoison - add freeze(truncate(x)) -> truncate(freeze(x)) support
David Spickett [Wed, 10 Aug 2022 10:20:15 +0000 (10:20 +0000)]
[LLDB][AArch64][NFC] Fix some clang-format annotations
We don't want the DEFINE_ macros or the array registers
being clang formatted.
RegisterInfos_arm64.h was missing the off annotation and
RegisterInfos_arm64_sve.h needed the off moving to before
the macro definitions.
David Truby [Thu, 4 Aug 2022 12:40:12 +0000 (12:40 +0000)]
[AArch64][SVE] Use SVE for VLS fcopysign for wide vectors
Currently fcopysign for VLS vectors lowers through NEON even when the
vector width is wider than a NEON vector, causing bad codegen as the
vectors are split. This patch causes SVE to be used for these vectors
instead, giving much better codegen on wide VLS vectors.
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D128642
Balázs Kéri [Wed, 10 Aug 2022 09:07:38 +0000 (11:07 +0200)]
[clang-tidy] Support C++14 in bugprone-signal-handler.
Check `bugprone-signal-handler` is improved to check for
C++-specific constructs in signal handlers. This check is
valid until C++17.
Reviewed By: whisperity
Differential Revision: https://reviews.llvm.org/D118996
Alex Bradbury [Wed, 10 Aug 2022 09:50:29 +0000 (10:50 +0100)]
[RISCV] Implement isUsedByReturnOnly TargetLowering hook in order to tailcall more libcalls
Prior to this patch, libcalls inserted by the SelectionDAG legalizer
could never be tailcalled. The eligibility of libcalls for tail calling
is is partly determined by checking TargetLowering::isInTailCallPosition
and comparing the return type of the libcall and the calleer.
isInTailCallPosition in turn calls TargetLowering::isUsedByReturnOnly
(which always returns false if not implemented by the target).
This patch provides a minimal implementation of
TargetLowering::isUsedByReturnOnly - enough to support tail calling
libcalls on hard float ABIs. Soft-float ABIs are left for a follow on
patch. libcall-tail-calls.ll also shows missed opportunities to tail
call integer libcalls, but this is due to issues outside of
the isUsedByReturnOnly hook.
Differential Revision: https://reviews.llvm.org/D131087
Nikita Popov [Wed, 10 Aug 2022 09:38:41 +0000 (11:38 +0200)]
Revert "[Flang] Use find_program() to find clang-tblgen"
This reverts commit
3bba12a81e242359580fa322cf1a9a0a43b461e6.
While this fixes doc generation in standalone flang builds, it
breaks them for combined builds of clang and flang.
Alex Bradbury [Wed, 10 Aug 2022 09:35:31 +0000 (10:35 +0100)]
[X86][NFCI] Remove target-specific branch optimisation that's handled in BranchFolding
This specific optimisation is handled in OptimizeBlock in BranchFolding
so is redundant. As discussed on the review thread, I've verified that
we have test coverage for that optimisation within test/CodeGen/X86 by
disabling the BranchFolding version of this transform after applying
this patch and rerunning the test suite.
Differential Revision: https://reviews.llvm.org/D129204
Alex Bradbury [Wed, 10 Aug 2022 09:34:10 +0000 (10:34 +0100)]
[WebAssembly] Produce error when encountering unlowerable Wasm global accesses
WebAssembly globals are represented as IR globals with the wasm_var
address space (AS1). Prior to this patch, a wasm global load that isn't
lowerable will produce a failure to select, while a wasm global store
will produced incorrect code. This patch ensures we consistently produce
a clear error.
As noted in the test cases, it's conceivable that a frontend or an
optimisation pass could produce similar IR even in the presence of the
semantic restrictions on pointers to Wasm globals in the frontend, which
is a separate problem to address.
Differential Revision: https://reviews.llvm.org/D131387
Phoebe Wang [Wed, 10 Aug 2022 09:26:11 +0000 (17:26 +0800)]
Revert "Reland "[compiler-rt][BF16] Provide __truncsfbf2 only when __bf16 is available""
This reverts commit
0729d00135ccbe1514a49168bbf26986823235c2.
Buildbot https://lab.llvm.org/buildbot/#/builders/19/builds/11989 failed
with "error: ISO C requires a translation unit to contain at least one declaration [-Werror,-Wempty-translation-unit]"
Simon Pilgrim [Wed, 10 Aug 2022 09:26:11 +0000 (10:26 +0100)]
[DAG] Use DAG.getFreeze() to create freeze node. NFC.
jacquesguan [Thu, 4 Aug 2022 09:24:45 +0000 (17:24 +0800)]
[RISCV] Add cost model for fp-mask cast op.
