This patch is part of the upstreaming effort from fir-dev branch.
It also ensures all descriptors created inline complies with LBOUND
requirement that the lower bound is `1` when the related dimension
extent is zero.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D128047
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
There's a few (3) cases where Fortran allows two distinct symbols to have
the same name in the same scope. Module file output copes with only two of
them. The third involves a separate module procedure that isn't separate:
both the procedure and its declared interface appear in the same (sub)module.
Fix to ensure that the interface is included in the module file output, so
that the module file reader doesn't suffer a bogus error about a "separate
module procedure without an interface".
Differential Revision: https://reviews.llvm.org/D127784
A message has a %s string substitution in it but somebody (probably me)
forgot to pass the argument that defines it.
Differential Revision: https://reviews.llvm.org/D127794
The predicate that determines image control statements needs
to distinguish STOP (which is) from ERROR STOP (which isn't).
Differential Revision: https://reviews.llvm.org/D127796
When generic resolution finds its specific procedure in a module,
and that specific procedure is not use-associated into the local scope
(perhaps because it was PRIVATE, perhaps because the generic was
use-associated with ONLY:), we create a new use-association with
a renaming. The name constructed for this renaming needs to be
additionally qualified with the module name of the specific procedure
in order to avoid clashing with another specific of the same name
that may have previously been use-associated in the same way from
a distinct module.
Differential Revision: https://reviews.llvm.org/D127790
Fix some mismatch in format used in the file and reduce the diff with fir-dev
to be able to finish the upstreaming on this file.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld, kiranchandramohan
Differential Revision: https://reviews.llvm.org/D127849
For BIND(C) statement, two common block with the same name can have the
same bind name. Fix the regression failure by adding this check. Also add
the regression tests.
Co-authored-by: Jean Perier <jperier@nvidia.com>
Reviewed By: clementval
Differential Revision: https://reviews.llvm.org/D127841
As OpenMP 5.0, for firstprivate, lastprivate, copyin, and copyprivate
clauses, if the list item is a polymorphic variable with the allocatable
attribute, the behavior is unspecified.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D127601
Add failed_images to the list of intrinsic functions. Add a
semantics test for failed_images.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D126805
num_images had previously been added to the list of intrinsics
before the type team_type had been implemented. Now that team_type
is implemented, add the num_images call that has the team argument.
Update the semantics tests for num_images.
Reviewed By: craig.rasmussen
Differential Revision: https://reviews.llvm.org/D126734
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D127738
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Lowering was crashing with "fatal internal error: node has not been analyzed"
if a PDT with initial component value was defined inside an internal
procedure. This is because the related expression cannot be analyzed
without the component values (which happens at the instatiation).
These expression do not need to be visited (the instantiations, if any
will be). Use the form of GetExpr that tolerates the parse tree expression
to not be analyzed into an evaluate::Expr when looking through the
symbols used in an internal procedure.
Note that the PDTs TODO will then fire (it happens after the PFT
analysis) as expected if the derived type is used.
Differential Revision: https://reviews.llvm.org/D127735
Codegen does not support fir.addressof of functions returning derived
types, arrays are descriptors inside GlobalOp region.
This is because the abstract-result-opt is required to rewrite such
functions (a hidden argument must be added), but this pass is meant to
run in GlobalOp currently.
Such fir.address_of may be created when lowering procedure pointers
initial value (TODO), or when creating derived type descriptors for
types with bindings.
Add a TODO in lowering until abstract-result-opt is modified to run
on GlobalOp too.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D127722
As Fortran 2018 C802 and C873, if bind name is specified, there can only
be only one entity. The check for common block is missed before. As
Fortran 2018 8.5.5 point 2, the bind name is one identifier, which is
unique. That is, one entity can not have multiple bind names. Also add
this check.
Reviewed By: klausler, Jean Perier
Differential Revision: https://reviews.llvm.org/D126961
f18 was treating "f() = 1" as a statement function definition
if it could be viewed as being in the specification part and
"f" was a USE-associated function returning a data pointer.
(The non-USE-associated case is fine.) Fix to allow for "f"
to be USE associated.
Differential Revision: https://reviews.llvm.org/D127430
Handle the case of a non-generic procedure that is USE associated
into a scope that has a generic interface of the same name with an
appropriate error rather than crashing.
Differential Revision: https://reviews.llvm.org/D127429
FIR models Fortran intrinsic types with deliberate KIND values. Like
Fortran, COMPLEX and REAL have related KINDs in FIR. Lowering now
converts REAL types to floating point (MLIR) up front. This patch moves
the code to convert from FIR RealType to MLIR FloatType out of codegen
and into the builder, allowing FIR ComplexTypes to have their element
type returned as an MLIR FloatType.
We should consider whether to replace fir::ComplexType with
mlir::ComplexType at some point. I believe these types are presently
used to convey distinctins in the target ABIs in the Tilikum bridge
however.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127636
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127634
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
The scope model used for branch target checking treats a label
on an END SELECT statement as if it were in the previous CASE block.
This makes it illegal to GO TO that label from within any earlier
CASE block in that statement. Fix by treating the CASE blocks as
nested scopes within the scope of the SELECT construct.
Also, add a "warning:" tag to related warning messages.
Differential Revision: https://reviews.llvm.org/D127425
Add TODO for KIND=2 so the user is notified correctly.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127619
Co-authored-by: Peter Steinfeld <psteinfeld@nvidia.com>
Loop variables of a worksharing loop and sequential loops in parallel
region are privatised by default. These variables are marked with
OmpPreDetermined. Skip explicit privatisation of these variables.
Note: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: Leporacanthicus
Differential Revision: https://reviews.llvm.org/D127249
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Mats Petersson <mats.petersson@arm.com>
Add TODO for half-precision for reduction.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127622
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
The pass was raising TODOs when a function both had a fir.boxproc<> argument
and a fir.type<> argument (even if the fir.type<> did not contain a
fir.boxproc itself).
Prevent the TODO from firing when a fir.type<> does not actually contain
a fir.boxproc. Add the location for the remaining TODO (it will be
needed when procedure pointer components are supported in lowering).
FYI, I actually tried to just implement the TODO, but I there is a funny
issue. When creating the new fir::RecordType, since the name and context
are the same as the type being translated, fir::RecordType:get just
returns the existing type, and there is no way to change it (finalize()
does nothing since it is already finalized). So this will require to add
the ability to mutate the existing type, and I am not sure what are the
MLIR constraints here, so I escaped and left the TODO for that case.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127633
Co-authored-by: Jean Perier <jperier@nvidia.com>
ALLOCATE statement allows reversed bounds (see Fortran 2018 9.7.1.2
point 1) in which case the extents are zero.
The same applies for the character length provided in the type spec that
can be negative. In which case the new length is zero.
Use genMaxWithZero to deal with these cases.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127617
Co-authored-by: Jean Perier <jperier@nvidia.com>
system_clock intrinsic calls with dynamically optional arguments
Modify intrinsic system_clock calls to allow for an argument that is optional
or a disassociated pointer or an unallocated allocatable. A call with such an
argument is the same as a call that does not specify that argument.
Rename (genIsNotNull -> genIsNotNullAddr) and (genIsNull -> genIsNullAddr)
and add a use of genIsNotNullAddr.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D127616
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
This patch adds code so that using bbc we are able to see an end-to-end lowering of simd construct in action.
Reviewed By: kiranchandramohan, peixin, shraiysh
Differential Revision: https://reviews.llvm.org/D125282
[flang]Add support for do concurrent
Upstreaming from fir-dev on https://github.com/flang-compiler/f18-llvm-project
Support for concurrent execution in do-loops.
A selection of tests are also added.
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D127240
This relatively small change will allow Flang's frontend driver,
`flang-new -fc1`, to consume and parse MLIR files. Semantically (i.e.
from user's perspective) this is identical to reading LLVM IR files.
Two file extensions are associated with MLIR files: .fir and .mlir. Note
that reading MLIR files makes only sense when running one of the
code-generation actions, i.e. when using one of the following action
flags: -S, -emit-obj, -emit-llvm, -emit-llvm-bc.
The majority of tests that required `tco` to run are updated to also run
with `flang-new -fc1`. A few tests are updated to use `fir-opt` instead
of `tco` (that's the preferred choice when testing a particular MLIR
pass). basic-program.fir is not updated as that test is intended to
verify the behaviour of `tco` specifically.
Differential Revision: https://reviews.llvm.org/D126890
Until now, `-x` wasn't really taken into account in Flang's compiler and
frontend drivers. `flang-new` and `flang-new -fc1` only recently gained
powers to consume inputs other than Fortran files and that's probably
why this hasn't been noticed yet.
This patch makes sure that `-x` is supported correctly and consistently
with Clang. To this end, verification is added when reading LLVM IR
files (i.e. IR modules are verified with `llvm::verifyModule`). This
way, LLVM IR parsing errors are correctly reported to Flang users. This
also aids testing.
With the new functionality, we can verify that `-x ir` breaks
compilation for e.g. Fortran files and vice-versa. Tests are updated
accordingly.
Differential Revision: https://reviews.llvm.org/D127207
Remove a backwards dependence from Optimizer -> Lower by moving Todo.h
to the optimizer and out of lowering.
This patch is part of the upstreaming effort from fir-dev branch.
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D127292
* Fix assertion strings.
* Fixes assertion that arose from bad FIR being constructed.
With the default member-wise component assignment, the LHS and RHS may
be compatible but distinct types. This change to lowering manages both
the LHS and RHS independently rather than assume the two types are
identical. This avoids creating bogus FIR and asserting/crashing in
codegen.
Update the tests with the member-wise copy code.
This patch is part of the upstreaming effort from fir-dev branch.
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D127297
Simply add a source and target materialization handler that do nothing
and that override the default handlers that would add illegal
LLVM::DialectCastOp otherwise.
This is the simplest workaround, but not an actual fix, something may be
inconsistent after D82831 (most likely fir lowering to llvm happens in a
way that mlir infrastructure is not expecting in D82831).
Here is a minimal reproducer of what the issue was:
```
func @foop(%a : !fir.real<4>) -> ()
func @bar(%a : !fir.real<2>) {
%1 = fir.convert %a : (!fir.real<2>) -> !fir.real<4>
call @foop(%1) : (!fir.real<4>) -> ()
return
}
```
tco -o - output was:
```
error: 'llvm.mlir.cast' op type must be non-index integer types, float types, or vector of mentioned types.
llvm.func @foop(!llvm.float)
llvm.func @bar(%arg0: !llvm.half) {
%0 = llvm.fpext %arg0 : !llvm.half to !llvm.float
%1 = llvm.mlir.cast %0 : !llvm.float to !fir.real<4>
llvm.call @foop(%1) : (!fir.real<4>) -> ()
llvm.return
}
```
This patch disable the introduction of the llvm.mlir.cast and preserve the previous behavior.
Also fixes https://github.com/llvm/llvm-project/issues/55210.
Note: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D127212
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Add support for lowering the schedule modifiers (simd, monotonic,
non-monotonic) in worksharing loops.
