1 // RUN: %clang_cc1 -triple arm64-apple-ios7 -target-feature +neon -ffreestanding -S -o - -emit-llvm %s | opt -S -mem2reg | FileCheck %s
2
3 // Test ARM64 SIMD copy vector element to vector element: vcopyq_lane*
4
5 #include <arm_neon.h>
6
7 // CHECK-LABEL: define <16 x i8> @test_vcopyq_laneq_s8(<16 x i8> %a1, <16 x i8> %a2) #0 {
8 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <16 x i8> %a2, i32 13
9 // CHECK: [[VSET_LANE:%.*]] = insertelement <16 x i8> %a1, i8 [[VGETQ_LANE]], i32 3
10 // CHECK: ret <16 x i8> [[VSET_LANE]]
test_vcopyq_laneq_s8(int8x16_t a1,int8x16_t a2)11 int8x16_t test_vcopyq_laneq_s8(int8x16_t a1, int8x16_t a2) {
12 return vcopyq_laneq_s8(a1, (int64_t) 3, a2, (int64_t) 13);
13 }
14
15 // CHECK-LABEL: define <16 x i8> @test_vcopyq_laneq_u8(<16 x i8> %a1, <16 x i8> %a2) #0 {
16 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <16 x i8> %a2, i32 13
17 // CHECK: [[VSET_LANE:%.*]] = insertelement <16 x i8> %a1, i8 [[VGETQ_LANE]], i32 3
18 // CHECK: ret <16 x i8> [[VSET_LANE]]
test_vcopyq_laneq_u8(uint8x16_t a1,uint8x16_t a2)19 uint8x16_t test_vcopyq_laneq_u8(uint8x16_t a1, uint8x16_t a2) {
20 return vcopyq_laneq_u8(a1, (int64_t) 3, a2, (int64_t) 13);
21
22 }
23
24 // CHECK-LABEL: define <8 x i16> @test_vcopyq_laneq_s16(<8 x i16> %a1, <8 x i16> %a2) #0 {
25 // CHECK: [[TMP0:%.*]] = bitcast <8 x i16> %a2 to <16 x i8>
26 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
27 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP1]], i32 7
28 // CHECK: [[TMP2:%.*]] = bitcast <8 x i16> %a1 to <16 x i8>
29 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <8 x i16>
30 // CHECK: [[VSET_LANE:%.*]] = insertelement <8 x i16> [[TMP3]], i16 [[VGETQ_LANE]], i32 3
31 // CHECK: ret <8 x i16> [[VSET_LANE]]
test_vcopyq_laneq_s16(int16x8_t a1,int16x8_t a2)32 int16x8_t test_vcopyq_laneq_s16(int16x8_t a1, int16x8_t a2) {
33 return vcopyq_laneq_s16(a1, (int64_t) 3, a2, (int64_t) 7);
34
35 }
36
37 // CHECK-LABEL: define <8 x i16> @test_vcopyq_laneq_u16(<8 x i16> %a1, <8 x i16> %a2) #0 {
38 // CHECK: [[TMP0:%.*]] = bitcast <8 x i16> %a2 to <16 x i8>
39 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
40 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[TMP1]], i32 7
41 // CHECK: [[TMP2:%.*]] = bitcast <8 x i16> %a1 to <16 x i8>
42 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <8 x i16>
43 // CHECK: [[VSET_LANE:%.*]] = insertelement <8 x i16> [[TMP3]], i16 [[VGETQ_LANE]], i32 3
44 // CHECK: ret <8 x i16> [[VSET_LANE]]
test_vcopyq_laneq_u16(uint16x8_t a1,uint16x8_t a2)45 uint16x8_t test_vcopyq_laneq_u16(uint16x8_t a1, uint16x8_t a2) {
46 return vcopyq_laneq_u16(a1, (int64_t) 3, a2, (int64_t) 7);
47
48 }
49
50 // CHECK-LABEL: define <4 x i32> @test_vcopyq_laneq_s32(<4 x i32> %a1, <4 x i32> %a2) #0 {
51 // CHECK: [[TMP0:%.*]] = bitcast <4 x i32> %a2 to <16 x i8>
52 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
53 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
54 // CHECK: [[TMP2:%.*]] = bitcast <4 x i32> %a1 to <16 x i8>
55 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
56 // CHECK: [[VSET_LANE:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[VGETQ_LANE]], i32 3
57 // CHECK: ret <4 x i32> [[VSET_LANE]]
test_vcopyq_laneq_s32(int32x4_t a1,int32x4_t a2)58 int32x4_t test_vcopyq_laneq_s32(int32x4_t a1, int32x4_t a2) {
59 return vcopyq_laneq_s32(a1, (int64_t) 3, a2, (int64_t) 3);
60 }
61
62 // CHECK-LABEL: define <4 x i32> @test_vcopyq_laneq_u32(<4 x i32> %a1, <4 x i32> %a2) #0 {
63 // CHECK: [[TMP0:%.*]] = bitcast <4 x i32> %a2 to <16 x i8>