The cost of convert from or to mask vector is different from other cases. We could not use PowDiff to calculate it. This patch set it to 3 as we use 3 instruction to make it.
Differential Revision: https://reviews.llvm.org/D131149
Dominik Adamski [Mon, 8 Aug 2022 09:51:54 +0000 (04:51 -0500)]
[Flang][OpenMP] Fix conversion of nested loops for SIMD directive
Flang was not able to convert simd directive which contains nested
Fortran loops. The nested Fortran loops inside SIMD directive
are modelled as FIR loops and they need to be translated into LLVM
MLIR dialect.
Differential Revision: https://reviews.llvm.org/D131402
Reviewed by: peixin
Signed-off-by: Dominik Adamski <dominik.adamski@amd.com>
David Spickett [Tue, 9 Aug 2022 08:38:17 +0000 (08:38 +0000)]
[LLDB] Add multi value test for const static enum
1438639a2f7eb9e9cba01454d3a9b1b16d179c9a removed a test
that was using undefined behaviour setting a non-typed enum
to a value outside its known range.
That test also checked if we formatted the value properly
when it could contain >1 valid enum value.
I don't think there's anything special about how we format
typed vs non-typed enums so I'm adding a test for ScopedEnum
that will expect to see 2 enum values plus extra.
Reviewed By: labath, Michael137, shafik
Differential Revision: https://reviews.llvm.org/D131472
Nikita Popov [Thu, 21 Jul 2022 07:53:26 +0000 (09:53 +0200)]
[Attributor] Check for noalias call in AAInstanceInfo
The relevant property of allocation functions of interest here is
their uniqueness (in the sense of disjoint provenance), which is
encoded by the noalias return attribute.
Differential Revision: https://reviews.llvm.org/D130225
Nikita Popov [Mon, 8 Aug 2022 10:40:49 +0000 (12:40 +0200)]
[Flang] Use find_program() to find clang-tblgen
Use find_program() to find the clang-tblgen executable. The current
code ends up using docs/clang-tblgen for me...
This is the same way that mlir-tblgen is obtained in:
https://github.com/llvm/llvm-project/blob/
27241435515554bc21105713c9d34cf886c5bced/flang/CMakeLists.txt#L95-L96.
Differential Revision: https://reviews.llvm.org/D131475
Tobias Hieta [Wed, 10 Aug 2022 08:02:12 +0000 (10:02 +0200)]
[workflow] Run release tasks for me (tru) as well.
Reviewed By: tstellar
Differential Revision: https://reviews.llvm.org/D131493
Tobias Hieta [Wed, 10 Aug 2022 08:01:13 +0000 (10:01 +0200)]
[release] Use threaded compression with xz
Use xz -T0 to use your threads
Reviewed By: tstellar
Differential Revision: https://reviews.llvm.org/D131470
Fangrui Song [Wed, 10 Aug 2022 07:32:40 +0000 (00:32 -0700)]
tsan: update Go rules to use -std=c++17
llvm-project has switched to require C++17.
Reviewed By: dvyukov
Differential Revision: https://reviews.llvm.org/D131536
Clement Courbet [Thu, 30 Jun 2022 07:34:20 +0000 (09:34 +0200)]
[clang][transformer] Fix crash on replacement-less ASTEdit.
Given that we provide an EditGenerator edit(ASTEdit), we can't ever be
sure that the user won't give us an empty replacement.
Differential Revision: https://reviews.llvm.org/D128887
Siva Chandra Reddy [Fri, 5 Aug 2022 20:53:46 +0000 (20:53 +0000)]
[libc] Add implementation of pthread_exit and thrd_exit.
Reviewed By: michaelrj
Differential Revision: https://reviews.llvm.org/D131451
Phoebe Wang [Wed, 10 Aug 2022 01:13:27 +0000 (09:13 +0800)]
Reland "[compiler-rt][BF16] Provide __truncsfbf2 only when __bf16 is available"
Disable the build on macOS due to the bot fail.
