fix: C++ empty record with align break aarch64 AAPCS64

Author: hstk30-hw

clang/docs/ReleaseNotes.rst

@@ -110,6 +110,9 @@ C++ Specific Potentially Breaking Changes
 ABI Changes in This Version
 ---------------------------
 
+- Let C++ empty record fall through to the standard argument-handling path instead of
+  always pass a single ``i8`` according to aarch64 AAPCS rules.
+
 AST Dumping Potentially Breaking Changes
 ----------------------------------------
 - How nested name specifiers are dumped and printed changes, keeping track of clang AST changes.

clang/lib/CodeGen/Targets/AArch64.cpp

@@ -422,6 +422,12 @@ ABIArgInfo AArch64ABIInfo::classifyArgumentType(QualType Ty, bool IsVariadicFn,
   }
 
   // Empty records:
+  // AAPCS64 does not say that empty records are ignored as arguments,
+  // but other compilers do so in certain situations, and we copy that behavior.
+  // Those situations are in fact language-mode-specific, which seems really
+  // unfortunate, but it's something we just have to accept. If this doesn't
+  // apply, just fall through to the standard argument-handling path.
+  // Darwin overrides the psABI here to ignore all empty records in all modes.
   uint64_t Size = getContext().getTypeSize(Ty);
   bool IsEmpty = isEmptyRecord(getContext(), Ty, true);
   if (!Ty->isSVESizelessBuiltinType() && (IsEmpty || Size == 0)) {
@@ -434,9 +440,6 @@ ABIArgInfo AArch64ABIInfo::classifyArgumentType(QualType Ty, bool IsVariadicFn,
     // behaviour here.
     if (Size == 0)
       return ABIArgInfo::getIgnore();
-
-    // Otherwise, they are passed as if they have a size of 1 byte.
-    return ABIArgInfo::getDirect(llvm::Type::getInt8Ty(getVMContext()));
   }
 
   // Homogeneous Floating-point Aggregates (HFAs) need to be expanded.

clang/test/CodeGen/AArch64/args.cpp

@@ -17,11 +17,29 @@ struct Empty {};
 
 // DARWIN: define{{.*}} i32 @empty_arg(i32 noundef %a)
 // C: define{{.*}} i32 @empty_arg(i32 noundef %a)
-// CXX: define{{.*}} i32 @empty_arg(i8 %e.coerce, i32 noundef %a)
+// CXX: define{{.*}} i32 @empty_arg(i64 %e.coerce, i32 noundef %a)
 EXTERNC int empty_arg(struct Empty e, int a) {
   return a;
 }
 
+// CXX: define{{.*}} i32 @empty_align8_arg(i64 %a.coerce, i32 noundef %b)
+struct EmptyAlign8 { int __attribute__((aligned(8))) : 0; };
+EXTERNC int empty_align8_arg(struct EmptyAlign8 a, int b) {
+  return b;
+}
+
+// CXX: define{{.*}} i32 @empty_align16_arg(i128 %a.coerce, i32 noundef %b)
+struct EmptyAlign16 { long long int __attribute__((aligned(16))) : 0; };
+EXTERNC int empty_align16_arg(struct EmptyAlign16 a, int b) {
+  return b;
+}
+
+// CXX: define{{.*}} i32 @empty_align32_arg(ptr dead_on_return noundef %a, i32 noundef %b)
+struct EmptyAlign32 { long long int __attribute__((aligned(32))) : 0; };
+EXTERNC int empty_align32_arg(struct EmptyAlign32 a, int b) {
+  return b;
+}
+
 // DARWIN: define{{.*}} void @empty_ret()
 // C: define{{.*}} void @empty_ret()
 // CXX: define{{.*}} void @empty_ret()

