1 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
2 // RUN:  -ffp-contract=fast -emit-llvm -o - %s | opt -S -mem2reg \
3 // RUN:  | FileCheck %s
4 
5 // Test new aarch64 intrinsics with poly64
6 
7 #include <arm_neon.h>
8 
9 // CHECK-LABEL: define <1 x i64> @test_vceq_p64(<1 x i64> %a, <1 x i64> %b) #0 {
10 // CHECK:   [[CMP_I:%.*]] = icmp eq <1 x i64> %a, %b
11 // CHECK:   [[SEXT_I:%.*]] = sext <1 x i1> [[CMP_I]] to <1 x i64>
12 // CHECK:   ret <1 x i64> [[SEXT_I]]
13 uint64x1_t test_vceq_p64(poly64x1_t a, poly64x1_t b) {
14   return vceq_p64(a, b);
15 }
16 
17 // CHECK-LABEL: define <2 x i64> @test_vceqq_p64(<2 x i64> %a, <2 x i64> %b) #0 {
18 // CHECK:   [[CMP_I:%.*]] = icmp eq <2 x i64> %a, %b
19 // CHECK:   [[SEXT_I:%.*]] = sext <2 x i1> [[CMP_I]] to <2 x i64>
20 // CHECK:   ret <2 x i64> [[SEXT_I]]
21 uint64x2_t test_vceqq_p64(poly64x2_t a, poly64x2_t b) {
22   return vceqq_p64(a, b);
23 }
24 
25 // CHECK-LABEL: define <1 x i64> @test_vtst_p64(<1 x i64> %a, <1 x i64> %b) #0 {
26 // CHECK:   [[TMP4:%.*]] = and <1 x i64> %a, %b
27 // CHECK:   [[TMP5:%.*]] = icmp ne <1 x i64> [[TMP4]], zeroinitializer
28 // CHECK:   [[VTST_I:%.*]] = sext <1 x i1> [[TMP5]] to <1 x i64>
29 // CHECK:   ret <1 x i64> [[VTST_I]]
30 uint64x1_t test_vtst_p64(poly64x1_t a, poly64x1_t b) {
31   return vtst_p64(a, b);
32 }
33 
34 // CHECK-LABEL: define <2 x i64> @test_vtstq_p64(<2 x i64> %a, <2 x i64> %b) #0 {
35 // CHECK:   [[TMP4:%.*]] = and <2 x i64> %a, %b
36 // CHECK:   [[TMP5:%.*]] = icmp ne <2 x i64> [[TMP4]], zeroinitializer
37 // CHECK:   [[VTST_I:%.*]] = sext <2 x i1> [[TMP5]] to <2 x i64>
38 // CHECK:   ret <2 x i64> [[VTST_I]]
39 uint64x2_t test_vtstq_p64(poly64x2_t a, poly64x2_t b) {
40   return vtstq_p64(a, b);
41 }
42 
43 // CHECK-LABEL: define <1 x i64> @test_vbsl_p64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) #0 {
44 // CHECK:   [[VBSL3_I:%.*]] = and <1 x i64> %a, %b
45 // CHECK:   [[TMP3:%.*]] = xor <1 x i64> %a, <i64 -1>
46 // CHECK:   [[VBSL4_I:%.*]] = and <1 x i64> [[TMP3]], %c
47 // CHECK:   [[VBSL5_I:%.*]] = or <1 x i64> [[VBSL3_I]], [[VBSL4_I]]
48 // CHECK:   ret <1 x i64> [[VBSL5_I]]
49 poly64x1_t test_vbsl_p64(poly64x1_t a, poly64x1_t b, poly64x1_t c) {
50   return vbsl_p64(a, b, c);
51 }
52 
53 // CHECK-LABEL: define <2 x i64> @test_vbslq_p64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) #0 {
54 // CHECK:   [[VBSL3_I:%.*]] = and <2 x i64> %a, %b
55 // CHECK:   [[TMP3:%.*]] = xor <2 x i64> %a, <i64 -1, i64 -1>
56 // CHECK:   [[VBSL4_I:%.*]] = and <2 x i64> [[TMP3]], %c
57 // CHECK:   [[VBSL5_I:%.*]] = or <2 x i64> [[VBSL3_I]], [[VBSL4_I]]
58 // CHECK:   ret <2 x i64> [[VBSL5_I]]
59 poly64x2_t test_vbslq_p64(poly64x2_t a, poly64x2_t b, poly64x2_t c) {
60   return vbslq_p64(a, b, c);
61 }
62 
63 // CHECK-LABEL: define i64 @test_vget_lane_p64(<1 x i64> %v) #0 {
64 // CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> %v to <8 x i8>
65 // CHECK:   [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
66 // CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x i64> [[TMP1]], i32 0
67 // CHECK:   ret i64 [[VGET_LANE]]
68 poly64_t test_vget_lane_p64(poly64x1_t v) {
69   return vget_lane_p64(v, 0);
70 }
71 
72 // CHECK-LABEL: define i64 @test_vgetq_lane_p64(<2 x i64> %v) #0 {
73 // CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %v to <16 x i8>
74 // CHECK:   [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
75 // CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x i64> [[TMP1]], i32 1
76 // CHECK:   ret i64 [[VGETQ_LANE]]
77 poly64_t test_vgetq_lane_p64(poly64x2_t v) {
78   return vgetq_lane_p64(v, 1);
79 }
80 
81 // CHECK-LABEL: define <1 x i64> @test_vset_lane_p64(i64 %a, <1 x i64> %v) #0 {
82 // CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> %v to <8 x i8>
83 // CHECK:   [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
84 // CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x i64> [[TMP1]], i64 %a, i32 0
85 // CHECK:   ret <1 x i64> [[VSET_LANE]]
86 poly64x1_t test_vset_lane_p64(poly64_t a, poly64x1_t v) {
87   return vset_lane_p64(a, v, 0);
88 }
89 
90 // CHECK-LABEL: define <2 x i64> @test_vsetq_lane_p64(i64 %a, <2 x i64> %v) #0 {
91 // CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %v to <16 x i8>
92 // CHECK:   [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
93 // CHECK:   [[VSET_LANE:%.*]] = insertelement <2 x i64> [[TMP1]], i64 %a, i32 1
