When the characteristics of a procedure depend on a procedure that
hasn't yet been defined, the compiler currently emits an unconditional
error message. This includes the case of a procedure whose
characteristics depend, perhaps indirectly, on itself. However, in the
case where the characteristics of a procedure are needed to resolve a
generic, we should not emit an error for a hitherto undefined procedure
-- either the call will resolve to another specific procedure, in which
case the error is spurious, or it won't, and then an error will issue
anyway.
Fixes https://github.com/llvm/llvm-project/issues/88677.
Currently, it is not possible to find back which fun.func is the host
procedure of some internal procedure because the mangling of the
internal procedure does not contain info about the BIND(C) name of the
host.
This info may be useful to ensure dwarf DW_TAG_subprogram of internal
procedures are nested under DW_TAG_subprogram of host procedures for
instance.
Fortran mandates "CHARACTER(1), VALUE" be passed as a C "char" in calls
to BIND(C) procedures (F'2023 18.3.7 (4)). Lowering passed them by
memory instead. Update call interface lowering code to pass them by
register. Fix related test and update it to use HLFIR.
Whenever lowering is checking if a function or global already exists in
the mlir::Module, it was doing module->lookup.
On big programs (~5000 globals and functions), this causes important
slowdowns because these lookups are linear. Use mlir::SymbolTable to
speed-up these lookups. The SymbolTable has to be created from the
ModuleOp and maintained in sync. It is therefore placed in the
converter, and FirOPBuilders can take a pointer to it to speed-up the
lookups.
This patch does not bring mlir::SymbolTable to FIR/HLFIR passes, but
some passes creating a lot of runtime calls could benefit from it too.
More analysis will be needed.
As an example of the speed-ups, this patch speeds-up compilation of
Whizard compare_amplitude_UFO.F90 from 5 mins to 2 mins on my machine
(there is still room for speed-ups).
Polymorphic entity lowering status is good. The main remaining TODO is
to allow lowering of vector subscripted polymorphic entity, but this
does not deserve blocking all application using polymorphism.
Remove experimental option and enable lowering of polymorphic entity by
default.
The current lowering did not handle sequence associated argument passed
by descriptor. This case is special because sequence association implies
that the actual and dummy argument need to to agree in rank and shape.
Usually, arguments that can be sequence associated are passed by raw
address, and the shape mistmatch is transparent. But there are three
cases of explicit and assumed-size arrays passed by descriptors:
- polymorphic arguments
- BIND(C) assumed-length arguments (F'2023 18.3.7 (5)).
- length parametrized derived types (TBD)
The callee side is expecting a descriptor containing the dummy rank and
shape. This was not the case. This patch fix that by evaluating the
dummy shape on the caller side using the interface (that has to be
available when arguments are passed by descriptors).
This PR adds a new attribute to carry over the information from
`cluster_dims`. The new attribute `CUDAClusterDimsAttr` holds 3 integer
attributes and is added to `func.func` operation.
This PR adds a new attribute to carry over the information from
`launch_bounds`. The new attribute `CUDALaunchBoundsAttr` holds 2 to 3
integer attrinbutes and is added to `func.func` operation.
This PR adds a new attribute to represent the CUDA attribute attached to
procedure. This attribute is attached to the func.func operation during
lowering.
Other procedures information such as `launch_bounds` and `cluster_dims`
will be added separately.
The newly introduced `CUDAAttribute` is meant for CUDA attributes
associated with variable. In order to not clash with the future
attribute for function/subroutine, rename `CUDAAttribute` to
`CUDADataAttribute`.
Runtime derived type info contains information to tell the runtime if
some argument in a user defined assignment must be passed with a
descriptor or not. This information was not properly build, it would
tell the runtime that TARGET argument must be passed via descriptor,
which is incorrect.
Share the logic between lowering and runtime info generation to
determine if an argument must be passed by descriptor or not.
Lower CUDA attribute for simple dummy argument. This is done in a
similar way than `TARGET`, `OPTIONAL` and so on.
This patch also move the `Fortran::common::CUDADataAttr` to
`fir::CUDAAttributeAttr` mapping to
`flang/include/flang/Optimizer/Support/Utils.h` so that it can be reused
where needed.