Note: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: peixin
Differential Revision: https://reviews.llvm.org/D127311
Co-authored-by: Mats Petersson <mats.petersson@arm.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
This is a small follow-up for https://reviews.llvm.org/D120051. It makes
sure that tables with "run-time type information for derived types" are
generated for code-gen actions. Originally, only non-code-gen actions
were updated (i.e. actions that were fully supported at that time).
Differential Revision: https://reviews.llvm.org/D127307
As Fortran 2018 R874, common block object must be one variable name, which
cannot be one named constant. Add this check.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D126762
As Fortran 2018 C1521, in procedure declaration statement, if
proc-language-binding-spec (bind(c)) is specified, the proc-interface
shall appear, it shall be an interface-name, and interface-name shall
be declared with a proc-language-binding-spec.
Reviewed By: klausler, Jean Perier
Differential Revision: https://reviews.llvm.org/D127121
Getting ompobject symbol is needed in multiple places and will be
needed later for the lowering of other constructs/clauses such as
copyin clause. Extract them into one function.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D127280
Move tthe function to allow its usage in the Optimizer/Builder functions.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D127295
Switch debub message to proper TODOs.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D127282
Given the flag `--always-execute-loop-body` the compiler emits code
to execute the body of the loop atleast once.
Note: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: awarzynski, schweitz
Differential Revision: https://reviews.llvm.org/D127128
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Co-authored-by: Valentin Clement <clementval@gmail.com>
Co-authored-by: Sameeran Joshi <sameeranjayant.joshi@amd.com>
This supports lowering parse-tree to MLIR for threadprivate directive
following the OpenMP 5.1 [2.21.2] standard. Take the following as an
example:
```
program m
integer, save :: i
!$omp threadprivate(i)
call sub(i)
!$omp parallel
call sub(i)
!$omp end parallel
end
```
```
func.func @_QQmain() {
%0 = fir.address_of(@_QFEi) : !fir.ref<i32>
%1 = omp.threadprivate %0 : !fir.ref<i32> -> !fir.ref<i32>
fir.call @_QPsub(%1) : (!fir.ref<i32>) -> ()
omp.parallel {
%2 = omp.threadprivate %0 : !fir.ref<i32> -> !fir.ref<i32>
fir.call @_QPsub(%2) : (!fir.ref<i32>) -> ()
omp.terminator
}
return
}
```
A threadprivate operation (omp.threadprivate) is created for all
references to a threadprivate variable. The runtime will appropriately
return a threadprivate var (%1 as above) or its copy (%2 as above)
depending on whether it is outside or inside a parallel region. For
threadprivate access outside the parallel region, the threadprivate
operation is created in instantiateVar. Inside the parallel region, it
is created in createBodyOfOp.
One new utility function collectSymbolSet is created for collecting
all the variables with a property within a evaluation, which may be one
Fortran, or OpenMP, or OpenACC construct.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D124226
The previous XArrayCoorOp conversion forgot to change getting the
operands from OpAdaptor for upper bound and step of slice. This leads to
the fail of incompatible of types of codegen when slices are index type.
Reviewed By: kiranchandramohan, schweitz
Differential Revision: https://reviews.llvm.org/D125967
The previous semantic analysis does not consider when the last part-ref
is scalar or complex part. Refactor the previous code and bring all the
checks into one place. The check starts from the designator by
extracting the dataref wrapped including the substring and complex part
and recursively check the base objects.
Co-authored-by: Peter Klausler <pklausler@nvidia.com>
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D126595
OpenMP 5.0 adds a new clause `in_reduction` on OpenMP directives.
This patch adds parser support for the same.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D124156
On targets with __float128 available and distinct from long double,
use it to support more kind=16 entry points. This affects mostly
x86-64 targets. This means that more runtime entry points are
defined for lowering to call.
Delete Common/long-double.h and its LONG_DOUBLE macro in favor of
testing the standard macro LDBL_MANT_DIG.
Differential Revision: https://reviews.llvm.org/D127025
For example, FINDLOC(A,X) should convert both A and X to COMPLEX(8)
if the operands are REAL(8) and COMPLEX(4), so that comparisons
can be done without losing inforation. The current implementation
unconditionally converts X to the type of the array A.
Differential Revision: https://reviews.llvm.org/D127030
Avoid calls to memcpy with zero byte counts if their address argument
calculations may not be valid expressions.
Differential Revision: https://reviews.llvm.org/D127027
F18 preserves lower bounds of explicit-shape named constant arrays, but
failed to also do so for implicit-shape named constants. Fix.
Differential Revision: https://reviews.llvm.org/D127021
In the USE statements that f18 emits to module files, ensure that symbols
from intrinsic modules are marked as such on their USE statements. And
ensure that the current working directory (".") cannot override the intrinsic
module search path when trying to locate an intrinsic module.
Differential Revision: https://reviews.llvm.org/D127019
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
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).
As per issue #1196, the loop induction variable, which is an argument
in the omp.wsloop operation, does not have a memory location, so when
passed to a function or subroutine, the reference to the value is not
a memory location, but the value of the induction variable. The callee
function/subroutine is then trying to dereference memory at address 1
or some other "not a good memory location".
This is fixed by creating a temporary memory location and storing the
value of the induction variable in that.
Test fixes as a consequence of the changed code generated.
Add checking for some of the omp-unstructured.f90 to check for alloca,
store and load operations, to ensure the correct flow. Add a test
for CYCLE inside a omp-do loop.
Also convert to use -emit-fir in the omp-unstructrued, and make
the symbol matching consistent in the omp-wsloop-variable test.
Reviewed By: peixin
Differential Revision: https://reviews.llvm.org/D126711
This patch supports the following checks:
```
[5.1] 2.21.2 THREADPRIVATE Directive
The threadprivate directive must appear in the declaration section of
a scoping unit in which the common block or variable is declared.
[5.1] 2.14.7 Declare Target Directive
The directive must appear in the declaration section of a scoping unit
in which the common block or variable is declared.
```
Reviewed By: kiranchandramohan, shraiysh, NimishMishra
Differential Revision: https://reviews.llvm.org/D125767
The basic infinite loop is lowered to a branch to the body of the
loop, and the body containing a back edge as its terminator.
Note: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: rovka
Differential Revision: https://reviews.llvm.org/D126697
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Upstream the code for handling loops with real control variables from
the fir-dev branch at
https://github.com/flang-compiler/f18-llvm-project/tree/fir-dev/
Also add a test.
Loops with real-valued control variables are always lowered to
unstructured loops. The real-valued control variables are handled the
same as integer ones, the only difference is that they need to use
floating point instructions instead of the integer equivalents.
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
Similar to procedure argument, the function result cannot be one
named constant.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D126693
As Fortran 2018 C819, a variable with the BIND attribute shall be declared
in the specification part of a module. Add the support for this check.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D126653
The following changes are made for OpenMP operations with unstructured region,
1. For combined constructs the outer operation is considered a structured
region and the inner one as the unstructured.
2. Added a condition to ensure that we create new blocks only once for nested
unstructured OpenMP constructs.
Tests are added for checking the structure of the CFG.
Note: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project. Code originally reviewed
at https://github.com/flang-compiler/f18-llvm-project/pull/1394.
Reviewed By: vdonaldson, shraiysh, peixin
Differential Revision: https://reviews.llvm.org/D126375
Upstream the code for handling while loops from the fir-dev branch at
https://github.com/flang-compiler/f18-llvm-project/tree/fir-dev/
Also add tests.
The while loop is lowered to a header block that checks the loop
condition and branches either to the exit block or to the body of the
loop. The body of the loop will unconditionally branch back to the
header.
Differential Revision: https://reviews.llvm.org/D126636
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: V Donaldson <vdonaldson@nvidia.com>
This patch implements the following semantic check:
A list-item cannot appear in more than one nontemporal clause.
Reviewed By: kiranchandramohan, shraiysh
Differential Revision: https://reviews.llvm.org/D110270
Test that chunk size is passed to the static init function.
Using three different variations:
1. Single constant.
2. Expression with constants.
3. Variable value.
Reviewed By: peixin, shraiysh
Differential Revision: https://reviews.llvm.org/D126383
As an extension for REAL literals, we allow an exponent letter which
matches an explicit kind-param. The standard requires the exponent
to be 'E' if a kind-param is present. This patch
- documents this extension in Extensions.md
- enables a portability warning if it is used with -pedantic
The test case for this, kinds05.f90, needs D125804, which makes
test_errors.py test warnings as well, to actually test the warnings.
I include it already now to keep things together, it will do no harm
(I hope ...).
We also add WARNING-directives to the test kinds04.f90 in preparation
for D125804. As the exponent-letter 'Q' does not imply the same kind
on all platforms, the emitted warnings are platform-dependent.
Therefore, the test is duplicated into two variants which are run
conditionally.
Finally, we promote the portability warning for when the exponent letter
is neither 'E' nor matching the kind-param to a standard warning.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D126459
When two or more generic interfaces are available by declaration or
by USE association at different scoping levels, we need to search
the outer generic interfaces as well as the inner ones, but only after
the inner ones have failed to produce a specific procedure that matches
a given set of actual arguments. This means that it is possible for
a specific procedure of a generic interface of an inner scope to override
a conflicting specific procedure of a generic interface of an outer
scope.
Also cope with forward references to derived types when a generic
interface is also in scope.
Fixes LLVM bug https://github.com/llvm/llvm-project/issues/55240 and
LLVM bug https://github.com/llvm/llvm-project/issues/55300.
Differential Revision: https://reviews.llvm.org/D126587
When processing the literal constants of the various kinds of
INTEGER that are too large by 1 (e.g., 2147483648_4) in expression
analysis, emit a portability warning rather than a fatal error if
the literal constant appears as the operand to a unary minus, since
the folded result will be in range. And don't emit any warning if
the negated literal is coming from a module file -- f18 wrote the
module file and the warning would simply be confusing, especially to
the programmer that wrote (-2147483647_4-1) in the first place.
Further, emit portability warnings for the canonical expressions for
infinities and NaN (-1./0., 0./0., & 1./0.), but not when they appear
in a module file, for the same reason. The Fortran language has no
syntax for these special values so we have to emit expressions that
fold to them.
Fixes LLVM bugs https://github.com/llvm/llvm-project/issues/55086 and
https://github.com/llvm/llvm-project/issues/55081.
Differential Revision: https://reviews.llvm.org/D126584
For pointer variables, using getSymbolAddress cannot get the coorect
address for atomic read/write operands. Use genExprAddr to fix it.
Reviewed By: shraiysh, NimishMishra
Differential Revision: https://reviews.llvm.org/D125793
If a line is over 72 characters long, flang's preprocessor cuts it there
and continues on the next line.
For this purpose it uses the standard way of continuing line with & on each line.
However, it doesn't work with long compiler directives, like OpenMP or OpenACC ones.
The line that continues the directive also has to
contain the corresponding sentinel at the beginning.
This change implements the described functionality.
