This patch adds lowering support for default clause.
1. During symbol resolution in semantics, should the enclosing context have
a default data sharing clause defined and a `parser::Name` is not attached
to an explicit data sharing clause, the
`semantics::Symbol::Flag::OmpPrivate` flag (in case of `default(private)`)
and `semantics::Symbol::Flag::OmpFirstprivate` flag (in case of
`default(firstprivate)`) is added to the symbol.
2. During lowering, all symbols having either
`semantics::Symbol::Flag::OmpPrivate` or
`semantics::Symbol::Flag::OmpFirstprivate` flag are collected and
privatised appropriately.
Co-authored-by: Peixin Qiao <qiaopeixin@huawei.com>
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D123930
Use pertinent names instead of numbered values, to make it easier to
update the test in future patches.
Differential Revision: https://reviews.llvm.org/D130474
This patch replaces x.getValue() with *x if the reference is obviously
protected by a presence check. Otherwise, it replaces x.getValue()
with x.value().
This patch adds an initial support to the lastprivate clause for worksharing loop. The patch creates necessary control flow to guarantee the store of the value from the logical last iteration of the workshare loop.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D130027
Firstly, we we make an additional GNUInstallDirs-style variable. With
NixOS, for example, this is crucial as we want those to go in
`${dev}/lib/cmake` not `${out}/lib/cmake` as that would a cmake subdir
of the "regular" libdir, which is installed even when no one needs to do
any development.
Secondly, we make *Config.cmake robust to absolute package install
paths. We for NixOS will in fact be passing them absolute paths to make
the `${dev}` vs `${out}` distinction mentioned above, and the
GNUInstallDirs-style variables are suposed to support absolute paths in
general so it's good practice besides the NixOS use-case.
Thirdly, we make `${project}_INSTALL_PACKAGE_DIR` CACHE PATHs like other
install dirs are.
Reviewed By: sebastian-ne
Differential Revision: https://reviews.llvm.org/D117973
The predicate "CanBeCalledViaImplicitInterface()" was returning false for
restricted specific intrinsic functions (e.g., SIN) because their procedure
characteristics have the elemental attribute; this leads to a bogus semantic
error when one attempts to use them as proc-targets in procedure pointer
assignment statements when the left-hand side of the assignment is a procedure
pointer with an implicit interface. However, these restricted specific intrinsic
functions have always been allowed as special cases for such usage -- it is
as if they are elemental when it is necessary for them to be so, but not
when it's a problem.
Differential Revision: https://reviews.llvm.org/D130386
An OPEN statement that implies closing a connection must invalidate
the unit's frame buffer so as to prevent stale data from the old
connection from being read into the newly-connected unit.
Differential Revision: https://reviews.llvm.org/D130430
Lower the Flang parse-tree containing OpenMP reductions to the OpenMP
dialect. The OpenMP dialect models reductions with,
1) A reduction declaration operation that specifies how to initialize, combine,
and atomically combine private reduction variables.
2) The OpenMP operation (like wsloop) that supports reductions has an array of
reduction accumulator variables (operands) and an array attribute of the same
size that points to the reduction declaration to be used for the reduction
accumulation.
3) The OpenMP reduction operation that takes a value and an accumulator.
This operation replaces the original reduction operation in the source.
(1) is implemented by the `createReductionDecl` in OpenMP.cpp,
(2) is implemented while creating the OpenMP operation,
(3) is implemented by the `genOpenMPReduction` function in OpenMP.cpp, and
called from Bridge.cpp. The implementation of (3) is not very robust.
NOTE 1: The patch currently supports only reductions for integer type addition.
NOTE 2: Only supports reduction in the worksharing loop.
NOTE 3: Does not generate atomic combination region.
NOTE 4: Other options for creating the reduction operation include
a) having the reduction operation as a construct containing an assignment
and then handling it appropriately in the Bridge.
b) we can modify `genAssignment` or `genFIR(AssignmentStmt)` in the Bridge to
handle OpenMP reduction but so far we have tried not to mix OpenMP
and non-OpenMP code and this will break that.
I will try (b) in a separate patch.