Finish plugging-in ASYNCHRONOUS IO in lowering (GetAsynchronousId was
not used yet).
Add a runtime implementation for GetAsynchronousId (only the signature
was defined). Always return zero since flang runtime "fakes"
asynchronous IO (data transfer are always complete, see
flang/docs/IORuntimeInternals.md).
Update all runtime integer argument and results for IDs to use the
AsynchronousId int alias for consistency.
In lowering, asynchronous attribute is added on the hlfir.declare of
ASYNCHRONOUS variable, but nothing else is done. This is OK given the
synchronous aspects of flang IO, but it would be safer to treat these
variable as volatile (prevent code motion of related store/loads) since
the asynchronous data change can also be done by C defined user
procedure (see 18.10.4 Asynchronous communication). Flang lowering
anyway does not give enough info for LLVM to do such code motions (the
variables that are passed in a call are not given the noescape
attribute, so LLVM will assume any later opaque call may modify the
related data and would not move load/stores of such variables
before/after calls even if it could from a pure Fortran point of view
without ASYNCHRONOUS).
This PR adds lowering the reference to a function that returns a
procedure pointer. It also fixed intrinsic ASSOCIATED to take such
argument.
---------
Co-authored-by: jeanPerier <jperier@nvidia.com>
Start implementing assumed-rank support as described in
https://github.com/llvm/llvm-project/blob/main/flang/docs/AssumedRank.md
This commit holds the minimal support for lowering calls to procedure
with assumed-rank arguments where the procedure implementation is done
in C.
The case for passing assumed-size to assumed-rank is left TODO since it
will be done a change in assumed-size lowering that is better done in
another patch.
Care is taken to set the lower bounds to zero when passing non allocatable no pointer as descriptor
to a BIND(C) procedure as required per 18.5.3 point 3. This was not done before while the requirements also applies to non assumed-rank descriptors. This change required special attention with IGNORE_TKR(t) to avoid emitting invalid fir.rebox operations (the actual argument type must be used in this case as the output type).
Implementation of Fortran procedure with assumed-rank arguments is still
TODO.
Lower procedure pointer components, except in the context of structure
constructor (left TODO).
Procedure pointer components lowering share most of the lowering logic
of procedure poionters with the following particularities:
- They are components, so an hlfir.designate must be generated to
retrieve the procedure pointer address from its derived type base.
- They may have a PASS argument. While there is no dispatching as with
type bound procedure, special care must be taken to retrieve the derived
type component base in this case since semantics placed it in the
argument list and not in the evaluate::ProcedureDesignator.
These components also bring a new level of recursive MLIR types since a
fir.type may now contain a component with an MLIR function type where
one of the argument is the fir.type itself. This required moving the
"derived type in construction" stackto the converter so that the object
and function type lowering utilities share the same state (currently the
function type utilty would end-up creating a new stack when lowering its
arguments, leading to infinite loops). The BoxedProcedurePass also
needed an update to deal with this recursive aspect.
This patch is fixing two issue relative to the dynamic dispatch for
polymorphic entities.
1. Fix the `requireDispatchCall` function. It was checking for the first
symbol of the component but this is not the one to be checked. Instead
the last symbol of the base of the component object is the one to check
to know if it is polymorphic object with a dispatch call or not. This is
demonstrated in the new added test in `flang/test/Lower/dispatch.f90`
where the first symbol would point to `q` which is monomorphic and would
result in a simple `fir.call`
2. Fix the pass object in a no pass situation. In a no pass situation
the pass object is lowered anyway to be able to do the lookup in the
binding table. It was previously lowered wrongly an lead to unresolved
lookup. The base of the component is the passed object and should be
lowered. To achieve this, the `gen(DataRef)` entry point is exposed form
`ConvertExprToHLFIR` through a `convertDataRefToValue` function. The
same test added in `flang/test/Lower/dispatch.f90` is checking for the
correct passed object.
In addition couple of tests were updated to HLFIR since the lowering
used only works with it.
**Scope of the PR:**
1. Lowering global and local procedure pointer declaration statement
with explicit or implicit interface. The explicit interface can be from
an interface block, a module procedure or an internal procedure.
