platform/upstream/llvm.git
2 years ago[AMDGPU] gfx11 vop3 and inherited vop instructions
Joe Nash [Mon, 16 May 2022 19:19:31 +0000 (15:19 -0400)]
[AMDGPU] gfx11 vop3 and inherited vop instructions

This patch includes MC layer support for VOP3 encoded instructions and generic VOP support
classes.
Some VOP1 and VOP2 instructions which share an encoding with gfx10 and are using
the AssemblerPredicate = isGFX10Plus are also enabled. That predicate
will be changed to isGFX10Only in a later patch.

Patch 15/N for upstreaming of AMDGPU gfx11 architecture.

Depends on D126468

Reviewed By: dp

Differential Revision: https://reviews.llvm.org/D126475

2 years ago[mlir:MultiOpDriver] Don't add ops which are not in the allowed list
Chia-hung Duan [Thu, 2 Jun 2022 18:27:36 +0000 (18:27 +0000)]
[mlir:MultiOpDriver] Don't add ops which are not in the allowed list

In strict mode, only the new inserted operation is allowed to add to the
worklist. Before this change, it would add the users of a replaced op
and it didn't check if the users are allowed to be pushed into the
worklist

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D126899

2 years ago[SLP]Improve shuffles cost estimation where possible.
Alexey Bataev [Thu, 9 Dec 2021 18:34:08 +0000 (10:34 -0800)]
[SLP]Improve shuffles cost estimation where possible.

Improved/fixed cost modeling for shuffles by providing masks, improved
cost model for non-identity insertelements.

Differential Revision: https://reviews.llvm.org/D115462

2 years agoscan-build-py: Change scripts to explicitly require python3
Anders Waldenborg [Mon, 9 May 2022 06:11:34 +0000 (08:11 +0200)]
scan-build-py: Change scripts to explicitly require python3

The "#!" line in all scan-build-py scripts were using just bare
"/usr/bin/python" which according to PEP-0394 can be either python3,
python2 or not exist at all.

E.g in latest debian and ubuntu releases "/usr/bin/python" does not
exist at all by default and user must install python-is-python2 or
python-is-python3 packages to get the bare version less "python"
command.

Until recently (70b06fe8a186 "scan-build-py: Force the opening in utf-8"
changed "libscanbuild") these scripts worked in both python2 and
python3, but now they (rightfully) are python3 only, and broke on
systems where the "python" command means python2.

By changing the "#!" to be "python3" it is not only explicit that the
scripts require python3 it also works on systems where "python" command
is python2 or nonexistent.

Differential Revision: https://reviews.llvm.org/D126804

2 years agoFix a buglet in remove_dots().
Paul Pluzhnikov [Thu, 2 Jun 2022 17:30:24 +0000 (10:30 -0700)]
Fix a buglet in remove_dots().

The function promises to canonicalize the path, but neglected to do so
for the root component.

For example, calling remove_dots("/tmp/foo.c", Style::windows_backslash)
resulted in "/tmp\foo.c". Now it produces "\tmp\foo.c".

Also fix FIXME in the corresponding test.

Reviewed By: rnk

Differential Revision: https://reviews.llvm.org/D126412

2 years ago[AMDGPU] gfx11 ds instructions
Joe Nash [Thu, 12 May 2022 17:32:19 +0000 (13:32 -0400)]
[AMDGPU] gfx11 ds instructions

MC layer support for ds instructions

Contributors:
Piotr Sobczak <Piotr.Sobczak@amd.com>

Patch 14/N for upstreaming of AMDGPU gfx11 architecture.

Depends on D126463

Reviewed By: arsenm, #amdgpu

Differential Revision: https://reviews.llvm.org/D126468

2 years ago[PS5] Make passing unions in registers match PS4 ABI
Paul Robinson [Thu, 2 Jun 2022 18:00:32 +0000 (11:00 -0700)]
[PS5] Make passing unions in registers match PS4 ABI

2 years ago[PS5] Classify __m64 as integer, matching PS4 ABI
Paul Robinson [Thu, 2 Jun 2022 17:53:40 +0000 (10:53 -0700)]
[PS5] Classify __m64 as integer, matching PS4 ABI

2 years ago[llvm][analyzer][NFC] Introduce SFINAE for specializing FoldingSetTraits
Balazs Benics [Thu, 2 Jun 2022 17:46:38 +0000 (19:46 +0200)]
[llvm][analyzer][NFC] Introduce SFINAE for specializing FoldingSetTraits

Reviewed By: martong

Differential Revision: https://reviews.llvm.org/D126803

2 years ago[analyzer][NFC] Uplift checkers after D126801
Balazs Benics [Thu, 2 Jun 2022 17:46:38 +0000 (19:46 +0200)]
[analyzer][NFC] Uplift checkers after D126801

Reviewed By: martong

Differential Revision: https://reviews.llvm.org/D126802

2 years ago[analyzer][NFC] Add partial specializations for ProgramStateTraits
Balazs Benics [Thu, 2 Jun 2022 17:46:38 +0000 (19:46 +0200)]
[analyzer][NFC] Add partial specializations for ProgramStateTraits

I'm also hoisting common code from the existing specializations into a
common trait impl to reduce code duplication.

Reviewed By: martong

Differential Revision: https://reviews.llvm.org/D126801

2 years ago[libc++] Fix typo in comment at __optional_storage_base
Will Hawkins [Thu, 2 Jun 2022 17:43:11 +0000 (19:43 +0200)]
[libc++] Fix typo in comment at __optional_storage_base

Small typo fix(es) for struct definition of __optional_storage_base for
a reference type.

