// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
// RUN: %clang_cc1 -fenable-matrix %s -emit-llvm -triple x86_64-unknown-linux -disable-llvm-passes -o - -std=c++11 | FileCheck %s

using i8x3 = _BitInt(8) __attribute__((ext_vector_type(3)));
using i8x3x3 = _BitInt(8) __attribute__((matrix_type(3, 3)));
using i32x3 = _BitInt(32) __attribute__((ext_vector_type(3)));
using i32x3x3 = _BitInt(32) __attribute__((matrix_type(3, 3)));
using i512x3 = _BitInt(512) __attribute__((ext_vector_type(3)));
using i512x3x3 = _BitInt(512) __attribute__((matrix_type(3, 3)));

// CHECK-LABEL: define dso_local i32 @_Z2v1Dv3_DB8_(
// CHECK-SAME: i32 [[A_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT:  [[ENTRY:.*:]]
// CHECK-NEXT:    [[RETVAL:%.*]] = alloca <3 x i8>, align 4
// CHECK-NEXT:    [[A:%.*]] = alloca <3 x i8>, align 4
// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca <3 x i8>, align 4
// CHECK-NEXT:    store i32 [[A_COERCE]], ptr [[A]], align 4
// CHECK-NEXT:    [[LOADVECN:%.*]] = load <4 x i8>, ptr [[A]], align 4
// CHECK-NEXT:    [[A1:%.*]] = shufflevector <4 x i8> [[LOADVECN]], <4 x i8> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[EXTRACTVEC:%.*]] = shufflevector <3 x i8> [[A1]], <3 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 poison>
// CHECK-NEXT:    store <4 x i8> [[EXTRACTVEC]], ptr [[A_ADDR]], align 4
// CHECK-NEXT:    [[LOADVECN2:%.*]] = load <4 x i8>, ptr [[A_ADDR]], align 4
// CHECK-NEXT:    [[EXTRACTVEC3:%.*]] = shufflevector <4 x i8> [[LOADVECN2]], <4 x i8> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[LOADVECN4:%.*]] = load <4 x i8>, ptr [[A_ADDR]], align 4
// CHECK-NEXT:    [[EXTRACTVEC5:%.*]] = shufflevector <4 x i8> [[LOADVECN4]], <4 x i8> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[ADD:%.*]] = add <3 x i8> [[EXTRACTVEC3]], [[EXTRACTVEC5]]
// CHECK-NEXT:    store <3 x i8> [[ADD]], ptr [[RETVAL]], align 4
// CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK-NEXT:    ret i32 [[TMP0]]
//
i8x3 v1(i8x3 a) {
  return a + a;
}

// CHECK-LABEL: define dso_local noundef <3 x i32> @_Z2v2Dv3_DB32_(
// CHECK-SAME: <3 x i32> noundef [[A:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK-NEXT:  [[ENTRY:.*:]]
// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca <3 x i32>, align 16
// CHECK-NEXT:    [[EXTRACTVEC:%.*]] = shufflevector <3 x i32> [[A]], <3 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 poison>
// CHECK-NEXT:    store <4 x i32> [[EXTRACTVEC]], ptr [[A_ADDR]], align 16
// CHECK-NEXT:    [[LOADVECN:%.*]] = load <4 x i32>, ptr [[A_ADDR]], align 16
// CHECK-NEXT:    [[EXTRACTVEC1:%.*]] = shufflevector <4 x i32> [[LOADVECN]], <4 x i32> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[LOADVECN2:%.*]] = load <4 x i32>, ptr [[A_ADDR]], align 16
// CHECK-NEXT:    [[EXTRACTVEC3:%.*]] = shufflevector <4 x i32> [[LOADVECN2]], <4 x i32> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[ADD:%.*]] = add <3 x i32> [[EXTRACTVEC1]], [[EXTRACTVEC3]]
// CHECK-NEXT:    ret <3 x i32> [[ADD]]
//
i32x3 v2(i32x3 a) {
  return a + a;
}