Also, some code was refactored in order to simplify and reuse existing code.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D126301
The CreateEntry() function in name resolution needs to allow for the name
of an alternate entry point already having been declared in the outer scope
as the homonymous specific procedure of a generic interface; e.g.,
interface foo
module procedure foo
end interface
subroutine bar
entry foo
end subroutine
Differential Revision: https://reviews.llvm.org/D126436
A recent change fixed the processing of BIND(C,NAME=expr) character
expressions so that they are evaluated as constants in the scope of
the subprogram. However, when the character name expression results
in an empty value after trimming, the compiler emits a warning message,
and this message is now causing a crash due to a lack of statement
context. To fix, extend the deferred processing of the BIND(C,NAME="")
so that a basic statement context exists.
Differential Revision: https://reviews.llvm.org/D126416
A recent fix beefed up semantics checking to catch the case of a call
to an external assumed-length character function; this check has false
positives in the case of an assumed-length character function that is
a dummy procedure. These do have a length that is passed in extra
compiler-created arguments. This patch refines the check and undoes some
changes to tests.
Differential Revision: https://reviews.llvm.org/D126390
This is a support for " #pragma omp atomic compare fail ". It has Parser & AST support for now.
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D123235
Since the FIR operations are mostly structured, it is only the functions
that could contain multiple blocks inside an operation. This changes
with OpenMP since OpenMP regions can contain multiple blocks. For
unstructured code, the blocks are created in advance and belong to the
top-level function. This caused code in OpenMP region to be placed under
the function level.
In this fix, if the OpenMP region is unstructured then new blocks are
created inside it.
Note1: This is part of upstreaming from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project. The code in this patch is a
subset of the changes in https://github.com/flang-compiler/f18-llvm-project/pull/1178.
Reviewed By: vdonaldson
Differential Revision: https://reviews.llvm.org/D126293
Co-authored-by: Val Donaldson <vdonaldson@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Valentin Clement <clementval@gmail.com>
The derived type information table construction code had a
crash whose root cause was replacing an expression with one
of its operands -- the deletion of the LHS of that assignment
led to the RHS being invalidated before it could be read.
Fix by cloning the RHS. Also update a TODO message to the
new "_todo_en_US" message class and add a comment about how
it should be resolved.
Differential Revision: https://reviews.llvm.org/D126154
The scalar-default-character-expression that defines the interoperable
name of a function or subroutine (or interface) must have its names
resolved within the context of the subprogram, despite its appearance
on a function-stmt or a subroutine-stmt. Failure to do so can lead
to bogus errors or to incorrect results.
The solution is to defer name resolution for function-stmt suffixes
(but not entry-stmt suffixes) and for subroutine-stmt language binding
specifications to EndSubprogram(). (Their resolution only need to be
deferred to the end of the specification part, but it's cleanest to
deal with it in EndSubprogram().)
Differential Revision: https://reviews.llvm.org/D126153
User-defined derived type I/O subroutines need to be unique for
a given type and operation in any scope, but it is acceptable
to have more than one defined I/O subroutine so long as only one
of them is visible.
Differential Revision: https://reviews.llvm.org/D126152
A utility predicate in semantics was incorrectly determining that
an INTERFACE ASSIGNMENT(=) (or other form of generic) could not have
a specific procedure with an unlimited polymorphic second argument.
This led to a crash later in expression analysis. Fix, and
extend tests.
Differential Revision: https://reviews.llvm.org/D126151
The purity or impurity of a call to a generic interface
depends on the attributes of the specific procedure or specific
binding. Change expression analysis of calls to generic interfaces
to replace the symbol in the parse tree with the specific procedure
or binding; this ensures that later checking for purity in
DO CONCURRENT and other contexts will be accurate.
Remove an "XFAIL" from a test that now passes again with this fix.
Differential Revision: https://reviews.llvm.org/D126150
Name resolution for subprograms checks whether the name is already
present in the enclosing scope as a generic interface, so that the
case of a generic with the same name as one of its specifics can be
handled. The particular means by which the enclosing scope is searched
for the name would resolve the name (bind a symbol to it) as a side
effect. This turns out to be the wrong thing to do when the subprogram
is going to have its symbol created in another scope to cope with its
BIND(C,NAME="name") name, and its Fortran name is already present in the
enclosing scope for a subprogram of the same name but without
BIND(C,NAME="name").
A very long explanation for a one-line fix, sorry. In short, change
the code to look up the name but not resolve it at that point.
Differential Revision: https://reviews.llvm.org/D126149
Semantics was allowing calls to CHARACTER(*) functions, which are odd
things -- they can be declared, and passed around, but can never actually
be called as such. They must be redeclared with an explicit length that
ends up being passed as a hidden argument. So check for these calls
and diagnose them, add tests, and clean up some existing tests that
were in error and now get caught.
Possible TODO for lowering: there were some test cases that used
bad calls to assumed-length CHARACTER*(*) functions and validated
their implementations. I've removed some, and adjusted another,
but the code that somehow implemented these calls may need to be
removed and replaced with an assert about bad semantics.
Differential Revision: https://reviews.llvm.org/D126148
The code in semantics that reinitializes symbol table pointers in
the parse tree of a parameterized derived type prior to a new
instantiation of the type was processing the symbols of the
derived type instantiation scope in arbitrary address order,
which could fail if a reference to a type parameter inherited from
an ancestor type was processed prior to the parent component sequence.
Fix by instantiating components of PDT instantiations in declaration
order.
Differential Revision: https://reviews.llvm.org/D126147
A dummy argument in an entry point of a subprogram with multiple
entry points need not be defined in other entry points. It is only
legal to reference such an argument when calling an entry point that
does have a definition. An entry point without such a definition
needs a local "substitute" definition sufficient to generate code.
It is nonconformant to reference such a definition at runtime.
Most such definitions and associated code will be deleted as dead
code at compile time. However, that is not always possible, as in
the following code. This code is conformant if all calls to entry
point ss set m=3, and all calls to entry point ee set n=3.
subroutine ss(a, b, m, d, k) ! no x, y, n
integer :: a(m), b(a(m)), m, d(k)
integer :: x(n), y(x(n)), n
integer :: k
1 print*, m, k
print*, a
print*, b
print*, d
if (m == 3) return
entry ee(x, y, n, d, k) ! no a, b, m
print*, n, k
print*, x
print*, y
print*, d
if (n /= 3) goto 1
end
integer :: xx(3), yy(5), zz(3)
xx = 5
yy = 7
zz = 9
call ss(xx, yy, 3, zz, 3)
call ss(xx, yy, 3, zz, 3)
end
Lowering currently generates fir::UndefOp's for all unused arguments.
This is usually ok, but cases such as the one here incorrectly access
unused UndefOp arguments for m and n from an entry point that doesn't
have a proper definition.
The problem is addressed by creating a more complete definition of an
unused argument in most cases. This is implemented in large part by
moving the definition of an unused argument from mapDummiesAndResults
to mapSymbolAttributes. The code in mapSymbolAttributes then chooses
one of three code generation options, depending on information
available there.
This patch deals with dummy procedures in alternate entries, and adds
a TODO for procedure pointers (the PFTBuilder is modified to analyze
procedure pointer symbol so that they are not silently ignored, and
instead hits proper TODOs).
BoxAnalyzer is also changed because assumed-sized arrays were wrongfully
categorized as constant shape arrays. This had no impact, except when
there were unused entry points.
Co-authored-by: jeanPerier <jperier@nvidia.com>
Differential Revision: https://reviews.llvm.org/D125867
When a Hollerith (or short character) literal is presented as an actual
argument that corresponds to a dummy argument for which a BOZ literal
would be acceptable, treat the Hollerith as if it had been a BOZ
literal in the same way -- and with the same code -- as f18 already
does for the similar extension in DATA statements.
Differential Revision: https://reviews.llvm.org/D126144
Now that the requirements and implementation of asynchronous I/O are
better understood, adjust their I/O runtime APIs. In particular:
1) Remove the BeginAsynchronousOutput/Input APIs; they're not needed,
since any data transfer statement might have ASYNCHRONOUS= and
(if ASYNCHRONOUS='YES') ID= control list specifiers that need to
at least be checked.
2) Add implementations for BeginWait(All) to check for the error
case of a bad unit number and nonzero ID=.
3) Rearrange and comment SetAsynchronous so that it's clear that
it can be called for READ/WRITE as well as for OPEN.
The implementation remains completely synchronous, but should be conforming.
Where opportunities make sense for true asynchronous implementations of
some big block transfers without SIZE= in the future, we'll need to add
a GetAsynchronousId API to capture ID= on a READ or WRITE; add sourceFile
and sourceLine arguments to BeginWait(All) for good error reporting;
track pending operations in unit.h; and add code to force synchronization
to non-asynchronous I/O operations.
Lowering should call SetAsynchronous when ASYNCHRONOUS= appears as
a control list specifier. It should also set ID=x variables to 0
until such time as we support asynchronous operations, if ever.
This patch only removes the removed APIs from lowering.
Differential Revision: https://reviews.llvm.org/D126143
Forward references to ENTRY names to pass them as actual procedure arguments
don't work in all cases, exposing some basic ordering problems in
name resolution for these symbols. Refactor; create all the
necessary procedure symbols, and either function result or host association
symbols (for subroutines), at the time that the subprogrma scope is
created, so that the names exist in the scope as text "before"
the ENTRY is processed in name resolution. Some processing
remains in PostEntryStmt() so that we can check that an ENTRY with
an explicit distinct RESULT doesn't also have declarations for the
ENTRY name.
Differential Revision: https://reviews.llvm.org/D126142
The binding names of inner procedures with BIND(C) are not exposed
to the loader and should be ignored for potential conflict errors.
Differential Revision: https://reviews.llvm.org/D126141
Intrinsic module names are not in the user's namespace, so they
are free to declare global names that conflict with intrinsic
modules.
Differential Revision: https://reviews.llvm.org/D126140
For the program provided as the test case flang fired the following
error:
error: Semantic errors in main.f90
error: 'foo' is not a procedure
This change fixes the error by postponing handling of `UseErrorDetails`
from `CharacterizeProcedure` to a later stage.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D125791
The types of lower bound, upper bound, and step are converted into the
type of the loop variable if necessary. OpenMP runtime requires 32-bit
or 64-bit loop variables. OpenMP loop iteration variable cannot have
more than 64 bits size and will be narrowed.
This patch is part of upstreaming code from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project. (#1256)
Co-authored-by: kiranchandramohan <kiranchandramohan@gmail.com>
Reviewed By: kiranchandramohan, shraiysh
Differential Revision: https://reviews.llvm.org/D125740
When parallel is used in a combined construct, then use a separate
function to create the parallel operation. It handles the parallel
specific clauses and leaves the rest for handling at the inner
operations.
Reviewed By: peixin, shraiysh
Differential Revision: https://reviews.llvm.org/D125465
Co-authored-by: Sourabh Singh Tomar <SourabhSingh.Tomar@amd.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Valentin Clement <clementval@gmail.com>
Co-authored-by: Nimish Mishra <neelam.nimish@gmail.com>
This patch basically extends https://reviews.llvm.org/D122008 with
support for MacOSX/Darwin.
