[CIR] Add basic thread-local storage (TLS) support #1996
+112
−8
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
lanza
requested review from
andykaylor,
bcardosolopes and
xlauko
as code owners
lanza
added a commit
that referenced
this pull request
Nov 20, 2025
This commit implements basic thread-local storage support in ClangIR: 1. Added EmitThreadLocalVarDeclLValue() method to CIRGenCXXABI interface to handle thread-local variable access in an ABI-specific way. 2. Implemented EmitThreadLocalVarDeclLValue() in CIRGenItaniumCXXABI that: - Accesses thread-local variables directly using their GlobalOp - Relies on the 'tls_dyn' attribute on GlobalOp for TLS semantics - Handles both value and reference types correctly - Defers complex TLS wrapper function support to future work 3. Modified CIRGenExpr.cpp to call the ABI method instead of asserting when encountering TLS_Dynamic variables. 4. Added CIRGenModule::getAddrOfGlobalVar() helper for retrieving global variable addresses. 5. Added comprehensive test coverage in thread-local.cpp that verifies: - Basic thread_local variables with constant initialization - GNU-style __thread variables - Function-local thread_local static variables - Reading from and writing to thread_local variables - Extern thread_local variable declarations - Correct CIR generation with tls_dyn attribute - Proper LLVM IR lowering to llvm.threadlocal.address intrinsic The implementation takes a simplified approach for now, relying on the GlobalOp's thread_local attribute rather than generating wrapper functions. This is sufficient for variables with trivial initialization. Future work may add full wrapper function support for complex dynamic initialization. ghstack-source-id: 65ab4e1 Pull-Request: #1996
lanza
added a commit
that referenced
this pull request
Nov 20, 2025
This commit implements basic thread-local storage support in ClangIR: 1. Added EmitThreadLocalVarDeclLValue() method to CIRGenCXXABI interface to handle thread-local variable access in an ABI-specific way. 2. Implemented EmitThreadLocalVarDeclLValue() in CIRGenItaniumCXXABI that: - Accesses thread-local variables directly using their GlobalOp - Relies on the 'tls_dyn' attribute on GlobalOp for TLS semantics - Handles both value and reference types correctly - Defers complex TLS wrapper function support to future work 3. Modified CIRGenExpr.cpp to call the ABI method instead of asserting when encountering TLS_Dynamic variables. 4. Added CIRGenModule::getAddrOfGlobalVar() helper for retrieving global variable addresses. 5. Added comprehensive test coverage in thread-local.cpp that verifies: - Basic thread_local variables with constant initialization - GNU-style __thread variables - Function-local thread_local static variables - Reading from and writing to thread_local variables - Extern thread_local variable declarations - Correct CIR generation with tls_dyn attribute - Proper LLVM IR lowering to llvm.threadlocal.address intrinsic The implementation takes a simplified approach for now, relying on the GlobalOp's thread_local attribute rather than generating wrapper functions. This is sufficient for variables with trivial initialization. Future work may add full wrapper function support for complex dynamic initialization. ghstack-source-id: 2a8fafc Pull-Request: #1996
lanza
added a commit
that referenced
this pull request
Nov 20, 2025
This commit implements thread-local storage support in ClangIR: 1. Added EmitThreadLocalVarDeclLValue() method to CIRGenCXXABI interface to handle thread-local variable access in an ABI-specific way. 2. Implemented EmitThreadLocalVarDeclLValue() in CIRGenItaniumCXXABI that: - Accesses thread-local variables using cir.get_global thread_local - Relies on LLVM lowering to insert @llvm.threadlocal.address intrinsics - Handles both value and reference types correctly - Uses ClangIR's declarative approach instead of wrapper functions 3. Modified CIRGenExpr.cpp to call the ABI method for TLS_Dynamic variables. 4. Added CIRGenModule::getAddrOfGlobalVar() helper for retrieving global variable addresses. 5. Added comprehensive test coverage in thread-local.cpp that verifies: - Basic thread_local variables with constant initialization - GNU-style __thread variables - Function-local thread_local static variables - Reading from and writing to thread_local variables - Extern thread_local variable declarations - Correct CIR generation with tls_dyn attribute - Proper LLVM IR lowering to llvm.threadlocal.address intrinsic ClangIR's approach differs from traditional CodeGen: instead of creating wrapper functions (_ZTW*) in the IR, ClangIR marks GlobalOps with tls_dyn and relies on LLVM lowering to insert @llvm.threadlocal.address intrinsics. This achieves correct TLS semantics while maintaining ClangIR's higher-level representation. ghstack-source-id: d912c08 Pull-Request: #1996