Reviewed By: #libc, jloser, philnik

Differential Revision: https://reviews.llvm.org/D126621

2 years ago[mlir] translate memref.reshape with static shapes but dynamic dims
Ashay Rane [Sun, 29 May 2022 00:33:04 +0000 (17:33 -0700)]
[mlir] translate memref.reshape with static shapes but dynamic dims

Prior to this patch, the lowering of memref.reshape operations to the
LLVM dialect failed if the shape argument had a static shape with
dynamic dimensions.  This patch adds the necessary support so that when
the shape argument has dynamic values, the lowering probes the dimension
at runtime to set the size in the `MemRefDescriptor` type.  This patch
also computes the stride for dynamic dimensions by deriving it from the
sizes of the inner dimensions.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D126604

2 years ago[RISCV] Add test case showing unnecessary extend after i32 smax on rv64. NFC
Craig Topper [Thu, 2 Jun 2022 16:37:05 +0000 (09:37 -0700)]
[RISCV] Add test case showing unnecessary extend after i32 smax on rv64. NFC

One of the operands of the smax is a positive value so computeKnownBits
determines the result of the smax must always be positive. This allows
DAG combiner to convert the sign extend to zero extend before type
legalization.

After type legalization the smax is promoted to i64 by sign extending
its inputs and the zero extend becomes an AND instruction. We are unable
to remove the AND at this point and it becomes a pair of shifts or a
zext.w.

The result of smax has as many sign bits as the minimum of its inputs.
Had we kept the sign extend instead of turning it into a zero extend
it would be removed by DAG combiner after type legalization.

2 years ago[lldb][NFC] Move non-clang specific method to the generic DWARF Parser
Luís Ferreira [Thu, 2 Jun 2022 16:01:04 +0000 (16:01 +0000)]
[lldb][NFC] Move non-clang specific method to the generic DWARF Parser

This patch renames DW_ACCESS_to_AccessType function and move it to the abstract
DWARFASTParser, since there is no clang-specific code there. This is useful for
plugins other than Clang.

Reviewed By: shafik, bulbazord

Differential Revision: https://reviews.llvm.org/D114719

2 years ago[OpenMP] omp_get_proc_id uses sched_getcpu fallback on FreeBSD 13.1 and above.
David CARLIER [Thu, 2 Jun 2022 16:10:29 +0000 (17:10 +0100)]
[OpenMP] omp_get_proc_id uses sched_getcpu fallback on FreeBSD 13.1 and above.

Reviewers: jlpeyton, jdoerfert

Reviewed-By: jlpeyton
Differential-Revision: https://reviews.llvm.org/D126408

2 years ago[OpenMP][libomp] Make LIBOMP_CONFIGURED_LIBFLAGS a list instead of string
Mikael Simberg [Thu, 2 Jun 2022 15:45:02 +0000 (10:45 -0500)]
[OpenMP][libomp] Make LIBOMP_CONFIGURED_LIBFLAGS a list instead of string

When configuring llvm with the openmp subproject, the build for the omp
target fails if LIBOMP_CONFIGURED_LIBFLAGS contains more than one item.
LIBOMP_CONFIGURED_LIBFLAGS should be a semicolon-separated list instead
of a string with items separated by spaces.

Differential Revision: https://reviews.llvm.org/D125370

2 years ago[llvm][ModuleInliner] Refactor InlineSizePriority and PriorityInlineOrder
Liqiang Tao [Thu, 2 Jun 2022 14:26:33 +0000 (22:26 +0800)]
[llvm][ModuleInliner] Refactor InlineSizePriority and PriorityInlineOrder

This patch introduces the abstract base class InlinePriority to serve as
the comparison function for the priority queue.  A derived class, such
as SizePriority, may choose to cache the priorities for different
functions for performance reasons.

This design shields the type used for the priority away from classes
outside InlinePriority and classes derived from it.  In turn,
PriorityInlineOrder no longer needs to be a template class.

Reviewed By: kazu

Differential Revision: https://reviews.llvm.org/D126300

2 years agoRevert "[llvm][ModuleInliner] Refactor InlineSizePriority and PriorityInlineOrder"
Liqiang Tao [Thu, 2 Jun 2022 15:15:42 +0000 (23:15 +0800)]
Revert "[llvm][ModuleInliner] Refactor InlineSizePriority and PriorityInlineOrder"

This reverts commit 50de7f1e7787bf327d1d04c4daab97c1c180b58c.

2 years ago[libc++] Lets to_chars use header implementation.
Mark de Wever [Mon, 16 May 2022 16:48:04 +0000 (18:48 +0200)]
[libc++] Lets to_chars use header implementation.

This removes the duplicated code from the dylib. Instead the dylib will
call the new functions in the header. Since this code is unneeded it's
removed from the unstable ABI.

Depends on D125704

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D125761

2 years ago[RISCV] Fix an inconsistency with compatible load/store handling
Philip Reames [Thu, 2 Jun 2022 14:56:06 +0000 (07:56 -0700)]
[RISCV] Fix an inconsistency with compatible load/store handling

Once we've computed the incoming predecessor state, we should use the same compatibility check with knowledge of MI as we did in phase 2 in order to be consistent across all phases.

Differential Revision: https://reviews.llvm.org/D126574

2 years ago[libc++][test] Enable some more string_view tests
Joe Loser [Thu, 2 Jun 2022 03:23:40 +0000 (21:23 -0600)]
[libc++][test] Enable some more string_view tests

Some test cases for `ends_with.ptr.pass` and `starts_with.ptr.pass` for
`string_view` are commented out, but work just fine. Uncomment them.

Differential Revision: https://reviews.llvm.org/D126849

2 years agoTidy up `pragma comment lib` handling and testing
Paul Robinson [Thu, 2 Jun 2022 14:50:52 +0000 (07:50 -0700)]
Tidy up `pragma comment lib` handling and testing

A bit of historical research shows that over the years:
Commit 99efc036 added `pragma comment lib` support for PS4.
Commit fd4db533 added `pragma comment lib` support for all ELF targets.
Commit 1d16515f reworked dependent-library support for all ELF targets.

The upshot is that some PS4-specific code became dead, and the
testing became somewhat fragmented.  I've removed the dead code and
combined the previous PS4-specific and linux-specific tests for the
diagnostics into one generic ELF test.
Also added a couple of PS5 runs while I was in there.