94 // CHECK:   ret <2 x i64> [[VSET_LANE]]
95 poly64x2_t test_vsetq_lane_p64(poly64_t a, poly64x2_t v) {
96   return vsetq_lane_p64(a, v, 1);
97 }
98 
99 // CHECK-LABEL: define <1 x i64> @test_vcopy_lane_p64(<1 x i64> %a, <1 x i64> %b) #0 {
100 // CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> %b to <8 x i8>
101 // CHECK:   [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
102 // CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x i64> [[TMP1]], i32 0
103 // CHECK:   [[TMP2:%.*]] = bitcast <1 x i64> %a to <8 x i8>
104 // CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x i64>
105 // CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x i64> [[TMP3]], i64 [[VGET_LANE]], i32 0
106 // CHECK:   ret <1 x i64> [[VSET_LANE]]
107 poly64x1_t test_vcopy_lane_p64(poly64x1_t a, poly64x1_t b) {
108   return vcopy_lane_p64(a, 0, b, 0);
109 
110 }
111 
112 // CHECK-LABEL: define <2 x i64> @test_vcopyq_lane_p64(<2 x i64> %a, <1 x i64> %b) #0 {
113 // CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> %b to <8 x i8>
114 // CHECK:   [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
115 // CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x i64> [[TMP1]], i32 0
116 // CHECK:   [[TMP2:%.*]] = bitcast <2 x i64> %a to <16 x i8>
117 // CHECK:   [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x i64>
118 // CHECK:   [[VSET_LANE:%.*]] = insertelement <2 x i64> [[TMP3]], i64 [[VGET_LANE]], i32 1
119 // CHECK:   ret <2 x i64> [[VSET_LANE]]
120 poly64x2_t test_vcopyq_lane_p64(poly64x2_t a, poly64x1_t b) {
121   return vcopyq_lane_p64(a, 1, b, 0);
122 }
123 
124 // CHECK-LABEL: define <2 x i64> @test_vcopyq_laneq_p64(<2 x i64> %a, <2 x i64> %b) #0 {
125 // CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %b to <16 x i8>
126 // CHECK:   [[TMP1:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
127 // CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x i64> [[TMP1]], i32 1
128 // CHECK:   [[TMP2:%.*]] = bitcast <2 x i64> %a to <16 x i8>
129 // CHECK:   [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x i64>
130 // CHECK:   [[VSET_LANE:%.*]] = insertelement <2 x i64> [[TMP3]], i64 [[VGETQ_LANE]], i32 1
131 // CHECK:   ret <2 x i64> [[VSET_LANE]]
132 poly64x2_t test_vcopyq_laneq_p64(poly64x2_t a, poly64x2_t b) {
133   return vcopyq_laneq_p64(a, 1, b, 1);
134 }
135 
136 // CHECK-LABEL: define <1 x i64> @test_vcreate_p64(i64 %a) #0 {
137 // CHECK:   [[TMP0:%.*]] = bitcast i64 %a to <1 x i64>
138 // CHECK:   ret <1 x i64> [[TMP0]]
139 poly64x1_t test_vcreate_p64(uint64_t a) {
140   return vcreate_p64(a);
141 }
142 
143 // CHECK-LABEL: define <1 x i64> @test_vdup_n_p64(i64 %a) #0 {
144 // CHECK:   [[VECINIT_I:%.*]] = insertelement <1 x i64> undef, i64 %a, i32 0
145 // CHECK:   ret <1 x i64> [[VECINIT_I]]
146 poly64x1_t test_vdup_n_p64(poly64_t a) {
147   return vdup_n_p64(a);
148 }
149 // CHECK-LABEL: define <2 x i64> @test_vdupq_n_p64(i64 %a) #0 {
150 // CHECK:   [[VECINIT_I:%.*]] = insertelement <2 x i64> undef, i64 %a, i32 0
151 // CHECK:   [[VECINIT1_I:%.*]] = insertelement <2 x i64> [[VECINIT_I]], i64 %a, i32 1
152 // CHECK:   ret <2 x i64> [[VECINIT1_I]]
153 poly64x2_t test_vdupq_n_p64(poly64_t a) {
154   return vdupq_n_p64(a);
155 }
156 
157 // CHECK-LABEL: define <1 x i64> @test_vmov_n_p64(i64 %a) #0 {
158 // CHECK:   [[VECINIT_I:%.*]] = insertelement <1 x i64> undef, i64 %a, i32 0
159 // CHECK:   ret <1 x i64> [[VECINIT_I]]
160 poly64x1_t test_vmov_n_p64(poly64_t a) {
161   return vmov_n_p64(a);
162 }
163 
164 // CHECK-LABEL: define <2 x i64> @test_vmovq_n_p64(i64 %a) #0 {
165 // CHECK:   [[VECINIT_I:%.*]] = insertelement <2 x i64> undef, i64 %a, i32 0
166 // CHECK:   [[VECINIT1_I:%.*]] = insertelement <2 x i64> [[VECINIT_I]], i64 %a, i32 1
167 // CHECK:   ret <2 x i64> [[VECINIT1_I]]
168 poly64x2_t test_vmovq_n_p64(poly64_t a) {
169   return vmovq_n_p64(a);
170 }
171 
172 // CHECK-LABEL: define <1 x i64> @test_vdup_lane_p64(<1 x i64> %vec) #0 {
173 // CHECK:   [[SHUFFLE:%.*]] = shufflevector <1 x i64> %vec, <1 x i64> %vec, <1 x i32> zeroinitializer
174 // CHECK:   ret <1 x i64> [[SHUFFLE]]
175 poly64x1_t test_vdup_lane_p64(poly64x1_t vec) {
176   return vdup_lane_p64(vec, 0);
177 }
178 
179 // CHECK-LABEL: define <2 x i64> @test_vdupq_lane_p64(<1 x i64> %vec) #0 {
180 // CHECK:   [[SHUFFLE:%.*]] = shufflevector <1 x i64> %vec, <1 x i64> %vec, <2 x i32> zeroinitializer