Reviewed By: bkramer
Differential Revision: https://reviews.llvm.org/D131147
Fangrui Song [Wed, 10 Aug 2022 04:52:08 +0000 (21:52 -0700)]
[ELF] Simplify llvm::enumerate with structured binding. NFC
Fangrui Song [Wed, 10 Aug 2022 04:46:28 +0000 (21:46 -0700)]
[ELF] De-template BitcodeFile::parse. NFC
Med Ismail Bennani [Wed, 10 Aug 2022 04:44:20 +0000 (21:44 -0700)]
Revert "Reland "[compiler-rt][BF16] Provide __truncsfbf2 only when __bf16 is available""
This reverts commit
ab4e5ed441d475ead43bc4ce1cdef842688292f4 since it
causes a build failure on macOS:
https://green.lab.llvm.org/green/job/lldb-cmake/45984/
Fangrui Song [Wed, 10 Aug 2022 04:43:35 +0000 (21:43 -0700)]
[ELF] De-template createBitcodeSymbol. NFC
Med Ismail Bennani [Tue, 9 Aug 2022 23:36:47 +0000 (16:36 -0700)]
[lldb/crashlog] Refactor the CrashLogParser logic
This patch changes the CrashLogParser class to be both the base class
and a Factory for the JSONCrashLogParser & TextCrashLogParser.
That should help remove some code duplication and ensure both class
have a parse method.
Differential Revision: https://reviews.llvm.org/D131085
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Wed, 10 Aug 2022 00:31:31 +0000 (17:31 -0700)]
[lldb/crashlog] Skip null image dsym fetching on interactive mode
Sometimes, it can happen that a crash report has null images in its list
of used binaries. This manifests like such:
```
0x0 - 0xffffffffffffffff ??? (*) <
00000000-0000-0000-0000-
000000000000> ???
```
When fetching debug symbols to symbolicate the crashlog stackframe,
having null images causes `dsymForUUID` to hang for few seconds.
This patch addresses that by skipping null images from being load by the
scripted process.
rdar://
97419487
Differential Revision: https://reviews.llvm.org/D131038
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Wed, 3 Aug 2022 00:29:01 +0000 (17:29 -0700)]
[lldb/crashlog] Add `-s|--skip-status` option to interactive mode
This patch introduces a new option for the interactive crashlog mode,
that will prevent it from dumping the `process status` & `thread backtrace`
output to the debugger console.
This is necessary when lldb in running from an IDE, to prevent flooding
the console with information that should be already present in the UI.
rdar://
96813296
Differential Revision: https://reviews.llvm.org/D131036
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Wed, 3 Aug 2022 23:36:56 +0000 (16:36 -0700)]
[lldb/crashlog] Remove 'process_path' parsing logic
In can happen when creating stackshot crash report that that key is missing.
Moreover, we try to parse that key but don't use it, or need it, since we
fetch images and symbolicate the stackframes using the binaries UUIDs.
This is why this patch removes everything that is related to the
`process_path`/`procPath` parsing.
rdar://
95054188
Differential Revision: https://reviews.llvm.org/D131033
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Thu, 4 Aug 2022 17:34:41 +0000 (10:34 -0700)]
[lldb/crashlog] Update frame regex matcher
This patch updates the regular expression matching stackframes in
crashlog to allow addresses that are 7 characters long and more (vs. 8
characters previously).
It changes the `0x[0-9a-fA-F]{7}[0-9a-fA-F]+` by `0x[0-9a-fA-F]{7,}`.
rdar://
97684839
Differential Revision: https://reviews.llvm.org/D131032
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Wed, 3 Aug 2022 23:45:48 +0000 (16:45 -0700)]
[lldb/crashlog] Surface error using SBCommandReturnObject argument
This patch allows the crashlog script to surface its errors to lldb by
using the provided SBCommandReturnObject argument.
rdar://
95048193
Differential Revision: https://reviews.llvm.org/D129614
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Wed, 3 Aug 2022 21:11:43 +0000 (14:11 -0700)]
[lldb/crashlog] Add '-t|--target' option to interactive mode
This patch introduces a new flag for the interactive crashlog mode, that
allow the user to specify, which target to use to create the scripted
process.
This can be very useful when lldb already have few targets created:
Instead of taking the first one (zeroth index), we will use that flag to
create a new target. If the user didn't provide a target path, we will rely
on the symbolicator to create a targer.If that fails and there are already
some targets loaded in lldb, we use the first one.
rdar://
94682869
Differential Revision: https://reviews.llvm.org/D129611
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
John Demme [Wed, 10 Aug 2022 03:07:33 +0000 (20:07 -0700)]
[MLIR] [Python] Fix the Windows build broken by d747a17
Windows builds require all control paths return. Since we don't have
`llvm_unreachable` in the Python bindings, just return `None`.