To facilitate this, I've added `MacOSX` to the list of supported OSes in
Target.cpp. Flang already supports `Darwin` and it doesn't really do
anything OS-specific there (it could probably safely skip checking the
OS for now).
Note that generating executables remains hidden behind the
`-flang-experimental-exec` flag. Also, we don't need to add `-lm` on
MacOSX as `libm` is effectively included in `libSystem` (which is linked
in unconditionally).
Differential Revision: https://reviews.llvm.org/D125628
Convert Fortran parse-tree into MLIR for collapse-clause.
Includes simple Fortran to LLVM-IR test, with auto-generated
check-lines (some of which have been edited by hand).
Reviewed By: kiranchandramohan, shraiysh, peixin
Differential Revision: https://reviews.llvm.org/D125302
As Fortran 2018 15.8.1(3), in a reference to an elemental procedure, if
any argument is an array, each actual argument that corresponds to an
INTENT (OUT) or INTENT (INOUT) dummy argument shall be an array. Add
this semantic check.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D125685
This supports the lowering parse-tree to MLIR for ordered clause in
worksharing-loop directive. Also add the test case for operation
conversion.
Part of this patch is from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Co-authored-by: Sourabh Singh Tomar <SourabhSingh.Tomar@amd.com>
Reviewed By: kiranchandramohan, NimishMishra
Differential Revision: https://reviews.llvm.org/D125456
This patch re-factors the driver code in LLVM Flang (frontend +
compiler) to use the MLIR style. For more context, please see:
https://discourse.llvm.org/t/rfc-coding-style-in-the-driver/
Most changes here are rather self-explanatory. Accessors are renamed to
be more consistent with the rest of LLVM (e.g. allSource -->
getAllSources). Additionally, MLIR clang-tidy files are added in the
affected directories.
clang-tidy and clang-format files were copied from MLIR. Small
additional changes are made to silence clang-tidy/clang-format
warnings.
[1] https://mlir.llvm.org/getting_started/DeveloperGuide/
Differential Revision: https://reviews.llvm.org/D125007
The warning caused build errors on a couple flang testers that are
building with -Werror. The diagnostic change makes the generated
error correct.
This is a followup to https://reviews.llvm.org/D125549
Differential Revision: https://reviews.llvm.org/D125587
The predicate that is used to detect an invalid assumed-size argument
to the intrinsic functions SHAPE, SIZE, & LBOUND gives false results
for arguments whose shapes are not calculatable at compilation time.
Replace with an explicit test for an assumed-size array dummy argument
symbol.
Differential Revision: https://reviews.llvm.org/D125342
Complex component references (z%RE, z%IM) of complex named constants
should be evaluated at compilation time.
Differential Revision: https://reviews.llvm.org/D125341
Inquiries into the bounds, size, and length of local variables (and function results)
are acceptable specification expressions. A recent change allowed them for dummy
arguments that are not OPTIONAL or INTENT(OUT), but didn't address other object
entities.
Differential Revision: https://reviews.llvm.org/D125343
Fortran 2018 requires that a compiler allow objects whose rank + corank
is 15, and that's our maximum; detect and diagnose violations.
Differential Revision: https://reviews.llvm.org/D125153
Evaluate real-valued references to the intrinsic functions MODULO
and MOD at compilation time without recourse to an external math
library.
Differential Revision: https://reviews.llvm.org/D125151
Fold references to the intrinsic function DIM with constant real
arguments. And clean up folding of comparisons with NaNs to address
a problem noticed in testing -- NaNs should successfully compare
unequal to all values, including themselves, instead of failing all
comparisons.
Differential Revision: https://reviews.llvm.org/D125146
This patch adds lowering for task construct from Fortran to
`omp.task` operation in OpenMPDialect Dialect (mlir). Also added tests
for the same.
Reviewed By: kiranchandramohan, peixin
Differential Revision: https://reviews.llvm.org/D124138
In a function, ENTRY E without an explicit RESULT() creates a
function result entity also named E that is storage associated with
the enclosing function's result. f18 was emitting an incorrect error
message if that function result E was referenced without any
declaration prior to its ENTRY statement when it should have been
implicitly declared instead.
Differential Revision: https://reviews.llvm.org/D125144
As is already supported for dummy procedures, we need to also accept
declarations of procedure pointers that consist of a POINTER attribute
statement followed by an INTERFACE block. (The case of an INTERFACE
block followed by a POINTER statement already works.)
While cleaning this case up, adjust the utility predicate IsProcedurePointer()
to recognize it (namely a SubprogramDetails symbol with Attr::POINTER)
and delete IsProcName(). Extend tests, and add better comments to
symbol.h to document the two ways in which procedure pointers are
represented.
Differential Revision: https://reviews.llvm.org/D125139
The semantic test for an intrinsic assignment to a polymorphic
derived type entity from a type that is an extension of its base
type was reversed, so it would allow assignments that it shouldn't
and disallowed some that it should; and the test case for it
incorectly assumed that the invalid semantics were correct.
Fix the code and the test, and add a new test for the invalid
case (LHS type is an extension of the RHS type).
Differential Revision: https://reviews.llvm.org/D125135
The rules in the Fortran standard for specification expressions
are full of special cases and exceptions, and semantics didn't get
them exactly right. It is valid to refer to an INTENT(OUT) dummy
argument in a specification expression in the context of a
specification inquiry function like SIZE(); it is not valid to
reference an OPTIONAL dummy argument outside of the context of
PRESENT. This patch makes the specification expression checker
a little context-sensitive about whether it's examining an actual
argument of a specification inquiry intrinsic function or not.
Differential Revision: https://reviews.llvm.org/D125131
DATA statements in inner procedures were not treating undeclared objects
as implicitly declared variables if the DATA statement appeared in a
specification part; they were treated as host-associated symbols instead.
This was incorrect. Fix DATA statement name resolution to always treat
DATA as if it had appeared in the executable part.
Differential Revision: https://reviews.llvm.org/D125129
A disassociated procedure pointer is allowed to be passed as an absent
actual argument that corresponds to an optional dummy procedure,
but not NULL(); accept that case as well.
Differential Revision: https://reviews.llvm.org/D125127
The type compatibility checks for the ARRAY= argument and the dummy
arguments and result of the OPERATION= argument to the REDUCE intrinsic
function need to allow for parameterized data types with LEN parameters.
(Their values are required to be identical but this is not a numbered
constraint requiring a compilation time check).
Differential Revision: https://reviews.llvm.org/D125124
Functions returning ALLOCATABLE or POINTER arrays have descriptor inquiries in
their results' shape expressions that won't compare equal. These functions
need only be checked for compatible ranks (& of course other characteristics).
Differential Revision: https://reviews.llvm.org/D125123
Structure contructors for instances of parameterized derived types
must have their components' values folded in the context of the values
of the type parameters.
Differential Revision: https://reviews.llvm.org/D125116
Actual arguments whose lengths are less than the expected length
of their corresponding dummy argument are errors; but this needs
to be refined. Short actual arguments that are variables remain
errors, but those that are expressions can be (again) extended on
the right with blanks.
Differential Revision: https://reviews.llvm.org/D125115
The construction of the dimension order vector used to populate the
result array was incorrect, leading to a scrambled-looking result
for rank-3 and higher results. Fix, and extend tests.
Differential Revision: https://reviews.llvm.org/D125113
The standard requires that the operands of the intrinsic function
SIGN() must have the same type (INTEGER or REAL), but they are not
required to have the same kind.
Differential Revision: https://reviews.llvm.org/D125105
The related real number system inquiry functions SPACING()
and RRSPACING() can be folded for constant arguments.
See 16.9.164 & 16.9.180 in Fortran 2018.
Differential Revision: https://reviews.llvm.org/D125100
When the result can be known at compilation time, fold it.
Success depends on whether the operands are polymorphic.
When neither one is polymorphic, the result is known and can
be either .TRUE. or .FALSE.; when either one is polymorphic,
a .FALSE. result still can be discerned.
Differential Revision: https://reviews.llvm.org/D125062
When processing an entry-stmt in name resolution, attrs_ was
reset before SetBindNameOn was called, causing the symbol to lose
the binding label information.
Differential Revision: https://reviews.llvm.org/D125097
As Fortran 2018 5.2.2 states, a program shall consist of exactly one
main program. Add this semantic check.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D125186
The code below causes flang to crash with an exception.
After fixing the crash flang with an internal error "no symbol found for 'bar'"
This change fixes all the issues.
program name
implicit none
integer, parameter :: bar = 1
integer foo(bar) /bar*2/
end program name
Reviewed By: kiranchandramohan, klausler
Differential Revision: https://reviews.llvm.org/D124914
As Fortran 2018 states, in each where-assignment-stmt, the mask-expr and
the variable being defined shall be arrays of the same shape. The
previous check does not consider checking if it is an array.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D125022
As Fortran 2018 15.5.2.9 point 2, the actual argument and dummy argument
have the same type and type parameters and an external function with
assumed character length may be associated with a dummy argument with
explicit character length. As Fortran 2018 15.5.2.9 point 7, if an
external procedure is used as an actual argument, it can be explicitly
declared to have the EXTERNAL attribute. This supports the external
procedure passed as actual argument with implicit character type, either
explicit character length or assumed character length.
Reviewed By: Jean Perier, klausler
Differential Revision: https://reviews.llvm.org/D124345
For arrays without a constant interior or arrays of character with
dynamic length arrays, the data types are converted to a pointer to the
element type, so the scale size of the constant extents needs to be
counted. The previous AllocaOp conversion does not consider the arrays
of character with dynamic length arrays, and the previous AllocMemOp
conversion does not consider arrays without a constant interior. This
fixes them and refactors the code so that it can be shared. Also add
the test cases.
Reviewed By: Jean Perier
Differential Revision: https://reviews.llvm.org/D124766
The OpenMP worksharing loop operation in the dialect is a proper loop
operation and not a container of a loop. So we have to lower the
parse-tree OpenMP loop construct and the do-loop inside the construct
to a omp.wsloop operation and there should not be a fir.do_loop inside
it. This is achieved by skipping fir.do_loop creation and calling genFIR
for the nested evaluations in the lowering of the do construct.
Note: Handling of more clauses, parallel do, storage of loop index variable etc will come in separate patches.
Part of the upstreaming effort to move LLVM Flang from fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project to the LLVM Project.
Reviewed By: peixin
Differential Revision: https://reviews.llvm.org/D125024
Co-authored-by: Sourabh Singh Tomar <SourabhSingh.Tomar@amd.com>
Co-authored-by: Shraiysh Vaishay <Shraiysh.Vaishay@amd.com>
The FIR `do_loop` is designed as a structured operation with a single
block inside it. Presence of unstructured constructs like jumps, exits
inside the loop will cause the loop to be marked as unstructured. These
loops are lowered using the `control-flow` dialect branch operations.
Fortran semantics do not allow the loop variable to be modified inside
the loop. To prevent accidental modification, the iteration of the
loop is modeled by two variables, trip-count and loop-variable.
