[CIR] Fix MLIRCIRInterfaces and clangFrontendTool build error #1866
+164,589
−44,614
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…m#1559) During the initial upstreaming of `cir.struct` support, a request was made to rename `cir.struct` to `cir.record` for consistency with its multiple uses. This change makes the modification along with renaming various other related classes. I've attempted to also update variable names and comments to keep everything consistent. This creates an unfortunate name collision between cir::RecordType and clang::RecordType, but the impact of that overlaps was relatively small. The only intended behavioral change is in the names used in the emitted CIR files and diagnostics.
…vm#1563) This change reverse the polarity of the cir::RecordType incomplete attribute, making the attribute name "complete" and reversing all checks in the code accordingly. The asm layout still prints "incomplete" for records that are not complete and treats the absence of this text as meaning that the definition provided is complete. No functional change is intended.
When we create a `cir.record` type, we track the `RecordKind` according to whether the source code declared the type using `struct` or `class` even though these are semantically equivalent. The distinction is used in naming the type when we lower to the LLVM dialect. This change updates the code to remove the last place where we were handling class and struct records differently.
Negative shift amounts are undefined behavior in C and C++. Because of that we can always zero-extend the shift amount which is slightly faster on certain architectures (e. g. x86). This also matches the behavior of the original clang Codegen. Backported from llvm/llvm-project#133405 Co-authored-by: Morris Hafner <[email protected]>
`cir.trap` corresponds to two operations, `call @llvm.trap` and `unreachable`. See the test case `Lowering/intrinsics.cir`. Co-authored-by: Yue Huang <[email protected]>
Currently, the following code snippet fails with a crash during CodeGen
```
class C {
public:
~C();
void operator=(C);
};
void d() {
C a, b;
a = b;
}
```
with error:
```
mlir::Block* clang::CIRGen::CIRGenFunction::getEHResumeBlock(bool, cir::TryOp): Assertion `tryOp && "expected available cir.try"' failed.
```
in CIRGenCleanup [these
lines](https://github.com/llvm/clangir/blob/204c03efbe898c9f64e477937d869767fdfb1310/clang/lib/CIR/CodeGen/CIRGenCleanup.cpp#L615C1-L617C6)
don't check if there is a TryOp when at the end of the scope chain
before
[getEHResumeBlock](https://github.com/llvm/clangir/blob/204c03efbe898c9f64e477937d869767fdfb1310/clang/lib/CIR/CodeGen/CIRGenException.cpp#L764)
is called causing the crash, because it contains an assertion.
This PR fixes this and adds a simple test for a case like this.
…lvm#1567) This is a just small fix that cover the case when the global union is declared with `static` keyword and one of the its users is an array
Add `TypeBuilderWithInferredContext` to each CIR type that supports MLIR context inference from its parameters.
We have been using RecordLayoutAttr to "cache" data layout information calculated for records. Unfortunately, it wasn't actually caching the information, and because each call was calculating more information than it needed, it was doing extra work. This replaces the previous implementation with a set of functions that compute only the information needed. Ideally, we would like to have a mechanism to properly cache this information, but until such a mechanism is implemented, these new functions should be a small step forward.
Backport support for zero init Vector type
Lower vqsubd_s64 and vqadds_u32
This changes the alias prefix for record types to make it less general.
Introduce common base class for attributes with single type parameter.
…nups (llvm#1581) The following code snippet crashes during CIR CodeGen using `clang tmp.cpp -Xclang -emit-cir -S -o -`: ``` #include <vector> void foo() { std::vector<int> v(1); } ``` The crash is: ``` mlir::Operation* mlir::Block::getTerminator(): Assertion `mightHaveTerminator()' failed. ``` What happens is the scope created [here](https://github.com/llvm/clangir/blob/c7b27ece0971c632f2ebec26bc855ee9b118ffcc/clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp#L2370C1-L2381C10) is malformed. The operations inside the scope using `--mlir-print-assume-verified` looks something like: ``` %0 = cir.alloca !cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>, !cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>>, ["ref.tmp0"] {alignment = 1 : i64} %1 = cir.load <<UNKNOWN SSA VALUE>> : !cir.ptr<!cir.int<u, 64>>, !cir.int<u, 64> %2 = cir.load <<UNKNOWN SSA VALUE>> : !cir.ptr<!cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>>>, !cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>> cir.call @_ZNSaIiEC1ERKS_(%0, %2) : (!cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>>, !cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>>) -> () extra(#cir<extra({nothrow = #cir.nothrow})>) %3 = cir.call @_ZNSt6vectorIiSaIiEE11_S_max_sizeERKS0_(%0) : (!cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>>) -> !cir.int<u, 64> extra(#cir<extra({nothrow = #cir.nothrow})>) %4 = cir.cmp(gt, %1, %3) : !cir.int<u, 64>, !cir.bool cir.yield %4 : !cir.bool cir.call @_ZNSaIiED1Ev(%0) : (!cir.ptr<!cir.record<class "std::allocator<int>" padded {!cir.int<u, 8>} #cir.record.decl.ast>>) -> () extra(#cir<extra({nothrow = #cir.nothrow})>) ``` `_ZNSaIiED1Ev` is `std::allocator<int>::~allocator()`. So, the destructor comes after the YieldOp has been created, which is wrong. The destructor comes from the destruction of [`RunCleanupScope`](https://github.com/llvm/clangir/blob/c7b27ece0971c632f2ebec26bc855ee9b118ffcc/clang/lib/CIR/CodeGen/CIRGenFunction.h#L1369) since `LexicalScope` inherits from it. RunCleanupsScope's dtor should come before the yield is created! This PR fixes this by calling `ForceCleanup` before creating the yield, so the YieldOp becomes the last operation in the scope. I found this bug using the code snippet above, but I have added a reduced version, using [creduce](https://github.com/csmith-project/creduce), as a test, because [std-cxx.h](https://github.com/llvm/clangir/blob/main/clang/test/CIR/Inputs/std-cxx.h) doesn't quite replicate `std::vector` correctly.
