1 // REQUIRES: aarch64-registered-target
2 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
3 // RUN: -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s
4
5 // Test new aarch64 intrinsics and types
6
7 #include <arm_neon.h>
8
9 // CHECK-LABEL: define <8 x i8> @test_vext_s8(<8 x i8> %a, <8 x i8> %b) #0 {
10 // CHECK: [[VEXT:%.*]] = shufflevector <8 x i8> %a, <8 x i8> %b, <8 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9>
11 // CHECK: ret <8 x i8> [[VEXT]]
test_vext_s8(int8x8_t a,int8x8_t b)12 int8x8_t test_vext_s8(int8x8_t a, int8x8_t b) {
13 return vext_s8(a, b, 2);
14 }
15
16 // CHECK-LABEL: define <4 x i16> @test_vext_s16(<4 x i16> %a, <4 x i16> %b) #0 {
17 // CHECK: [[TMP0:%.*]] = bitcast <4 x i16> %a to <8 x i8>
18 // CHECK: [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
19 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
20 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
21 // CHECK: [[VEXT:%.*]] = shufflevector <4 x i16> [[TMP2]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
22 // CHECK: ret <4 x i16> [[VEXT]]
test_vext_s16(int16x4_t a,int16x4_t b)23 int16x4_t test_vext_s16(int16x4_t a, int16x4_t b) {
24 return vext_s16(a, b, 3);
25 }
26
27 // CHECK-LABEL: define <2 x i32> @test_vext_s32(<2 x i32> %a, <2 x i32> %b) #0 {
28 // CHECK: [[TMP0:%.*]] = bitcast <2 x i32> %a to <8 x i8>
29 // CHECK: [[TMP1:%.*]] = bitcast <2 x i32> %b to <8 x i8>
30 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
31 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
32 // CHECK: [[VEXT:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 1, i32 2>
33 // CHECK: ret <2 x i32> [[VEXT]]
test_vext_s32(int32x2_t a,int32x2_t b)34 int32x2_t test_vext_s32(int32x2_t a, int32x2_t b) {
35 return vext_s32(a, b, 1);
36 }
37
38 // CHECK-LABEL: define <1 x i64> @test_vext_s64(<1 x i64> %a, <1 x i64> %b) #0 {
39 // CHECK: [[TMP0:%.*]] = bitcast <1 x i64> %a to <8 x i8>
40 // CHECK: [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
41 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
42 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
43 // CHECK: [[VEXT:%.*]] = shufflevector <1 x i64> [[TMP2]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
44 // CHECK: ret <1 x i64> [[VEXT]]
test_vext_s64(int64x1_t a,int64x1_t b)45 int64x1_t test_vext_s64(int64x1_t a, int64x1_t b) {
46 return vext_s64(a, b, 0);
47 }
48
49 // CHECK-LABEL: define <16 x i8> @test_vextq_s8(<16 x i8> %a, <16 x i8> %b) #0 {
50 // CHECK: [[VEXT:%.*]] = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17>
51 // CHECK: ret <16 x i8> [[VEXT]]
test_vextq_s8(int8x16_t a,int8x16_t b)52 int8x16_t test_vextq_s8(int8x16_t a, int8x16_t b) {
53 return vextq_s8(a, b, 2);
54 }
55
56 // CHECK-LABEL: define <8 x i16> @test_vextq_s16(<8 x i16> %a, <8 x i16> %b) #0 {
57 // CHECK: [[TMP0:%.*]] = bitcast <8 x i16> %a to <16 x i8>
58 // CHECK: [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
59 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
60 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
61 // CHECK: [[VEXT:%.*]] = shufflevector <8 x i16> [[TMP2]], <8 x i16> [[TMP3]], <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
62 // CHECK: ret <8 x i16> [[VEXT]]
test_vextq_s16(int16x8_t a,int16x8_t b)63 int16x8_t test_vextq_s16(int16x8_t a, int16x8_t b) {
64 return vextq_s16(a, b, 3);
65 }
66
67 // CHECK-LABEL: define <4 x i32> @test_vextq_s32(<4 x i32> %a, <4 x i32> %b) #0 {
