Summary:
This is the interface to the conversion pass that converts the FIR dialect (and standard dialects) of MLIR to the LLVM-IR dialect. The implementation of this bridge will be forthcoming in subsequent diffs.
The purpose of the code gen bridge is to take the IR, which will be expressed as a mix of higher-level Fortran constructs as expressed in the FIR dialect and the MLIR standard dialect, and convert the representation into the LLVM-IR dialect. At that point, the LLVM-IR dialect can be converted into LLVM IR (proper) by a standard MLIR pass.
Reviewers: jeanPerier, vjayathirtha-nv, sscalpone, jdoerfert, DavidTruby, kiranchandramohan
Reviewed By: sscalpone, kiranchandramohan
Tags: #llvm, #flang
Differential Revision: https://reviews.llvm.org/D79177
This range allows for performing many different operations on successor operands, including erasing/adding/setting. This removes the need for the explicit canEraseSuccessorOperand and eraseSuccessorOperand methods.
Differential Revision: https://reviews.llvm.org/D79077
As we start defining more complex Ops, we increasingly see the need for
Ops-with-regions to be able to construct Ops within their regions in
their ::build methods. However, these methods only have access to
Builder, and not OpBuilder. Creating a local instance of OpBuilder
inside ::build and using it fails to trigger the operation creation
hooks in derived builders (e.g., ConversionPatternRewriter). In this
case, we risk breaking the logic of the derived builder. At the same
time, OpBuilder::create, which is by far the largest user of ::build
already passes "this" as the first argument, so an OpBuilder instance is
already available.
Update all ::build methods in all Ops in MLIR and Flang to take
"OpBuilder &" instead of "Builder *". Note the change from pointer and
to reference to comply with the common style in MLIR, this also ensures
all other users must change their ::build methods.
Differential Revision: https://reviews.llvm.org/D78713
This provides a much cleaner interface into Symbols, and allows for users to start injecting op-specific information. For example, derived op can now inject when a symbol can be discarded if use_empty. This would let us drop unused external functions, which generally have public visibility.
This revision also adds a new `extraTraitClassDeclaration` field to ODS OpInterface to allow for injecting declarations into the trait class that gets attached to the operations.
Differential Revision: https://reviews.llvm.org/D78522
Many ops with this trait have `getBody()` and `getBodyBuilder()` methods defined in `extraClassDeclaration` in tablegen. `getBody()` implementation is the same accross all these ops, but `getBodyBuilder()` can return builders with varying insertion points set. In this PR, `getBody()` is moved into `SingleImplicitBlockTerminator` struct and `getBodyBuilder()` is replaced with `OpBuilder::atBlock(End|Terminator)(op.getBody);`.
Differential Revision: https://reviews.llvm.org/D78864
Summary:
This includes a refactor of the existing combined construct checks
that were present, as well as adding the remaining combined constructs
that had not been implemented yet.
Reviewers: jdoerfert
Subscribers: yaxunl, guansong, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D77812
Summary:
Updates recent work on DATA statement semantic checking in
flang/lib/Semantics/check-data.{h,cpp} to use the compiler's
internal representation for typed expressions rather than working
on the raw parse tree. Saves the analyzed expressions for DATA
statement values as parse tree decorations because they'll soon be
needed in lowering. Corrects wording of some error messages.
Fixes a bug in constant expression checking: structure constructors
are not constant expressions if they set an allocatable component
to anything other than NULL.
Includes infrastructure changes to make this work, some renaming
to reflect the fact that the implied DO loop indices tracked by
expression analysis are not (just) from array constructors, remove
some dead code, and improve some comments.
Reviewers: tskeith, sscalpone, jdoerfert, DavidTruby, anchu-rajendran, schweitz
Reviewed By: tskeith, anchu-rajendran, schweitz
Subscribers: llvm-commits, flang-commits
Tags: #flang, #llvm
Differential Revision: https://reviews.llvm.org/D78834
Summary:
Add size and offset properties to symbols, representing their byte size
and offset within their enclosing scope.
Add size and align properties to scopes so that they are available for
scopes representing derived types.
Add ComputeOffsets pass over the symbol table to fill in those fields.