This is a backport from llvm/llvm-project#137184 Co-authored-by: Morris Hafner <[email protected]>
Co-authored-by: Morris Hafner <[email protected]>
llvm#1591) Standardize element type name
In this PR I extended the incubator implementation to support global init for VectorType
Backport `VectorType::verify` implementation with test for invalid cases
- Adds `CIR_` prefixes to integer type constraints types to disambiguate their names from other dialects. - Renames `PrimitiveInt` to `CIR_AnyFundamentalIntType` to align more with constraint conventions. - Adds bunch of helper constraint classes to be able to define base types to reduce clutter of necessary type casts. - Reworks constraints to use `CIR_ConfinedType` to avoid repeating validation checks. - Adds `IntOfWidths` variadic bitwidth constraint to reduce boilerplate code needed to handle multi-bitwidth parameters. - Constraints are moved into a separate file, which starts decoupling of constraints and types to remove the cyclic dependency between types and attributes and will eventually help fix several outstanding TODOs.
This refactors cir floating point type constraints, and fixes long double verifier to use constraints directly.
Apply `CIR_AnyIntOrFloatType` type constraint on element type and rename `elementTy` to `elementType`.
Backport the Comma Operator for ComplexType from the upstream
This PR fixes an issue when parsing `TBAAAttr` derived attributes. The root cause of the issue is that the RTTI system is not capable of downcasting `mlir::Attribute` to the `BaseType`, which is `cir::TBAAAttr`, since each attribute has its unique TypeID. As such, the default generated parser fails, because the `dyn_cast` to `cir::TBAAAttr` fails (this happens for example with the attached test case). If anyone can think of a solution that wouldn't require explicitly listing all the attributes I'm happy to modify it.
Implements Issue llvm#1802
This PR adds support for the **discrete bit-field layout**, which is used in the Microsoft bit-field ABI. Since ClangIR does not currently support the Microsoft ABI, this uses the compiler flag ```bash -mms-bitfields Set the default structure layout to be compatible with the Microsoft compiler standard ``` Fixes llvm#1798
Adding these so work can be done in parallel for any PTX related intrinsic. related llvm#1813
This PR backports support for volatile bit-fields in AAPCS. The corresponding upstream PRs are: - llvm/llvm-project#151252 - llvm/llvm-project#151875
…vm#1859) 1. **Catch Entry Parsing**: `parseCatchEntry` in `CIRDialect.cpp` now properly handles `#cir.unwind` attributes 2. **Resume Operation**: Fix record definition of `cir.resume` in order to handle locations correctly ## Issue: - `error: expected '{' to begin a region` when parsing `catch [#cir.unwind {` - `error: undefined symbol alias id` when parsing `cir.resume loc(#loc)`
Implemented missing feature for issue llvm#1793. Also implemented LLVM lowering for the hot attribute, and added a new test case.
wannacu
requested review from
andykaylor,
bcardosolopes,
lanza and
xlauko
as code owners
wannacu
force-pushed
the
fix_build_errors
branch
from
edad603 to
725ed34
Compare
xlauko
reviewed
Sep 2, 2025
Collaborator
xlauko
left a comment
xlauko
left a comment
There was a problem hiding this comment.
Can you update PR name to contain [CIR] prefix.
Also it would be nice to provide in the PR description the actual bug.
bcardosolopes
approved these changes
Sep 3, 2025
bcardosolopes
changed the title
lanza
force-pushed
the
main
branch
2 times, most recently
from
8acaf96 to
58135ea
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
27 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Remove redundant call while here