NOTE 5: OpenMP dialect gained support for reduction with the patches:
D105358, D107343. See https://discourse.llvm.org/t/rfc-openmp-reduction-support/3367
for more details.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D130077
Co-authored-by: Peixin-Qiao <qiaopeixin@huawei.com>
When a procedure pointer is associated with a procedure target, f18
will emit hard error messages if the pointer is incompatible with the
target in a way that would cause the program to not work if the
procedure pointer were actually called. However, in the context of
the ASSOCIATED() intrinsic, the compiler should not consider a query
to be a compilation-time error if the pointer and the target are
incompatible. The standard gives us sufficient wiggle room here
to treat these cases as warnings.
Attempts to use ASSOCIATED to test whether a procedure pointer is
associated with a data object, or vice versa, remain errors.
Differential Revision: https://reviews.llvm.org/D130384
In flang pipeline, the inliner calls createCanonicalizerPass with the region
simplification disabled. The inliner pass does canonicalization even if
no inlining happens. After canonicalization, FIR lite region simplification
must be called to get rid of unreachable regions.
This code exposes the need to run SimplifyRegionLitePass after the inliner is
called with FIR pipeline.
Differential Revision: https://reviews.llvm.org/D130484
f18 intentionally does not support the spottily-implemented language extension
in which one can pass NULL() for an allocatable dummy argument. This is perhaps
a sanctioned side effect in other compilers of the fact that they pass distinct
"base address" and "descriptor address" physical arguments.
Make the error message in this case more specific to the circumstances, and
add a note to Extensions.md to clarify that this behavior is intended.
(We could, with some effort in lowering, support passing NULL for an INTENT(IN)
allocatable dummy, but let's see whether such nonconforming usage appears
in a real application before spending any more time on it.)
Differential Revision: https://reviews.llvm.org/D130380
As Fortran 2018 16.9.169, the argument of selected_int_kind is integer
scalar, and result is default integer scalar. The constant expression in
this intrinsic has been supported by folding the constant expression.
This supports lowering and runtime for variables in this intrinsic.
Reviewed By: Jean Perier
Differential Revision: https://reviews.llvm.org/D129959
This change is required for release builds - see
https://reviews.llvm.org/D130185 for more context.
I'm sending this without a review as this is rather straightforward and
identical to the changes from https://reviews.llvm.org/D130185 (the test
updated in this patch wasn't yet in-tree when D130185 landed).
As Fortran 2018 16.9.170, the argument of `selected_real_kind` is integer
scalar, and result is default integer scalar. The constant expression in
this intrinsic has been supported by folding the constant expression.
This supports lowering this intrinsic for variables using runtime.
Reviewed By: Jean Perier
Differential Revision: https://reviews.llvm.org/D130183
Flang C++ Style Guide tells us to use *X when the reference is
protected by a presense test. However, (*X).foo() is a little harder
to read, especially when X is a complicated expression.
This patch slightly deviates from the guide (but retains the spirit)
by using X->foo() instead.
Differential Revision: https://reviews.llvm.org/D130413
Flang C++ Style Guide tells us to use x.value() when no presence test
is obviously protecting the reference. Since a failure in EXPECT_TRUE
doesn't terminate a given test, I don't count it as "protection" here.
Differential Revision: https://reviews.llvm.org/D130410
The predicate IsHostAssocited() was implemented in a way that would
return true only for cases of host association into a module or inner
subprogram. Technically, the use of a name in a BLOCK construct
that is not declared therein is considered in the Fortran standard
to also be a form of host association, and this matters when doing
error checking on DATA statements.
Differential Revision: https://reviews.llvm.org/D130388
When a procedure pointer with no interface is associated with
an EXTERNAL name with no interface information, but it is later
inferred that the procedure pointer must be a subroutine because it
appears in a CALL statement, don't complain that the EXTERNAL name
is not also known to be a subroutine.
Subroutine vs. function errors are still caught in procedure pointer
assignment compatibility checking; this fix simply ensures that those
more nuanced tests are not overridded by the attribute set equality test.
Also, leave in some code for dumping the differing attributes in legitimate
error cases that was added in the coures of debugging the specific problem.
