[clang] 8918319 - [CIR] Implement non-scalar lvalue return values (#190795)

[via cfe-commits]

Author: Erich Keane
Date: 2026-04-07T14:27:08-07:00
New Revision: 8918319853fbdf9e6f6cb69e96848f913a22bc31

URL: https://github.com/llvm/llvm-project/commit/8918319853fbdf9e6f6cb69e96848f913a22bc31
DIFF: https://github.com/llvm/llvm-project/commit/8918319853fbdf9e6f6cb69e96848f913a22bc31.diff

LOG: [CIR] Implement non-scalar lvalue return values (#190795)

I could only get these to happen in C++03 (as we do a
materialize-temporary-expr in later standards), but this does appear in
a number of benchmarks. The implementation ends up being pretty trivial,
as we just have to lower the aggregate correctly.

Added: 
    clang/test/CIR/CodeGen/non-scalar-lval-return.cpp

Modified: 
    clang/lib/CIR/CodeGen/CIRGenExpr.cpp

Removed: 
    


################################################################################
diff  --git a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
index 81737aeb4c847..a306cc68dff8e 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExpr.cpp
@@ -1895,10 +1895,9 @@ LValue CIRGenFunction::emitCompoundLiteralLValue(const CompoundLiteralExpr *e) {
 LValue CIRGenFunction::emitCallExprLValue(const CallExpr *e) {
   RValue rv = emitCallExpr(e);
 
-  if (!rv.isScalar()) {
-    cgm.errorNYI(e->getSourceRange(), "emitCallExprLValue: non-scalar return");
-    return {};
-  }
+  if (!rv.isScalar())
+    return makeAddrLValue(rv.getAggregateAddress(), e->getType(),
+                          AlignmentSource::Decl);
 
   assert(e->getCallReturnType(getContext())->isReferenceType() &&
          "Can't have a scalar return unless the return type is a "

diff  --git a/clang/test/CIR/CodeGen/non-scalar-lval-return.cpp b/clang/test/CIR/CodeGen/non-scalar-lval-return.cpp
new file mode 100644
index 0000000000000..df8186fc1f6ff
--- /dev/null
+++ b/clang/test/CIR/CodeGen/non-scalar-lval-return.cpp
@@ -0,0 +1,42 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -std=c++03 -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s -check-prefix=CIR
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -std=c++03 -fclangir -emit-llvm %s -o %t-cir.ll
+// RUN: FileCheck --input-file=%t-cir.ll %s -check-prefix=LLVM,LLVMCIR
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -std=c++03 -emit-llvm %s -o %t.ll
+// RUN: FileCheck --input-file=%t.ll %s -check-prefix=LLVM,OGCG
+
+struct Struct {
+  int member;
+  Struct(int);
+};
+
+extern "C" Struct getStruct(int i) { return i; }
+
+extern "C" void use() {
+  int g = getStruct(0).member;
+
+  // CIR-LABEL: @use()
+  // CIR: %[[G_ALLOCA:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["g", init]
+  // CIR: %[[TEMP_ALLOCA:.*]] = cir.alloca !rec_Struct, !cir.ptr<!rec_Struct>
+  // CIR: %[[ZERO:.*]] = cir.const #cir.int<0> : !s32i
+  // CIR: %[[GET_STRUCT_CALL:.*]] = cir.call @getStruct(%[[ZERO]])
+  // CIR: cir.store{{.*}} %[[GET_STRUCT_CALL]], %[[TEMP_ALLOCA]]
+  // CIR: %[[GET_MEMBER:.*]] = cir.get_member %[[TEMP_ALLOCA]][0] {name = "member"}
+  // CIR: %[[LOAD_MEM:.*]] = cir.load{{.*}}%[[GET_MEMBER]]
+  // CIR: cir.store{{.*}} %[[LOAD_MEM]], %[[G_ALLOCA]] : !s32i, !cir.ptr<!s32i>
+  //
+  // LLVM-LABEL: @use()
+  // LLVM: %[[G_ALLOCA:.*]] = alloca i32
+  // LLVM: %[[TEMP_ALLOCA:.*]] = alloca %struct.Struct
+  //
+  // LLVMCIR: %[[GET_STRUCT_CALL:.*]] = call %struct.Struct @getStruct(i32 noundef 0)
+  // LLVMCIR: store %struct.Struct %[[GET_STRUCT_CALL]], ptr %[[TEMP_ALLOCA]]
+  // OGCG: %[[GET_STRUCT_CALL_BEFORE:.*]] = call i32 @getStruct(i32 noundef 0)
+  // OGCG: %[[GET_STRUCT_CALL:.*]] = getelementptr{{.*}}%struct.Struct, ptr %[[TEMP_ALLOCA]], i32 0, i32 0
+  // OGCG: store i32 %[[GET_STRUCT_CALL_BEFORE]], ptr %[[GET_STRUCT_CALL]]
+  //
+  // LLVM: %[[GET_MEMBER:.*]] = getelementptr {{.*}}%struct.Struct, ptr %[[TEMP_ALLOCA]], i32 0, i32 0
+  // LLVM: %[[LOAD_MEM:.*]] = load i32, ptr %[[GET_MEMBER]]
+  // LLVM: store i32 %[[LOAD_MEM]], ptr %[[G_ALLOCA]]
+}
+