2 years ago[gn build] Port f9b8f422e711
LLVM GN Syncbot [Thu, 2 Jun 2022 14:47:37 +0000 (14:47 +0000)]
[gn build] Port f9b8f422e711

2 years ago[lldb][NFC] Move generic DWARFASTParser code out of Clang-specific code
Luís Ferreira [Thu, 2 Jun 2022 14:32:36 +0000 (14:32 +0000)]
[lldb][NFC] Move generic DWARFASTParser code out of Clang-specific code

This patch moves ParseChildArrayInfo out of DWARFASTParserClang in order
to decouple Clang-specific logic from DWARFASTParser.

Reviewed By: clayborg

Differential Revision: https://reviews.llvm.org/D114668

Signed-off-by: Luís Ferreira <contact@lsferreira.net>
2 years ago[llvm][ModuleInliner] Refactor InlineSizePriority and PriorityInlineOrder
Liqiang Tao [Thu, 2 Jun 2022 14:26:33 +0000 (22:26 +0800)]
[llvm][ModuleInliner] Refactor InlineSizePriority and PriorityInlineOrder

This patch introduces the abstract base class InlinePriority to serve as
the comparison function for the priority queue.  A derived class, such
as SizePriority, may choose to cache the priorities for different
functions for performance reasons.

This design shields the type used for the priority away from classes
outside InlinePriority and classes derived from it.  In turn,
PriorityInlineOrder no longer needs to be a template class.

Reviewed By: kazu

Differential Revision: https://reviews.llvm.org/D126300

2 years ago[MCParser] Set default alignment value when meeting invalid align
serge-sans-paille [Mon, 16 May 2022 14:02:44 +0000 (16:02 +0200)]
[MCParser] Set default alignment value when meeting invalid align

Upon invalid alignment value, still set a default valid alignment value to avoid
hitting later asserts.

Fix #55273

Differential Revision: https://reviews.llvm.org/D125688

2 years ago[SVE] Fixed custom lowering of ISD::INSERT_SUBVECTOR.
Paul Walker [Sun, 22 May 2022 19:32:10 +0000 (20:32 +0100)]
[SVE] Fixed custom lowering of ISD::INSERT_SUBVECTOR.

LowerINSERT_SUBVECTOR emits AArch64ISD::UUNPK## when lowering
scalable vector floating point INSERT_SUBVECTOR. However, these
nodes only make sense for integer types and thus isel patterns do
not exist for floating point, which leads to isel failures.

This patch ensures floating point operands are cast to integer
before the core lowering takes place.

Fixes: #55037

Differential Revision: https://reviews.llvm.org/D126487

2 years ago[Binary] Remove OffloadBinary from the Objects enumeration
Joseph Huber [Thu, 2 Jun 2022 13:32:32 +0000 (09:32 -0400)]
[Binary] Remove OffloadBinary from the Objects enumeration

Summary:
We use the beginning and end of this enumeration to determine what is
and isn't an object format. The enumeration for the OffloadBinary was
put here by mistake which led to it being mistakenly classified as an
Object file.

2 years agoUpdate more DR status information for C.
Aaron Ballman [Thu, 2 Jun 2022 13:31:34 +0000 (09:31 -0400)]
Update more DR status information for C.

This adds new files to track DRs 100-199 and 400-499, but the file
contents are still a work in progress. It also updates the associated
status in the DR tracking page.

2 years agoRevert "unbreak Modules/cxx20-export-import.cpp with LLVM_APPEND_VC_REV after 8c8a267...
Nico Weber [Thu, 2 Jun 2022 13:31:57 +0000 (09:31 -0400)]
Revert "unbreak Modules/cxx20-export-import.cpp with LLVM_APPEND_VC_REV after 8c8a2679a20f"

This reverts commit 4463bd0f89181234e0cef982e21de2e96038f873.
8c8a2679a20f was reverted in d42fe9aa8420.

2 years ago[PS5] Support r and y specifiers of freebsd_kernel_printf format strings
Paul Robinson [Thu, 2 Jun 2022 13:25:40 +0000 (06:25 -0700)]
[PS5] Support r and y specifiers of freebsd_kernel_printf format strings

2 years ago[mlir] add decompose and generalize to structured transform ops
Alex Zinenko [Tue, 31 May 2022 13:49:02 +0000 (15:49 +0200)]
[mlir] add decompose and generalize to structured transform ops

These ops complement the tiling/padding transformations by transforming
higher-level named structured operations such as depthwise convolutions into
lower-level and/or generic equivalents that are better handled by some
downstream transformations.

Differential Revision: https://reviews.llvm.org/D126698

2 years agoRevert "[clang][AIX] add option mdefault-visibility-export-mapping"
Hans Wennborg [Thu, 2 Jun 2022 13:09:39 +0000 (15:09 +0200)]
Revert "[clang][AIX] add option mdefault-visibility-export-mapping"

This caused assertions, see comment on the code review:

llvm/clang/lib/AST/Decl.cpp:1510:
clang::LinkageInfo clang::LinkageComputer::getLVForDecl(const clang::NamedDecl *, clang::LVComputationKind):
Assertion `D->getCachedLinkage() == LV.getLinkage()' failed.

> The option mdefault-visibility-export-mapping is created to allow
> mapping default visibility to an explicit shared library export
> (e.g. dllexport). Exactly how and if this is manifested is target
> dependent (since it depends on how they map dllexport in the IR).
>
> Three values are provided for the option:
>
> * none: the default and behavior without the option, no additional export linkage information is created.
> * explicit: add the export for entities with explict default visibility from the source, including RTTI
> * all: add the export for all entities with default visibility
>
> This option is useful for targets which do not export symbols as part of
> their usual default linkage behaviour (e.g. AIX), such targets
> traditionally specified such information in external files (e.g. export
> lists), but this mapping allows them to use the visibility information
> typically used for this purpose on other (e.g. ELF) platforms.
>
> Reviewed By: MaskRay
>
> Differential Revision: https://reviews.llvm.org/D126340

This reverts commit 8c8a2679a20f621994fa904bcfc68775e7345edc.