lanza
added a commit
that referenced
this pull request
Nov 20, 2025
This commit implements thread-local storage support in ClangIR: 1. Added EmitThreadLocalVarDeclLValue() method to CIRGenCXXABI interface to handle thread-local variable access in an ABI-specific way. 2. Implemented EmitThreadLocalVarDeclLValue() in CIRGenItaniumCXXABI that: - Accesses thread-local variables using cir.get_global thread_local - Relies on LLVM lowering to insert @llvm.threadlocal.address intrinsics - Handles both value and reference types correctly - Uses ClangIR's declarative approach instead of wrapper functions 3. Modified CIRGenExpr.cpp to call the ABI method for TLS_Dynamic variables. 4. Added CIRGenModule::getAddrOfGlobalVar() helper for retrieving global variable addresses. 5. Added comprehensive test coverage in thread-local.cpp that verifies: - Basic thread_local variables with constant initialization - GNU-style __thread variables - Function-local thread_local static variables - Reading from and writing to thread_local variables - Extern thread_local variable declarations - Correct CIR generation with tls_dyn attribute - Proper LLVM IR lowering to llvm.threadlocal.address intrinsic ClangIR's approach differs from traditional CodeGen: instead of creating wrapper functions (_ZTW*) in the IR, ClangIR marks GlobalOps with tls_dyn and relies on LLVM lowering to insert @llvm.threadlocal.address intrinsics. This achieves correct TLS semantics while maintaining ClangIR's higher-level representation. ghstack-source-id: 267f175 Pull-Request: #1996
bcardosolopes
requested changes
Nov 20, 2025
Member
bcardosolopes
left a comment
bcardosolopes
left a comment
There was a problem hiding this comment.
Overall good, thanks! Some nits and ready to go
| llvm::DenseSet<clang::GlobalDecl> DiagnosedConflictingDefinitions; | ||
|
|
||
| /// thread_local variables defined or used in this TU. | ||
| std::vector<const clang::VarDecl *> CXXThreadLocals; |
Member
There was a problem hiding this comment.
this should be lowercase too!
Member
There was a problem hiding this comment.
this one needs changing too (even though a few others on this file haven't been migrated just yet)
lanza
added a commit
that referenced
this pull request
Nov 21, 2025
This commit implements thread-local storage support in ClangIR: 1. Added EmitThreadLocalVarDeclLValue() method to CIRGenCXXABI interface to handle thread-local variable access in an ABI-specific way. 2. Implemented EmitThreadLocalVarDeclLValue() in CIRGenItaniumCXXABI that: - Accesses thread-local variables using cir.get_global thread_local - Relies on LLVM lowering to insert @llvm.threadlocal.address intrinsics - Handles both value and reference types correctly - Uses ClangIR's declarative approach instead of wrapper functions 3. Modified CIRGenExpr.cpp to call the ABI method for TLS_Dynamic variables. 4. Added CIRGenModule::getAddrOfGlobalVar() helper for retrieving global variable addresses. 5. Added comprehensive test coverage in thread-local.cpp that verifies: - Basic thread_local variables with constant initialization - GNU-style __thread variables - Function-local thread_local static variables - Reading from and writing to thread_local variables - Extern thread_local variable declarations - Correct CIR generation with tls_dyn attribute - Proper LLVM IR lowering to llvm.threadlocal.address intrinsic ClangIR's approach differs from traditional CodeGen: instead of creating wrapper functions (_ZTW*) in the IR, ClangIR marks GlobalOps with tls_dyn and relies on LLVM lowering to insert @llvm.threadlocal.address intrinsics. This achieves correct TLS semantics while maintaining ClangIR's higher-level representation. ghstack-source-id: 0a94a40 Pull-Request: #1996
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Stack from ghstack (oldest at bottom):
This commit implements basic thread-local storage support in ClangIR:
Added EmitThreadLocalVarDeclLValue() method to CIRGenCXXABI interface
to handle thread-local variable access in an ABI-specific way.
Implemented EmitThreadLocalVarDeclLValue() in CIRGenItaniumCXXABI that:
Modified CIRGenExpr.cpp to call the ABI method instead of asserting
when encountering TLS_Dynamic variables.
Added CIRGenModule::getAddrOfGlobalVar() helper for retrieving
global variable addresses.
Added comprehensive test coverage in thread-local.cpp that verifies:
The implementation takes a simplified approach for now, relying on the
GlobalOp's thread_local attribute rather than generating wrapper functions.
This is sufficient for variables with trivial initialization. Future work
may add full wrapper function support for complex dynamic initialization.