There were system scope tests only for integer min/max. Expand this to
cover all of the integer operations.

There were incomplete generated checks plus some dead manual checks.

The folding of conversions of character arrays from one kind to
another was not preserving the value of the character length expression
on the intermediate array constructor, causing spurious errors
when defining named constant arrays of one kind of character with
arrays of another kind.

Fixes llvm-test-suite/Fortran/gfortran/regression/widechar_10.f90.

Differential Revision: https://reviews.llvm.org/D157340

An incorrect "Implicit declaration of function '...' has a different result type
than in previous declaration" is being emitted for ENTRY names used
recursively.  The predicate used to check for recursive use only allowed
for scopes of functions, not ENTRYs.

Fixes llvm-test-suite/Fortran/gfortran/regression/whole_file_9.f90.

Differential Revision: https://reviews.llvm.org/D157337

When folding the TRANSFER() intrinsic function, there's a CHECK()
that the initial image was built without error.  However, there is one
condition that elicits a warning -- a derived type SOURCE= with an
allocatable or dynamic component.  Remove this CHECK().

Fixes llvm-test-suite/Fortran/gfortran/regression/transfer_null_1.f90.

Differential Revision: https://reviews.llvm.org/D157336

All supported compilers have support for coroutines in C++20 now.

Differential Revision: https://reviews.llvm.org/D145086

The builders are in a broken state, which causes all the CI builds
to appear as red. Make them soft fail until we've managed to reach
the build bot owners.

Differential Revision: https://reviews.llvm.org/D157421

[flang] Don't complain about implicitly-typed dummy arguments inherited into an IMPLICIT NONE submodule

When a module procedure's interface is defined in an ancestor (sub)module
with an implicitly typed dummy argument, don't emit a bogus error in
the separate module procedure later if its submodule has IMPLICIT NONE.

Fixes llvm-test-suite/Fortran/gfortran/regression/submodule_26.f90.

Differential Revision: https://reviews.llvm.org/D157335

At least one other Fortran compiler supports the use of unrestricted intrinsic
functions as specific procedures in generic interfaces, and the usage seems
to be both useful and unambiguous.  Support it with a portability warning.

Fixes llvm-test-suite/Fortran/gfortran/regression/pr95500.f90.

Differential Revision: https://reviews.llvm.org/D157333

Macro is used to avoid repeating same pattern many times.
Also fixed the ordering of ops to be consistent.

Reviewed By: K-Wu

Differential Revision: https://reviews.llvm.org/D157419

When a constant array value has a non-default lower bound, the current
expression formatting code uses a non-Fortran syntax to dump the lower
bounds. (There's no way in Fortran to explicitly specify such a constant value,
but they can be created through the use of named constants.)  But we
don't want this lower bounds syntax from expression dumping to show up
in module files, since it can't be parsed back in.  So disable that
part of expression formatting by default.

Fixes https://github.com/llvm/llvm-project/issues/64391.

Differential Revision: https://reviews.llvm.org/D157330

Prevents regression on some future work to improve codegen for concat_vectors(extract_subvector(),extract_subvector()) patterns.

X86ISD::SHUF128 optimization is still pretty poor (especially the zmm variant), not optimizing the shuffle demanded elts like we do for SHUFPS.

The shape of an expanded scalar MASK= was incorrect in the template
code that folds FINDLOC/MAXLOC/MINLOC, leading to a vector indexing
crash when a DIM= argument is also present.

Fixes https://github.com/llvm/llvm-project/issues/64286.

Differential Revision: https://reviews.llvm.org/D157329

This reverts commit 8e7f0320.
As requested on https://reviews.llvm.org/rG8e7f0320ad7fb760f#1236457

Support for scalable vectors in vector.multi_reduction is added by
simply updating MultiDimReductionOp::verify.

Also, the conversion pattern for reducing n-D vector.multi_reduction to
2D vector.multi_reduction is updated.

Differential Revision: https://reviews.llvm.org/D157092

The parenthesis' operand is used as a mold for the generated
hlfir.elemental.

Depends on D157316

Reviewed By: tblah, clementval

Differential Revision: https://reviews.llvm.org/D157318

Pass the first argument as the polymorphic mold for the generated
hlfir.elemental.

Depends on D157316

Reviewed By: tblah, clementval

Differential Revision: https://reviews.llvm.org/D157317

The polymorphic temporary array is created using the provided mold
and the shape of the hlfir.elemental. The array is allocated right away,
because it is going to be initialized element per element.

Depends on D157315

Reviewed By: clementval, tblah

Differential Revision: https://reviews.llvm.org/D157316

To properly create temporary array for a polymorphic result
of hlfir.elemental we need to keep the mold as its operand.
This patch adds just the basic support.

Reviewed By: clementval, tblah

Differential Revision: https://reviews.llvm.org/D157315

This is a follow up to D156552.

clang::StreamingDiagnostic doesn't have a uint8_t overload,
which makes MSVC without /permissive- decide it's ambiguous.
Pre-approved in replies to D153276.

* Remove the incorrect patterns from AArch64fmla_p/AArch64fmls_p
* Add correct patterns to AArch64fmla_m1/AArch64fmls_m1
* Refactor fma_patfrags for the sake of PatFrags

Fixes https://github.com/llvm/llvm-project/issues/64419

Differential Revision: https://reviews.llvm.org/D157095

create_thread_loop can time out sometimes on macOS CI.

Differential Revision: https://reviews.llvm.org/D157291

This reverts commit 74c141a4.

Looks like this breaks a whole bunch of unexpected tests:
https://lab.llvm.org/buildbot/#/builders/109/builds/70813

Differential Revision: https://reviews.llvm.org/D157074

This target goes over the limit in some build modes.

Discourse discussion: https://discourse.llvm.org/t/rfc-update-lldb-code-ownership/72253

Differential revision: https://reviews.llvm.org/D156949

Differential Revision: https://reviews.llvm.org/D157073

Failure to vectorize this led to a revert of D156532, so add
a PhaseOrdering test to prevent this from happening again.

The issue is actually related to ScatterVectorize nodes. If such node
gets reordered during bottom-to-top reordering, it may have associated
non-empty ReorderIndices. In this case, such nodes need to be handled
the same way as regular Vectorize nodes, not NeedToGather nodes. In this
case we need to reorder ReorderIndices array rather than scalars.

**For an explanation of these patches see D154153.**

Commit message:
This patch adds the utility base class `ModuleToObject`. This class provides an
interface for compiling module operations into binary strings, by default this
class serialize modules to LLVM bitcode.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D154100

This reverts commit 1f370886 as it causes a
large regression in x264, and some other regressions in downstream embedded
benchmarks under LTO.