-> The trip-count and loop-variable are initialized in the pre-header.
The trip-count is set as (end-start+step)/step where end, start and
step have the usual meanings. The loop-variable is initialized to start.
-> The header block contains a conditional branch instruction which
selects between branching to the body of the loop or the exit block
depending on the value of the trip-count.
-> Inside the body, the trip-count is decremented and the loop-variable
incremented by the step value. Finally it branches to the header of the
loop.
Part of the upstreaming effort to move LLVM Flang from fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project to the LLVM Project.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D124837
Co-authored-by: Val Donaldson <vdonaldson@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Peter Klausler <pklausler@nvidia.com>
This change makes sure that Flang's driver recognises LLVM IR and BC as
supported file formats. To this end, `isFortran` is extended and renamed
as `isSupportedByFlang` (the latter better reflects the new
functionality).
New tests are added to verify that the target triple is correctly
overridden by the frontend driver's default value or the value specified
with `-triple`. Strictly speaking, this is not a functionality that's
new in this patch (it was added in D124664). This patch simply enables
us to write such tests and hence I'm including them here.
Differential Revision: https://reviews.llvm.org/D124667
All frontend actions that generate code (MLIR, LLVM IR/BC,
Assembly/Object Code) are re-factored as essentially one action,
`CodeGenAction`, with minor specialisations. To facilate all this,
`CodeGenAction` is extended to hold `TargetMachine` and backend action
type (MLIR vs LLVM IR vs LLVM BC vs Assembly vs Object Code).
`CodeGenAction` is no longer a pure abstract class and the
corresponding `ExecuteAction` is implemented so that it covers all use
cases. All this allows a much better code re-use.
Key functionality is extracted into some helpful hooks:
* `SetUpTargetMachine`
* `GetOutputStream`
* `EmitObjectCodeHelper`
* `EmitBCHelper`
I hope that this clarifies the overall structure. I suspect that we may
need to revisit this again as the functionality grows in complexity.
Differential Revision: https://reviews.llvm.org/D124665
*SUMMARY*
Currently, the frontend driver assumes that a target triple is either:
* provided by the frontend itself (e.g. when lowering and generating
code),
* specified through the `-triple/-target` command line flags.
If `-triple/-target` is not used, the frontend will simply use the host
triple.
This is going to be insufficient when e.g. consuming an LLVM IR file
that has no triple specified (reading LLVM files is WIP, see D124667).
We shouldn't require the triple to be specified via the command line in
such situation. Instead, the frontend driver should contain a good
default, e.g. the host triple.
This patch updates Flang's `CompilerInvocation` to do just that, i.e.
defines its default target triple. Similarly to Clang:
* the default `CompilerInvocation` triple is set as the host triple,
* the value specified with `-triple` takes precedence over the frontend
driver default and the current module triple,
* the frontend driver default takes precedence over the module triple.
*TESTS*
This change requires 2 unit tests to be updated. That's because relevant
frontend actions are updated to assume that there's always a valid
triple available in the current `CompilerInvocation`. This update is
required because the unit tests bypass the regular `CompilerInvocation`
set-up (in particular, they don't call
`CompilerInvocation::CreateFromArgs`). I've also taken the liberty to
disable the pre-precossor formatting in the affected unit tests as well
(it is not required).
No new tests are added. As `flang-new -fc1` does not support consuming
LLVM IR files just yet, it is not possible to compile an LLVM IR file
without a triple. More specifically, atm all LLVM IR files are generated
and stored internally and the driver makes sure that these contain a
valid target triple. This is about to change in D124667 (which adds
support for reading LLVM IR/BC files) and that's where tests for
exercising the default frontend driver triple will be added.
*WHAT DOES CLANG DO?*
For reference, the default target triple for Clang's
`CompilerInvocation` is set through option marshalling infra [1] in
Options.td. Please check the definition of the `-triple` flag:
```
def triple : Separate<["-"], "triple">,
HelpText<"Specify target triple (e.g. i686-apple-darwin9)">,
MarshallingInfoString<TargetOpts<"Triple">, "llvm::Triple::normalize(llvm::sys::getDefaultTargetTriple())">,
AlwaysEmit, Normalizer<"normalizeTriple">;
```
Ideally, we should re-use the marshalling infra in Flang.
[1] https://clang.llvm.org/docs/InternalsManual.html#option-marshalling-infrastructure
Differential Revision: https://reviews.llvm.org/D124664
When we pass an alternate return specifier to a regular (not an asterisk)
dummy argument, flang would throw an internal compiler error of
derefencing a null pointer.
To avoid the ICE, a check was added.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D123947
The RETURN statement is allowed in functions and subroutines, but not
in main programs. It is however a common extension, which we also
implement, to allow RETURN from main programs -- we only issue a
portability warning when -pedantic or -std=f2018 are set.
This patch fixes false positives for this portability warning, where it
was triggered also when RETURN was present in functions or subroutines.
Fixexs #55080
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D124732
This patch restricts the value of `if` clause expression to an I1 value.
It also restricts the value of `num_threads` clause expression to an I32
value.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D124142
As Fortran 2018 C1533, a nonintrinsic elemental procedure shall not be
used as an actual argument. The semantic check for implicit iterface is
missed.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D124379
MLIR has a common pattern for "arguments" that uses syntax
like `%x : i32 {attrs} loc("sourceloc")` which is implemented
in adhoc ways throughout the codebase. The approach this uses
is verbose (because it is implemented with parallel arrays) and
inconsistent (e.g. lots of things drop source location info).
Solve this by introducing OpAsmParser::Argument and make addRegion
(which sets up BlockArguments for the region) take it. Convert the
world to propagating this down. This means that we correctly
capture and propagate source location information in a lot more
cases (e.g. see the affine.for testcase example), and it also
simplifies much code.
Differential Revision: https://reviews.llvm.org/D124649
A recent change is eliciting a valid warning from the out-of-tree
flang build bot; fix by using a reference in a range-based for().
Differential Revision: https://reviews.llvm.org/D124682
Semantics is not preventing a named common block to appear with
different size in a same file (named common block should always have
the same storage size (see Fortran 2018 8.10.2.5), but it is a common
extension to accept different sizes).
Lowering was not coping with this well, since it just use the first
common block appearance, starting with BLOCK DATAs to define common
blocks (this also was an issue with the blank common block, which can
legally appear with different size in different scoping units).
Semantics is also not preventing named common from being initialized
outside of a BLOCK DATA, and lowering was dealing badly with this,
since it only gave an initial value to common blocks Globals if the
first common block appearance, starting with BLOCK DATAs had an initial
value.
Semantics is also allowing blank common to be initialized, while
lowering was assuming this would never happen, and was never creating
an initial value for it.
Lastly, semantics was not complaining if a COMMON block was initialized
in several scoping unit in a same file, while lowering can only generate
one of these initial value.
To fix this, add a structure to keep track of COMMON block properties
(biggest size, and initial value if any) at the Program level. Once the
size of a common block appearance is know, the common block appearance
is checked against this information. It allows semantics to emit an error
in case of multiple initialization in different scopes of a same common
block, and to warn in case named common blocks appears with different
sizes. Lastly, this allows lowering to use the Program level info about
common blocks to emit the right GlobalOp for a Common Block, regardless
of the COMMON Block appearances order: It emits a GlobalOp with the
biggest size, whose lowest bytes are initialized with the initial value
if any is given in a scope where the common block appears.
Lowering is updated to go emit the common blocks before anything else so
that the related GlobalOps are available when lowering the scopes where
common block appear. It is also updated to not assume that blank common
are never initialized.
Differential Revision: https://reviews.llvm.org/D124622
The asm parser had a notional distinction between parsing an
operand (like "%foo" or "%4#3") and parsing a region argument
(which isn't supposed to allow a result number like #3).
Unfortunately the implementation has two problems:
1) It didn't actually check for the result number and reject
it. parseRegionArgument and parseOperand were identical.
2) It had a lot of machinery built up around it that paralleled
operand parsing. This also was functionally identical, but
also had some subtle differences (e.g. the parseOptional
stuff had a different result type).
I thought about just removing all of this, but decided that the
missing error checking was important, so I reimplemented it with
a `allowResultNumber` flag on parseOperand. This keeps the
codepaths unified and adds the missing error checks.
Differential Revision: https://reviews.llvm.org/D124470
This patch adds code to lower simple Fortran Do loops with loop control.
Lowering is performed by the the `genFIR` function when called with a
`Fortran::parser::DoConstruct`. `genFIR` function calls `genFIRIncrementLoopBegin`
then calls functions to lower the body of the loop and finally calls
the function `genFIRIncrementLoopEnd`. `genFIRIncrementLoopBegin` is
responsible for creating the FIR `do_loop` as well as storing the value of
the loop index to the loop variable. `genFIRIncrementLoopEnd` returns
the incremented value of the loop index and also stores the index value
outside the loop. This is important since the loop variable can be used
outside the loop. Information about a loop is collected in a structure
`IncrementLoopInfo`.
Note 1: Future patches will bring in lowering for unstructured,
infinite, while loops
Note 2: This patch is part of upstreaming code from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: awarzynski
Differential Revision: https://reviews.llvm.org/D124277
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Val Donaldson <vdonaldson@nvidia.com>
Co-authored-by: Peter Klausler <pklausler@nvidia.com>
Co-authored-by: Valentin Clement <clementval@gmail.com>
Previously MASK= elements were accessed in assumption that mask is an array of
input argument rank (and in combination with explicit DIM= argument we had
out-of-bounds access), but for MAXLOC/MINLOC/FINDLOC mask should be be
conformable and could be scalar.
Add new regression tests with scalar mask for verification.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D124408
This patch provides the basic infrastructure for lowering declarative
constructs for OpenMP and OpenACC.
This is part of the upstreaming effort from the fir-dev branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Reviewed By: kiranchandramohan, shraiysh, clementval
Differential Revision: https://reviews.llvm.org/D124225
A non-CHARACTER expression in a CASE statement is allowed to have
a distinct kind (not type) from the expression in its SELECT CASE.
If a value in a CASE statement is out of range for the SELECT CASE
type, emit a warning, but it should not be a fatal error.
Differential Revision: https://reviews.llvm.org/D124544
Name resolution fails with a bogus "is not a variable" error message
when a host-associated object appears in a NAMELIST group. The root
cause is that ConvertToObjectEntity() returns false for host-associated
objects. Fix that, and also apply a similar fix to ConvertToProcEntity()
nearby.
Differential Revision: https://reviews.llvm.org/D124541
Similarly to LBOUND in https://reviews.llvm.org/D123237, fix UBOUND() folding
for constant arrays (for both w/ and w/o DIM=): convert
GetConstantArrayLboundHelper into common helper class for both lower/upper
bounds.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D123520
Lowering of FailImage statement generates a runtime call and the
unreachable operation. The unreachable operation cannot terminate
a structured operation like the IF operation, hence mark as
unstructured.
Note: This patch is part of upstreaming code from the fir-dev branch of
https://github.com/flang-compiler/f18-llvm-project.