2. Lowering procedure pointer assignment, where the target procedure
could be external, module or internal procedures.
3. Lowering reference to procedure pointers so that it works end to end.
**PR notes:**
1. The first commit of the PR does not include testing. I would like to
collect some comments first, which may alter the output. Once I confirm
the implementation, I will add some testing as a follow up commit to
this PR.
2. No special handling of the host-associated entities when an internal
procedure is the target of a procedure pointer assignment in this PR.
**Implementation notes:**
1. The implementation is using the HLFIR path.
2. Flang currently uses `getUntypedBoxProcType` to get the
`fir::BoxProcType` for `ProcedureDesignator` when getting the address of
a procedure in order to pass it as an actual argument. This PR inherits
the same design decision for procedure pointer as the `fir::StoreOp`
requires the same memory type.
Note: this commit is actually resubmitting the original commit from
PR #70461 that was reverted. See PR #73221.
**Scope of the PR:**
1. Lowering global and local procedure pointer declaration statement
with explicit or implicit interface. The explicit interface can be from
an interface block, a module procedure or an internal procedure.
2. Lowering procedure pointer assignment, where the target procedure
could be external, module or internal procedures.
3. Lowering reference to procedure pointers so that it works end to end.
**PR notes:**
1. The first commit of the PR does not include testing. I would like to
collect some comments first, which may alter the output. Once I confirm
the implementation, I will add some testing as a follow up commit to
this PR.
2. No special handling of the host-associated entities when an internal
procedure is the target of a procedure pointer assignment in this PR.
**Implementation notes:**
1. The implementation is using the HLFIR path.
2. Flang currently uses `getUntypedBoxProcType` to get the
`fir::BoxProcType` for `ProcedureDesignator` when getting the address of
a procedure in order to pass it as an actual argument. This PR inherits
the same design decision for procedure pointer as the `fir::StoreOp`
requires the same memory type.
Semantics is emitting an error when an actual argument to a procedure
that has an implicit interface has a polymorphic type. This is too
general; while TYPE(*) and CLASS(*) unlimited polymorphic items require
the presence of an explicit procedure interface, CLASS(T) data can be
passed over an implicit interface to a procedure expecting a
corresponding dummy argument with TYPE(T), so long as T is not
parameterized.
(Only XLF handles this usage correctly among other Fortran compilers.)
(Making this work in the case of an actual CLASS(T) array may well
require additional changes in lowering to copy data to/from a temporary
buffer to ensure contiguity when the actual type of the array is an
extension of T.)
Update evaluate::ActualArgument to propagate the %VAL and %REF markers
until lowering.
Semantic checks are added to %VAL to ensure the argument is a numerical
or logical scalar.
I did not push these markers into the characteristics because other
compilers do not complain about inconsistent usages (e.g. using %VAL in
a call on a procedure with an interface without VALUE dummies is not
flagged by any compilers I tested, and it is not an issue for lowering,
so I decided to stay simple here and minimize the footprint of these
legacy features).
Lowering retrieves these markers and does the right thing: pass %VAL in
registers and pass %REF by address without adding any extra arguments
for characters.
Currently, if the first usage of a procedure not defined in the file was
inside a procedure designator reference (not a call to it), the lowered
func.func lacked the argument attributes if any.
Fix this by using `CallInterface<T>::declare` too in SignatureBuilder to
create a new func.func instead of using custom code.
Note: this problem was made worse by the fact that module variables
fir.global are currently lowered before the module procedures func.func
are created. I will try to fix that in a later patch (the debug location
may still be wrong in certain cases) because there is quite some test
fallout when changing the order of globals/funcop in the output.
1. Deal with BIND(C,NAME="")
BIND(C,NAME="") is different from BIND(C). The latter implies that there
us a binding label which is the Fortran symbol name (no Fortran mangling
must be added like underscores). The former implies there is no binding
label (the name in the object file must be the same as if it there was
no BIND(C) attribute at all).
This is correctly implemented in the front-end, but lowering mistakenly
overrode this in the code dealing with the case where BIND(C) is
inherited from a procedure interface. Handling of this last case is moved into name
resolution.