64 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
65 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
66 // CHECK: [[TMP2:%.*]] = bitcast <4 x i32> %a1 to <16 x i8>
67 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
68 // CHECK: [[VSET_LANE:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[VGETQ_LANE]], i32 3
69 // CHECK: ret <4 x i32> [[VSET_LANE]]
test_vcopyq_laneq_u32(uint32x4_t a1,uint32x4_t a2)70 uint32x4_t test_vcopyq_laneq_u32(uint32x4_t a1, uint32x4_t a2) {
71 return vcopyq_laneq_u32(a1, (int64_t) 3, a2, (int64_t) 3);
72 }
73
74 // CHECK-LABEL: define <2 x i64> @test_vcopyq_laneq_s64(<2 x i64> %a1, <2 x i64> %a2) #0 {
75 // CHECK: [[TMP0:%.*]] = bitcast <2 x i64> %a2 to <16 x i8>
76 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
77 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <2 x i64> [[TMP1]], i32 1
78 // CHECK: [[TMP2:%.*]] = bitcast <2 x i64> %a1 to <16 x i8>
79 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x i64>
80 // CHECK: [[VSET_LANE:%.*]] = insertelement <2 x i64> [[TMP3]], i64 [[VGETQ_LANE]], i32 0
81 // CHECK: ret <2 x i64> [[VSET_LANE]]
test_vcopyq_laneq_s64(int64x2_t a1,int64x2_t a2)82 int64x2_t test_vcopyq_laneq_s64(int64x2_t a1, int64x2_t a2) {
83 return vcopyq_laneq_s64(a1, (int64_t) 0, a2, (int64_t) 1);
84 }
85
86 // CHECK-LABEL: define <2 x i64> @test_vcopyq_laneq_u64(<2 x i64> %a1, <2 x i64> %a2) #0 {
87 // CHECK: [[TMP0:%.*]] = bitcast <2 x i64> %a2 to <16 x i8>
88 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
89 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <2 x i64> [[TMP1]], i32 1
90 // CHECK: [[TMP2:%.*]] = bitcast <2 x i64> %a1 to <16 x i8>
91 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x i64>
92 // CHECK: [[VSET_LANE:%.*]] = insertelement <2 x i64> [[TMP3]], i64 [[VGETQ_LANE]], i32 0
93 // CHECK: ret <2 x i64> [[VSET_LANE]]
test_vcopyq_laneq_u64(uint64x2_t a1,uint64x2_t a2)94 uint64x2_t test_vcopyq_laneq_u64(uint64x2_t a1, uint64x2_t a2) {
95 return vcopyq_laneq_u64(a1, (int64_t) 0, a2, (int64_t) 1);
96 }
97
98 // CHECK-LABEL: define <4 x float> @test_vcopyq_laneq_f32(<4 x float> %a1, <4 x float> %a2) #0 {
99 // CHECK: [[TMP0:%.*]] = bitcast <4 x float> %a2 to <16 x i8>
100 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
101 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <4 x float> [[TMP1]], i32 3
102 // CHECK: [[TMP2:%.*]] = bitcast <4 x float> %a1 to <16 x i8>
103 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x float>
104 // CHECK: [[VSET_LANE:%.*]] = insertelement <4 x float> [[TMP3]], float [[VGETQ_LANE]], i32 0
105 // CHECK: ret <4 x float> [[VSET_LANE]]
test_vcopyq_laneq_f32(float32x4_t a1,float32x4_t a2)106 float32x4_t test_vcopyq_laneq_f32(float32x4_t a1, float32x4_t a2) {
107 return vcopyq_laneq_f32(a1, 0, a2, 3);
108 }
109
110 // CHECK-LABEL: define <2 x double> @test_vcopyq_laneq_f64(<2 x double> %a1, <2 x double> %a2) #0 {
111 // CHECK: [[TMP0:%.*]] = bitcast <2 x double> %a2 to <16 x i8>
112 // CHECK: [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
113 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
114 // CHECK: [[TMP2:%.*]] = bitcast <2 x double> %a1 to <16 x i8>
115 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
116 // CHECK: [[VSET_LANE:%.*]] = insertelement <2 x double> [[TMP3]], double [[VGETQ_LANE]], i32 0
117 // CHECK: ret <2 x double> [[VSET_LANE]]
test_vcopyq_laneq_f64(float64x2_t a1,float64x2_t a2)118 float64x2_t test_vcopyq_laneq_f64(float64x2_t a1, float64x2_t a2) {
119 return vcopyq_laneq_f64(a1, 0, a2, 1);
120 }
121
122