Reviewed By: clementval
Differential Revision: https://reviews.llvm.org/D124520
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
This patch basically implements [1] in ExecuteCompilerInvocation.cpp. It
also:
* replaces `CreateFrontendBaseAction` with `CreateFrontendAction`
(only one method is needed ATM, this change removes the extra
indirection)
* removes `InvalidAction` from the `ActionKind` enum (I don't think it
adds much and keeping it would mean adding a new void case in
`CreateFrontendAction`)
* sets the default frontend action in FrontendOptions.h to
`ParseSyntaxOnly` (note that this is still overridden independently
in `ParseFrontendArg` in CompilerInvocation.cpp)
No new functionality is added, hence no tests.
[1] https://llvm.org/docs/CodingStandards.html#don-t-use-default-labels-in-fully-covered-switches-over-enumerations
Differential Revision: https://reviews.llvm.org/D124245
Semantics now needs to preserve the parse trees from module files,
in case they contain parameterized derived type definitions with
component initializers that may require re-analysis during PDT
instantiation. Save them in the SemanticsContext.
Differential Revision: https://reviews.llvm.org/D124467
A recent change assumed that the native C++ "long double" maps to
a Fortran data type; but this turns out to not be true for ppc64le,
which uses "double-double" for "long double".
This is a quick patch to get the ppc64le flang build bot back up.
A better fix that either uses HostTypeExists<> or replaces "long double"
with "ieee128_t" (or some other solution) is expected to follow soon.
Differential Revision: https://reviews.llvm.org/D124423
At the top level of program units in a source file, two subprograms
are allowed to have the same name if at least one of them has a
distinct interoperable binding name. F18's symbol table requires
(most) symbols in a scope to have distinct names, though. Solve
by using compiler-created names for the symbols of global scope
subprograms that have interoperable binding names.
Differential Revision: https://reviews.llvm.org/D124295
F(X)=Y may be initially parsed as a statement function definition; an
existing pass will detect statement functions that should be rewritten
into assignment statemets with array element references as their
left-hand side variables. However, F() may also be a reference to a
function that returns a data pointer, and f18 did not handle this
case correctly.
The right fix is to rewrite the parse tree for F(X)=Y into an assignment
to a function reference result. The cases that are actually assignments
to array elements -- including all of the cases previously handled --
will have their left-hand sides converted to array element references
later by another existing rewriting step.
Differential Revision: https://reviews.llvm.org/D124299
The lowering code was mistakenly assuming that the second argument
in the signature provided by semantics is the DIM argument. This
caused calls with a KIND argument but no DIM to be lowered as if the
KIND argument was DIM.
Differential Revision: https://reviews.llvm.org/D124243
In some case the lowering of `ichar` is generating an `arith.extui` operation
with the same from/to type. This operation do not accept from/to types to be
the same. If the from/to types are identical, we do not generate the extra
operation.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D124107
Blanks are allowed in more places than I allowed for, and
"NAN(foobar)" is allowed to have any parenthesis-balanced
characters in parentheses.
Update: Fix up old sanity test, then avoid usage of "limit" when null.
Differential Revision: https://reviews.llvm.org/D124294
A recent change that corrected the name resolution of a generic interface
when the same name was visible in scope incorrectly prevented a local
generic from shadowing an outer name that is not a generic, subprogram,
or derived type -- e.g., a simple variable -- leading to an inappropriate
error message.
Differential Revision: https://reviews.llvm.org/D124276
This patch adds a few new member methods in the `PluginParseTreeAction`
frontend action base class. With these new methods, the plugin API
becomes independent of the driver internals. In particular, plugin
writers no longer require the `CompilerInstance.h` header file to access
various driver data structures (instead, they can use newly added
hooks).
This change is desirable as `CompilerInstance.h` includes various
headers from Clang (both explicitly and implicitly). Some of these
header files are generated at build time (through TableGen) and
including them creates a dependency on some of Clang's build targets.
However, plugins in Flang should not depend on Clang build targets.
Note that plugins might still work fine most of the time, even without
this change and without adding Clang build targets as dependency in
plugin's CMake definition. Indeed, these Clang build targets are often
generated early in the build process. However, that's not guaranteed and
we did notice that on occasions plugins would fail to build.
Differential Revision: https://reviews.llvm.org/D120999
Transformational bessel intrinsic functions require the same math runtime
as elemental bessel intrinsics.
Currently elemental bessels could be folded if f18 was linked with pgmath
(cmake -DLIBPGMATH_DIR option). `j0`, `y0`, ... C libm functions were not
used because they are not standard C functions: they are Posix
extensions.
This patch enable:
- Using the Posix bessel host runtime functions when available.
- folding the transformational bessel using the elemental version.
Differential Revision: https://reviews.llvm.org/D124167
The following code causes the compiler to ICE in several places due to
lack of support of recursive procedure definitions through the function
result.
function foo() result(r)
procedure(foo), pointer :: r
end function foo
This patch adds lowering support for atomic read and write constructs.
Also added is pointer modelling code to allow FIR pointer like types to
be inferred and converted while lowering.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D122725
Co-authored-by: Kiran Chandramohan <kiran.chandramohan@arm.com>
Set LBOUND() constant folding for parentheses expr. as ones
Array bounds should not propagate throught omitted bounds specifications or
temporary variables - fix constant folding in case of Parentheses<T> expression
by explicitly returning array of ones (or scalar in case of DIM=).
Add set of tests for (x) bounds checks (w/ and w/o 'parameter' arrays)
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D123838
Previously constant folding uses 'dim' without checks which leads to ICE if we
do not have DIM= parameter. And for inputs without DIM= we need to form an
array of rank size with computed bounds instead of single value.
Add additional PackageConstant function to simplify 'if (dim)' handling since we
need to distinguish between scalar initialization in case of DIM= argument and
rank=1 array.
Also add a few more tests with 'parameter' type to verify folding for constant
arrays.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D123237
A missing "!" in the call interface lowering caused all derived type
arguments without length parameters that require and explicit interface
to be passed via fir.box (runtime descriptor).
This was not the intent: there is no point passing a simple derived type
scalars or explicit shapes by descriptor just because they have an attribute
like TARGET. This would actually be problematic with existing code that is
not always 100% compliant: some code implicitly calls procedures with
TARGET dummy attributes (this is not something a compiler can enforce
if the call and procedure definition are not in the same file).
Add a Scope::IsDerivedTypeWithLengthParameter to avoid passing derived
types with only kind parameters by descriptor. There is no point, the
callee knows about the kind parameter values.
Differential Revision: https://reviews.llvm.org/D123990
When resolving a procedure reference, do not allow a successful
intrinsic procedure probe result to override an existing
symbol.
Differential Revision: https://reviews.llvm.org/D123905
Adds flang/include/flang/Common/log2-visit.h, which defines
a Fortran::common::visit() template function that is a drop-in
replacement for std::visit(). Modifies most use sites in
the front-end and runtime to use common::visit().
The C++ standard mandates that std::visit() have O(1) execution
time, which forces implementations to build dispatch tables.
This new common::visit() is O(log2 N) in the number of alternatives
in a variant<>, but that N tends to be small and so this change
produces a fairly significant improvement in compiler build
memory requirements, a 5-10% improvement in compiler build time,
and a small improvement in compiler execution time.
Building with -DFLANG_USE_STD_VISIT causes common::visit()
to be an alias for std::visit().
Calls to common::visit() with multiple variant arguments
are referred to std::visit(), pending further work.
This change is enabled only for GCC builds with GCC >= 9;
an earlier attempt (D122441) ran into bugs in some versions of
clang and was reverted rather than simply disabled; and it is
not well tested with MSVC. In non-GCC and older GCC builds,
common::visit() is simply an alias for std::visit().
The x%KIND inquiry needs to be supported when 'x' is itself
a complex part reference or a type parameter inquiry.
Differential Revision: https://reviews.llvm.org/D123733
A POINTER attribute statement is allowed to add the POINTER attribute
to a procedure entity that has already been declared, e.g. with an
INTERFACE block.
Differential Revision: https://reviews.llvm.org/D123732
f18 was emitting a warning about short character actual arguments to
subprograms and statement functions; every other compiler considers this
case to be an error.
Differential Revision: https://reviews.llvm.org/D123731
For parameterized derived type component initializers whose
expressions' types depend on parameter values, f18's current
scheme of analyzing the initialization expression once during
name resolution fails. For example,
type :: pdt(k)
integer, kind :: k
real :: component = real(0.0, kind=k)
end type
To handle such cases, it is necessary to re-analyze the parse
trees of these initialization expressions once for each distinct
initialization of the type.
This patch adds code to wipe an expression parse tree of its
typed expressions, and update those of its symbol table pointers
that reference type parameters, and then re-analyze that parse
tree to generate the properly typed component initializers.
Differential Revision: https://reviews.llvm.org/D123728
Prior to this patch, the semantics utility GetExpr() will crash
unconditionally if it encounters a typed expression in the parse
tree that has not been set by expression semantics. This is the
right behavior when called from lowering, by which time it is known
that the program had no fatal user errors, since it signifies a
fatal internal error. However, prior to lowering, in the statement
semantics checking code, a more nuanced test should be used before
crashing -- specifically, we should not crash in the face of a
missing typed expression when in error recovery mode.
Getting this right requires GetExpr() and its helper class to have
access to the semantics context, so that it can check AnyFatalErrors()
before crashing. So this patch touches nearly all of its call sites.
Differential Revision: https://reviews.llvm.org/D123873
When a procedure pointer or procedure dummy argument has a
defined interface, the rank of the pointer (or dummy) is the
rank of the interface.
Also tweak code discovered in shape analysis when investigating
this problam so that it returns a vector of emptied extents rather
than std::nullopt when the extents are not scope-invariant, so that
the rank can at least be known.
Differential Revision: https://reviews.llvm.org/D123727
To avoid clashing with names of user derived types, the redundant
syntax TYPE(intrinsic type spec) must be interpreted as a monomorphic
derived type when "intrinsic type spec" is a single word. This
affects TYPE(BYTE) and TYPE(DOUBLECOMPLEX), but not TYPE(DOUBLE COMPLEX)
in free form source.
Differential Revision: https://reviews.llvm.org/D123724
Items in NAMELIST groups might be host-associated implicitly-typed
variables, but name resolution can't know that when the NAMELIST
appears in a specification part and the host's execution part has
not yet been analyzed. So defer NAMELIST group item name resolution
to the end of the execution part. This is safe because nothing
else in name resolution depends on whether a variable is in a
NAMELIST group or not.
Differential Revision: https://reviews.llvm.org/D123723
Construct entities from ASSOCIATE, SELECT TYPE, and SELECT RANK
are modifiable if the are associated with modifiable variables
without vector subscripts. Update WhyNotModifiable() to accept
construct entities that are appropriate.
A need for more general error reporting from one overload of
WhyNotModifiable() caused its result type to change to
std::optional<parser::Message> instead of ::MessageFixedText,
and this change had some consequences that rippled through
call sites.
Some test results that didn't allow for modifiable construct
entities needed to be updated.