68 // CHECK: [[TMP0:%.*]] = bitcast <4 x i32> %a to <16 x i8>
69 // CHECK: [[TMP1:%.*]] = bitcast <4 x i32> %b to <16 x i8>
70 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
71 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
72 // CHECK: [[VEXT:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> <i32 1, i32 2, i32 3, i32 4>
73 // CHECK: ret <4 x i32> [[VEXT]]
test_vextq_s32(int32x4_t a,int32x4_t b)74 int32x4_t test_vextq_s32(int32x4_t a, int32x4_t b) {
75 return vextq_s32(a, b, 1);
76 }
77
78 // CHECK-LABEL: define <2 x i64> @test_vextq_s64(<2 x i64> %a, <2 x i64> %b) #0 {
79 // CHECK: [[TMP0:%.*]] = bitcast <2 x i64> %a to <16 x i8>
80 // CHECK: [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
81 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
82 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
83 // CHECK: [[VEXT:%.*]] = shufflevector <2 x i64> [[TMP2]], <2 x i64> [[TMP3]], <2 x i32> <i32 1, i32 2>
84 // CHECK: ret <2 x i64> [[VEXT]]
test_vextq_s64(int64x2_t a,int64x2_t b)85 int64x2_t test_vextq_s64(int64x2_t a, int64x2_t b) {
86 return vextq_s64(a, b, 1);
87 }
88
89 // CHECK-LABEL: define <8 x i8> @test_vext_u8(<8 x i8> %a, <8 x i8> %b) #0 {
90 // CHECK: [[VEXT:%.*]] = shufflevector <8 x i8> %a, <8 x i8> %b, <8 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9>
91 // CHECK: ret <8 x i8> [[VEXT]]
test_vext_u8(uint8x8_t a,uint8x8_t b)92 uint8x8_t test_vext_u8(uint8x8_t a, uint8x8_t b) {
93 return vext_u8(a, b, 2);
94 }
95
96 // CHECK-LABEL: define <4 x i16> @test_vext_u16(<4 x i16> %a, <4 x i16> %b) #0 {
97 // CHECK: [[TMP0:%.*]] = bitcast <4 x i16> %a to <8 x i8>
98 // CHECK: [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
99 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
100 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
101 // CHECK: [[VEXT:%.*]] = shufflevector <4 x i16> [[TMP2]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
102 // CHECK: ret <4 x i16> [[VEXT]]
test_vext_u16(uint16x4_t a,uint16x4_t b)103 uint16x4_t test_vext_u16(uint16x4_t a, uint16x4_t b) {
104 return vext_u16(a, b, 3);
105 }
106
107 // CHECK-LABEL: define <2 x i32> @test_vext_u32(<2 x i32> %a, <2 x i32> %b) #0 {
108 // CHECK: [[TMP0:%.*]] = bitcast <2 x i32> %a to <8 x i8>
109 // CHECK: [[TMP1:%.*]] = bitcast <2 x i32> %b to <8 x i8>
110 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
111 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
112 // CHECK: [[VEXT:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 1, i32 2>
113 // CHECK: ret <2 x i32> [[VEXT]]
test_vext_u32(uint32x2_t a,uint32x2_t b)114 uint32x2_t test_vext_u32(uint32x2_t a, uint32x2_t b) {
115 return vext_u32(a, b, 1);
116 }
117
118 // CHECK-LABEL: define <1 x i64> @test_vext_u64(<1 x i64> %a, <1 x i64> %b) #0 {
119 // CHECK: [[TMP0:%.*]] = bitcast <1 x i64> %a to <8 x i8>
120 // CHECK: [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
121 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
122 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
123 // CHECK: [[VEXT:%.*]] = shufflevector <1 x i64> [[TMP2]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
124 // CHECK: ret <1 x i64> [[VEXT]]
test_vext_u64(uint64x1_t a,uint64x1_t b)125 uint64x1_t test_vext_u64(uint64x1_t a, uint64x1_t b) {
126 return vext_u64(a, b, 0);
127 }
128
129 // CHECK-LABEL: define <16 x i8> @test_vextq_u8(<16 x i8> %a, <16 x i8> %b) #0 {