2 years ago[libunwind][ci][AIX] Add libunwind to buildbot CI
Xing Xue [Thu, 2 Jun 2022 13:03:10 +0000 (09:03 -0400)]
[libunwind][ci][AIX] Add libunwind to buildbot CI

Summary:
This patch changes scripts to add libunwind CI on AIX. Test config file ibm-libunwind-shared.cfg.in is introduced for testing on AIX.

Reviewed by: ldionne, MaskRay, libunwind, ibc++abi

Differential Revision: https://reviews.llvm.org/D126017

2 years ago[flang] Add check for conflict of BIND(C) and Parameter attributes
PeixinQiao [Thu, 2 Jun 2022 12:58:05 +0000 (20:58 +0800)]
[flang] Add check for conflict of BIND(C) and Parameter attributes

The entity with BIND(C) attribute cannot be a named constant, so the
BIND(C) and parameter attributes are conflicted. Add check for it.

Reviewed By: klausler

Differential Revision: https://reviews.llvm.org/D126654

2 years agoRevert "cmake: use llvm dir variables for clang/utils/hmaptool"
Nikita Popov [Thu, 2 Jun 2022 12:51:55 +0000 (14:51 +0200)]
Revert "cmake: use llvm dir variables for clang/utils/hmaptool"

As discussed on the review, this change breaks the standalone
clang build. When building against an installed LLVM, the
LLVM_TOOLS_BINARY_DIR cmake variable points to the location of
the installed LLVM tools, not to the cmake build directory. This
means that we would end up trying to move hmaptool into something
like /usr/bin as part of the normal build, while this should only
be happening when running an install target.

This reverts commit bf1ab1f0eb9578914343f48096229ecccd0ecf52.

2 years agoRevert "Drop qualifiers from return types in C (DR423)"
Aaron Ballman [Thu, 2 Jun 2022 12:26:38 +0000 (08:26 -0400)]
Revert "Drop qualifiers from return types in C (DR423)"

This reverts commit d374b65f2da1bdd3d9a7e9ac8ed4ad5467c882f9.

The changes lose AST fidelity (reported in #55778), but also may be
improperly dropping _Atomic qualifiers. I am rolling the changes back
until I've finished discussions in WG14 about the proper resolution to
DR423.

2 years ago[flang][docs] Document CMake support
Andrzej Warzynski [Tue, 31 May 2022 18:29:52 +0000 (19:29 +0100)]
[flang][docs] Document CMake support

One out-of-date section is also deleted.

Differential Revision: https://reviews.llvm.org/D126712

2 years ago[mlir][SCF] Add scf.foreach_thread.parallel_insert_slice canonicalization.
Nicolas Vasilache [Thu, 2 Jun 2022 11:52:45 +0000 (11:52 +0000)]
[mlir][SCF] Add scf.foreach_thread.parallel_insert_slice canonicalization.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D126761

2 years agoFix llvm.memset semantics description
Guillaume Chatelet [Thu, 2 Jun 2022 11:25:03 +0000 (13:25 +0200)]
Fix llvm.memset semantics description

The description was referring to a ``src`` parameter probably copied over from ``llvm.memcpy``

2 years ago[VPlan] Remove dead VPlan-native special case from BranchOnCount (NFC).
Florian Hahn [Thu, 2 Jun 2022 11:07:54 +0000 (12:07 +0100)]
[VPlan] Remove dead VPlan-native special case from BranchOnCount (NFC).

After 05776122b682684ad this special case doesn't exist any longer.

2 years ago[clang-tidy] Remove extra ";" in ModernizeModuleTest.cpp
Mikael Holmen [Thu, 2 Jun 2022 10:50:00 +0000 (12:50 +0200)]
[clang-tidy] Remove extra ";" in ModernizeModuleTest.cpp

Without this fix we get

../../clang-tools-extra/unittests/clang-tidy/ModernizeModuleTest.cpp:270:2: error: extra ';' outside of a function is incompatible with C++98 [-Werror,-Wc++98-compat-extra-semi]
};
 ^
1 error generated.

when compiling with -Werror.

2 years ago[Docs] Update default in opaque pointer docs (NFC)
Nikita Popov [Thu, 2 Jun 2022 10:28:12 +0000 (12:28 +0200)]
[Docs] Update default in opaque pointer docs (NFC)

Also mention a relevant C API.

2 years ago[SVE] Remove custom lowering of scalable vector MGATHER & MSCATTER operations.
Paul Walker [Mon, 7 Feb 2022 17:55:42 +0000 (17:55 +0000)]
[SVE] Remove custom lowering of scalable vector MGATHER & MSCATTER operations.

Differential Revision: https://reviews.llvm.org/D126255

2 years ago[OrcV2Examples] Fix test with opaque pointers
Nikita Popov [Thu, 2 Jun 2022 10:17:56 +0000 (12:17 +0200)]
[OrcV2Examples] Fix test with opaque pointers

The other two modules now use opaque pointers, so make sure this
one does as well.

2 years ago[flang] Update memcpy intrinsic mangling for opaque pointers
Nikita Popov [Thu, 2 Jun 2022 10:07:37 +0000 (12:07 +0200)]
[flang] Update memcpy intrinsic mangling for opaque pointers

Flang is manually mangling names for memset/memcpy/memmove
intrinsics, so we need to update the mangling to use the opaque
pointer format (p0 instead of p0i8).

2 years ago[flang] Update tests for opaque pointers
Nikita Popov [Thu, 2 Jun 2022 09:53:00 +0000 (11:53 +0200)]
[flang] Update tests for opaque pointers

There is still one remaining failure in Lower/forall/character-1.f90.

2 years ago[ARM] Fix a test case typo. NFC.
Martin Storsjö [Thu, 2 Jun 2022 09:57:51 +0000 (12:57 +0300)]
[ARM] Fix a test case typo. NFC.

The test looked for the wrong string, but it happened to match as
it was a substring of the actual output.

This fixes a typo from d8e67c1cccd8fcb62230166caea744592288da17.