Differential Revision: https://reviews.llvm.org/D130385
In some contexts the code looks cleaner if we fold zero-length substring
references into empty character constants, but those don't work in
variable definition contexts and can lead to bogus semantic error
messages.
Differential Revision: https://reviews.llvm.org/D130383
For aint/anint, LLVM conversion operations llvm.trunc and llvm.round
can support the edge case of aint(-0.) and anint(-0.). The output is -0.
and it is the same of `gfortran` and `classic flang`, while the output
of `ifort` is 0.. The `real(10)/real(16)` is not supported before.
Support it and remove the runtime functions for aint/anint.
For nint, `gfortran`, `ifort`, and LLVM Flang using llvm.lround have
different results when the magnitude of argument is more than the max of
result value range. So delay its support in lowering after more
investigations.
Reviewed By: vzakhari
Differential Revision: https://reviews.llvm.org/D130024
The integer arithmetic template supports ISHFTC() but the
integer intrinsic folding code had yet to call it; finish
the job.
Differential Revision: https://reviews.llvm.org/D130379
This patch fixes:
llvm-project/flang/lib/Semantics/expression.cpp:405:12: error:
moving a local object in a return statement prevents copy elision
[-Werror,-Wpessimizing-move]
The code that copies data from a constant source array into a character
array constant result was failing to copy its last element if it was
only partially defined due to misalignment.
Differential Revision: https://reviews.llvm.org/D130376
A type-param-inquiry of %KIND or %LEN applies to a designator, and
so must also be allowed for a substring. F18 presently (mis)parses
instances of a type-param-inquiry as structure component references
and then fixes them in expression semantics when types are known and
we can distinguish them. But when the base of a type-param-inquiry is
a substring of an array element, as in "charArray(i)(j:k)%len",
parsing fails.
Adjust the grammar to parse these cases, and extend expression semantics
to process the new production.
Differential Revision: https://reviews.llvm.org/D130375
Fortran specifically allows character-valued I/O specifiers
to have trailing blanks, e.g. OPEN(666,STATUS='SCRATCH ').
The runtime I/O library already handles them, but the I/O
static checks in semantics do not.
Differential Revision: https://reviews.llvm.org/D130381
This patch aims to create a webpage to document
Flang's command line options on https://flang.llvm.org/docs/
in a similar way to Clang's
https://clang.llvm.org/docs/ClangCommandLineReference.html
This is done by using clang_tablegen to generate an .rst
file from Options.td (which is current shared with Clang)
For this to work, ClangOptionDocEmitter.cpp was updated
to allow specific Flang flags to be included,
rather than bulk excluding clang flags.
Note:
Some headings in the generated documentation will incorrectly
contain references to Clang, e.g.
"Flags controlling the behaviour of Clang during compilation"
This is because Options.td (Which is shared between both Clang and Flang)
contains hard-coded DocBrief sections. I couldn't find a non-intrusive way
to make this target-dependant, as such I've left this as is, and it will need revisiting later.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D129864
This commit changes how math intrinsics are lowered: we, first,
try to lower them into MLIR operations or libm calls via
mathOperations table and only then fallback to pgmath runtime calls.
The pgmath fallback is needed, because mathOperations does not
support all intrinsics that pgmath supports. The main purpose
of this change is to get rid of llvmIntrinsics table so that
we do not have to update both llvmIntrinsics and mathOperations
when adding new intrinsic support.
mathOperations lowering should phase out pgmath lowering, when
more operations are available (e.g. power operations being
added in D129809 and D129811; complex type operations from
Complex dialect).
Differential Revision: https://reviews.llvm.org/D130129
Convert arith.cmpi to the canonical form with constants on the right side
to simplify further optimizations and open more opportunities for CSE.
Differential Revision: https://reviews.llvm.org/D129929
Lower F08 shift (shiftl, shiftr, shifta) and combined shift (dshiftl, dshiftr)
intrinsics. The combined shift intrinsics are implemented using the
definitions of shiftl and shiftr as described by the standard.
For non-conformant arguments to the shift intrinsics, the implementation tries
to replicate the behavior of other compilers if most of the other behave
consistently.
Differential Revision: https://reviews.llvm.org/D129316
Lower F08 bit population count intrinsics popcnt, poppar, leadz and trailz. popcnt, leadz and trailz are implemented using the corresponding MLIR math intrinsics. poppar is implemented in terms of popcnt.