Differential Revision: https://reviews.llvm.org/D123722
TYPE IS and CLASS IS guards in SELECT TYPE constructs are
allowed to specify the same type as the type of the selector
but f18's implementation of that predicate required strict
equality of the derived type representations. We need to
allow for assumed values of LEN type parameters to match
explicit and deferred type parameter values in the selector
and require equality for KIND type parameters. Implement
DerivedTypeSpec::Match() to perform this more relaxed type
comparison, and use it in check-select-type.cpp.
Differential Revision: https://reviews.llvm.org/D123721
It is possible for generic interfaces of equivalent (but not necessarily
identical -- operator(.eq.) is equivalent to operator(==)) names to
be declared in a host scope and a nested scope, and the nested declaration
should function as an extension of the host's.
Differential Revision: https://reviews.llvm.org/D123719
A generic interface (however spelled) can have the same name as
an intrinsic procedure in the same scope. When an explicit INTRINSIC
attribute statement appears in a nested scope, semantics was
unconditionally declaring a new symbol that hid the generic entirely.
Catch this case and create instead a host association symbol for
the generic that can then be decorated with the INTRINSIC attribute.
Differential Revision: https://reviews.llvm.org/D123718
Fortran allows a generic interface to have he same name as an
intrinsic procedure. If the intrinsic is explicitly marked with
the INTRINSIC attribute, restrictions apply (C848) - the generic
must contain only functions or subroutines, depending on the
intrinsic. Explicit or not, the generic overrides the intrinsic,
but the intrinsic behavior must still be available for calls
whose actual arguments do not match any of the specific procedures.
Semantics was not checking constraint C848, and it didn't allow
an explicit INTRINSIC attribute on a name of a generic interface.
Differential Revision: https://reviews.llvm.org/D123713
Shape analysis of RESHAPE(..., SHAPE=s) should of course return
the SHAPE= actual argument when it is constant; but when it is
not, its length is still known, and thus so is the rank of the
result of RESHAPE(), and shape analysis should at least return
a shape vector of the right length rather than a result that
makes the result appear to be a scalar, which can lead to some
bogus error messages.
Also, while here: rename a private GetShapeHelper::AsShape()
routine so that it can't be confused with the ones in the API
of shape.h.
Differential Revision: https://reviews.llvm.org/D123712
When a type specification appears in the prefix of a FUNCTION statement,
defer its processing as late as possible so that any symbols in the
tpe specification can be resolved in the function's scope to local
declarations, including use-associated symbols. f18 was already doing
this deferral in a limited form for derived types, and this patch
makes it work for intrinsic type parameter values as well.
In short, "real(kind(x)) function foo(x)" now works as it should.
"As late as possible" means the end of the specification part, or
the first appearance of the function result name in the specification
part.
Differential Revision: https://reviews.llvm.org/D123705
Fortran admits a few ways to have multiple symbols with the
same name in the same scope. Two of them involve generic
interfaces (from INTERFACE or GENERIC, the syntax doesn't matter);
these are allowed to inhabit a scope with either a derived type or
a subprogram that is also a specific procedure of the generic.
(But not both a derived type and a subprogram; they could not
cohabit a scope anyway, generic or not.)
In cases of USE association, f18 needs to be capable of combining
use-associated generic interfaces with other use-associated entities.
Two generics get merged (this case was nearly correct); a generic
and a derived type can merge into a GenericDetails with a shadowed
derivedType(); and a generic can replace or ignore a use-associated
procedure of the same name so long as that procedure is already
one of its specifics.
Further, these modifications to the use-associated generic
interface must be made to a local copy of the symbol. The previous
code was messing directly with the symbol in the module's scope.
The fix is basically a reimplementation of the member function
DoAddUse() in name resolution.
Differential Revision: https://reviews.llvm.org/D123704
Error messages can have a list of attachments; these are used to point
to related source locations, supply additional information, and to
encapsulate error messages that were *not* emitted in a given context
to explain why a warning was justified.
This patch adds a message severity ("Because") for that last case,
and extends to AttachTo() API to provide a means for overriding
the severity of an attached message.
Some existing message attachments had their severities adjusted,
now that we're printing them. And operator==() for Message was
cleaned up while debugging after I noticed that it was recursively
O(N**2) and subject to returning a false positive.
Differential Revision: https://reviews.llvm.org/D123710
The intrinsics DREAL, DIMAG, and DCONJG are from Fortran 77 extensions.
For DREAL, the type of argument is extended to any complex. For DIMAG
and DCONJG, the type of argument for them should be complex(8). For DIMAG,
the result type should be real(8). For DCONJG, the result type should be
complex(8). Fix the intrinsic interface for them and add test cases for
the semantic checks and the lowering.
Reviewed By: Jean Perier
Differential Revision: https://reviews.llvm.org/D123459
The float number is represented as (-1)^s * 1.f * 2^(-127) for 32-bit,
where s is the signed flag, f is the mantissa. When the exponent bits
are all zeros, the float number is represented as (-1)^s * 0.f *2^(-126)
for 32-bit, in which case, the intPart is '0'.
Reviewed By: Jean Perier
https://reviews.llvm.org/D123673
During real range reduction to [0.5, 4) with
SQRT(2**(2a) * x) = SQRT(2**(2a)) * SQRT(x) = 2**a * SQRT(x)
we fall into inf. recursion if IsZero() == true.
Explicitly handle SQRT(0.0) instead of additional checks during folding. Also
add helpers for +0.0/-0.0 generation to clean up a bit.
Reviewed By: klausler
Differential Revision: https://reviews.llvm.org/D123131
The semantics of `-mmlir` are identical to `-mllvm`. The only notable
difference is that `-mmlir` options should be forwarded to MLIR rather
than LLVM.
Note that MLIR llvm::cl options are lazily constructed on demand (see
the definition of options in PassManagerOptions.cpp). This means that:
* MLIR global options are only visible when explicitly initialised and
displayed only when using `-mmlir --help`,
* Flang and LLVM global options are always visible and displayed when
using either `-mllvm -help` or `-mmlir --help`.
In other words, `-mmlir --help` is a superset of `-mllvm --help`. This is not
ideal, but we'd need to refactor all option definitions in Flang and
LLVM to improve this. I suggesting leaving this for later.
Differential Revision: https://reviews.llvm.org/D123297
https://reviews.llvm.org/D123211 broke builds that set both
`LLVM_BUILD_LLVM_DYLIB` and `LLVM_LINK_LLVM_DYLIB` (see [1]). This patch
fixes that.
The build failure was caused by the fact that the LLVMPasses library,
which is an LLVM "component", was listed directly as link-time
dependency. Instead, one should use `LINK_COMPONENTS` in CMake files. I
made an identical mistake recently and then subsequently fixed it in
https://reviews.llvm.org/D121461 - please visit that revision for more
detail.
I'm merging this without a review. The change is straightforward, we
recently discussed it and I was able to confirm locally that it fixes
the build issue.
[1] https://lab.llvm.org/buildbot/#/builders/177/builds/4619
Catch and nicely describe errors in CASE range values
that are out of range for the type of the SELECT CASE.
Differential Revision: https://reviews.llvm.org/D123708
When padding is required in a COMMON block to ensure alignment
of a component, emit a portability warning.
Differential Revision: https://reviews.llvm.org/D123706
This patch supports lowering parse-tree to MLIR of ordered threads
directive following Section 2.19.9 of the OpenMP 5.1 standard.
This is part of the upstreaming effort from the fir-dev branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Reviewed By: shraiysh
Differential Revision: https://reviews.llvm.org/D123590
This patch supports the following checks for ORDERED construct:
```
[5.1] 2.19.9 ORDERED Construct
The worksharing-loop or worksharing-loop SIMD region to which an ordered
region corresponding to an ordered construct without a depend clause
binds must have an ordered clause without the parameter specified on the
corresponding worksharing-loop or worksharing-loop SIMD directive.
The worksharing-loop region to which an ordered region that corresponds
to an ordered construct with any depend clauses binds must have an
ordered clause with the parameter specified on the corresponding
worksharing-loop directive.
An ordered construct with the depend clause specified must be closely
nested inside a worksharing-loop (or parallel worksharing-loop)
construct.
An ordered region that corresponds to an ordered construct with the simd
clause specified must be closely nested inside a simd or
worksharing-loop SIMD region.
```
Reviewed By: kiranchandramohan, shraiysh, NimishMishra
Differential Revision: https://reviews.llvm.org/D113399
Support for generating LLVM BC files is added in Flang's compiler and
frontend drivers. This requires the `BitcodeWriterPass` pass to be run
on the input LLVM IR module and is implemented as a dedicated frontend
aciton. The new functionality as seen by the user (compiler driver):
```
flang-new -c -emit-llvm file.90
```
or (frontend driver):
```
flang-new -fc1 -emit-llvm-bc file.f90
```
The new behaviour is consistent with `clang` and `clang -cc1`.
Differential Revision: https://reviews.llvm.org/D123211
Privatisation creates local copies of variables in the OpenMP region.
Two functions `createHostAssociateVarClone` and `copyHostAssociateVar`
are added to create a clone of the variable for basic privatisation and to
copy the contents for first-privatisation.
Note: Tests for more data-types will be added when the fir.do_loop is
upstreamed.
This is part of the upstreaming effort from the fir-dev branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Reviewed By: peixin, NimishMishra
Differential Revision: https://reviews.llvm.org/D122595
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Peter Klausler <pklausler@nvidia.com>
Co-authored-by: Valentin Clement <clementval@gmail.com>
Co-authored-by: Sourabh Singh Tomar <SourabhSingh.Tomar@amd.com>
Co-authored-by: Nimish Mishra <neelam.nimish@gmail.com>
Co-authored-by: Peixin-Qiao <qiaopeixin@huawei.com>
Handle dynamic optional argument in GET_COMMAND_ARGUMENT and GET_ENVIRONMENT_VARIABLE
(previously compiled but caused segfaults). The previous code
handled static presence/absence aspects, but not when an absent dummy optional was
passed to one of the optional intrinsic arguments.
Simplify the runtime call lowering to simply lower the runtime call without
dealing with optionality there. This keeps the optional handling logic in
IntrinsicCall.cpp.
Note that the new code will generate some extra "if (not null addr )/then/else"
when the actual arguments are always there at runtime. That makes the implementation
a lot simpler/safer, and I think it is OK for now (I do not expect these runtime
function to be called in hot loop nests).
Differential Revision: https://reviews.llvm.org/D123388
The information about OpenMP/OpenACC declarative directives in modules
should be carried in export mod files. This supports export OpenMP
Threadprivate directive and import OpenMP declarative directives.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D120396
This patch avoids to fold `fir.box_addr` when the defining `fir.embox` op
has a slice. If the op is folded the slice information are lost.
This kind of problem occurred with code like:
```
call check(y(half+1:))
```
where `y` is an array.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D123392
As Fortran 2018 3.18 states, the branch target statement can be
`forall-construct-stmt`, but cannot be `forall-stmt`. `forall-stmt` is
wrapped by `Statement` in `action-stmt` and `action-stmt` can be one
branch target statement. Fix the semantic analysis and add two
regression test cases in lowering.