2 years ago[llvm-ocaml] Add and use opaque pointer compatible bindings
Nikita Popov [Thu, 2 Jun 2022 08:53:32 +0000 (10:53 +0200)]
[llvm-ocaml] Add and use opaque pointer compatible bindings

Add bindings for LLVMConstGEP2, LLVMAddAlias2, LLVMBuildLoad2,
LLVMBuildInvoke2, LLVMBuildGEP2, LLVMBuildInBoundsGEP2,
LLVMBuildStructGEP2, LLVMBuildPtrDiff2 and use these in tests.

2 years ago[analyzer][NFC] Move overconstrained check from reAssume to assumeDualImpl
Gabor Marton [Tue, 31 May 2022 15:49:52 +0000 (17:49 +0200)]
[analyzer][NFC] Move overconstrained check from reAssume to assumeDualImpl

Depends on D126406. Checking of the overconstrained property is much
better suited here.

Differential Revision: https://reviews.llvm.org/D126707

2 years ago[clang] [ARM] Add __builtin_sponentry like on aarch64
Martin Storsjö [Thu, 19 May 2022 10:17:24 +0000 (13:17 +0300)]
[clang] [ARM] Add __builtin_sponentry like on aarch64

This is used for calling the SEH aware setjmp on MinGW.

Differential Revision: https://reviews.llvm.org/D126764

2 years ago[ARM] Implement lowering of the sponentry intrinsic
Martin Storsjö [Thu, 19 May 2022 10:16:53 +0000 (13:16 +0300)]
[ARM] Implement lowering of the sponentry intrinsic

This is needed for SEH based setjmp on Windows.

Differential Revision: https://reviews.llvm.org/D126763

2 years ago[ARM] Adjust the frame pointer when it's needed for SEH unwinding
Martin Storsjö [Fri, 26 Nov 2021 13:17:47 +0000 (15:17 +0200)]
[ARM] Adjust the frame pointer when it's needed for SEH unwinding

For functions that require restoring SP from FP (e.g. that need to
align the stack, or that have variable sized allocations), the prologue
and epilogue previously used to look like this:

    push {r4-r5, r11, lr}
    add r11, sp, #8
    ...
    sub r4, r11, #8
    mov sp, r4
    pop {r4-r5, r11, pc}

This is problematic, because this unwinding operation (restoring sp
from r11 - offset) can't be expressed with the SEH unwind opcodes
(probably because this unwind procedure doesn't map exactly to
individual instructions; note the detour via r4 in the epilogue too).

To make unwinding work, the GPR push is split into two; the first one
pushing all other registers, and the second one pushing r11+lr, so that
r11 can be set pointing at this spot on the stack:

    push {r4-r5}
    push {r11, lr}
    mov r11, sp
    ...
    mov sp, r11
    pop {r11, lr}
    pop {r4-r5}
    bx lr

For the same setup, MSVC generates code that uses two registers;
r11 still pointing at the {r11,lr} pair, but a separate register
used for restoring the stack at the end:

    push {r4-r5, r7, r11, lr}
    add r11, sp, #12
    mov r7, sp
    ...
    mov sp, r7
    pop {r4-r5, r7, r11, pc}

For cases with clobbered float/vector registers, they are pushed
after the GPRs, before the {r11,lr} pair.

Differential Revision: https://reviews.llvm.org/D125649

2 years ago[ARM] Add SEH opcodes in frame lowering
Martin Storsjö [Fri, 26 Nov 2021 12:28:16 +0000 (14:28 +0200)]
[ARM] Add SEH opcodes in frame lowering

Skip inserting regular CFI instructions if using WinCFI.

This is based a fair amount on the corresponding ARM64 implementation,
but instead of trying to insert the SEH opcodes one by one where
we generate other prolog/epilog instructions, we try to walk over the
whole prolog/epilog range and insert them. This is done because in
many cases, the exact number of instructions inserted is abstracted
away deeper.

For some cases, we manually insert specific SEH opcodes directly where
instructions are generated, where the automatic mapping of instructions
to SEH opcodes doesn't hold up (e.g. for __chkstk stack probes).

Skip Thumb2SizeReduction for SEH prologs/epilogs, and force
tail calls to wide instructions (just like on MachO), to make sure
that the unwind info actually matches the width of the final
instructions, without heuristics about what later passes will do.

Mark SEH instructions as scheduling boundaries, to make sure that they
aren't reordered away from the instruction they describe by
PostRAScheduler.

Mark the SEH instructions with the NoMerge flag, to avoid doing
tail merging of functions that have multiple epilogs that all end
with the same sequence of "b <other>; .seh_nop_w, .seh_endepilogue".

Differential Revision: https://reviews.llvm.org/D125648

2 years ago[LegalizeTypes][VP] Add widen and split support for VP FP integer casting op.
jacquesguan [Thu, 2 Jun 2022 03:00:02 +0000 (03:00 +0000)]
[LegalizeTypes][VP] Add widen and split support for VP FP integer casting op.

This patch adds widen and split support for VP_FPTOSI, VP_FPTOUI, VP_SITOFP and VP_UITOFP.

Differential Revision: https://reviews.llvm.org/D126847

2 years ago[mlir][complex] Lower complex.sin/cos to libm
lewuathe [Thu, 2 Jun 2022 08:38:20 +0000 (10:38 +0200)]
[mlir][complex] Lower complex.sin/cos to libm

Lower sin/cos operation in complex dialect to libm as a baseline. This follows up to https://reviews.llvm.org/D125550.

Reviewed By: pifon2a

Differential Revision: https://reviews.llvm.org/D126755

2 years ago[llvm-ocaml] Add LLVMBuildCall2 binding
Nikita Popov [Thu, 2 Jun 2022 08:22:59 +0000 (10:22 +0200)]
[llvm-ocaml] Add LLVMBuildCall2 binding

Add binding for the opaque pointer compatible LLVMBuildCall2
API and use it in tests.