Differential Revision: https://reviews.llvm.org/D129584
First of all, `LLVM_TOOLS_INSTALL_DIR` put there breaks our NixOS
builds, because `LLVM_TOOLS_INSTALL_DIR` defined the same as
`CMAKE_INSTALL_BINDIR` becomes an *absolute* path, and then when
downstream projects try to install there too this breaks because our
builds always install to fresh directories for isolation's sake.
Second of all, note that `LLVM_TOOLS_INSTALL_DIR` stands out against the
other specially crafted `LLVM_CONFIG_*` variables substituted in
`llvm/cmake/modules/LLVMConfig.cmake.in`.
@beanz added it in d0e1c2a550 to fix a
dangling reference in `AddLLVM`, but I am suspicious of how this
variable doesn't follow the pattern.
Those other ones are carefully made to be build-time vs install-time
variables depending on which `LLVMConfig.cmake` is being generated, are
carefully made relative as appropriate, etc. etc. For my NixOS use-case
they are also fine because they are never used as downstream install
variables, only for reading not writing.
To avoid the problems I face, and restore symmetry, I deleted the
exported and arranged to have many `${project}_TOOLS_INSTALL_DIR`s.
`AddLLVM` now instead expects each project to define its own, and they
do so based on `CMAKE_INSTALL_BINDIR`. `LLVMConfig` still exports
`LLVM_TOOLS_BINARY_DIR` which is the location for the tools defined in
the usual way, matching the other remaining exported variables.
For the `AddLLVM` changes, I tried to copy the existing pattern of
internal vs non-internal or for LLVM vs for downstream function/macro
names, but it would good to confirm I did that correctly.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D117977
Try 2 to merge 4fbd1d6c87.
Flang algebraic simplification pass will run algebraic simplification
rewrite patterns for Math/Complex/etc. dialects. It is enabled
under opt-for-speed optimization levels (i.e. for O1/O2/O3; Os/Oz will not
enable it).
With this change the FIR/MLIR optimization pipeline becomes affected
by the -O* optimization level switches. Until now these switches
only affected the middle-end and back-end.
Differential Revision: https://reviews.llvm.org/D130035
This patch aims to create a webpage to document
Flang's command line options on https://flang.llvm.org/docs/
in a similar way to Clang's
https://clang.llvm.org/docs/ClangCommandLineReference.html
This is done by using clang_tablegen to generate an .rst
file from Options.td (which is current shared with Clang)
For this to work, ClangOptionDocEmitter.cpp was updated
to allow specific Flang flags to be included,
rather than bulk excluding clang flags.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D129864
Tests that use `--mlir-pass-statistics-display=` from MLIR require the
following condition to hold: (extracted from LLVM's Statistics.h):
```
#define LLVM_ENABLE_STATS 1
```
This is normally enforced with `REQUIRES: asserts`. This patch updates
relevant Flang tests accordingly.
For "Release" builds (with assertions disabled), the affected tests will
be failing without this change.
Differential Revision: https://reviews.llvm.org/D130185
Flang algebraic simplification pass will run algebraic simplification
rewrite patterns for Math/Complex/etc. dialects. It is enabled
under opt-for-speed optimization levels (i.e. for O1/O2/O3; Os/Oz will not
enable it).
With this change the FIR/MLIR optimization pipeline becomes affected
by the -O* optimization level switches. Until now these switches
only affected the middle-end and back-end.
Differential Revision: https://reviews.llvm.org/D130035
Flang C++ Style Guide tells us to use x.value() when no presence test
is obviously protecting the reference. Since assert can be disabled,
I don't count it as "protection" here.
Differential Revision: https://reviews.llvm.org/D130144
This update makes sure that `flang-new -O2 -O0` will run at `-O0` rather
than `-O2`.
This bug was identified and originally fixed by Vyacheslav Zakharin in
https://reviews.llvm.org/D130035. I've extracted the fix into a separate
patch.
co-authored by: Vyacheslav Zakharin <vyacheslav.p.zakharin@intel.com>
Differential Revision: https://reviews.llvm.org/D130104
