1 // RUN: %clang_cc1 -triple arm64-apple-ios7 -target-feature +neon -ffreestanding -S -o - -disable-O0-optnone -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: [[VGETQ_LANE:%.*]] = extractelement <8 x i16> %a2, i32 7
26 // CHECK: [[VSET_LANE:%.*]] = insertelement <8 x i16> %a1, i16 [[VGETQ_LANE]], i32 3
27 // CHECK: ret <8 x i16> [[VSET_LANE]]
test_vcopyq_laneq_s16(int16x8_t a1,int16x8_t a2)28 int16x8_t test_vcopyq_laneq_s16(int16x8_t a1, int16x8_t a2) {
29 return vcopyq_laneq_s16(a1, (int64_t) 3, a2, (int64_t) 7);
30
31 }
32
33 // CHECK-LABEL: define <8 x i16> @test_vcopyq_laneq_u16(<8 x i16> %a1, <8 x i16> %a2) #0 {
34 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <8 x i16> %a2, i32 7
35 // CHECK: [[VSET_LANE:%.*]] = insertelement <8 x i16> %a1, i16 [[VGETQ_LANE]], i32 3
36 // CHECK: ret <8 x i16> [[VSET_LANE]]
test_vcopyq_laneq_u16(uint16x8_t a1,uint16x8_t a2)37 uint16x8_t test_vcopyq_laneq_u16(uint16x8_t a1, uint16x8_t a2) {
38 return vcopyq_laneq_u16(a1, (int64_t) 3, a2, (int64_t) 7);
39
40 }
41
42 // CHECK-LABEL: define <4 x i32> @test_vcopyq_laneq_s32(<4 x i32> %a1, <4 x i32> %a2) #0 {
43 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <4 x i32> %a2, i32 3
44 // CHECK: [[VSET_LANE:%.*]] = insertelement <4 x i32> %a1, i32 [[VGETQ_LANE]], i32 3
45 // CHECK: ret <4 x i32> [[VSET_LANE]]
test_vcopyq_laneq_s32(int32x4_t a1,int32x4_t a2)46 int32x4_t test_vcopyq_laneq_s32(int32x4_t a1, int32x4_t a2) {
47 return vcopyq_laneq_s32(a1, (int64_t) 3, a2, (int64_t) 3);
48 }
49
50 // CHECK-LABEL: define <4 x i32> @test_vcopyq_laneq_u32(<4 x i32> %a1, <4 x i32> %a2) #0 {
51 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <4 x i32> %a2, i32 3
52 // CHECK: [[VSET_LANE:%.*]] = insertelement <4 x i32> %a1, i32 [[VGETQ_LANE]], i32 3
53 // CHECK: ret <4 x i32> [[VSET_LANE]]
test_vcopyq_laneq_u32(uint32x4_t a1,uint32x4_t a2)54 uint32x4_t test_vcopyq_laneq_u32(uint32x4_t a1, uint32x4_t a2) {
55 return vcopyq_laneq_u32(a1, (int64_t) 3, a2, (int64_t) 3);
56 }
57
58 // CHECK-LABEL: define <2 x i64> @test_vcopyq_laneq_s64(<2 x i64> %a1, <2 x i64> %a2) #0 {
59 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <2 x i64> %a2, i32 1
60 // CHECK: [[VSET_LANE:%.*]] = insertelement <2 x i64> %a1, i64 [[VGETQ_LANE]], i32 0
61 // CHECK: ret <2 x i64> [[VSET_LANE]]
test_vcopyq_laneq_s64(int64x2_t a1,int64x2_t a2)62 int64x2_t test_vcopyq_laneq_s64(int64x2_t a1, int64x2_t a2) {
63 return vcopyq_laneq_s64(a1, (int64_t) 0, a2, (int64_t) 1);
64 }
65
66 // CHECK-LABEL: define <2 x i64> @test_vcopyq_laneq_u64(<2 x i64> %a1, <2 x i64> %a2) #0 {
67 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <2 x i64> %a2, i32 1
68 // CHECK: [[VSET_LANE:%.*]] = insertelement <2 x i64> %a1, i64 [[VGETQ_LANE]], i32 0
69 // CHECK: ret <2 x i64> [[VSET_LANE]]
test_vcopyq_laneq_u64(uint64x2_t a1,uint64x2_t a2)70 uint64x2_t test_vcopyq_laneq_u64(uint64x2_t a1, uint64x2_t a2) {
71 return vcopyq_laneq_u64(a1, (int64_t) 0, a2, (int64_t) 1);
72 }
73
74 // CHECK-LABEL: define <4 x float> @test_vcopyq_laneq_f32(<4 x float> %a1, <4 x float> %a2) #0 {
75 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <4 x float> %a2, i32 3
76 // CHECK: [[VSET_LANE:%.*]] = insertelement <4 x float> %a1, float [[VGETQ_LANE]], i32 0
77 // CHECK: ret <4 x float> [[VSET_LANE]]
test_vcopyq_laneq_f32(float32x4_t a1,float32x4_t a2)78 float32x4_t test_vcopyq_laneq_f32(float32x4_t a1, float32x4_t a2) {
79 return vcopyq_laneq_f32(a1, 0, a2, 3);
80 }
81
82 // CHECK-LABEL: define <2 x double> @test_vcopyq_laneq_f64(<2 x double> %a1, <2 x double> %a2) #0 {
83 // CHECK: [[VGETQ_LANE:%.*]] = extractelement <2 x double> %a2, i32 1
84 // CHECK: [[VSET_LANE:%.*]] = insertelement <2 x double> %a1, double [[VGETQ_LANE]], i32 0
85 // CHECK: ret <2 x double> [[VSET_LANE]]
test_vcopyq_laneq_f64(float64x2_t a1,float64x2_t a2)86 float64x2_t test_vcopyq_laneq_f64(float64x2_t a1, float64x2_t a2) {
87 return vcopyq_laneq_f64(a1, 0, a2, 1);
88 }
89
90