John Demme [Wed, 10 Aug 2022 02:37:04 +0000 (19:37 -0700)]
[MLIR] [Python] Fix `Value.owner` to handle BlockArgs
Previously, calling `Value.owner()` would C++ assert in debug builds if
`Value` was a block argument. Additionally, the behavior was just wrong
in release builds. This patch adds support for BlockArg Values.
Phoebe Wang [Wed, 10 Aug 2022 01:13:27 +0000 (09:13 +0800)]
Reland "[compiler-rt][BF16] Provide __truncsfbf2 only when __bf16 is available"
Fix the mising change for truncdfbf2.c
Reviewed By: bkramer
Differential Revision: https://reviews.llvm.org/D131147
Phoebe Wang [Wed, 10 Aug 2022 02:05:12 +0000 (10:05 +0800)]
Revert "[compiler-rt][BF16] Provide __truncsfbf2 only when __bf16 is available"
This reverts commit
0ae2a41d803d715975d710ec289a4bd90ca534e2.
Revert due to buildbot fail.
Vitaly Buka [Wed, 10 Aug 2022 01:44:00 +0000 (18:44 -0700)]
[sanitizer] Let internal symbolizer to use posix_memalign
Phoebe Wang [Wed, 10 Aug 2022 01:13:27 +0000 (09:13 +0800)]
[compiler-rt][BF16] Provide __truncsfbf2 only when __bf16 is available
#56854 shows a backwards compatibility problem when builtins of compiler-rt don't follow ABI. We need to prevent to fall into the trap again for BF16.
Reviewed By: bkramer
Differential Revision: https://reviews.llvm.org/D131147
Freddy Ye [Wed, 10 Aug 2022 00:59:21 +0000 (08:59 +0800)]
[X86][BF16] Enable __bf16 for x86 targets.
X86 psABI has updated to support __bf16 type, the ABI of which is the
same as FP16. See https://discourse.llvm.org/t/patch-add-optional-bfloat16-support/63149
Reviewed By: pengfei
Differential Revision: https://reviews.llvm.org/D130964
Phoebe Wang [Wed, 10 Aug 2022 00:58:41 +0000 (08:58 +0800)]
[X86][BF16] Make backend type bf16 to follow the psABI
X86 psABI has updated to support __bf16 type, the ABI of which is the
same as FP16. See https://discourse.llvm.org/t/patch-add-optional-bfloat16-support/63149
This is an alternative of D129858, which has less code modification and
supports the vector type as well.
Reviewed By: LuoYuanke
Differential Revision: https://reviews.llvm.org/D130832
Fangrui Song [Wed, 10 Aug 2022 00:50:42 +0000 (17:50 -0700)]
[COFF] Change a llvm::StringSet<> to llvm::DenseSet<StringRef>. NFC
The former stores strings and is therefore more expensive.
YingChi Long [Tue, 9 Aug 2022 17:11:41 +0000 (01:11 +0800)]
[clang] add APValue type check in `TryPrintAsStringLiteral`
Fixes https://github.com/llvm/llvm-project/issues/57013
https://reviews.llvm.org/D115031 improved printing of non-type template
parameter args. But checking if the end of Inits is 0 without checking
if APValue is an integer, causes clang to segfault. This patch adds
the code to check the type. (May not be a proper bugfix.)
Reviewed By: aaron.ballman, lichray
Differential Revision: https://reviews.llvm.org/D131466
malavikasamak [Wed, 10 Aug 2022 00:25:19 +0000 (17:25 -0700)]
[analyzer] Fix false positive in use-after-move checker
Differential Revision: https://reviews.llvm.org/D131525
Aart Bik [Tue, 9 Aug 2022 22:48:00 +0000 (15:48 -0700)]
[mlir][sparse] fix doc
Reviewed By: Peiming
Differential Revision: https://reviews.llvm.org/D131527
Adrian Prantl [Tue, 9 Aug 2022 23:35:10 +0000 (16:35 -0700)]
Fix modeline
Adrian Vogelsgesang [Sun, 7 Aug 2022 15:01:04 +0000 (08:01 -0700)]
[libc++] Implement `operator<=>` for `type_index`
Implements part of P1614R2 "The Mothership has Landed"
Differential Revision: https://reviews.llvm.org/D131357
Med Ismail Bennani [Thu, 28 Jul 2022 18:29:49 +0000 (11:29 -0700)]
[lldb/crashlog] Fix interactive crashlog test
This patch should fix the interactive crashlog test by checking in the
binary as a yaml to regeneate the binary with the addresses and offsets
when running the test.