2 years ago[mlir][complex] Sanity check for tan operation in complex dialect
lewuathe [Thu, 2 Jun 2022 08:33:13 +0000 (10:33 +0200)]
[mlir][complex] Sanity check for tan operation in complex dialect

Add a sanity check for newly added tan operation in complex dialect. It follows-up to https://reviews.llvm.org/D126685.

Differential Revision: https://reviews.llvm.org/D126858

2 years ago[clang-format] Handle do-while loops for RemoveBracesLLVM
owenca [Wed, 1 Jun 2022 08:06:02 +0000 (01:06 -0700)]
[clang-format] Handle do-while loops for RemoveBracesLLVM

Also updates the unit tests to match the updated LLVM Coding
Standards.

Differential Revision: https://reviews.llvm.org/D126758

2 years ago[clang][dataflow] Add transfer functions for structured bindings
Stanislav Gatev [Thu, 24 Feb 2022 17:07:41 +0000 (17:07 +0000)]
[clang][dataflow] Add transfer functions for structured bindings

This is part of the implementation of the dataflow analysis framework.
See "[RFC] A dataflow analysis framework for Clang AST" on cfe-dev.

Differential Revision: https://reviews.llvm.org/D120495

Reviewed-by: ymandel, xazax.hun
2 years ago[IRTests] With opaque ptrs we can test that there is no bitcast inserted
Benjamin Kramer [Thu, 2 Jun 2022 07:56:11 +0000 (09:56 +0200)]
[IRTests] With opaque ptrs we can test that there is no bitcast inserted

2 years ago[IR] Enable opaque pointers by default
Nikita Popov [Tue, 31 May 2022 09:58:35 +0000 (11:58 +0200)]
[IR] Enable opaque pointers by default

This enabled opaque pointers by default in LLVM. The effect of this
is twofold:

* If IR that contains *neither* explicit ptr nor %T* types is passed
  to tools, we will now use opaque pointer mode, unless
  -opaque-pointers=0 has been explicitly passed.
* Users of LLVM as a library will now default to opaque pointers.
  It is possible to opt-out by calling setOpaquePointers(false) on
  LLVMContext.

A cmake option to toggle this default will not be provided. Frontends
or other tools that want to (temporarily) keep using typed pointers
should disable opaque pointers via LLVMContext.

Differential Revision: https://reviews.llvm.org/D126689

2 years ago[BOLT][NFC] Use colors in CFG dumps
Amir Ayupov [Thu, 2 Jun 2022 07:26:23 +0000 (00:26 -0700)]
[BOLT][NFC] Use colors in CFG dumps

Use color coding to distinguish nodes:
- Entry nodes have bold border
- Scalar (non-loopy) code is milk white
- Outer loops are light yellow
- Innermost loops are light blue

`-print-loops` needs to be enabled to provide BinaryLoopInfo.
Examples:
{F23170673}
{F23170680}

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D126248

2 years ago[lldb, test] Fix typos in the lldb tests
Venkata Ramanaiah Nalamothu [Thu, 2 Jun 2022 06:32:44 +0000 (12:02 +0530)]
[lldb, test]  Fix typos in the lldb tests

Reviewed By: JDevlieghere

Differential Revision: https://reviews.llvm.org/D126596

2 years ago[LoongArch] Add the missing ld.d instruction definition
Weining Lu [Thu, 2 Jun 2022 06:57:23 +0000 (14:57 +0800)]
[LoongArch] Add the missing ld.d instruction definition

2 years ago[BOLT][NFC] Print block instructions in dumpGraph as part of node label
Amir Ayupov [Thu, 2 Jun 2022 06:40:55 +0000 (23:40 -0700)]
[BOLT][NFC] Print block instructions in dumpGraph as part of node label

Reuse the option `-dot-tooltip-code` to put block instructions into the label.
This way, the instructions are displayed by default when used with dot viewer.

When the .dot file is used with dot2html, instructions are hidden by default,
and are shown by clicking on a node.

{F23169510}

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D126237

2 years ago[BOLT][UTILS] Add dot2html helper tool to embed dot into html
Amir Ayupov [Thu, 2 Jun 2022 06:37:04 +0000 (23:37 -0700)]
[BOLT][UTILS] Add dot2html helper tool to embed dot into html

To be rendered in browser using d3-graphviz.
Example: {F23169510}

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D126218

2 years ago[mlir][Arithmetic] Add constant folder for RemF.
jacquesguan [Fri, 20 May 2022 07:31:07 +0000 (07:31 +0000)]
[mlir][Arithmetic] Add constant folder for RemF.

This patch adds the constant folder for RemF.

Differential Revision: https://reviews.llvm.org/D126045

2 years ago[LoongArch] Fix register type in FCMP_*_D instructions definition
Weining Lu [Thu, 2 Jun 2022 06:18:08 +0000 (14:18 +0800)]
[LoongArch] Fix register type in FCMP_*_D instructions definition

The correct type is FPR64 but not FPR32.

2 years ago[NFC][workflow] Fix issue where the workflow would say all PR's already exists
Tobias Hieta [Thu, 2 Jun 2022 06:20:21 +0000 (08:20 +0200)]
[NFC][workflow] Fix issue where the workflow would say all PR's already exists

2 years ago[gn build] Port a15ae4139cea
LLVM GN Syncbot [Thu, 2 Jun 2022 06:04:25 +0000 (06:04 +0000)]
[gn build] Port a15ae4139cea

2 years ago[libc++][doc] Fixes unwanted list nesting.
Mark de Wever [Thu, 2 Jun 2022 06:02:19 +0000 (08:02 +0200)]
[libc++][doc] Fixes unwanted list nesting.

2 years ago[libc++] Make to_chars base 10 header only.
Mark de Wever [Mon, 16 May 2022 15:12:18 +0000 (17:12 +0200)]
[libc++] Make to_chars base 10 header only.

The functions to_chars and from_chars should offer 128-bit support. This
is the first step to implement 128-bit version of to_chars. Before
implementing 128-bit support the current code will be polished.