Reviewed By: Jean Perier
Differential Revision: https://reviews.llvm.org/D123373
This patch adds translation for allocate clause for parallel and single
constructs.
Also added tests for block constructs.
This patch also adds tests for parallel construct which were not added earlier.
Reviewed By: NimishMishra, peixin
Differential Revision: https://reviews.llvm.org/D122483
Co-authored-by: Sourabh Singh Tomar <SourabhSingh.Tomar@amd.com>
The actual argument passed to STATUS may be a dummy OPTIONAL or a
disassociated POINTER/unallocated ALLOCATABLE.
Differential Revision: https://reviews.llvm.org/D123380
isAbsent/isPresent helpers only give information about static presence
of intrinsic arguments. Many intrinsic arguments optionality is dynamic
(an absent dummy can legally be passed to these intrinsics). This
requires a different handling (like `handleDynamicOptional`).
Rename the helpers to avoid misleading coder/reader into thinking all
optionality cases are covered by them.
Differential Revision: https://reviews.llvm.org/D123378
This patch revamps the BranchOpInterface a bit and allows a proper implementation of what was previously `getMutableSuccessorOperands` for operations, which internally produce arguments to some of the block arguments. A motivating example for this would be an invoke op with a error handling path:
```
invoke %function(%0)
label ^success ^error(%1 : i32)
^error(%e: !error, %arg0 : i32):
...
```
The advantages of this are that any users of `BranchOpInterface` can still argue over remaining block argument operands (such as `%1` in the example above), as well as make use of the modifying capabilities to add more operands, erase an operand etc.
The way this patch implements that functionality is via a new class called `SuccessorOperands`, which is now returned by `getSuccessorOperands`. It basically contains an `unsigned` denoting how many operator produced operands exist, as well as a `MutableOperandRange`, which are the usual forwarded operands we are used to. The produced operands are assumed to the first few block arguments, followed by the forwarded operands afterwards. The role of `SuccessorOperands` is to provide various utility functions to modify and query the successor arguments from a `BranchOpInterface`.
Differential Revision: https://reviews.llvm.org/D123062
Fix https://github.com/flang-compiler/f18-llvm-project/issues/1416.
The `constRows` variable was being decremented too soon, causing the
last constant interior dimension extent being used to multiply the GEP
offset. This lead to wrong address computation and caused segfaults.
Note: also upstream fir.embox tests that can be upstreamed.
Differential Revision: https://reviews.llvm.org/D123130
Do not use the shift of a fir.embox to set lower bounds if there is
a fir.slice operand. This matches Fortran semantics where lower bounds
of array sections are ones.
Note that in case there is a fir.slice, the array shift may be provided
because it is used to calculate the origin/base address of an array slice.
Add a TODO for substring codegen since I noticed it was not upstreamed
yet and would cause some program to silently compile incorrectly.
Differential Revision: https://reviews.llvm.org/D123123
Any header or module file in the Flang source directory is of no use to
the compiler unless it is copied into the build directory. Indeed, all
compiler search paths are relative to the compiler executable (flang-new
in our case). Hence, "omp_lib.h" should be copied into the build
directory alongside other compiler-provided files that can be "included"
(header files) or "used" (module files).
For now, "omp_lib.h" is copied into "<build-dir>/include/flang/OpenMP".
We may decide to change this in future. For example, Clang copies a
bunch of runtime headers into “<build-dir>/lib/clang/<version-number>”.
We could also consider using a similar header from a different
sub-project.
Flang's driver search path is updated accordingly. A rule for
"installing" the "omp_lib.h" header is _yet to be added_ (we will also
need to determine the suitable location for this).
Differential Revision: https://reviews.llvm.org/D122015
This commit restructures how TypeID is implemented to ideally avoid
the current problems related to shared libraries. This is done by changing
the "implicit" fallback path to use the name of the type, instead of using
a static template variable (which breaks shared libraries). The major downside to this
is that it adds some additional initialization costs for the implicit path. Given the
use of type names for uniqueness in the fallback, we also no longer allow types
defined in anonymous namespaces to have an implicit TypeID. To simplify defining
an ID for these classes, a new `MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID` macro
was added to allow for explicitly defining a TypeID directly on an internal class.
To help identify when types are using the fallback, `-debug-only=typeid` can be
used to log which types are using implicit ids.
This change generally only requires changes to the test passes, which are all defined
in anonymous namespaces, and thus can't use the fallback any longer.
Differential Revision: https://reviews.llvm.org/D122775
The actual argument shall have deferred the same type parameters as
the dummy argument if the argument is allocatable or pointer variable.
Currently programs not following this get one crash during execution.
Reviewed By: Jean Perier
Differential Revision: https://reviews.llvm.org/D122779
These are mostly small changes to make the code a bit clearer and more
consistent. Summary of changes:
* add missing namespace qualifiers (that's the preference in Flang)
* replace const member methods with static methods (to avoid passing
the *this pointer unnecessarily)
* rename `currentObjTy` (current object type) as `cpnTy` (component
type) - the latter feels more fitting
* remove redundant `return failure();` calls (` return
mlir::emitError` gives the same result)
* updated a few comments
Differential Revision: https://reviews.llvm.org/D122799
In case a character component PDT length only depends on kind parameters,
fold it while instantiating the PDT. This is especially important if the
component has an initializer because later semantic phases (offset
computation or runtime type info generation) might get confused and
generate offset/type info that will lead to crashes in lowering.
Differential Revision: https://reviews.llvm.org/D122938
This patch addes some global initialization and global
box initialization tests.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: schweitz
Differential Revision: https://reviews.llvm.org/D122881
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Previously, some semantic checks that are checking if an entity is an
allocatable were relying on the expression being a designator whose
last symbol has the allocatable attribute.
This is wrong since this was considering substrings and array sections of
allocatables as being allocatable. This is wrong (see NOTE 2 in
Fortran 2018 section 9.5.3.1).
Add evaluate::IsAllocatableDesignator to correctly test this.
Also add some semantic tests for ALLOCATED to test the newly added helper.
Note that ifort and nag are rejecting coindexed-named-object in
ALLOCATED (`allocated(coarray_scalar_alloc[2])`).
I think it is wrong given allocated argument is intent(in) as per
16.2.1 point 3.
So 15.5.2.6 point 4 regarding allocatable dummy is not violated (If the actual
argument is a coindexed object, the dummy argument shall have the INTENT (IN)
attribute.) and I think this is valid. gfortran accepts it.
The need for this helper was exposed in https://reviews.llvm.org/D122779.
Differential Revision: https://reviews.llvm.org/D122899
Co-authored-by: Peixin-Qiao <qiaopeixin@huawei.com>
When including debug lines as code, the `D` should be considered as
a white space. Currently an error was raised about bad labels because
it the `D` remained a `D` when considering the source line as code.
Differential Revision: https://reviews.llvm.org/D122711
When folding MAXLOC/MINLOC, the current element being compared was moved twice
in row in case it became the new extremum. With numeric and logical types, it
made no difference (std::move is a no-op for them), but for characters
where the string storage is actually moved, it caused the new extremum to
be set to the empty string, leading to wrong results.
Note: I could have left the first std::move relating to logical Findloc, but it
brings nothing and makes the code less auditable, so I also removed it.
Differential Revision: https://reviews.llvm.org/D122590
This patch adds translation for parallel sections from PFT to MLIR.
Reviewed By: kiranchandramohan, NimishMishra
Differential Revision: https://reviews.llvm.org/D122464
Implement constant folding for the intrinsic function NEAREST()
and the related functions IEEE_NEXT_AFTER(), IEEE_NEXT_UP(), and
IEEE_NEXT_DOWN().
Differential Revision: https://reviews.llvm.org/D122510
- Adds default implementations of `isDefinedOutsideOfLoop` and `moveOutOfLoop` since 99% of all implementations of these functions were identical
- `moveOutOfLoop` takes one operation and doesn't return anything anymore. 100% of all implementations of this function would always return `success` and uses would either respond with a pass failure or an `llvm_unreachable`.
Enable lowering to the relaxed and precise variants in the pgmath
library.
This is part of the upstreaming effort from the fir-dev branch in [1].
[1] https://github.com/flang-compiler/f18-llvm-project
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Peter Klausler <pklausler@nvidia.com>
Reviewed By: clementval
Differential Revision: https://reviews.llvm.org/D122484
This patch adds the lowering of coarray statements to the runtime
functions. The runtime functions are currently not implemented.
This patch is part of the upstreaming effort from fir-dev branch.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D122466
This patch adds the ReductionClauseInterface and also adds reduction
support for `omp.parallel` operation.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D122402
Adds flang/include/flang/Common/visit.h, which defines
a Fortran::common::visit() template function that is a drop-in
replacement for std::visit(). Modifies most use sites in
the front-end and runtime to use common::visit().
The C++ standard mandates that std::visit() have O(1) execution
time, which forces implementations to build dispatch tables.
This new common::visit() is O(log2 N) in the number of alternatives
in a variant<>, but that N tends to be small and so this change
produces a fairly significant improvement in compiler build
memory requirements, a 5-10% improvement in compiler build time,
and a small improvement in compiler execution time.
Building with -DFLANG_USE_STD_VISIT causes common::visit()
to be an alias for std::visit().
Calls to common::visit() with multiple variant arguments
are referred to std::visit(), pending further work.
Differential Revision: https://reviews.llvm.org/D122441
Assignment semantics was coughing up bad errors and crashes for
intrinsic assignments to unlimited polymorphic entities while
looking for any (impossible) user defined ASSIGNMENT(=) generic
or intrinsic type conversion.
Differential Revision: https://reviews.llvm.org/D122440
Follow up of https://reviews.llvm.org/D121488. Ensure lower bounds
are `1` when the related dimension extent is zero. Note that lower
bounds from descriptors are now guaranteed to fulfill this property
after the runtime/codegen patches.
Also fixes explicit shape array extent lowering when instantiating
variables to deal with negative extent cases (issue found while testing
LBOUND edge case). This notably caused allocation crashes when dealing
with automatic arrays with reversed bounds or negative size
specification expression. The standard specifies that the extent of such
arrays is zero. This change has some ripple effect in the current lit
tests.
Add move two helpers as part of this change:
- Add a helper to tell if a fir::ExtendedValue describes an assumed size
array (last dimension extent is unknown to the compiler, both at compile
time and runtime).
- Move and share getIntIfConstant from Character.cpp so that it can be
used elsewhere (NFC).
Differential Revision: https://reviews.llvm.org/D122467
The "seenProcs" sets passed as arguments to the procedure and dummy
procedure characterization routines need to be passed by value so that
local updates to those sets do not become permanent. They are
presently passed by reference and that has led to bogus errors about
recursively defined procedures in testing.
(It might be faster to pass the sets by reference and undo those local
updates in these functions, but that's error-prone, and the performance
difference is not expected to be detectable in practice.)
Differential Revision: https://reviews.llvm.org/D122439
Eric Schweitz