rdar://
93655633
Differential Revision: https://reviews.llvm.org/D129603
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Adrian Vogelsgesang [Thu, 4 Aug 2022 22:43:59 +0000 (15:43 -0700)]
[libc++][NFC] Remove TEST_HAS_NO_SPACESHIP_OPERATOR
The corresponding _LIBCPP_HAS_NO_SPACESHIP_OPERATOR macro was already
removed in commit
c0f87e8382
Differential Revision: https://reviews.llvm.org/D131215
Kazu Hirata [Tue, 9 Aug 2022 23:24:53 +0000 (16:24 -0700)]
[llvm-profdata] Support JSON as as an output-only format
This patch teaches llvm-profdata to output the sample profile in the
JSON format. The new option is intended to be used for research and
development purposes. For example, one can write a Python script to
take a JSON file and analyze how similar different inline instances of
a given function are to each other.
I've chosen JSON because Python can parse it reasonably fast, and it
just takes a couple of lines to read the whole data:
import json
with open ('profile.json') as f:
profile = json.load(f)
Differential Revision: https://reviews.llvm.org/D130944
Adrian Vogelsgesang [Wed, 27 Jul 2022 20:55:58 +0000 (13:55 -0700)]
[libunwind] Use `_dl_find_object` if available
As shown in P2544R0 [1] and the accompanying benchmark [2], the
current unwinding logic does not scale for multi-threaded programs.
This is because `dl_iterate_phdr` takes a global lock.
glibc 2.35 added `_dl_find_object` which directly returns the unwind
info for a given target address. `_dl_find_object` is fully lock-free
and hence allows parallel exception unwinding on multiple threads.
With this commit, libunwind now takes advantage of `_dl_find_object`.
Thereby, this commit improves libunwind's performance on benchmark [2]
for unwinding exception on 20 threads from 1103ms to 78ms.
(measured on Intel Xeon Silver 4114 with 20 physical cores)
[1] https://isocpp.org/files/papers/P2544R0.html
[2] https://github.com/neumannt/exceptionperformance
Detailed performance numbers from the benchmark:
Before:
> Testing unwinding performance: sqrt computation with occasional errors
>
> testing baseline using 1 2 4 8 16 20 threads
> failure rate 0%: 34 35 34 35 35 36
> testing exceptions using 1 2 4 8 16 20 threads
> failure rate 0%: 16 32 33 34 35 36
> failure rate 0.1%: 16 32 34 36 35 36
> failure rate 1%: 20 40 40 43 90 113
> failure rate 10%: 59 92 140 304 880 1103
> [...]
>
> Testing invocation overhead: recursive fib with occasional errors
>
> testing exceptions using 1 2 4 8 16 20 threads
> failure rate 0%: 19 32 37 38 39 36
> failure rate 0.1%: 22 32 40 40 39 34
> failure rate 1%: 20 28 38 39 48 40
> failure rate 10%: 25 39 44 50 92 113
After:
> Testing unwinding performance: sqrt computation with occasional errors
>
> testing baseline using 1 2 4 8 16 20 threads
> failure rate 0%: 19 30 35 38 39 35
> testing baseline using 1 2 4 8 16 20 threads
> failure rate 0%: 32 35 33 34 34 36
> testing exceptions using 1 2 4 8 16 20 threads
> failure rate 0%: 16 35 33 37 35 35
> failure rate 0.1%: 16 32 36 33 34 37
> failure rate 1%: 21 37 39 40 40 41
> failure rate 10%: 72 75 76 80 80 78
> [...]
>
> Testing invocation overhead: recursive fib with occasional errors
>
> testing baseline using 1 2 4 8 16 20 threads
> failure rate 0%: 18 35 37 34 38 37
> testing exceptions using 1 2 4 8 16 20 threads
> failure rate 0%: 19 33 40 40 41 39
> failure rate 0.1%: 21 33 39 38 39 38
> failure rate 1%: 20 36 39 40 41 40
> failure rate 10%: 25 45 41 42 44 43
Differential Revision: https://reviews.llvm.org/D130668
alex-t [Sat, 6 Aug 2022 23:47:35 +0000 (01:47 +0200)]
[AMDGPU] SIFixSGPRCopies refactoring
This change finalizes the series of patches aiming to replace old
strategy of VGPR to SGPR copies loweriong. Following the
https://reviews.llvm.org/D128252 and https://reviews.llvm.org/D130367 code
parts that are no longer used were removed. Pass main loop is no longer used
for the MIR changes but collect information for further analysis. Actual MIR
lowering happens further according the analysys result in the set of separate
functions. Another important change concerns the order of lowering: VGPR to
SGPR copies lowering is done first to have priority on the rest of the MIR
changes.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D131246
Evgenii Stepanov [Tue, 9 Aug 2022 22:31:52 +0000 (15:31 -0700)]
[sanitizer] Use C++17 to build the symbolizer.
