Skip to content

[flang][cuda] Fix crash in semantic #88577

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Apr 12, 2024
Merged

Conversation

clementval
Copy link
Contributor

Fix for #88451

Do not perform semantic check about data transfer on assignment statement in device context.

@clementval clementval requested a review from klausler April 12, 2024 20:57
@llvmbot llvmbot added flang Flang issues not falling into any other category flang:semantics labels Apr 12, 2024
@llvmbot
Copy link
Member

llvmbot commented Apr 12, 2024

@llvm/pr-subscribers-flang-semantics

Author: Valentin Clement (バレンタイン クレメン) (clementval)

Changes

Fix for #88451

Do not perform semantic check about data transfer on assignment statement in device context.


Full diff: https://github.com/llvm/llvm-project/pull/88577.diff

2 Files Affected:

  • (modified) flang/lib/Semantics/check-cuda.cpp (+6-1)
  • (modified) flang/test/Semantics/cuf11.cuf (+12)
diff --git a/flang/lib/Semantics/check-cuda.cpp b/flang/lib/Semantics/check-cuda.cpp
index e0a796972441ba..49679c1a9304f3 100644
--- a/flang/lib/Semantics/check-cuda.cpp
+++ b/flang/lib/Semantics/check-cuda.cpp
@@ -480,10 +480,15 @@ void CUDAChecker::Enter(const parser::CUFKernelDoConstruct &x) {
 }
 
 void CUDAChecker::Enter(const parser::AssignmentStmt &x) {
+  auto lhsLoc{std::get<parser::Variable>(x.t).GetSource()};
+  const auto &scope{context_.FindScope(lhsLoc)};
+  const Scope &progUnit{GetProgramUnitContaining(scope)};
+  if (IsCUDADeviceContext(&progUnit))
+    return; // Data transfer with assignment is only perform on host.
+
   const evaluate::Assignment *assign{semantics::GetAssignment(x)};
   int nbLhs{evaluate::GetNbOfCUDASymbols(assign->lhs)};
   int nbRhs{evaluate::GetNbOfCUDASymbols(assign->rhs)};
-  auto lhsLoc{std::get<parser::Variable>(x.t).GetSource()};
 
   // device to host transfer with more than one device object on the rhs is not
   // legal.
diff --git a/flang/test/Semantics/cuf11.cuf b/flang/test/Semantics/cuf11.cuf
index 96108e2b245560..b915c7246ed947 100644
--- a/flang/test/Semantics/cuf11.cuf
+++ b/flang/test/Semantics/cuf11.cuf
@@ -1,5 +1,17 @@
 ! RUN: %python %S/test_errors.py %s %flang_fc1
 
+module mod1
+contains
+  attributes(global) subroutine sub1(adev)
+    real :: adev(10)
+    integer :: tid
+    tid = threadIdx%x
+! Use to crash the compiler. Make sure we have the proper semantic error.
+!ERROR: Actual argument for 'i=' has bad type 'REAL(4)'
+    adev(tid + 1) = scale(real(tid), 2.0) 
+  end subroutine sub1
+end module
+
 subroutine sub1()
   real, device :: adev(10), bdev(10)
   real :: ahost(10)

@clementval clementval merged commit 7b6b023 into llvm:main Apr 12, 2024
@clementval clementval deleted the cuda_crash branch April 12, 2024 21:28
bazuzi pushed a commit to bazuzi/llvm-project that referenced this pull request Apr 15, 2024
Fix for llvm#88451

Do not perform semantic check about data transfer on assignment
statement in device context.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
flang:semantics flang Flang issues not falling into any other category
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants