…utdown'
This patch emits the lowering for 'device_type' on an 'init' or
'shutdown'. This one is fairly unique, as these directives have it as an
attribute, rather than as a component of the individual operands, like
the rest of the constructs.
So this patch implements the lowering as an attribute.
In order to do tis, a few refactorings had to happen: First, the
'emitOpenACCOp' functions needed to pick up th edirective kind/location
so that the NYI diagnostic could be reasonable.
Second, and most impactful, the `applyAttributes` function ends up
needing to encode some of the appertainment rules, thanks to the way the
OpenACC-MLIR operands get their attributes attached. Since they each use
a special function (rather than something that can be legalized at
runtime), the forms of 'setDefaultAttr' is only valid for some ops. SO
this patch uses some `if constexpr` and a small type-trait to help
legalize these.
This patch upstreams initial support for making function calls in CIR.
Function arguments and return values are not included to keep the patch
small for review.
Related to #132487
Previous implementations that used the cir::LValue class omitted hanling
of the LValueBaseInfo class, which tracks information about the basis
for the LValue's alignment. As more code was upstreamed from the
incubator, we were accumulating technical debt by adding more places
where this wasn't handled correctly. This change puts the interfaces in
place to track this information.
The information being tracked isn't used yet, so no functional change is
intended. The tracking is being added now because it will become more
difficult to add it as more features are implemented.
These two are very simple. they don't require any clauses and don't
have an associated statement, so they have very simple output. This
patch implements them, but none of the associated clauses.
This patch does the lowering of the OpenACC 'data' construct, which
requires getting the `default` clause (as `data` requires at least 1 of
a list of clauses, and this is the easiest one). The lowering of the
clauses appears to happen in 1 of 2 ways: a- as an operand. or b- as an
attribute.
This patch adds infrastructure to lower as an attribute, as that is how
'data' works.
In addition to that, it changes the OpenACCClauseVisitor a bit, which
previously just required that each of the derived classes have all of
the clauses covered. This patch modifies it so that the visitor directly
calls the derived class from its visitor function, which leaves the
base-class ones the ability to defer to a generic function. This was
previously like this because I had some use cases that I didn't end up
using, and the 'generic' function here seems much more useful.
This patch adds some lowering code for Compute Constructs, plus the
infrastructure to someday do clauses. Doing this requires adding the
dialect to the CIRGenerator.
This patch does not however implement/correctly initialize lowering from
OpenACC-Dialect to anything lower however.
The handling for NullStmt was going to an error saying the statement
handling wasn't implemented. It doesn't need any implementation. It is
sufficient for emitSimpleStmt to just return success for that statement
class. This change does that.
This adds support for handling the address of and dereference unary
operations in ClangIR code generation. This also adds handling for
nullptr and proper initialization via the NullToPointer cast.
This is the first of a few patches that will do infrastructure work to
enable the OpenACC lowering via the OpenACC dialect.
At the moment this just gets the various function calls that will end up
generating OpenACC, plus some tests to validate that we're doing the
diagnostics in OpenACC specific locations.
Additionally, this adds Stmt and Decl files for CIRGen.
This patch adds support for comparison operators with ClangIR, both
integral and floating point.
---------
Co-authored-by: Morris Hafner <mhafner@nvidia.com>
Co-authored-by: Henrich Lauko <xlauko@mail.muni.cz>
Co-authored-by: Andy Kaylor <akaylor@nvidia.com>
When unary operation support was initially upstreamed, the cir.cast
operation hadn't been upstreamed yet, so logical not wasn't included.
Since casts have now been added, this change adds support for logical
not.
This adds ClangIR support for break and continue statements in loops.
Because only loops are currently implemented upstream in CIR, only
breaks in loops are supported here, but this same code will work (with
minor changes to the verification and cfg flattening) when switch
statements are upstreamed.
The initial upstreaming of unary operations left promoted types
unhandled for the unary plus and minus operators. This change implements
support for promoted types and performs a bit of related code cleanup.
A previous checkin used a workaround to generate the nsw flag where
needed for unary ops. This change upstreams a subsequent change that was
made in the incubator to generate the flag correctly.
This fixes unused variable warnings that have crept into the ClangIR
code. In some cases the variable will be needed later, but all unused
variables are being removed here. They can be reintroduced when they are
needed.
To make rebasing the incubator project easier, we've been trying to
place upstreamed code in the same order in which it appears in the
incubator project. However, while the upstream implementation is still
relatively sparse, it is often difficult to find points of reference for
placement of new declarations. To help with that, I refactored
CIRGenFunction.h in the incubator to put all the emit* functions in one
place and sort them alphabetically.
This change reorganizes the upstream CIRGenFunction.h to match the new
incubator ordering.
Lowering of int-to-bool casts had been left as NYI because the incubator
implemented it by lowering to cir.cmp, which hasn't been upstreamed yet,
but there is no reason this cast can't be lowered directly to LLVM's
compare operation when we're lowering directly to the LLVM dialect.