2. Deal with BIND(C) internal procedure
Also according to 18.10.2, BIND(C) does not give a p prevent name
resolution from adding a label to them, otherwise,
bindc_internal_proc.f90 was not going through semantics (bogus error
about conflicting global names). Nothing TODO in lowering other than
removing the TODO.
Outside of BIND(C), assumed length character scalar and explicit shape
are passed by address + an extra length argument (fir.boxchar in FIR).
The standard mandates that they be passed via CFI descriptor in BIND(C)
interface (fir.box in FIR). This patch fix the handling for this case.
The following PowerPC vector type syntax is added:
VECTOR ( element-type-spec )
where element-type-sec is integer-type-spec, real-type-sec or unsigned-type-spec.
Two opaque types (__VECTOR_PAIR and __VECTOR_QUAD) are also added.
A finite set of functionalities are implemented in order to support the new types:
1. declare objects
2. declare function result
3. declare type dummy arguments
4. intrinsic assignment between the new type objects (e.g. v1=v2)
5. reference functions that return the new types
Submit on behalf of @tislam @danielcchen
Authors: @tislam @danielcchen
Differential Revision: https://reviews.llvm.org/D150876
There are several observations regarding the copy-in/copy-out:
* Actual argument associated with INTENT(OUT) dummy argument that
requires finalization (7.5.6.3 p. 7) may be read by the finalization
function, so a copy-in is required.
* A temporary created for the copy-in/copy-out must be destroyed
without finalization after the call (or after the corresponding copy-out),
otherwise, memory leaks may occur.
* The copy-out assignment must not perform finalization for the LHS.
* The copy-out assignment from the temporary to the actual argument
may or may not need to initialize the LHS.
This change-set introduces new runtime methods: CopyOutAssign and
DestroyWithoutFinalization. They are called by the compiler generated
code to match the behavior described above.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D151135
Reboxing of the actual argument according to the type of the dummy
argument has to be aware of the potential rank mismatch, when
IGNORE_TKR(R) is used. This change only adds support for the mismatching
rank when the dummy argument has unlimited polymorphic type.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D151016
- Fix the BIND(C) assumed-shape case: TYPE(*) assumed shape are passed
via CFI_cdesc_t according to Fortran 2018 standard 18.3.6 point 2 (5).
- Align the none BIND(C) case with the BIND(C) case. There is little
point passing TYPE(*) assumed size via descriptor, use a simple
address. C710 ensures there is no way the knowledge of the actual
type will be required when manipulating the dummy.
Differential Revision: https://reviews.llvm.org/D148130
When interfacing with C code, assumed type should be passed as
basic pointer.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D146300
A block construct is an execution control construct that supports
declaration scopes contained within a parent subprogram scope or another
block scope. (blocks may be nested.) This is implemented by applying
basic scope processing to the block level.
Name uniquing/mangling is extended to support this. The term "block" is
heavily overloaded in Fortran standards. Prior name uniquing used tag `B`
for common block objects. Existing tag choices were modified to free up `B`
for block construct entities, and `C` for common blocks, and resolve
additional issues with other tags. The "old tag -> new tag" changes can
be summarized as:
-> B -- block construct -> new
B -> C -- common block
C -> YI -- intrinsic type descriptor; not currently generated
CT -> Y -- nonintrinsic type descriptor; not currently generated
G -> N -- namelist group
L -> -- block data; not needed -> deleted
Existing name uniquing components consist of a tag followed by a name
from user source code, such as a module, subprogram, or variable name.
Block constructs are different in that they may be anonymous. (Like other
constructs, a block may have a `block-construct-name` that can be used
in exit statements, but this name is optional.) So blocks are given a
numeric compiler-generated preorder index starting with `B1`, `B2`,
and so on, on a per-procedure basis.
Name uniquing is also modified to include component names for all
containing procedures rather than for just the immediate host. This
fixes an existing name clash bug with same-named entities in same-named
host subprograms contained in different-named containing subprograms,
and variations of the bug involving modules and submodules.
F18 clause 9.7.3.1 (Deallocation of allocatable variables) paragraph 1
has a requirement that an allocated, unsaved allocatable local variable
must be deallocated on procedure exit. The following paragraph 2 states:
When a BLOCK construct terminates, any unsaved allocated allocatable
local variable of the construct is deallocated.