130 // CHECK: [[VEXT:%.*]] = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17>
131 // CHECK: ret <16 x i8> [[VEXT]]
test_vextq_u8(uint8x16_t a,uint8x16_t b)132 uint8x16_t test_vextq_u8(uint8x16_t a, uint8x16_t b) {
133 return vextq_u8(a, b, 2);
134 }
135
136 // CHECK-LABEL: define <8 x i16> @test_vextq_u16(<8 x i16> %a, <8 x i16> %b) #0 {
137 // CHECK: [[TMP0:%.*]] = bitcast <8 x i16> %a to <16 x i8>
138 // CHECK: [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
139 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
140 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
141 // CHECK: [[VEXT:%.*]] = shufflevector <8 x i16> [[TMP2]], <8 x i16> [[TMP3]], <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
142 // CHECK: ret <8 x i16> [[VEXT]]
test_vextq_u16(uint16x8_t a,uint16x8_t b)143 uint16x8_t test_vextq_u16(uint16x8_t a, uint16x8_t b) {
144 return vextq_u16(a, b, 3);
145 }
146
147 // CHECK-LABEL: define <4 x i32> @test_vextq_u32(<4 x i32> %a, <4 x i32> %b) #0 {
148 // CHECK: [[TMP0:%.*]] = bitcast <4 x i32> %a to <16 x i8>
149 // CHECK: [[TMP1:%.*]] = bitcast <4 x i32> %b to <16 x i8>
150 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
151 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
152 // CHECK: [[VEXT:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> <i32 1, i32 2, i32 3, i32 4>
153 // CHECK: ret <4 x i32> [[VEXT]]
test_vextq_u32(uint32x4_t a,uint32x4_t b)154 uint32x4_t test_vextq_u32(uint32x4_t a, uint32x4_t b) {
155 return vextq_u32(a, b, 1);
156 }
157
158 // CHECK-LABEL: define <2 x i64> @test_vextq_u64(<2 x i64> %a, <2 x i64> %b) #0 {
159 // CHECK: [[TMP0:%.*]] = bitcast <2 x i64> %a to <16 x i8>
160 // CHECK: [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
161 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
162 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
163 // CHECK: [[VEXT:%.*]] = shufflevector <2 x i64> [[TMP2]], <2 x i64> [[TMP3]], <2 x i32> <i32 1, i32 2>
164 // CHECK: ret <2 x i64> [[VEXT]]
test_vextq_u64(uint64x2_t a,uint64x2_t b)165 uint64x2_t test_vextq_u64(uint64x2_t a, uint64x2_t b) {
166 return vextq_u64(a, b, 1);
167 }
168
169 // CHECK-LABEL: define <2 x float> @test_vext_f32(<2 x float> %a, <2 x float> %b) #0 {
170 // CHECK: [[TMP0:%.*]] = bitcast <2 x float> %a to <8 x i8>
171 // CHECK: [[TMP1:%.*]] = bitcast <2 x float> %b to <8 x i8>
172 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x float>
173 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
174 // CHECK: [[VEXT:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> [[TMP3]], <2 x i32> <i32 1, i32 2>
175 // CHECK: ret <2 x float> [[VEXT]]
test_vext_f32(float32x2_t a,float32x2_t b)176 float32x2_t test_vext_f32(float32x2_t a, float32x2_t b) {
177 return vext_f32(a, b, 1);
178 }
179
180 // CHECK-LABEL: define <1 x double> @test_vext_f64(<1 x double> %a, <1 x double> %b) #0 {
181 // CHECK: [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
182 // CHECK: [[TMP1:%.*]] = bitcast <1 x double> %b to <8 x i8>
183 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
184 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
185 // CHECK: [[VEXT:%.*]] = shufflevector <1 x double> [[TMP2]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
186 // CHECK: ret <1 x double> [[VEXT]]
test_vext_f64(float64x1_t a,float64x1_t b)187 float64x1_t test_vext_f64(float64x1_t a, float64x1_t b) {