181 // CHECK:   ret <2 x i64> [[SHUFFLE]]
182 poly64x2_t test_vdupq_lane_p64(poly64x1_t vec) {
183   return vdupq_lane_p64(vec, 0);
184 }
185 
186 // CHECK-LABEL: define <2 x i64> @test_vdupq_laneq_p64(<2 x i64> %vec) #0 {
187 // CHECK:   [[SHUFFLE:%.*]] = shufflevector <2 x i64> %vec, <2 x i64> %vec, <2 x i32> <i32 1, i32 1>
188 // CHECK:   ret <2 x i64> [[SHUFFLE]]
189 poly64x2_t test_vdupq_laneq_p64(poly64x2_t vec) {
190   return vdupq_laneq_p64(vec, 1);
191 }
192 
193 // CHECK-LABEL: define <2 x i64> @test_vcombine_p64(<1 x i64> %low, <1 x i64> %high) #0 {
194 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <1 x i64> %low, <1 x i64> %high, <2 x i32> <i32 0, i32 1>
195 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
196 poly64x2_t test_vcombine_p64(poly64x1_t low, poly64x1_t high) {
197   return vcombine_p64(low, high);
198 }
199 
200 // CHECK-LABEL: define <1 x i64> @test_vld1_p64(i64* %ptr) #0 {
201 // CHECK:   [[TMP0:%.*]] = bitcast i64* %ptr to i8*
202 // CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to <1 x i64>*
203 // CHECK:   [[TMP2:%.*]] = load <1 x i64>, <1 x i64>* [[TMP1]]
204 // CHECK:   ret <1 x i64> [[TMP2]]
205 poly64x1_t test_vld1_p64(poly64_t const * ptr) {
206   return vld1_p64(ptr);
207 }
208 
209 // CHECK-LABEL: define <2 x i64> @test_vld1q_p64(i64* %ptr) #0 {
210 // CHECK:   [[TMP0:%.*]] = bitcast i64* %ptr to i8*
211 // CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to <2 x i64>*
212 // CHECK:   [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* [[TMP1]]
213 // CHECK:   ret <2 x i64> [[TMP2]]
214 poly64x2_t test_vld1q_p64(poly64_t const * ptr) {
215   return vld1q_p64(ptr);
216 }
217 
218 // CHECK-LABEL: define void @test_vst1_p64(i64* %ptr, <1 x i64> %val) #0 {
219 // CHECK:   [[TMP0:%.*]] = bitcast i64* %ptr to i8*
220 // CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %val to <8 x i8>
221 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP0]] to <1 x i64>*
222 // CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
223 // CHECK:   store <1 x i64> [[TMP3]], <1 x i64>* [[TMP2]]
224 // CHECK:   ret void
225 void test_vst1_p64(poly64_t * ptr, poly64x1_t val) {
226   return vst1_p64(ptr, val);
227 }
228 
229 // CHECK-LABEL: define void @test_vst1q_p64(i64* %ptr, <2 x i64> %val) #0 {
230 // CHECK:   [[TMP0:%.*]] = bitcast i64* %ptr to i8*
231 // CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %val to <16 x i8>
232 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP0]] to <2 x i64>*
233 // CHECK:   [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
234 // CHECK:   store <2 x i64> [[TMP3]], <2 x i64>* [[TMP2]]
235 // CHECK:   ret void
236 void test_vst1q_p64(poly64_t * ptr, poly64x2_t val) {
237   return vst1q_p64(ptr, val);
238 }
239 
240 // CHECK-LABEL: define %struct.poly64x1x2_t @test_vld2_p64(i64* %ptr) #0 {
241 // CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x2_t, align 8
242 // CHECK:   [[__RET:%.*]] = alloca %struct.poly64x1x2_t, align 8
243 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
244 // CHECK:   [[TMP1:%.*]] = bitcast i64* %ptr to i8*
245 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to <1 x i64>*
246 // CHECK:   [[VLD2:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2.v1i64.p0v1i64(<1 x i64>* [[TMP2]])
247 // CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64> }*
248 // CHECK:   store { <1 x i64>, <1 x i64> } [[VLD2]], { <1 x i64>, <1 x i64> }* [[TMP3]]
249 // CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x1x2_t* [[RETVAL]] to i8*
250 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
251 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP4]], i8* [[TMP5]], i64 16, i32 8, i1 false)
252 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[RETVAL]], align 8
253 // CHECK:   ret %struct.poly64x1x2_t [[TMP6]]
254 poly64x1x2_t test_vld2_p64(poly64_t const * ptr) {
255   return vld2_p64(ptr);
256 }
257 
258 // CHECK-LABEL: define %struct.poly64x2x2_t @test_vld2q_p64(i64* %ptr) #0 {
259 // CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x2_t, align 16
260 // CHECK:   [[__RET:%.*]] = alloca %struct.poly64x2x2_t, align 16
261 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
262 // CHECK:   [[TMP1:%.*]] = bitcast i64* %ptr to i8*
263 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to <2 x i64>*