Compute descriptor size based on rank and length parameters. Extract
DerivedTypeSpec::NumLengthParameters from DynamicType::RequiresDescriptor
to share the code.
Add Scope::GetSymbols to get symbols in canonical order.
compute-offsets.cpp and mod-file.cpp both need to process symbols in the
order in which they are declared. Move the collecting of those symbols
into Scope so that it can be shared.
Add symbol size and offset to output of `-fdebug-dump-symbols` and use
that in some tests.
Still to do:
- make size and alignment rules configurable based on target
- use offsets to check EQUIVALENCE statements
Differential Revision: https://reviews.llvm.org/D78680
Summary:
Move InstantiateComponent and InstantiateIntrinsicType from symbol.cpp
to type.cpp as that is where they are called.
Put both in InstantiateHelper class to better isolate them.
Add CreateDerivedTypeSpec in InstantiateHelper.
Add non-const forms for AsIntrinsic and AsDerived to avoid const_cast.
No functional changes.
Reviewers: DavidTruby, klausler, PeteSteinfeld, jdoerfert
Reviewed By: klausler
Subscribers: llvm-commits
Tags: #flang, #llvm
Differential Revision: https://reviews.llvm.org/D78678
These have proved incredibly useful for interleaving values between a range w.r.t to streams. After this revision, the mlir/Support/STLExtras.h is empty. A followup revision will remove it from the tree.
Differential Revision: https://reviews.llvm.org/D78067
There is no printf formatting string for std::int64_t. Instead we have
to cast to std::intmax_t and use `%jd`. This change simplifies that by
automatically converting std::int64_t to std::intmax_t when formatting
messages.
Original-commit: flang-compiler/f18@8a2343dfff
Reviewed-on: https://github.com/flang-compiler/f18/pull/1101
Tree-same-pre-rewrite: false
In most cases, I just added the contraint names to the code and tests.
I implemented the following checks:
- C736 A child type with a coarray ultimate component must have a parent with
a coarray ultimate component.
- C737 A child type with and EVENT_TYPE or LOCK_TYPE component must have a
parent either which is EVENT_TYPE or LOCK_TYPE or a type with an EVENT_TYPE
or LOCK_TYPE component.
- C740 Sequence types must contain at least on component
- C740 Data components of sequence types must either be of an intrinsic type
or a sequenced derived type.
After implementing these checks, some tests had new errors unrelated to their
original purpose, so I fixed them.
Original-commit: flang-compiler/f18@098f01bc47
Reviewed-on: https://github.com/flang-compiler/f18/pull/1097
[mlir rebase] Add MLIR config and react to MLIR name changes
Similar to flang-compiler/f18#1085. Now use the MLIR package to set up paths for include files and libraries. Three MLIR names changed:
* VectorOpsDialect to VectorDialect
* AffineOpsDialect to AffineDialect
* createVectorizePass to createSuperVectorizePass
Update README.md to explain how to link with MLIR. Update the example gcc to version 8.3.
Update drone.io config to define -DMLIR_DIR
Co-authored-by: Jean Perier <jperier@nvidia.com>
Original-commit: flang-compiler/f18@116f64315f
Reviewed-on: https://github.com/flang-compiler/f18/pull/1090
In general all the basic functionality seems to work and removes some redundancy
and more complicated features in favor of borrowing infrastructure from LLVM
build configurations. Here's a quick summary of details and remaining issues:
* Testing has spanned Ubuntu 18.04 & 19.10, CentOS 7, RHEL 8, and
MacOS/darwin. Architectures include x86_64 and Arm. Without
access to Window nothing has been tested there yet.
* As we change file and directory naming schemes (i.e.,
capitalization) some odd things can occur on MacOS systems with
case preserving but not case senstive file system configurations.
Can be painful and certainly something to watch out for as any
any such changes continue.
* Testing infrastructure still needs to be tuned up and worked on.
Note that there do appear to be cases of some tests hanging (on
MacOS in particular). They appear unrelated to the build
process.
* Shared library configurations need testing (and probably fixing).