Siva Chandra Reddy [Tue, 9 Aug 2022 22:27:46 +0000 (22:27 +0000)]
[libc][NFC] Fix a few compiler warnings.
Philip Reames [Tue, 9 Aug 2022 21:58:26 +0000 (14:58 -0700)]
[RISCV] Split check lines for fpclamptosat_vec test
This is currently exercising scalarization code path; with vectors enabled, we hit a different code path. Explicitly exercise both so that both configurations have testing.
Pengxuan Zheng [Mon, 8 Aug 2022 19:43:30 +0000 (12:43 -0700)]
[ARM] Do not use LOAD_STACK_GUARD with ROPI/RWPI
ROPI/RWPI are not supported with LOAD_STACK_GUARD currently.
Reviewed By: nickdesaulniers, rengolin
Differential Revision: https://reviews.llvm.org/D131427
Jim Ingham [Tue, 9 Aug 2022 21:54:18 +0000 (14:54 -0700)]
@skipIfAsan for the other long test in TestMemoryRegion.py
These two tests were each of them too long when running under ASAN,
so we have to skip both to get a clean ASAN bot run.
Siva Chandra Reddy [Tue, 2 Aug 2022 07:14:12 +0000 (07:14 +0000)]
[libc] Add __cxa_atexit support to the atexit function.
Reviewed By: abrachet
Differential Revision: https://reviews.llvm.org/D131219
Siva Chandra Reddy [Sat, 6 Aug 2022 00:19:51 +0000 (00:19 +0000)]
[libc] Add a utility data structure named FixedVector.
This data structure uses a backing cpp::array object and supports a
vector like push_back API. In comparison with a traditional vector
data structure, it is of a fixed capacity and cannot be resized.
Differential Revision: https://reviews.llvm.org/D131377
Dinar Temirbulatov [Tue, 9 Aug 2022 21:10:17 +0000 (22:10 +0100)]
[AArch64][LoopVectorize] Introduce trip count minimal value threshold to ignore tail-folding.
After D121595 was commited, I noticed regressions assosicated with small trip
count numbersvectorisation by tail folding with scalable vectors. As a solution
for those issues I propose to introduce the minimal trip count threshold value.
Differential Revision: https://reviews.llvm.org/D130755
Jeff Niu [Tue, 9 Aug 2022 20:10:21 +0000 (16:10 -0400)]
[flang] Revert changes (NFC)
Jeff Niu [Tue, 9 Aug 2022 19:58:49 +0000 (15:58 -0400)]
[mlir] Fix build of toyc-ch6 (NFC)
Jim Ingham [Tue, 9 Aug 2022 19:49:43 +0000 (12:49 -0700)]
The memory region tests have been consistently timing on the ASAN
bot. This happens because they are building a long result into
a Python string, and the asan checker is making that very slow.
The last two tests here are both slow, but the 'test_command' is the
really slow one. I'm going to start disabling just that one and see
if that gets the ASAN bots clean.
Haojian Wu [Tue, 9 Aug 2022 19:46:56 +0000 (21:46 +0200)]
[pseudo] Fix a suspicious usage of `sizeof(this)`.
It should be `sizeof(*this)`.
Jeff Niu [Mon, 8 Aug 2022 22:04:05 +0000 (18:04 -0400)]
[mlir] Cleanup DenseArrayAttrBase definition and expose raw API
This patch cleans up the definition of `DenseArrayAttrBase` by relying
more on ODS-generated methods. It also exposes an API for using the raw
data of a dense array, similar to `DenseIntOrFPElementsAttr::getRaw`.
