[flang-commits] [flang] [flang][cuda] Do not initialize shared variable (PR #178489)

[via flang-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

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

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

<details>
<summary>Changes</summary>



---
Full diff: https://github.com/llvm/llvm-project/pull/178489.diff


2 Files Affected:

- (modified) flang/lib/Optimizer/Transforms/CUDA/CUFComputeSharedMemoryOffsetsAndSize.cpp (-7) 
- (modified) flang/test/Fir/CUDA/cuda-shared-offset.mlir (+6-6) 


``````````diff
diff --git a/flang/lib/Optimizer/Transforms/CUDA/CUFComputeSharedMemoryOffsetsAndSize.cpp b/flang/lib/Optimizer/Transforms/CUDA/CUFComputeSharedMemoryOffsetsAndSize.cpp
index 7bae0602fe5ca..87dc27e0382ef 100644
--- a/flang/lib/Optimizer/Transforms/CUDA/CUFComputeSharedMemoryOffsetsAndSize.cpp
+++ b/flang/lib/Optimizer/Transforms/CUDA/CUFComputeSharedMemoryOffsetsAndSize.cpp
@@ -70,13 +70,6 @@ static void createSharedMemoryGlobal(fir::FirOpBuilder &builder,
                                   cuf::DataAttribute::Shared)));
 
   mlir::DenseElementsAttr init = {};
-  mlir::Type i8Ty = builder.getI8Type();
-  if (size > 0) {
-    auto vecTy = mlir::VectorType::get(
-        static_cast<fir::SequenceType::Extent>(size), i8Ty);
-    mlir::Attribute zero = mlir::IntegerAttr::get(i8Ty, 0);
-    init = mlir::DenseElementsAttr::get(vecTy, llvm::ArrayRef(zero));
-  }
   auto sharedMem =
       fir::GlobalOp::create(builder, loc, sharedMemGlobalName, false, false,
                             sharedMemType, init, linkage, attrs);
diff --git a/flang/test/Fir/CUDA/cuda-shared-offset.mlir b/flang/test/Fir/CUDA/cuda-shared-offset.mlir
index 1a39fefe85cda..59ab040529ce4 100644
--- a/flang/test/Fir/CUDA/cuda-shared-offset.mlir
+++ b/flang/test/Fir/CUDA/cuda-shared-offset.mlir
@@ -51,12 +51,12 @@ module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<!llvm.ptr, dense<
 // CHECK: cuf.shared_memory[%c0{{.*}} : i32] f32 align 4 {bindc_name = "r", isStatic, uniq_name = "_QFshared_staticEr"} -> !fir.ref<f32>
 // CHECK: gpu.return
 // CHECK: }
-// CHECK: fir.global internal @_QPshared_static__shared_mem__a(dense<0> : vector<4xi8>) {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
-// CHECK: fir.global internal @_QPshared_static__shared_mem__b(dense<0> : vector<4xi8>) {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
-// CHECK: fir.global internal @_QPshared_static__shared_mem__c(dense<0> : vector<4xi8>) {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
-// CHECK: fir.global internal @_QPshared_static__shared_mem__d(dense<0> : vector<4xi8>) {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
-// CHECK: fir.global internal @_QPshared_static__shared_mem__e(dense<0> : vector<8xi8>) {alignment = 8 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<8xi8>
-// CHECK: fir.global internal @_QPshared_static__shared_mem__r(dense<0> : vector<4xi8>) {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
+// CHECK: fir.global internal @_QPshared_static__shared_mem__a {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
+// CHECK: fir.global internal @_QPshared_static__shared_mem__b {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
+// CHECK: fir.global internal @_QPshared_static__shared_mem__c {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
+// CHECK: fir.global internal @_QPshared_static__shared_mem__d {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
+// CHECK: fir.global internal @_QPshared_static__shared_mem__e {alignment = 8 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<8xi8>
+// CHECK: fir.global internal @_QPshared_static__shared_mem__r {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<4xi8>
 // CHECK: }
 // CHECK: }
 

``````````

</details>


https://github.com/llvm/llvm-project/pull/178489