264 // CHECK:   [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2.v2i64.p0v2i64(<2 x i64>* [[TMP2]])
265 // CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
266 // CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]]
267 // CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x2x2_t* [[RETVAL]] to i8*
268 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
269 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP4]], i8* [[TMP5]], i64 32, i32 16, i1 false)
270 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[RETVAL]], align 16
271 // CHECK:   ret %struct.poly64x2x2_t [[TMP6]]
272 poly64x2x2_t test_vld2q_p64(poly64_t const * ptr) {
273   return vld2q_p64(ptr);
274 }
275 
276 // CHECK-LABEL: define %struct.poly64x1x3_t @test_vld3_p64(i64* %ptr) #0 {
277 // CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x3_t, align 8
278 // CHECK:   [[__RET:%.*]] = alloca %struct.poly64x1x3_t, align 8
279 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
280 // CHECK:   [[TMP1:%.*]] = bitcast i64* %ptr to i8*
281 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to <1 x i64>*
282 // CHECK:   [[VLD3:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3.v1i64.p0v1i64(<1 x i64>* [[TMP2]])
283 // CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
284 // CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP3]]
285 // CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x1x3_t* [[RETVAL]] to i8*
286 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
287 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP4]], i8* [[TMP5]], i64 24, i32 8, i1 false)
288 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[RETVAL]], align 8
289 // CHECK:   ret %struct.poly64x1x3_t [[TMP6]]
290 poly64x1x3_t test_vld3_p64(poly64_t const * ptr) {
291   return vld3_p64(ptr);
292 }
293 
294 // CHECK-LABEL: define %struct.poly64x2x3_t @test_vld3q_p64(i64* %ptr) #0 {
295 // CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x3_t, align 16
296 // CHECK:   [[__RET:%.*]] = alloca %struct.poly64x2x3_t, align 16
297 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
298 // CHECK:   [[TMP1:%.*]] = bitcast i64* %ptr to i8*
299 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to <2 x i64>*
300 // CHECK:   [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3.v2i64.p0v2i64(<2 x i64>* [[TMP2]])
301 // CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
302 // CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
303 // CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x2x3_t* [[RETVAL]] to i8*
304 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
305 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP4]], i8* [[TMP5]], i64 48, i32 16, i1 false)
306 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[RETVAL]], align 16
307 // CHECK:   ret %struct.poly64x2x3_t [[TMP6]]
308 poly64x2x3_t test_vld3q_p64(poly64_t const * ptr) {
309   return vld3q_p64(ptr);
310 }
311 
312 // CHECK-LABEL: define %struct.poly64x1x4_t @test_vld4_p64(i64* %ptr) #0 {
313 // CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x4_t, align 8
314 // CHECK:   [[__RET:%.*]] = alloca %struct.poly64x1x4_t, align 8
315 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
316 // CHECK:   [[TMP1:%.*]] = bitcast i64* %ptr to i8*
317 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to <1 x i64>*
318 // CHECK:   [[VLD4:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4.v1i64.p0v1i64(<1 x i64>* [[TMP2]])
319 // CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
320 // CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP3]]
321 // CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x1x4_t* [[RETVAL]] to i8*
322 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
323 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP4]], i8* [[TMP5]], i64 32, i32 8, i1 false)
324 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[RETVAL]], align 8
325 // CHECK:   ret %struct.poly64x1x4_t [[TMP6]]
326 poly64x1x4_t test_vld4_p64(poly64_t const * ptr) {
327   return vld4_p64(ptr);
328 }
329 
330 // CHECK-LABEL: define %struct.poly64x2x4_t @test_vld4q_p64(i64* %ptr) #0 {
331 // CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x4_t, align 16
332 // CHECK:   [[__RET:%.*]] = alloca %struct.poly64x2x4_t, align 16
333 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
334 // CHECK:   [[TMP1:%.*]] = bitcast i64* %ptr to i8*