* Tested both standalone and 'in-mono repo' builds. Changes for
supporting the mono repo builds will require LLVM-level changes that
are straightforward when the time comes.
* The configuration contains a work-around for LLVM's C++ standard mode
passing down into Flang/F18 builds (i.e., LLVM CMake configuration would
force a -std=c++11 flag to show up in command line arguments. The
current configuration removes that automatically and is more strict in
following new CMake guidelines for enforcing C++17 mode across all the
CMake files.
* Cleaned up a lot of repetition in the command line arguments. It
is likely that more work is still needed to both allow for
customization and working around CMake defailts (or those
inherited from LLVM's configuration files). On some platforms agressive
optimization flags (e.g. -O3) can actually break builds due to the inlining
of templates in .cpp source files that then no longer are available for use
cases outside those source files (shows up as link errors). Sticking at -O2
appears to fix this. Currently this CMake configuration forces this in
release mode but at the cost of stomping on any CMake, or user customized,
settings for the release flags.
* Made the lit tests non-source directory dependent where appropriate. This is
done by configuring certain test shell files to refer to the correct paths
whether an in or out of tree build is being performed. These configured
files are output in the build directory. A %B substitution is introduced in
lit to refer to the build directory, mirroring the %S substitution for the
source directory, so that the tests can refer to the configured shell scripts.
Co-authored-by: David Truby <david.truby@arm.com>
Original-commit: flang-compiler/f18@d1c7184159
Reviewed-on: https://github.com/flang-compiler/f18/pull/1045
The full list of constraints is C727, C728, C729, C730, C743, C755, C759, C778,
and C1543.
I added a function to tools.cpp to check to see if a symbol name is the name
of an intrinsic type.
The biggest change was to resolve-names.cpp to check to see if attributes were
either duplicated or in conflict with each other. I changed all locations
where attributes were set to check for duplicates or conflicts.
I also added tests for all checks and annotated the tests and code with the
numbers of the constraints being tested/checked.
Original-commit: flang-compiler/f18@3f30e8a61e
Reviewed-on: https://github.com/flang-compiler/f18/pull/1084
The previous code had handling for cases when too many file descriptors may be
opened; this is not necessary with MemoryBuffer as the file descriptors are
closed after the mapping occurs. MemoryBuffer also internally handles the case
where a file is small and therefore an mmap is bad for performance; such files
are simply copied to memory after being opened.
Many places elsewhere in the code assume that the buffer is not empty, and the
old file opening code handles this by replacing an empty file with a buffer
containing a single newline. That behavior is now kept in the new MemoryBuffer
based code.
Original-commit: flang-compiler/f18@d34df84351
Reviewed-on: https://github.com/flang-compiler/f18/pull/1032
This patch replaces the occurrence of std::ostream by llvm::raw_ostream.
In LLVM Coding Standards[1] "All new code should use raw_ostream
instead of ostream".[1]
As a consequence, this patch also replaces the use of:
std::stringstream by llvm::raw_string_ostream or llvm::raw_ostream*
std::ofstream by llvm::raw_fd_ostream
std::endl by '\n' and flush()[2]
std::cout by llvm::outs() and
std::cerr by llvm::errs()
It also replaces std::strerro by llvm::sys::StrError** , but NOT in Fortran
runtime libraries
*std::stringstream were replaced by llvm::raw_ostream in all methods that
used std::stringstream as a parameter. Moreover, it removes the pointers to
these streams.
[1]https://llvm.org/docs/CodingStandards.html
[2]https://releases.llvm.org/2.5/docs/CodingStandards.html#ll_avoidendl
Signed-off-by: Caroline Concatto <caroline.concatto@arm.com>
Running clang-format-7
Signed-off-by: Caroline Concatto <caroline.concatto@arm.com>
Removing residue of ostream library
Signed-off-by: Caroline Concatto <caroline.concatto@arm.com>
Original-commit: flang-compiler/f18@a3507d44b8
Reviewed-on: https://github.com/flang-compiler/f18/pull/1047
When a module subprogram has the MODULE prefix the following must match
with the corresponding separate module procedure interface body:
- C1549: characteristics and dummy argument names
- C1550: binding label
- C1551: NON_RECURSIVE prefix
SubprogramMatchHelper performs all of these checks.