This moves the code from the dylib to the header in prepartion of

P2291 "Add Constexpr Modifiers to Functions to_chars and from_chars for
Integral Types in <charconv> Header"

Note some more cleanups will be done in follow-up commits
- Remove the _LIBCPP_AVAILABILITY_TO_CHARS from to_chars. With all code
  in the header the availablilty macro is no longer needed. This
  requires enabling the unit tests for additional platforms.
- The code in the dylib can switch to using the header implementation.
  This allows removing the code duplicated in the header and the dylib.

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D125704

2 years ago[clang-tidy] Reject invalid enum initializers in C files
Richard [Sat, 14 May 2022 23:57:10 +0000 (17:57 -0600)]
[clang-tidy] Reject invalid enum initializers in C files

C requires that enum values fit into an int.  Scan the macro tokens
present in an initializing expression and reject macros that contain
tokens that have suffixes making them larger than int.

C forbids the comma operator in enum initializing expressions, so
optionally reject comma operator.

Differential Revision: https://reviews.llvm.org/D125622

Fixes #55467

2 years ago[LSR] Early exit for RateFormula when it is already losing. NFC
eopXD [Thu, 2 Jun 2022 03:39:38 +0000 (20:39 -0700)]
[LSR] Early exit for RateFormula when it is already losing. NFC

This patch does not effect any behavior of the current code.

The codebase implicitly implies that `Cost::RateFormula` is only called
when the `Cost` is not in losing status, or else there may be possible
to trigger the assertion of `Cost::isValid`.

The intention here is to prevent mis-use where future development
allow `Cost` that is already loser to call `Cost::RateFormula` - Early
exit when `Cost` is already losing.

Reviewed By: Meinersbur, #loopoptwg

Differential Revision: https://reviews.llvm.org/D125670

2 years ago[mlir] Support convert token type from LLVM IR.
jacquesguan [Wed, 1 Jun 2022 06:29:09 +0000 (06:29 +0000)]
[mlir] Support convert token type from LLVM IR.

This patch supports the token type for converting from LLVM IR.

Differential Revision: https://reviews.llvm.org/D126756

2 years ago[test] Fix LLVMgold.so tests after D125847
Fangrui Song [Thu, 2 Jun 2022 03:05:01 +0000 (20:05 -0700)]
[test] Fix LLVMgold.so tests after D125847

2 years ago[libunwind] Add more information to eh_frame_hdr version error
Florian Mayer [Wed, 1 Jun 2022 23:43:59 +0000 (16:43 -0700)]
[libunwind] Add more information to eh_frame_hdr version error

This makes it easier to find the offending ELF file.

Reviewed By: #libunwind, MaskRay

Differential Revision: https://reviews.llvm.org/D126841

2 years agounbreak Modules/cxx20-export-import.cpp with LLVM_APPEND_VC_REV after 8c8a2679a20f
Nico Weber [Thu, 2 Jun 2022 02:33:36 +0000 (22:33 -0400)]
unbreak Modules/cxx20-export-import.cpp with LLVM_APPEND_VC_REV after 8c8a2679a20f

See revision b8b7a9dcdcbc for prior art.

2 years ago[memprof] Print out the segment information in YAML format.
Snehasish Kumar [Wed, 1 Jun 2022 22:20:53 +0000 (22:20 +0000)]
[memprof] Print out the segment information in YAML format.

This change prints out the segment information in the raw profile in
YAML format for testing. Since we don't capture build ids yet, we print
out <None> for now.

Reviewed By: tejohnson

Differential Revision: https://reviews.llvm.org/D126840

2 years ago[LegalizeTypes][VP] Add widen and split support for VP_SIGN_EXTEND and VP_ZERO_EXTEND.
jacquesguan [Thu, 26 May 2022 03:14:27 +0000 (03:14 +0000)]
[LegalizeTypes][VP] Add widen and split support for VP_SIGN_EXTEND and VP_ZERO_EXTEND.

Differential Revision: https://reviews.llvm.org/D126442

2 years ago[memprof] Update summary output.
Snehasish Kumar [Wed, 1 Jun 2022 22:20:53 +0000 (22:20 +0000)]
[memprof] Update summary output.

Update the YAML format print out of the profile to include a summary
instead of displaying the headers in the raw file buffer. This allows us
to release the raw buffer early saving memory.

Reviewed By: tejohnson

Differential Revision: https://reviews.llvm.org/D126834

2 years agoFix compilation in gold-plugin triggered by D125847
Matthias Braun [Thu, 2 Jun 2022 01:51:35 +0000 (18:51 -0700)]
Fix compilation in gold-plugin triggered by D125847

2 years ago[libc] Add pthread_create and pthread_join functions.
Siva Chandra Reddy [Wed, 25 May 2022 21:07:23 +0000 (21:07 +0000)]
[libc] Add pthread_create and pthread_join functions.

They do not yet support all the feature/attributes in pthread_attr_t.
Future changes will add such support.

Reviewed By: lntue

Differential Revision: https://reviews.llvm.org/D126718

2 years ago[libc++][test] Enable constexpr string comparison tests
Joe Loser [Tue, 31 May 2022 00:21:56 +0000 (18:21 -0600)]
[libc++][test] Enable constexpr string comparison tests

Some tests in `string.view.comparison` are not enabled due to previous lack of
support for `constexpr std::string`. Now that it is implemented, we can enable
these tests.

Differential Revision: https://reviews.llvm.org/D126737

2 years agoFix DXContainer test on ppcbe
Chris Bieneman [Thu, 2 Jun 2022 01:18:41 +0000 (20:18 -0500)]
Fix DXContainer test on ppcbe

Silly mistake, the code here was byteswapping in-place which corrupts
the data structure that is used later to write the correct file.

2 years ago[BOLT][DWARF] Fix TU Index handling for DWARF4/5
Alexander Yermolovich [Thu, 2 Jun 2022 01:15:41 +0000 (18:15 -0700)]
[BOLT][DWARF] Fix TU Index handling for DWARF4/5

When we generate split dwarf with -fdebug-types-section we will have
.debug_types.dwo sections. These go into TU Index when we run llvm-dwp. BOLT was
not handling DWP input correctly with this section.