335 // CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to <2 x i64>*
336 // CHECK:   [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4.v2i64.p0v2i64(<2 x i64>* [[TMP2]])
337 // CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
338 // CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
339 // CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x2x4_t* [[RETVAL]] to i8*
340 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
341 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP4]], i8* [[TMP5]], i64 64, i32 16, i1 false)
342 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[RETVAL]], align 16
343 // CHECK:   ret %struct.poly64x2x4_t [[TMP6]]
344 poly64x2x4_t test_vld4q_p64(poly64_t const * ptr) {
345   return vld4q_p64(ptr);
346 }
347 
348 // CHECK-LABEL: define void @test_vst2_p64(i64* %ptr, [2 x <1 x i64>] %val.coerce) #0 {
349 // CHECK:   [[VAL:%.*]] = alloca %struct.poly64x1x2_t, align 8
350 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x2_t, align 8
351 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[VAL]], i32 0, i32 0
352 // CHECK:   store [2 x <1 x i64>] [[VAL]].coerce, [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
353 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__S1]] to i8*
354 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x1x2_t* [[VAL]] to i8*
355 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP0]], i8* [[TMP1]], i64 16, i32 8, i1 false)
356 // CHECK:   [[TMP2:%.*]] = bitcast i64* %ptr to i8*
357 // CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[__S1]], i32 0, i32 0
358 // CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 0
359 // CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
360 // CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
361 // CHECK:   [[VAL2:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[__S1]], i32 0, i32 0
362 // CHECK:   [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL2]], i64 0, i64 1
363 // CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX3]], align 8
364 // CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
365 // CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
366 // CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
367 // CHECK:   call void @llvm.aarch64.neon.st2.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i8* [[TMP2]])
368 // CHECK:   ret void
369 void test_vst2_p64(poly64_t * ptr, poly64x1x2_t val) {
370   return vst2_p64(ptr, val);
371 }
372 
373 // CHECK-LABEL: define void @test_vst2q_p64(i64* %ptr, [2 x <2 x i64>] %val.coerce) #0 {
374 // CHECK:   [[VAL:%.*]] = alloca %struct.poly64x2x2_t, align 16
375 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x2_t, align 16
376 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[VAL]], i32 0, i32 0
377 // CHECK:   store [2 x <2 x i64>] [[VAL]].coerce, [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
378 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__S1]] to i8*
379 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x2x2_t* [[VAL]] to i8*
380 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP0]], i8* [[TMP1]], i64 32, i32 16, i1 false)
381 // CHECK:   [[TMP2:%.*]] = bitcast i64* %ptr to i8*
382 // CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[__S1]], i32 0, i32 0
383 // CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 0
384 // CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
385 // CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
386 // CHECK:   [[VAL2:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[__S1]], i32 0, i32 0
387 // CHECK:   [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL2]], i64 0, i64 1
388 // CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX3]], align 16
389 // CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
390 // CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
391 // CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
392 // CHECK:   call void @llvm.aarch64.neon.st2.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i8* [[TMP2]])
393 // CHECK:   ret void
394 void test_vst2q_p64(poly64_t * ptr, poly64x2x2_t val) {
395   return vst2q_p64(ptr, val);
396 }
397 
398 // CHECK-LABEL: define void @test_vst3_p64(i64* %ptr, [3 x <1 x i64>] %val.coerce) #0 {
399 // CHECK:   [[VAL:%.*]] = alloca %struct.poly64x1x3_t, align 8
400 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x3_t, align 8