Reviewed By: lattner, mehdi_amini
Differential Revision: https://reviews.llvm.org/D131450
yundiqian [Tue, 9 Aug 2022 19:10:08 +0000 (12:10 -0700)]
fix mlgo regalloc test model generation for tflite
To move from TF C API to TFLite, we found that the argmax op in TFLite does not work for int64 inputs, so cast the int64 inputs to int32 inputs to make TFLite argmax op work
Differential Revision: https://reviews.llvm.org/D131462
Jeff Niu [Tue, 9 Aug 2022 18:40:07 +0000 (14:40 -0400)]
[mlir][LLVMIR] (NFC) Add convenience builders for ConstantOp
And clean up some of the user code
Pengxuan Zheng [Tue, 9 Aug 2022 00:49:08 +0000 (17:49 -0700)]
[LLD][COFF] Identify /GL object files which are inside libraries
With D26647, we can already identify input object files compiled by cl.exe with
/GL. It seems to be helpful to do the same and print an error message for those
object files compiled with /GL but are inside libraries/archives too.
Reviewed By: rnk, thieta
Differential Revision: https://reviews.llvm.org/D131458
Dylan Fleming [Tue, 9 Aug 2022 19:22:16 +0000 (19:22 +0000)]
[Flang][OpenMP] Add support for integer multiplication reduction in worksharing-loop
Adds support for reduction of multiplcation
by extending OpenMP.cpp::genOpenMPReduction()
and altering the identity constant emitted in
OpenMP.cpp::createReductionDelc()
This patch builds D130077 and as such,
only supports reductions for interger types in
worksharping loops.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D130767
Peter Klausler [Tue, 9 Aug 2022 16:01:50 +0000 (09:01 -0700)]
[flang] Extend characterization & checking for procedure bindings
Procedure bindings with explicit interfaces don't work when the
interface is shadowed by a generic interface of the same name,
and can produce spurious semantic error messages. Extend the
characterization and checking code for such things, and the utility
functionns on which they depend, to dig through generics when they
occlude interface-defining subprograms. This is done on demand in
checking code, not once during name resolution, because the
procedures in question may also be forward-referenced.
Differential Revision: https://reviews.llvm.org/D131105
diff --git a/flang/include/flang/Semantics/symbol.h b/flang/include/flang/Semantics/symbol.h
index
e79f8ab6503e..
0b03bf06eb73 100644
--- a/flang/include/flang/Semantics/symbol.h
Aart Bik [Tue, 9 Aug 2022 16:39:58 +0000 (09:39 -0700)]
[mlir][sparse] fix switch statement bug on two binary ops
They appeared at the wrong place in the switch, treating
them as unary op rather than binary op.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D131509
Aart Bik [Tue, 9 Aug 2022 18:00:07 +0000 (11:00 -0700)]
[mlir][sparse] update bibliography of sparse tensor dialect
Reviewed By: Peiming
Differential Revision: https://reviews.llvm.org/D131513
Lei Zhang [Tue, 9 Aug 2022 18:32:22 +0000 (14:32 -0400)]
[mlir][spirv] Use functors for default memory space mappings
This makes it easier to use as a utility function to query the
mappings, including the reverse.
This commit also drops some storage classes that aren't needed
for now.
Reviewed By: kuhar
Differential Revision: https://reviews.llvm.org/D131411
Lei Zhang [Tue, 9 Aug 2022 18:25:38 +0000 (14:25 -0400)]
[mlir][spirv] Detach memory space mapping from type conversion
This commit moves MemRef memory space to SPIR-V storage class
conversion out of the main SPIR-V type converter. Now the mapping
should happen as a prelimiary step before performing the final
conversion to SPIR-V. Flows are expect to write their own memory
space mappings like the `MapMemRefStorageClassPass` to handle
memory space mappings according to their needs.
This is needed because SPIR-V is serving multiple client APIs,
including Vulkan and OpenCL. Different client APIs might want
to use different storage classes for buffers in a particular
memory space, e.g., `StorageBuffer` for Vulkan vs. `CrossWorkgroup`
for OpenCL when converting the default 0 memory space. Hardcoding
a specific mapping makes that hard. While it's possible to embed
selection logic further inside the main type converter, it will
make the main type converter even complicated. So it's better to
separate the concerns, as mapping the memory space is really
concretizing the meaning of those numeric memory spaces in the
particular context of SPIR-V lowering.
Reviewed By: kuhar
Differential Revision: https://reviews.llvm.org/D131410
Philip Reames [Tue, 9 Aug 2022 18:23:09 +0000 (11:23 -0700)]
[RISCV] Refresh two autogened tests to avoid future whitespace diffs [nfc]
Lei Zhang [Tue, 9 Aug 2022 18:15:55 +0000 (14:15 -0400)]
[mlir][spirv] Make MemRef memory space mapping pass more flexible
* Avoid restricting the pass to to builtin module ops. The pass
should be able to run on any region ops.
