// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// RUN: %clang_cc1 -triple s390x-linux-gnu -O1 -emit-llvm %s -o - | FileCheck %s
//
// Test GNU atomic builtins for __int128 (with default alignment of 8 bytes
// only), resulting in libcalls.

#include <stdatomic.h>
#include <stdint.h>

__int128 Ptr;
__int128 Ret;
__int128 Val;
__int128 Exp;
__int128 Des;

// TODO: This test and several more below have the unnecessary use of an alloca
// remaining. This is due to 369c9b7, which changes the behavior of the MemCpyOpt
// pass. It seems that a 'writable' attribute should now be added to the argument
// in order for this optimization to proceed.

// CHECK-LABEL: @f1(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    tail call void @__atomic_load(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull [[AGG_RESULT:%.*]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f1() {
  return __atomic_load_n(&Ptr, memory_order_seq_cst);
}

// CHECK-LABEL: @f2(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    tail call void @__atomic_load(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull @Ret, i32 noundef signext 5)
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Ret, align 8, !tbaa [[TBAA2:![0-9]+]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f2() {
  __atomic_load(&Ptr, &Ret, memory_order_seq_cst);
  return Ret;
}

// CHECK-LABEL: @f3(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[DOTATOMICTMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_store(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull [[DOTATOMICTMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
void f3() {
  __atomic_store_n(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f4(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    tail call void @__atomic_store(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull @Val, i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
void f4() {
  __atomic_store(&Ptr, &Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f5(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[DOTATOMICTMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_exchange(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull [[DOTATOMICTMP]], ptr noundef nonnull [[AGG_RESULT:%.*]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f5() {
  return __atomic_exchange_n(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f6(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    tail call void @__atomic_exchange(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull @Val, ptr noundef nonnull @Ret, i32 noundef signext 5)
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Ret, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f6() {
  __atomic_exchange(&Ptr, &Val, &Ret, memory_order_seq_cst);
  return Ret;
}

// CHECK-LABEL: @f7(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[DOTATOMICTMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Des, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[DOTATOMICTMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[CALL:%.*]] = call zeroext i1 @__atomic_compare_exchange(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull @Exp, ptr noundef nonnull [[DOTATOMICTMP]], i32 noundef signext 5, i32 noundef signext 5)
// CHECK-NEXT:    ret i1 [[CALL]]
//
_Bool f7() {
  return __atomic_compare_exchange_n(&Ptr, &Exp, Des, 0,
                                     memory_order_seq_cst, memory_order_seq_cst);
}

// CHECK-LABEL: @f8(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[CALL:%.*]] = tail call zeroext i1 @__atomic_compare_exchange(i64 noundef 16, ptr noundef nonnull @Ptr, ptr noundef nonnull @Exp, ptr noundef nonnull @Des, i32 noundef signext 5, i32 noundef signext 5)
// CHECK-NEXT:    ret i1 [[CALL]]
//
_Bool f8() {
  return __atomic_compare_exchange(&Ptr, &Exp, &Des, 0,
                                   memory_order_seq_cst, memory_order_seq_cst);
}

// CHECK-LABEL: @f9(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_add_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[TMP]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    [[TMP1:%.*]] = load i128, ptr [[TMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[TMP2:%.*]] = add i128 [[TMP1]], [[TMP0]]
// CHECK-NEXT:    store i128 [[TMP2]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f9() {
  return __atomic_add_fetch(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f10(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_sub_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[TMP]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    [[TMP1:%.*]] = load i128, ptr [[TMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[TMP2:%.*]] = sub i128 [[TMP1]], [[TMP0]]
// CHECK-NEXT:    store i128 [[TMP2]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f10() {
  return __atomic_sub_fetch(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f11(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_and_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[TMP]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    [[TMP1:%.*]] = load i128, ptr [[TMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[TMP2:%.*]] = and i128 [[TMP1]], [[TMP0]]
// CHECK-NEXT:    store i128 [[TMP2]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f11() {
  return __atomic_and_fetch(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f12(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_xor_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[TMP]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    [[TMP1:%.*]] = load i128, ptr [[TMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[TMP2:%.*]] = xor i128 [[TMP1]], [[TMP0]]
// CHECK-NEXT:    store i128 [[TMP2]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f12() {
  return __atomic_xor_fetch(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f13(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_or_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[TMP]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    [[TMP1:%.*]] = load i128, ptr [[TMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[TMP2:%.*]] = or i128 [[TMP1]], [[TMP0]]
// CHECK-NEXT:    store i128 [[TMP2]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f13() {
  return __atomic_or_fetch(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f14(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[TMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_nand_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[TMP]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    [[TMP1:%.*]] = load i128, ptr [[TMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    [[TMP2:%.*]] = and i128 [[TMP1]], [[TMP0]]
// CHECK-NEXT:    [[TMP3:%.*]] = xor i128 [[TMP2]], -1
// CHECK-NEXT:    store i128 [[TMP3]], ptr [[AGG_RESULT:%.*]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    ret void
//
__int128 f14() {
  return __atomic_nand_fetch(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f15(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_add_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[AGG_RESULT:%.*]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f15() {
  return __atomic_fetch_add(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f16(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_sub_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[AGG_RESULT:%.*]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f16() {
  return __atomic_fetch_sub(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f17(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_and_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[AGG_RESULT:%.*]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f17() {
  return __atomic_fetch_and(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f18(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_xor_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[AGG_RESULT:%.*]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f18() {
  return __atomic_fetch_xor(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f19(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_or_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[AGG_RESULT:%.*]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f19() {
  return __atomic_fetch_or(&Ptr, Val, memory_order_seq_cst);
}

// CHECK-LABEL: @f20(
// CHECK-NEXT:  entry:
// CHECK-NEXT:    [[INDIRECT_ARG_TEMP:%.*]] = alloca i128, align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load i128, ptr @Val, align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    store i128 [[TMP0]], ptr [[INDIRECT_ARG_TEMP]], align 8, !tbaa [[TBAA2]]
// CHECK-NEXT:    call void @__atomic_fetch_nand_16(ptr dead_on_unwind nonnull writable sret(i128) align 8 [[AGG_RESULT:%.*]], ptr noundef nonnull @Ptr, ptr noundef nonnull [[INDIRECT_ARG_TEMP]], i32 noundef signext 5)
// CHECK-NEXT:    ret void
//
__int128 f20() {
  return __atomic_fetch_nand(&Ptr, Val, memory_order_seq_cst);
}