// CHECK-LABEL: define dso_local noundef <3 x i512> @_Z2v3Dv3_DB512_(
// CHECK-SAME: ptr noundef byval(<3 x i512>) align 256 [[TMP0:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK-NEXT:  [[ENTRY:.*:]]
// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca <3 x i512>, align 256
// CHECK-NEXT:    [[LOADVECN:%.*]] = load <4 x i512>, ptr [[TMP0]], align 256
// CHECK-NEXT:    [[A:%.*]] = shufflevector <4 x i512> [[LOADVECN]], <4 x i512> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[EXTRACTVEC:%.*]] = shufflevector <3 x i512> [[A]], <3 x i512> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 poison>
// CHECK-NEXT:    store <4 x i512> [[EXTRACTVEC]], ptr [[A_ADDR]], align 256
// CHECK-NEXT:    [[LOADVECN1:%.*]] = load <4 x i512>, ptr [[A_ADDR]], align 256
// CHECK-NEXT:    [[EXTRACTVEC2:%.*]] = shufflevector <4 x i512> [[LOADVECN1]], <4 x i512> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[LOADVECN3:%.*]] = load <4 x i512>, ptr [[A_ADDR]], align 256
// CHECK-NEXT:    [[EXTRACTVEC4:%.*]] = shufflevector <4 x i512> [[LOADVECN3]], <4 x i512> poison, <3 x i32> <i32 0, i32 1, i32 2>
// CHECK-NEXT:    [[ADD:%.*]] = add <3 x i512> [[EXTRACTVEC2]], [[EXTRACTVEC4]]
// CHECK-NEXT:    ret <3 x i512> [[ADD]]
//
i512x3 v3(i512x3 a) {
  return a + a;
}

// CHECK-LABEL: define dso_local noundef <9 x i8> @_Z2m1u11matrix_typeILm3ELm3EDB8_E(
// CHECK-SAME: <9 x i8> noundef [[A:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK-NEXT:  [[ENTRY:.*:]]
// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca [9 x i8], align 1
// CHECK-NEXT:    store <9 x i8> [[A]], ptr [[A_ADDR]], align 1
// CHECK-NEXT:    [[TMP0:%.*]] = load <9 x i8>, ptr [[A_ADDR]], align 1
// CHECK-NEXT:    [[TMP1:%.*]] = load <9 x i8>, ptr [[A_ADDR]], align 1
// CHECK-NEXT:    [[TMP2:%.*]] = add <9 x i8> [[TMP0]], [[TMP1]]
// CHECK-NEXT:    ret <9 x i8> [[TMP2]]
//
i8x3x3 m1(i8x3x3 a) {
  return a + a;
}

// CHECK-LABEL: define dso_local noundef <9 x i32> @_Z2m2u11matrix_typeILm3ELm3EDB32_E(
// CHECK-SAME: <9 x i32> noundef [[A:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK-NEXT:  [[ENTRY:.*:]]
// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca [9 x i32], align 4
// CHECK-NEXT:    store <9 x i32> [[A]], ptr [[A_ADDR]], align 4
// CHECK-NEXT:    [[TMP0:%.*]] = load <9 x i32>, ptr [[A_ADDR]], align 4
// CHECK-NEXT:    [[TMP1:%.*]] = load <9 x i32>, ptr [[A_ADDR]], align 4
// CHECK-NEXT:    [[TMP2:%.*]] = add <9 x i32> [[TMP0]], [[TMP1]]
// CHECK-NEXT:    ret <9 x i32> [[TMP2]]
//
i32x3x3 m2(i32x3x3 a) {
  return a + a;
}

// CHECK-LABEL: define dso_local noundef <9 x i512> @_Z2m3u11matrix_typeILm3ELm3EDB512_E(
// CHECK-SAME: <9 x i512> noundef [[A:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK-NEXT:  [[ENTRY:.*:]]
// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca [9 x i512], align 8
// CHECK-NEXT:    store <9 x i512> [[A]], ptr [[A_ADDR]], align 8
// CHECK-NEXT:    [[TMP0:%.*]] = load <9 x i512>, ptr [[A_ADDR]], align 8
// CHECK-NEXT:    [[TMP1:%.*]] = load <9 x i512>, ptr [[A_ADDR]], align 8
// CHECK-NEXT:    [[TMP2:%.*]] = add <9 x i512> [[TMP0]], [[TMP1]]
// CHECK-NEXT:    ret <9 x i512> [[TMP2]]
//
i512x3x3 m3(i512x3x3 a) {
  return a + a;
}