401 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[VAL]], i32 0, i32 0
402 // CHECK:   store [3 x <1 x i64>] [[VAL]].coerce, [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
403 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__S1]] to i8*
404 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x1x3_t* [[VAL]] to i8*
405 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP0]], i8* [[TMP1]], i64 24, i32 8, i1 false)
406 // CHECK:   [[TMP2:%.*]] = bitcast i64* %ptr to i8*
407 // CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
408 // CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 0
409 // CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
410 // CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
411 // CHECK:   [[VAL2:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
412 // CHECK:   [[ARRAYIDX3:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL2]], i64 0, i64 1
413 // CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX3]], align 8
414 // CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
415 // CHECK:   [[VAL4:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
416 // CHECK:   [[ARRAYIDX5:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL4]], i64 0, i64 2
417 // CHECK:   [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX5]], align 8
418 // CHECK:   [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
419 // CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
420 // CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
421 // CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
422 // CHECK:   call void @llvm.aarch64.neon.st3.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i8* [[TMP2]])
423 // CHECK:   ret void
424 void test_vst3_p64(poly64_t * ptr, poly64x1x3_t val) {
425   return vst3_p64(ptr, val);
426 }
427 
428 // CHECK-LABEL: define void @test_vst3q_p64(i64* %ptr, [3 x <2 x i64>] %val.coerce) #0 {
429 // CHECK:   [[VAL:%.*]] = alloca %struct.poly64x2x3_t, align 16
430 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x3_t, align 16
431 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[VAL]], i32 0, i32 0
432 // CHECK:   store [3 x <2 x i64>] [[VAL]].coerce, [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
433 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__S1]] to i8*
434 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x2x3_t* [[VAL]] to i8*
435 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP0]], i8* [[TMP1]], i64 48, i32 16, i1 false)
436 // CHECK:   [[TMP2:%.*]] = bitcast i64* %ptr to i8*
437 // CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
438 // CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 0
439 // CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
440 // CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
441 // CHECK:   [[VAL2:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
442 // CHECK:   [[ARRAYIDX3:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL2]], i64 0, i64 1
443 // CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX3]], align 16
444 // CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
445 // CHECK:   [[VAL4:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
446 // CHECK:   [[ARRAYIDX5:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL4]], i64 0, i64 2
447 // CHECK:   [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX5]], align 16
448 // CHECK:   [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
449 // CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
450 // CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
451 // CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
452 // CHECK:   call void @llvm.aarch64.neon.st3.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i8* [[TMP2]])
453 // CHECK:   ret void
454 void test_vst3q_p64(poly64_t * ptr, poly64x2x3_t val) {
455   return vst3q_p64(ptr, val);
456 }
457 
458 // CHECK-LABEL: define void @test_vst4_p64(i64* %ptr, [4 x <1 x i64>] %val.coerce) #0 {
459 // CHECK:   [[VAL:%.*]] = alloca %struct.poly64x1x4_t, align 8
460 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x4_t, align 8
461 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[VAL]], i32 0, i32 0
462 // CHECK:   store [4 x <1 x i64>] [[VAL]].coerce, [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
463 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__S1]] to i8*