* Avoid hardcoding func FuncOp when handling functions. Instead,
use the function op interface.
* Assigns the default mapping in the constructor. So for cases
where we are using the pass in a pipeline, we still have a
meaningful default.
Along the way, dropped uncessary unrealized conversion casts and
use full conversion. The pass should be able to convert all sorts
of ops; there is really no need to have such bridages.
Reviewed By: kuhar
Differential Revision: https://reviews.llvm.org/D131409
Jerry Wu [Tue, 9 Aug 2022 18:19:05 +0000 (11:19 -0700)]
[MLIR] Extend vector.gather to accept tensor as base
In addition to memref, accept ranked tensor as the base operand of vector.gather, similar to vector.trasnfer_read.
This will allow us to vectorize noncontiguous tensor.extract into vector.gather. Full discussion can be found here: https://github.com/iree-org/iree/issues/9198
Reviewed By: hanchung, dcaballe
Differential Revision: https://reviews.llvm.org/D130097
Jerry Wu [Tue, 9 Aug 2022 18:12:00 +0000 (11:12 -0700)]
[MLIR] Support lowering n-D arith.index_cast to LLVM
Previously we can only lower arith.index_cast with 1-D vectors to LLVM. This change added the support for n-D vectors.
Reviewed By: ftynse, hanchung
Differential Revision: https://reviews.llvm.org/D129907
Lei Zhang [Tue, 9 Aug 2022 18:03:54 +0000 (14:03 -0400)]
[mlir][spirv] Migrate to use specalized enum attributes
Previously we are using IntegerAttr to back all SPIR-V enum
attributes. Therefore we all such attributes are showed like
IntegerAttr in IRs, which is barely readable and breaks
roundtripability of the IR. This commit changes to use
`EnumAttr` as the base directly so that we can have separate
attribute definitions and better IR printing.
Reviewed By: kuhar
Differential Revision: https://reviews.llvm.org/D131311
Aart Bik [Tue, 9 Aug 2022 17:29:42 +0000 (10:29 -0700)]
[mlir][sparse] improve semi-ring doc
Spell out SparseVector instead of just using SparseVec
Reviewed By: jim22k, bixia
Differential Revision: https://reviews.llvm.org/D131511
Filipp Zhinkin [Tue, 9 Aug 2022 17:55:50 +0000 (20:55 +0300)]
[X86][ARM] Update tests for bitwise logic trees of shifts; NFC
Baseline tests for D131189.
Jorge Gorbe Moya [Tue, 9 Aug 2022 17:44:09 +0000 (10:44 -0700)]
Move FormattersMatchCandidate flags to a struct.
This removes some error-prone repetition in
FormatManager::GetPossibleMatches, where the same three boolean flags
are passed in a row multiple times as arguments to recursive calls to
GetPossibleMatches.
Instead of:
```
// same flags, but with did_strip_typedef set to true.
GetPossibleMatches(..., did_strip_ptr, did_strip_ref, true);
```
we can now say
```
GetPossibleMatches(..., current_flags.WithStrippedTypedef());
```
which hopefully makes the intent clearer, and more readable in case we
add another flag.
Reviewed by: DavidSpickett, labath
Differential Revision: https://reviews.llvm.org/D131459
Archibald Elliott [Mon, 8 Aug 2022 10:44:26 +0000 (11:44 +0100)]
[docs][AArch64] Label Features with Arm ARM Names
This patch adds the names of the Arm Architecture Reference Manual (ARM)
features to the corresponding Subtarget Features in the AArch64 backend
and target parser.
The aim of this is to make it clearer what architectural features a
subtarget feature might enable (so, which features a CPU must provide to
support that subtarget feature), and so make it easier to add new CPUs
in the future.
Differential Revision: https://reviews.llvm.org/D131257
Adrian Prantl [Tue, 9 Aug 2022 17:34:57 +0000 (10:34 -0700)]
LiveDebugValues: Fix another crash related to unreachable blocks
This is a follow-up patch to D130999. In the test, the MIR contains an
unreachable MBB but the code attempts to look it up in MLocs. This
patch fixes this issue by checking for the default-constructed value.
rdar://
97226240
Differential Revision: https://reviews.llvm.org/D131453
Sanjay Patel [Tue, 9 Aug 2022 17:03:23 +0000 (13:03 -0400)]
[InstCombine] fold usub.with.overflow to icmp when there's no use of the math value
https://alive2.llvm.org/ce/z/UE48FH
This is part of solving issue #56926.