[flang-commits] [flang] [flang] [cuda] Fix CUDA implicit data transfer entity creation (PR #139414)

[via flang-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-flang-fir-hlfir

Author: Zhen Wang (wangzpgi)

<details>
<summary>Changes</summary>

Fixed an issue in `genCUDAImplicitDataTransfer` where creating an `hlfir::Entity` from a symbol address could fail when the address comes from a `hlfir.declare` operation. Fix is to check if the address comes from a `hlfir.declare` operation. If so, use the base value from the declare op when available. Falling back to the original address otherwise.

---
Full diff: https://github.com/llvm/llvm-project/pull/139414.diff


1 Files Affected:

- (modified) flang/lib/Lower/Bridge.cpp (+7-1) 


``````````diff
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 43375e84f21fa..bfe8898ebff3d 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -4778,7 +4778,13 @@ class FirConverter : public Fortran::lower::AbstractConverter {
               nbDeviceResidentObject <= 1 &&
               "Only one reference to the device resident object is supported");
           auto addr = getSymbolAddress(sym);
-          hlfir::Entity entity{addr};
+          mlir::Value baseValue;
+          if (auto declareOp = llvm::dyn_cast<hlfir::DeclareOp>(addr.getDefiningOp()))
+            baseValue = declareOp.getBase();
+          else
+            baseValue = addr;
+
+          hlfir::Entity entity{baseValue};
           auto [temp, cleanup] =
               hlfir::createTempFromMold(loc, builder, entity);
           auto needCleanup = fir::getIntIfConstant(cleanup);

``````````

</details>


https://github.com/llvm/llvm-project/pull/139414