clang/test/CodeGen/AArch64/struct-coerce-using-ptr.cpp

@@ -575,19 +575,21 @@ void TSpp_align16(SSpp_align16 s) { *s.a.x = 1; }
 struct Sempty {
 };
 // CHECK-A64-LABEL: define dso_local void @_Z6Tempty6Sempty(
-// CHECK-A64-SAME: i8 [[S_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-A64-SAME: i64 [[S_COERCE:%.*]]) #[[ATTR0]] {
 // CHECK-A64-NEXT:  [[ENTRY:.*:]]
 // CHECK-A64-NEXT:    [[S:%.*]] = alloca [[STRUCT_SEMPTY:%.*]], align 1
 // CHECK-A64-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_SEMPTY]], ptr [[S]], i32 0, i32 0
-// CHECK-A64-NEXT:    store i8 [[S_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-A64-NEXT:    [[COERCE_VAL_II:%.*]] = trunc i64 [[S_COERCE]] to i8
+// CHECK-A64-NEXT:    store i8 [[COERCE_VAL_II]], ptr [[COERCE_DIVE]], align 1
 // CHECK-A64-NEXT:    ret void
 //
 // CHECK-A64_32-LABEL: define void @_Z6Tempty6Sempty(
-// CHECK-A64_32-SAME: i8 [[S_COERCE:%.*]]) #[[ATTR0]] {
+// CHECK-A64_32-SAME: i64 [[S_COERCE:%.*]]) #[[ATTR0]] {
 // CHECK-A64_32-NEXT:  [[ENTRY:.*:]]
 // CHECK-A64_32-NEXT:    [[S:%.*]] = alloca [[STRUCT_SEMPTY:%.*]], align 1
 // CHECK-A64_32-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_SEMPTY]], ptr [[S]], i32 0, i32 0
-// CHECK-A64_32-NEXT:    store i8 [[S_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-A64_32-NEXT:    [[COERCE_VAL_II:%.*]] = trunc i64 [[S_COERCE]] to i8
+// CHECK-A64_32-NEXT:    store i8 [[COERCE_VAL_II]], ptr [[COERCE_DIVE]], align 1
 // CHECK-A64_32-NEXT:    ret void
 //
 void Tempty(Sempty s) { }

clang/test/CodeGen/arm64-microsoft-arguments.cpp

@@ -57,7 +57,7 @@ S4 f4() {
 
 // Pass and return from instance method called from instance method.
 // CHECK: define {{.*}} void @{{.*}}bar@Q1{{.*}}(ptr {{[^,]*}} %this, ptr dead_on_unwind inreg noalias writable sret(%class.P1) align 1 %agg.result)
-// CHECK: call void {{.*}}foo@P1{{.*}}(ptr noundef{{[^,]*}} %ref.tmp, ptr dead_on_unwind inreg writable sret(%class.P1) align 1 %agg.result, i8 %0)
+// CHECK: call void {{.*}}foo@P1{{.*}}(ptr noundef{{[^,]*}} %ref.tmp, ptr dead_on_unwind inreg writable sret(%class.P1) align 1 %agg.result, i64 %coerce.val.ii)
 
 class P1 {
 public:
@@ -76,7 +76,7 @@ P1 Q1::bar() {
 
 // Pass and return from instance method called from free function.
 // CHECK: define {{.*}} void {{.*}}bar{{.*}}()
-// CHECK: call void {{.*}}foo@P2{{.*}}(ptr noundef{{[^,]*}} %ref.tmp, ptr dead_on_unwind inreg writable sret(%class.P2) align 1 %retval, i8 %0)
+// CHECK: call void {{.*}}foo@P2{{.*}}(ptr noundef{{[^,]*}} %ref.tmp, ptr dead_on_unwind inreg writable sret(%class.P2) align 1 %retval, i64 %coerce.val.ii)
 class P2 {
 public:
   P2 foo(P2 x);

clang/test/CodeGenCXX/aarch64-arguments.cpp

@@ -1,5 +1,5 @@
 // RUN: %clang_cc1 -triple arm64-none-linux -emit-llvm -w -o - %s | FileCheck -check-prefix=PCS %s
 
-// PCS: define{{.*}} void @{{.*}}(i8 %a
+// PCS: define{{.*}} void @{{.*}}(i64 %a.coerce)
 struct s0 {};
 void f0(s0 a) {}

clang/test/CodeGenCXX/arm64-darwinpcs.cpp

@@ -7,7 +7,7 @@ void test_extensions(bool a, char b, short c) {}
 
 struct Empty {};
 void test_empty(Empty e) {}
-// CHECK: define{{.*}} void @_Z10test_empty5Empty(i8
+// CHECK: define{{.*}} void @_Z10test_empty5Empty(i64 %e.coerce)
 // CHECK-DARWIN: define{{.*}} void @_Z10test_empty5Empty()
 
 struct HFA {