464 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x1x4_t* [[VAL]] to i8*
465 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP0]], i8* [[TMP1]], i64 32, i32 8, i1 false)
466 // CHECK:   [[TMP2:%.*]] = bitcast i64* %ptr to i8*
467 // CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
468 // CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 0
469 // CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
470 // CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
471 // CHECK:   [[VAL2:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
472 // CHECK:   [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL2]], i64 0, i64 1
473 // CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX3]], align 8
474 // CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
475 // CHECK:   [[VAL4:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
476 // CHECK:   [[ARRAYIDX5:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL4]], i64 0, i64 2
477 // CHECK:   [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX5]], align 8
478 // CHECK:   [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
479 // CHECK:   [[VAL6:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
480 // CHECK:   [[ARRAYIDX7:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL6]], i64 0, i64 3
481 // CHECK:   [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX7]], align 8
482 // CHECK:   [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
483 // CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
484 // CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
485 // CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
486 // CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
487 // CHECK:   call void @llvm.aarch64.neon.st4.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i8* [[TMP2]])
488 // CHECK:   ret void
489 void test_vst4_p64(poly64_t * ptr, poly64x1x4_t val) {
490   return vst4_p64(ptr, val);
491 }
492 
493 // CHECK-LABEL: define void @test_vst4q_p64(i64* %ptr, [4 x <2 x i64>] %val.coerce) #0 {
494 // CHECK:   [[VAL:%.*]] = alloca %struct.poly64x2x4_t, align 16
495 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x4_t, align 16
496 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[VAL]], i32 0, i32 0
497 // CHECK:   store [4 x <2 x i64>] [[VAL]].coerce, [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
498 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__S1]] to i8*
499 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x2x4_t* [[VAL]] to i8*
500 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[TMP0]], i8* [[TMP1]], i64 64, i32 16, i1 false)
501 // CHECK:   [[TMP2:%.*]] = bitcast i64* %ptr to i8*
502 // CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
503 // CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 0
504 // CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
505 // CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
506 // CHECK:   [[VAL2:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
507 // CHECK:   [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL2]], i64 0, i64 1
508 // CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX3]], align 16
509 // CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
510 // CHECK:   [[VAL4:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
511 // CHECK:   [[ARRAYIDX5:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL4]], i64 0, i64 2
512 // CHECK:   [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX5]], align 16
513 // CHECK:   [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
514 // CHECK:   [[VAL6:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
515 // CHECK:   [[ARRAYIDX7:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL6]], i64 0, i64 3
516 // CHECK:   [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX7]], align 16
517 // CHECK:   [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
518 // CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
519 // CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
520 // CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
521 // CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
522 // CHECK:   call void @llvm.aarch64.neon.st4.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i8* [[TMP2]])
523 // CHECK:   ret void
524 void test_vst4q_p64(poly64_t * ptr, poly64x2x4_t val) {
525   return vst4q_p64(ptr, val);
526 }