Added support for handling DWP with TU Index as an input and output for DWARF4.
Added support for handling DWP with TU Index as an input for DWARF5

Reviewed By: maksfb

Differential Revision: https://reviews.llvm.org/D126087

2 years agoBranchFolder: Require NoPHIs
Matt Arsenault [Fri, 22 Apr 2022 23:15:58 +0000 (19:15 -0400)]
BranchFolder: Require NoPHIs

The pass doesn't handle SSA and breaks any phis.

2 years agoAMDGPU: Add release notes about atomic load and store
Matt Arsenault [Wed, 4 May 2022 12:12:51 +0000 (13:12 +0100)]
AMDGPU: Add release notes about atomic load and store

2 years agoAMDGPU: Fix missing c++ mode comment
Matt Arsenault [Wed, 18 May 2022 15:00:41 +0000 (16:00 +0100)]
AMDGPU: Fix missing c++ mode comment

2 years agoLTO: Decide upfront whether to use opaque/non-opaque pointer types
Matthias Braun [Wed, 18 May 2022 03:33:22 +0000 (20:33 -0700)]
LTO: Decide upfront whether to use opaque/non-opaque pointer types

LTO code may end up mixing bitcode files from various sources varying in
their use of opaque pointer types. The current strategy to decide
between opaque / typed pointers upon the first bitcode file loaded does
not work here, since we could be loading a non-opaque bitcode file first
and would then be unable to load any files with opaque pointer types
later.

So for LTO this:
- Adds an `lto::Config::OpaquePointer` option and enforces an upfront
  decision between the two modes.
- Adds `-opaque-pointers`/`-no-opaque-pointers` options to the gold
  plugin; disabled by default.
- `--opaque-pointers`/`--no-opaque-pointers` options with
  `-plugin-opt=-opaque-pointers`/`-plugin-opt=-no-opaque-pointers`
  aliases to lld; disabled by default.
- Adds an `-lto-opaque-pointers` option to the `llvm-lto2` tool.
- Changes the clang driver to pass `-plugin-opt=-opaque-pointers` to
  the linker in LTO modes when clang was configured with opaque
  pointers enabled by default.

This fixes https://github.com/llvm/llvm-project/issues/55377

Differential Revision: https://reviews.llvm.org/D125847

2 years ago[ValueTypes] Define MVTs for v128i2/v64i4 as well as i2 and i4.
Hendrik Greving [Thu, 2 Jun 2022 00:46:50 +0000 (00:46 +0000)]
[ValueTypes] Define MVTs for v128i2/v64i4 as well as i2 and i4.

Adds MVT::v128i2, MVT::v64i4, and implied MVT::i2, MVT::i4.

Keeps MVT::i2, MVT::i4 lowering actions as expand, which should be
removed once targets set this explicitly.

Adjusts 11 lit tests to reflect slightly different behavior during
DAG combine.

Differential Revision: https://reviews.llvm.org/D125247

2 years ago[Polly] Fix -Wreorder-ctor. NFC
Fangrui Song [Thu, 2 Jun 2022 00:33:14 +0000 (17:33 -0700)]
[Polly] Fix -Wreorder-ctor. NFC

2 years ago[RegisterClassInfo] Invalidate cached information if ignoreCSRForAllocationOrder...
Quentin Colombet [Thu, 2 Jun 2022 00:14:03 +0000 (17:14 -0700)]
[RegisterClassInfo] Invalidate cached information if ignoreCSRForAllocationOrder changes

Even if CSR list is same between functions, we could have had a different
allocation order if ignoreCSRForAllocationOrder is evaluated differently.
Hence invalidate cached register class information if
ignoreCSRForAllocationOrder changes.

Patch by Srividya Karumuri <srividya_karumuri@apple.com>

Differential Revision: https://reviews.llvm.org/D126565

2 years ago[Object] Fix namespace style issues in D122069
Fangrui Song [Thu, 2 Jun 2022 00:04:58 +0000 (17:04 -0700)]
[Object] Fix namespace style issues in D122069

https://llvm.org/docs/CodingStandards.html#use-namespace-qualifiers-to-implement-previously-declared-functions

2 years ago[OMPIRBuilder] Add the support for compare capture
Shilei Tian [Wed, 1 Jun 2022 23:53:36 +0000 (19:53 -0400)]
[OMPIRBuilder] Add the support for compare capture

This patch adds the support for `compare capture` in `OMPIRBuilder`.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D120007

2 years ago[RISCV] Split fixed-vector-strided-load-store.ll so it can be autogened
Philip Reames [Wed, 1 Jun 2022 23:08:19 +0000 (16:08 -0700)]
[RISCV] Split fixed-vector-strided-load-store.ll so it can be autogened

I've gotten tired of updating register allocation changes by hand, let's just autogen this even if we have to duplicate it.

2 years agoRevert "[SLP]Improve shuffles cost estimation where possible."
Alexey Bataev [Wed, 1 Jun 2022 22:40:00 +0000 (15:40 -0700)]
Revert "[SLP]Improve shuffles cost estimation where possible."

This reverts commit fd5a6ce9dcb77b7821c95355d73af0b3b2020647 to fix
a crash detected by a buildbot
https://lab.llvm.org/buildbot/#/builders/179/builds/3805/steps/11/logs/stdio.

2 years ago[Binary] Promote OffloadBinary to inherit from Binary
Joseph Huber [Wed, 1 Jun 2022 19:28:34 +0000 (15:28 -0400)]
[Binary] Promote OffloadBinary to inherit from Binary

We use the `OffloadBinary` to create binary images of offloading files
and their corresonding metadata. This patch changes this to inherit from
the base `Binary` class. This allows us to create and insepect these
more generically. This patch includes all the necessary glue to
implement this as a new binary format, along with added the magic bytes
we use to distinguish the offloading binary to the `file_magic`
implementation.

Reviewed By: tra

Differential Revision: https://reviews.llvm.org/D126812