[CIR][AArch64][Lowering] Support fields with structs containing const…

…ant arrays or pointers (#1136)

This PR adds support for function arguments with structs that contain
constant arrays or pointers for AArch64.

For example, 
```
typedef struct {
  int a[42];
} CAT;

void pass_cat(CAT a) {}
```

As usual, the main ideas are gotten from the original
[CodeGen](https://github.com/llvm/clangir/blob/3aed38cf52e72cb51a907fad9dd53802f6505b81/clang/lib/AST/ASTContext.cpp#L1823),
and I have added a couple of tests.

clang/lib/CIR/Dialect/Transforms/TargetLowering/CIRLowerContext.cpp

@@ -94,6 +94,11 @@ clang::TypeInfo CIRLowerContext::getTypeInfoImpl(const mlir::Type T) const {
       Align = Target->getDoubleAlign();
       break;
     }
+    if (mlir::isa<PointerType>(T)) {
+      Width = Target->getPointerWidth(clang::LangAS::Default);
+      Align = Target->getPointerAlign(clang::LangAS::Default);
+      break;
+    }
     cir_cconv_unreachable("Unknown builtin type!");
     break;
   }
@@ -167,9 +172,28 @@ int64_t CIRLowerContext::toBits(clang::CharUnits CharSize) const {
   return CharSize.getQuantity() * getCharWidth();
 }
 
+/// Performing the computation in CharUnits
+/// instead of in bits prevents overflowing the uint64_t for some large arrays.
+clang::TypeInfoChars getConstantArrayInfoInChars(const CIRLowerContext &ctx,
+                                                 cir::ArrayType arrTy) {
+  clang::TypeInfoChars eltInfo = ctx.getTypeInfoInChars(arrTy.getEltType());
+  uint64_t tySize = arrTy.getSize();
+  assert((tySize == 0 || static_cast<uint64_t>(eltInfo.Width.getQuantity()) <=
+                             (uint64_t)(-1) / tySize) &&
+         "Overflow in array type char size evaluation");
+  uint64_t width = eltInfo.Width.getQuantity() * tySize;
+  unsigned align = eltInfo.Align.getQuantity();
+  if (!ctx.getTargetInfo().getCXXABI().isMicrosoft() ||
+      ctx.getTargetInfo().getPointerWidth(clang::LangAS::Default) == 64)
+    width = llvm::alignTo(width, align);
+  return clang::TypeInfoChars(clang::CharUnits::fromQuantity(width),
+                              clang::CharUnits::fromQuantity(align),
+                              eltInfo.AlignRequirement);
+}
+
 clang::TypeInfoChars CIRLowerContext::getTypeInfoInChars(mlir::Type T) const {
   if (auto arrTy = mlir::dyn_cast<ArrayType>(T))
-    cir_cconv_unreachable("NYI");
+    return getConstantArrayInfoInChars(*this, arrTy);
   clang::TypeInfo Info = getTypeInfo(T);
   return clang::TypeInfoChars(toCharUnitsFromBits(Info.Width),
                               toCharUnitsFromBits(Info.Align),

clang/test/CIR/CallConvLowering/AArch64/aarch64-cc-structs.c

@@ -234,3 +234,20 @@ S_PAD ret_s_pad() {
   S_PAD s;
   return s;
 }
+
+typedef struct {
+  int a[42];
+} CAT;
+
+// CHECK: cir.func @pass_cat(%arg0: !cir.ptr<!ty_CAT>
+// CHECK: %[[#V0:]]  = cir.alloca !cir.ptr<!ty_CAT>, !cir.ptr<!cir.ptr<!ty_CAT>>, [""] {alignment = 8 : i64}
+// CHECK: cir.store %arg0, %[[#V0]]  : !cir.ptr<!ty_CAT>, !cir.ptr<!cir.ptr<!ty_CAT>>
+// CHECK: %[[#V1:]]  = cir.load %[[#V0]]  : !cir.ptr<!cir.ptr<!ty_CAT>>, !cir.ptr<!ty_CAT>
+// CHECK: cir.return
+
+// LLVM: void @pass_cat(ptr %[[#V0:]])
+// LLVM: %[[#V2:]] = alloca ptr, i64 1, align 8
+// LLVM: store ptr %[[#V0]], ptr %[[#V2]], align 8
+// LLVM: %[[#V3:]] = load ptr, ptr %[[#V2]], align 8
+// LLVM: ret void
+void pass_cat(CAT a) {}