Similarly, F18 clause 7.5.6.3 (When finalization occurs) paragraph 3
has a requirement that a nonpointer, nonallocatable object must be
finalized on procedure exit. The following paragraph 4 states:
A nonpointer nonallocatable local variable of a BLOCK construct
is finalized immediately before it would become undefined due to
termination of the BLOCK construct.
These deallocation and finalization requirements, along with stack
restoration requirements, require knowledge of block exits. In addition
to normal block termination at an end-block-stmt, a block may be
terminated by executing a branching statement that targets a statement
outside of the block. This includes
Single-target branch statements:
- goto
- exit
- cycle
- return
Bounded multiple-target branch statements:
- arithmetic goto
- IO statement with END, EOR, or ERR specifiers
Unbounded multiple-target branch statements:
- call with alternate return specs
- computed goto
- assigned goto
Lowering code is extended to determine if one of these branches exits
one or more relevant blocks or other constructs, and adds a mechanism to
insert any necessary deallocation, finalization, or stack restoration
code at the source of the branch. For a single-target branch it suffices
to generate the exit code just prior to taking the indicated branch.
Each target of a multiple-target branch must be analyzed individually.
Where necessary, the code must first branch to an intermediate basic
block that contains exit code, followed by a branch to the original target
statement.
This patch implements an `activeConstructStack` construct exit mechanism
that queries a new `activeConstruct` PFT bit to insert stack restoration
code at block exits. It ties in to existing code in ConvertVariable.cpp
routine `instantiateLocal` which has code for finalization, making block
exit finalization on par with subprogram exit finalization. Deallocation
is as yet unimplemented for subprograms or blocks. This may result in
memory leaks for affected objects at either the subprogram or block level.
Deallocation cases can be addressed uniformly for both scopes in a future
patch, presumably with code insertion in routine `instantiateLocal`.
The exit code mechanism is not limited to block construct exits. It is
also available for use with other constructs. In particular, it is used
to replace custom deallocation code for a select case construct character
selector expression where applicable. This functionality is also added
to select type and associate constructs. It is available for use with
other constructs, such as select rank and image control constructs,
if that turns out to be necessary.
Overlapping nonfunctional changes include eliminating "FIR" from some
routine names and eliminating obsolete spaces in comments.
There is a lot of Fortran code that takes advantage of F77 implicit
interface to pass arguments with a different type than those from
the subprogram definition (which is well defined if the storage
and passing convention are the same or compatible).
When the definition and calls are in different files, there is nothing
special to do: the actual arguments are already used to compute the
call interface.
The trouble for lowering comes when the definition is in the same
compilation unit (Semantics raises warning). Then, lowering will
be provided with the interface from the definition to prepare the
argument, and this leads to many ad-hoc handling (see
builder.convertWithSemantics) in the current lowering to cope
with the dummy/actual mismatches on a case by case basis. The
current lowering to FIR is not even complete for all mismatch cases that
can be found in the wild (see https://github.com/llvm/llvm-project/issues/60550),
it is crashing or hitting asserts for many of the added tests.
For HLFIR, instead of coping on a case by case basis, the call
interface will be recomputed according to the actual arguments when
calling an external procedure that can be called with an explicit
interface.
One extra case still has to be handled manually because it may happen
in calls with explicit interfaces: passing a character procedure
designator to a non character procedure dummy (and vice-versa) is widely
accepted even with explicit interfaces (and flang semantic accepts it).
Yet, this "mismatch" cannot be dealt with a simple fir.convert because
character dummy procedure are passed with a different passing
convention: an extra argument is hoisted for the result length (in FIR,
there is no extra argument yet, but the MLIR func argument is a
tuple<fir.boxproc, len>).
Differential Revision: https://reviews.llvm.org/D143636
- Add a convertProcedureDesignatorToHLFIR that converts the
fir::ExtendedValue from the current lowering to a
fir.boxproc/tuple<fir.boxproc, len> mlir::Value.