527 
528 // CHECK-LABEL: define <1 x i64> @test_vext_p64(<1 x i64> %a, <1 x i64> %b) #0 {
529 // CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> %a to <8 x i8>
530 // CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
531 // CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
532 // CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
533 // CHECK:   [[VEXT:%.*]] = shufflevector <1 x i64> [[TMP2]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
534 // CHECK:   ret <1 x i64> [[VEXT]]
535 poly64x1_t test_vext_p64(poly64x1_t a, poly64x1_t b) {
536   return vext_u64(a, b, 0);
537 
538 }
539 
540 // CHECK-LABEL: define <2 x i64> @test_vextq_p64(<2 x i64> %a, <2 x i64> %b) #0 {
541 // CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to <16 x i8>
542 // CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
543 // CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
544 // CHECK:   [[TMP3:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
545 // CHECK:   [[VEXT:%.*]] = shufflevector <2 x i64> [[TMP2]], <2 x i64> [[TMP3]], <2 x i32> <i32 1, i32 2>
546 // CHECK:   ret <2 x i64> [[VEXT]]
547 poly64x2_t test_vextq_p64(poly64x2_t a, poly64x2_t b) {
548   return vextq_p64(a, b, 1);
549 }
550 
551 // CHECK-LABEL: define <2 x i64> @test_vzip1q_p64(<2 x i64> %a, <2 x i64> %b) #0 {
552 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 0, i32 2>
553 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
554 poly64x2_t test_vzip1q_p64(poly64x2_t a, poly64x2_t b) {
555   return vzip1q_p64(a, b);
556 }
557 
558 // CHECK-LABEL: define <2 x i64> @test_vzip2q_p64(<2 x i64> %a, <2 x i64> %b) #0 {
559 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 1, i32 3>
560 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
561 poly64x2_t test_vzip2q_p64(poly64x2_t a, poly64x2_t b) {
562   return vzip2q_u64(a, b);
563 }
564 
565 // CHECK-LABEL: define <2 x i64> @test_vuzp1q_p64(<2 x i64> %a, <2 x i64> %b) #0 {
566 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 0, i32 2>
567 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
568 poly64x2_t test_vuzp1q_p64(poly64x2_t a, poly64x2_t b) {
569   return vuzp1q_p64(a, b);
570 }
571 
572 // CHECK-LABEL: define <2 x i64> @test_vuzp2q_p64(<2 x i64> %a, <2 x i64> %b) #0 {
573 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 1, i32 3>
574 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
575 poly64x2_t test_vuzp2q_p64(poly64x2_t a, poly64x2_t b) {
576   return vuzp2q_u64(a, b);
577 }
578 
579 // CHECK-LABEL: define <2 x i64> @test_vtrn1q_p64(<2 x i64> %a, <2 x i64> %b) #0 {
580 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 0, i32 2>
581 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
582 poly64x2_t test_vtrn1q_p64(poly64x2_t a, poly64x2_t b) {
583   return vtrn1q_p64(a, b);
584 }
585 
586 // CHECK-LABEL: define <2 x i64> @test_vtrn2q_p64(<2 x i64> %a, <2 x i64> %b) #0 {
587 // CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> <i32 1, i32 3>
588 // CHECK:   ret <2 x i64> [[SHUFFLE_I]]
589 poly64x2_t test_vtrn2q_p64(poly64x2_t a, poly64x2_t b) {
590   return vtrn2q_u64(a, b);
591 }
592 
593 // CHECK-LABEL: define <1 x i64> @test_vsri_n_p64(<1 x i64> %a, <1 x i64> %b) #0 {
594 // CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> %a to <8 x i8>
595 // CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
596 // CHECK:   [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
597 // CHECK:   [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
598 // CHECK:   [[VSRI_N2:%.*]] = call <1 x i64> @llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 33)
599 // CHECK:   ret <1 x i64> [[VSRI_N2]]
600 poly64x1_t test_vsri_n_p64(poly64x1_t a, poly64x1_t b) {
601   return vsri_n_p64(a, b, 33);
602 }
603 
604 // CHECK-LABEL: define <2 x i64> @test_vsriq_n_p64(<2 x i64> %a, <2 x i64> %b) #0 {
605 // CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to <16 x i8>
606 // CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
607 // CHECK:   [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
608 // CHECK:   [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
609 // CHECK:   [[VSRI_N2:%.*]] = call <2 x i64> @llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 64)
610 // CHECK:   ret <2 x i64> [[VSRI_N2]]
611 poly64x2_t test_vsriq_n_p64(poly64x2_t a, poly64x2_t b) {
612   return vsriq_n_p64(a, b, 64);
613 }
614 
615