This change lowers the cast directly to an LLVM compare to zero.
Alloca operations were being emitted into the entry block of the current
function unconditionally, even if the variable they represented was
declared in a different scope. This change upstreams the code for
handling
insertion of the alloca into the proper lexcial scope. It also adds a
CIR-to-CIR transformation to hoist allocas to the function entry block,
which is necessary to produce the expected LLVM IR during lowering.
This patch adds upstreams support for BinOp including lvalue
assignments. Note that this does not include ternary ops,
BinOpOverflowOp, pointer arithmetic, ShiftOp and SelectOp which are
required for logical binary operators.
---------
Co-authored-by: Morris Hafner <mhafner@nvidia.com>
Co-authored-by: Andy Kaylor <akaylor@nvidia.com>
This change adds support for empty for-loops. This will be the
infrastructre needed for complete for loops, but that depends on compare
operations, which have not been upstreamed yet.
Unary ops had previously been omitted from the list of ops handled in
the CIRCanonicalizePass. Although the incubator code doesn't use them
directly, the mlir folding code does.
This change enables folding of unary ops by adding them to the list.
Upstream the parts of class `CIRGenFunction::LexicalScope` that
implement function return values. There is a bit of other functionality
in here, such as the implicit `cir.yield` at the end of a non-function
scope, but this is mostly about function returns.
The parts of `LexicalScope` that handle calling destructors, switch
statements, ternary expressions, and exception handling still need to be
upstreamed.
There is a change in the generated ClangIR (which is why many of the
tests needed updating). Return values are stored in the
compiler-generated variable `__retval` rather than being passed to the
`cir.return` op directly. But there should be no change in the behavior
of the generated code.
This change introduces the cir-canonicalize pass. This is a simple
cir-to-cir transformation that eliminates empty scopes and redundant
branches. It will be expanded in future changes to simplify other
redundant instruction sequences.
MLIR verification and mlir-specific command-line option handling is also
introduced here.
This patch upstreams ClangIR's CastOp with the following exceptions:
- No Fixed/FP conversions
- No casts between value categories
- No complex casts
- No array_to_ptrdecay
- No address_space
- No casts involving record types (member pointers, base/derived casts)
- No casts specific to ObjC or OpenCL
---------
Co-authored-by: Morris Hafner <mhafner@nvidia.com>
Co-authored-by: Erich Keane <ekeane@nvidia.com>
This adds support for emitting ClangIR for statements whose value is
ignored. The test case being added (CIR/CodeGen/basic.c) tests a few
more things. The "f1" test case is the only part that's immediately
relevant to this change, but the other cases were part of the same test
in the incubator and they are supported so I brought in the entire test.
The ClangIR CFG has to be flat before it can be lowered to LLVM IR. That
is, there can be no nested regions and all blocks in a region must
belong to the parent region. Currently only cir.scope operations violate
these rules, so the initial implementation of the cir-flatten-cfg pass
only has to transform scope operations.
GCC, unlike clang, issues a warning when one virtual function is
overridden in a derived class but one or more other virtual functions
with the same name and different signature from a base class are not
overridden. This leads to many warnings in the MLIR and ClangIR code
when using the OpenConversionPattern<>::matchAndRewrite() function in
the ordinary way. The "hiding" behavior is what we want, so we're just
disabling the warning here.
Local variable initialization was previously being ignored. This change
adds support for initialization of scalar variables with constant values
and introduces the constant emitter framework.
This change implements variable linkage types in ClangIR except for
common linkage which requires Comdat support.
---------
Co-authored-by: Morris Hafner <mhafner@nvidia.com>
Co-authored-by: Henrich Lauko <xlauko@mail.muni.cz>
Previous CIR commits have introduced a few warnings. This change fixes
those.
There are still warnings present when building with GCC because GCC
warns about virtual functions being hidden in the mlir::OpConversion
classes. A separate discussion will be required to decide what should be
done about those.
The previously upstreamed lowering from ClangIR to LLVM IR diverged from
the incubator implementation, but when the incubator was updated to
incorporate these changes some issues arose which require the upstream
implementation to be modified to re-align with the incubator.
First, in the earlier upstream implementation a CIRAttrVisitor class was
introduced with the intention that an mlir-tblgen based extension would
be created to automatically add all CIR attributes to the visitor. When
I proposed this in mlir-tblgen a reviewer suggested that what I wanted
could be better accomplished with TypeSwitch.
See https://github.com/llvm/llvm-project/pull/126332
This was done in the incubator, and here I am bringing that
implementation upstream.
The other issue was that the global op initialization in the incubator
had more cases than I had accounted for in my previous upstream
refactoring. I did still refactor the incubator code, but not in quite
the same way as the upstream code. This change re-aligns the two.