188 return vext_f64(a, b, 0);
189 }
190
191 // CHECK-LABEL: define <4 x float> @test_vextq_f32(<4 x float> %a, <4 x float> %b) #0 {
192 // CHECK: [[TMP0:%.*]] = bitcast <4 x float> %a to <16 x i8>
193 // CHECK: [[TMP1:%.*]] = bitcast <4 x float> %b to <16 x i8>
194 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x float>
195 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
196 // CHECK: [[VEXT:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> [[TMP3]], <4 x i32> <i32 1, i32 2, i32 3, i32 4>
197 // CHECK: ret <4 x float> [[VEXT]]
test_vextq_f32(float32x4_t a,float32x4_t b)198 float32x4_t test_vextq_f32(float32x4_t a, float32x4_t b) {
199 return vextq_f32(a, b, 1);
200 }
201
202 // CHECK-LABEL: define <2 x double> @test_vextq_f64(<2 x double> %a, <2 x double> %b) #0 {
203 // CHECK: [[TMP0:%.*]] = bitcast <2 x double> %a to <16 x i8>
204 // CHECK: [[TMP1:%.*]] = bitcast <2 x double> %b to <16 x i8>
205 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x double>
206 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
207 // CHECK: [[VEXT:%.*]] = shufflevector <2 x double> [[TMP2]], <2 x double> [[TMP3]], <2 x i32> <i32 1, i32 2>
208 // CHECK: ret <2 x double> [[VEXT]]
test_vextq_f64(float64x2_t a,float64x2_t b)209 float64x2_t test_vextq_f64(float64x2_t a, float64x2_t b) {
210 return vextq_f64(a, b, 1);
211 }
212
213 // CHECK-LABEL: define <8 x i8> @test_vext_p8(<8 x i8> %a, <8 x i8> %b) #0 {
214 // CHECK: [[VEXT:%.*]] = shufflevector <8 x i8> %a, <8 x i8> %b, <8 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9>
215 // CHECK: ret <8 x i8> [[VEXT]]
test_vext_p8(poly8x8_t a,poly8x8_t b)216 poly8x8_t test_vext_p8(poly8x8_t a, poly8x8_t b) {
217 return vext_p8(a, b, 2);
218 }
219
220 // CHECK-LABEL: define <4 x i16> @test_vext_p16(<4 x i16> %a, <4 x i16> %b) #0 {
221 // CHECK: [[TMP0:%.*]] = bitcast <4 x i16> %a to <8 x i8>
222 // CHECK: [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
223 // CHECK: [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
224 // CHECK: [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
225 // CHECK: [[VEXT:%.*]] = shufflevector <4 x i16> [[TMP2]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
226 // CHECK: ret <4 x i16> [[VEXT]]
test_vext_p16(poly16x4_t a,poly16x4_t b)227 poly16x4_t test_vext_p16(poly16x4_t a, poly16x4_t b) {
228 return vext_p16(a, b, 3);
229 }
230
231 // CHECK-LABEL: define <16 x i8> @test_vextq_p8(<16 x i8> %a, <16 x i8> %b) #0 {
232 // CHECK: [[VEXT:%.*]] = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17>
233 // CHECK: ret <16 x i8> [[VEXT]]
test_vextq_p8(poly8x16_t a,poly8x16_t b)234 poly8x16_t test_vextq_p8(poly8x16_t a, poly8x16_t b) {
235 return vextq_p8(a, b, 2);
236 }
237
238 // CHECK-LABEL: define <8 x i16> @test_vextq_p16(<8 x i16> %a, <8 x i16> %b) #0 {
239 // CHECK: [[TMP0:%.*]] = bitcast <8 x i16> %a to <16 x i8>
240 // CHECK: [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
241 // CHECK: [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
242 // CHECK: [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
243 // CHECK: [[VEXT:%.*]] = shufflevector <8 x i16> [[TMP2]], <8 x i16> [[TMP3]], <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
244 // CHECK: ret <8 x i16> [[VEXT]]
test_vextq_p16(poly16x8_t a,poly16x8_t b)245 poly16x8_t test_vextq_p16(poly16x8_t a, poly16x8_t b) {
246 return vextq_p16(a, b, 3);
247 }
248