Rename separate-module-procs.f90 to separate-mp01.f90 so we can have
separate-mp02.f90 (etc).
Make ShapesAreCompatible public in characteristics.h.
Add Scope::IsSubmodule.
Original-commit: flang-compiler/f18@d121578af1
Reviewed-on: https://github.com/flang-compiler/f18/pull/1080
A separate-module-subprogram is declared as `module procedure ...`
and gets its characteristics from the declaration of that name as
a separate module procedure. When we encounter one, we need to create
symbols in the new subprogram scope for the dummy arguments and
function return (if any).
The failure to create these symbols led to the bug in issue flang-compiler/f18#1054:
when a dummy argument was referenced, the compiler interpreted it as
an implicit declaration because there was no symbol for the argument.
Fixesflang-compiler/f18#1054.
Original-commit: flang-compiler/f18@4d3c4bac84
Reviewed-on: https://github.com/flang-compiler/f18/pull/1080
Tree-same-pre-rewrite: false
Adds FIR library that implements an MLIR dialect to which Fortran
parse-tree will be lowered to.
FIR is defined and documented inside FIROps.td added in this commit.
It is possible to generate a more readable description FIRLangRef.md
from FIROps.td following the related instructions added to the README.md
by this commit.
This patch adds FIR definition and implementation that allow parsing,
printing, and verifying FIR. FIR transformations and lowering to Standard
and LLVM dialects are not part of this patch. The FIR verifiers are verifying
the basic properties of FIR operations in order to provide a sufficient
frame for lowering. Verifiers for more advanced FIR properties can be added
as needed.
Coarrays are not covered by FIR defined in this patch.
This patch also adds tco tool that is meant to process FIR input files and
drives transformations on it. The tco tool is used for testing.
In this patch, it is only used to demonstrate parsing/verifying/
and dumping FIR with round-trip tests.
Note:
This commit does not reflect an actual work log, it is a feature-based split of the
changes done in the FIR experimental branch. The related work log can be found in the
commits between:
742edde572
and
2ff5524212
Changes on top of these original commits were made during this patch review.
Original-commit: flang-compiler/f18@30b428a51e
Reviewed-on: https://github.com/flang-compiler/f18/pull/1035
Scan FORMAT strings locally to avoid C++ binary runtime dependence when computing deepest parenthesis nesting
Remove a dependency on ostream from runtime
Remove remaining direct external references from runtime to C++ library binaries
Remove runtime dependences on lib/common
SetPos() and SetRec()
Instantiate templates for input
Begin input; rearrange locking, deal with CLOSE races
View()
Update error message in test to agree with compiler change
First cut at real input
More robust I/O runtime error handling
Debugging of REAL input
Add iostat.{h,cpp}
Rename runtime/numeric-* to runtime/edit-*
Move templates around, templatize integer output editing
Move LOGICAL and CHARACTER output from io-api.cpp to edit-output.cpp
Change pointer argument to reference
More list-directed input
Complex list-directed input
Use enum class Direction rather than bool for templates
Catch up with changes to master
Undo reformatting of Lower code
Use record number instead of subscripts for internal unit
Unformatted sequential backspace
Testing and debugging
Dodge bogus GCC warning
Add <cstddef> for std::size_t to fix CI build
Address review comments
Original-commit: flang-compiler/f18@50406b3496
Reviewed-on: https://github.com/flang-compiler/f18/pull/1053
The check for whether a private component is accessible was depending on
determining whether the source range of the current scope was within the
source range of the module that the component was declared in. This
could fail if the current scope was of kind `ImpliedDos` and had no
source range.
The fix is to add `Scope::Contains` to check the relationship by
traversing the parent links. These are created when the Scope is so are
always reliable. The source range of a scope is built up over time.
Original-commit: flang-compiler/f18@d787108637
Reviewed-on: https://github.com/flang-compiler/f18/pull/1060
If the pure context check succeeds, call `WhyNotModifiable` to verify
the LHS can be modified.
Detect assignment to whole assumed-size array.
Change `IsVariable` to return false for a parameter or a component or
array reference whose base it a parameter.