clang/test/CIR/CallConvLowering/AArch64/ptr-fields.c

@@ -0,0 +1,49 @@
+// RUN: %clang_cc1 -triple aarch64-unknown-linux-gnu  -fclangir -fclangir-call-conv-lowering -emit-cir-flat -mmlir --mlir-print-ir-after=cir-call-conv-lowering %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s -check-prefix=CIR
+// RUN: %clang_cc1 -triple aarch64-unknown-linux-gnu  -fclangir -emit-llvm %s -o %t.ll -fclangir-call-conv-lowering
+// RUN: FileCheck --input-file=%t.ll %s -check-prefix=LLVM
+
+typedef int (*myfptr)(int);
+
+typedef struct {
+  myfptr f;
+} A;
+
+int foo(int x) { return x; }
+
+// CIR: cir.func @passA(%arg0: !u64i
+// CIR: %[[#V0:]] = cir.alloca !ty_A, !cir.ptr<!ty_A>, [""] {alignment = 4 : i64}
+// CIR: %[[#V1:]] = cir.cast(bitcast, %[[#V0]] : !cir.ptr<!ty_A>), !cir.ptr<!u64i>
+// CIR: cir.store %arg0, %[[#V1]] : !u64i, !cir.ptr<!u64i>
+// CIR: %[[#V2:]] = cir.get_global @foo : !cir.ptr<!cir.func<!s32i (!s32i)>>
+// CIR: %[[#V3:]] = cir.get_member %[[#V0]][0] {name = "f"} : !cir.ptr<!ty_A> -> !cir.ptr<!cir.ptr<!cir.func<!s32i (!s32i)>>>
+// CIR: cir.store %[[#V2]], %[[#V3]] : !cir.ptr<!cir.func<!s32i (!s32i)>>, !cir.ptr<!cir.ptr<!cir.func<!s32i (!s32i)>>>
+// CIR: cir.return
+
+// LLVM: void @passA(i64 %[[#V0:]])
+// LLVM: %[[#V2:]] = alloca %struct.A, i64 1, align 4
+// LLVM: store i64 %[[#V0]], ptr %[[#V2]], align 8
+// LLVM: %[[#V3:]] = getelementptr %struct.A, ptr %[[#V2]], i32 0, i32 0
+// LLVM: store ptr @foo, ptr %[[#V3]], align 8
+// LLVM: ret void
+void passA(A a) { a.f = foo; }
+
+typedef struct {
+  int a;
+} S_1;
+
+typedef struct {
+  S_1* s;
+} S_2;
+
+// CIR: cir.func @passB(%arg0: !u64i
+// CIR: %[[#V0:]]  = cir.alloca !ty_S_2_, !cir.ptr<!ty_S_2_>, [""] {alignment = 4 : i64}
+// CIR: %[[#V1:]]  = cir.cast(bitcast, %[[#V0]]  : !cir.ptr<!ty_S_2_>), !cir.ptr<!u64i>
+// CIR: cir.store %arg0, %[[#V1]]  : !u64i, !cir.ptr<!u64i>
+// CIR: cir.return
+
+// LLVM: void @passB(i64 %[[#V0:]])
+// LLVM: %[[#V2:]]  = alloca %struct.S_2, i64 1, align 4
+// LLVM: store i64 %[[#V0]], ptr %[[#V2]], align 8
+// LLVM: ret void
+void passB(S_2 s) {}