- Allow fir.boxproc/tuple<fir.boxproc, len> as hlfir::Entity values
(a function is an address, but from a Fortran entity point of view,
procedure that are not procedure pointers cannot be assigned to, so
it makes a lot more sense to consider those as values).
- Modify symbol association to not generate an hlfir.declare for dummy
procedures. They are not needed and allowing hlfir.declare to declare
function values would make its verifier and handling overly complex
for little benefits (maybe an hlfir.declare_proc could be added if it
turnout out useful later for debug info and attributes storing
purposes).
- Allow translation from hlfir::Entity to fir::ExtendedValue.
convertToBox return type had to be relaxed because some intrinsics
handles both object and procedure arguments and need to lower their
object arguments "asBox". fir::BoxValue is not intended to carry
dummy procedures (all its member functions would make little sense
and its verifier does not accept such type).
Note that AsAddr, AsValue and AsBox will always return the same MLIR
value for procedure designators because they are always handled the
same way in FIR.
Differential Revision: https://reviews.llvm.org/D143585
In lowering to HLFIR, deal with user calls involving a mix of:
- dummy with VALUE
- Polymorphism
- contiguous dummy
- assumed shape dummy
- OPTIONAL arguments
- NULL() passed to OPTIONAL arguments.
- elemental calls
Does not deal with assumed ranked dummy arguments.
This patch unifies the preparation of all arguments that must be passed
in memory and are not passed as allocatable/pointers.
For optionals, the same argument preparation is done, except the utility
that generates the IR for the argument preparation is called inside a
fir.if.
The addressing of array arguments in elemental calls is delayed so that
it can also happen during this argument preparation, and be placed in
the fir.if when the array may be absent.
Structure helpers are added to convey a prepared dummy argument and the
data that may be needed to do the clean-up after the call (temporary
storage deallocation or copy-out). And a utility is added to wrap
the preparation code inside a fir.if and convey these values through
the fir.if.
Certain aspects of this patch brings the HLFIR lowering support beyond
what the current lowering to FIR supports (e.g. handling of NULL(), handling
of optional in elemental calls, handling of copy-in/copy-out involving
polymorphic entities).
Differential Revision: https://reviews.llvm.org/D142695
Addresses and properties (bounds, length parameters) of host
variables associated in an internal procedure were all passed via
an extra tuple argument of the internal procedure.
This extra tuple is in general an overhead: it must be created and
passed, and require creating thunks when taking the address of the
internal procedure.
This patch allows not using the tuple for host global variables
(from modules, common block, or local saved variables) since they can
be instantiated from the fir.global symbol in the internal procedure
instead.
Add a fir.internal_proc attribute to mlir::FuncOp for internal procedures
so that ArrayValueCopy can still detect internal procedures even if they
do not have a tuple argument.
Differential Revision: https://reviews.llvm.org/D140288
Lower procedure ref to user defined elemental procedure when:
- there are no arguments that may be dynamically optional
- for functions, the result has no length parameters
- the reference can be unordered
- there are not character by value arguments
This uses the recently added hlfir.elemental operation and tools.
The "core" of the argument preparation is shared between elemental
and non elemental calls (genUserCalls is code moved without any
functional changes)
Differential Revision: https://reviews.llvm.org/D140118
This patch mechanically replaces None with std::nullopt where the
compiler would warn if None were deprecated. The intent is to reduce
the amount of manual work required in migrating from Optional to
std::optional.
This is part of an effort to migrate from llvm::Optional to
std::optional:
https://discourse.llvm.org/t/deprecating-llvm-optional-x-hasvalue-getvalue-getvalueor/63716
Take into account the result passed as arguments when computing
the pass object index.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D138712
The dynamic type of an unlimited polymorphic entity has the
derived category but does not have derived type spec. This leads
to a crash for a nullptr dereference. This patch avoids this crash
by checking if that the dynamic type is not unlimited polymorphic
before dereferencing the derived type spec.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D138691
When the `type(c_ptr/c_funptr)` argument has value attribute in non-BIND(C)
procedure, it is passed by VALUE in gfortran. ifort does not do this. Be
consistent with gfortran.
Fix#58756.
Reviewed By: PeteSteinfeld, jeanPerier
Differential Revision: https://reviews.llvm.org/D137237