When analyzing an assignment statement, report an error if the LHS is
a constant expression. Otherwise it might get folded and when we detect
the problem later the error will be confusing.
Handle Substring on LHS of assignment. Change ExtractDataRef and IsVariable
to work on a Substring.
Fix IsImpliedShape and IsAssumedSize predicates in ArraySpec.
Fix C709 check in check-declarations.cpp.
Original-commit: flang-compiler/f18@f2d2657aab
Reviewed-on: https://github.com/flang-compiler/f18/pull/1050
One overload of WhyNotModifiable returned an optional message while the
other returns a unique_ptr. Change the latter to be consistent with the
former and other message-returning functions in this file.
Also, reorder the if clauses to reduce the level of indentation.
Original-commit: flang-compiler/f18@864e9cfc7e
Reviewed-on: https://github.com/flang-compiler/f18/pull/1050
Tree-same-pre-rewrite: false
When a misparsed FunctionReference was converted to a StructureConstructor,
the components accessed were not checked for accessibility.
The conversion happens in expression analysis so that where the accessibity
must be checked. So move `CheckAccessibleComponent` to `tools.h` so that it
can be shared by `resolve-names.cpp` and `expression.cpp`.
Add FindModuleContaining to help implement this and use it other places.
Check that an access-spec can only appear in a module.
Remove some unnecessary "semantics::" qualifiers.
Original-commit: flang-compiler/f18@99ce156e49
Reviewed-on: https://github.com/flang-compiler/f18/pull/1046
Data statements contains expressions but they are not wrapped in one of
the kinds of parse tree nodes that are analyzed, like `parser::Expr`.
So potential errors were not discovered.
Change `ExprChecker` to handle `DataStmtConstant` and analyze any
expressions that are contained in it. Note that the analyzed form of
the expression is not yet saved in the parse tree.
Original-commit: flang-compiler/f18@8bdaf0a521
Reviewed-on: https://github.com/flang-compiler/f18/pull/1044
I've updated the compiler and test source with references to the contraints at
the points where they were enforced and tested. Many of these were already
implemented and required no code change. A few constraint checks were both
implemented and tested, and I only added references to the constraint
numbers in the compiler source and tests. Here are the things I had to
implement:
Constraint C716 states that, in a REAL constant, if both a kind-param and an
exponent letter appear, the exponent letter must be 'E'.
Constraints C715 and C719 require that a KIND value be actually implemented.
Constraint C722 requires that functions that return assumed-length character
types are external.
Constraint C726 disallows assumed lenght charater types for dummy arguments and
return types.
Original-commit: flang-compiler/f18@45998741e5
Reviewed-on: https://github.com/flang-compiler/f18/pull/1031
Tree-same-pre-rewrite: false
In a data statement like `data x / a(1) /`, `a(1)` may be an array
element or a structure constructor. It is parsed as an array element
so if it turns out `a` is a derived type it must be rewritten as a
strucutre constructor.
Original-commit: flang-compiler/f18@a2b2a330e7
Reviewed-on: https://github.com/flang-compiler/f18/pull/1024
Check that masks and LHS of assignments in WHERE statements and
constructs have consistent shapes. They must all have the same rank and
any extents that are compile-time constants must match.
Emit a warning for assignments in FORALL statements and constructs where
the LHS does not reference each of the index variables.
Original-commit: flang-compiler/f18@8b04dbebcf
Reviewed-on: https://github.com/flang-compiler/f18/pull/1009
This DATA statement was getting a parsing error:
`data x /a(i)%b/`
The parser was expecting the ending '/' where the '%' was. The problem
was parsing `a(i)` as a structure constructor. Instead, move the
constant subobject case before structure constructor, but match it only
if not followed by '('. That is because in `data x /a(1)(2)/`, `a(1)` is
a valid structure constructor.
Also, remove the NamedConstant alternative from DataStmtRepeat. A named
constant is always parsed as a ConstantSubobject so it can never occur.
Original-commit: flang-compiler/f18@04a76b2726
Reviewed-on: https://github.com/flang-compiler/f18/pull/1012
Eric Schweitz