1 ; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon | FileCheck %s
2 ; Intrinsic wrangling & arm64 does it differently.
4 define i64 @test_vshrd_n_s64(i64 %a) {
5 ; CHECK: test_vshrd_n_s64
6 ; CHECK: sshr {{d[0-9]+}}, {{d[0-9]+}}, #63
8 %vsshr = insertelement <1 x i64> undef, i64 %a, i32 0
9 %vsshr1 = call <1 x i64> @llvm.aarch64.neon.vshrds.n(<1 x i64> %vsshr, i32 63)
10 %0 = extractelement <1 x i64> %vsshr1, i32 0
14 declare <1 x i64> @llvm.aarch64.neon.vshrds.n(<1 x i64>, i32)
16 define i64 @test_vshrd_n_u64(i64 %a) {
17 ; CHECK: test_vshrd_n_u64
18 ; CHECK: ushr {{d[0-9]+}}, {{d[0-9]+}}, #63
20 %vushr = insertelement <1 x i64> undef, i64 %a, i32 0
21 %vushr1 = call <1 x i64> @llvm.aarch64.neon.vshrdu.n(<1 x i64> %vushr, i32 63)
22 %0 = extractelement <1 x i64> %vushr1, i32 0
26 declare <1 x i64> @llvm.aarch64.neon.vshrdu.n(<1 x i64>, i32)
28 define i64 @test_vrshrd_n_s64(i64 %a) {
29 ; CHECK: test_vrshrd_n_s64
30 ; CHECK: srshr {{d[0-9]+}}, {{d[0-9]+}}, #63
32 %vsrshr = insertelement <1 x i64> undef, i64 %a, i32 0
33 %vsrshr1 = call <1 x i64> @llvm.aarch64.neon.vsrshr.v1i64(<1 x i64> %vsrshr, i32 63)
34 %0 = extractelement <1 x i64> %vsrshr1, i32 0
38 declare <1 x i64> @llvm.aarch64.neon.vsrshr.v1i64(<1 x i64>, i32)
40 define i64 @test_vrshrd_n_u64(i64 %a) {
41 ; CHECK: test_vrshrd_n_u64
42 ; CHECK: urshr {{d[0-9]+}}, {{d[0-9]+}}, #63
44 %vurshr = insertelement <1 x i64> undef, i64 %a, i32 0
45 %vurshr1 = call <1 x i64> @llvm.aarch64.neon.vurshr.v1i64(<1 x i64> %vurshr, i32 63)
46 %0 = extractelement <1 x i64> %vurshr1, i32 0
50 declare <1 x i64> @llvm.aarch64.neon.vurshr.v1i64(<1 x i64>, i32)
52 define i64 @test_vsrad_n_s64(i64 %a, i64 %b) {
53 ; CHECK: test_vsrad_n_s64
54 ; CHECK: ssra {{d[0-9]+}}, {{d[0-9]+}}, #63
56 %vssra = insertelement <1 x i64> undef, i64 %a, i32 0
57 %vssra1 = insertelement <1 x i64> undef, i64 %b, i32 0
58 %vssra2 = call <1 x i64> @llvm.aarch64.neon.vsrads.n(<1 x i64> %vssra, <1 x i64> %vssra1, i32 63)
59 %0 = extractelement <1 x i64> %vssra2, i32 0
63 declare <1 x i64> @llvm.aarch64.neon.vsrads.n(<1 x i64>, <1 x i64>, i32)
65 define i64 @test_vsrad_n_u64(i64 %a, i64 %b) {
66 ; CHECK: test_vsrad_n_u64
67 ; CHECK: usra {{d[0-9]+}}, {{d[0-9]+}}, #63
69 %vusra = insertelement <1 x i64> undef, i64 %a, i32 0
70 %vusra1 = insertelement <1 x i64> undef, i64 %b, i32 0
71 %vusra2 = call <1 x i64> @llvm.aarch64.neon.vsradu.n(<1 x i64> %vusra, <1 x i64> %vusra1, i32 63)
72 %0 = extractelement <1 x i64> %vusra2, i32 0
76 declare <1 x i64> @llvm.aarch64.neon.vsradu.n(<1 x i64>, <1 x i64>, i32)
78 define i64 @test_vrsrad_n_s64(i64 %a, i64 %b) {
79 ; CHECK: test_vrsrad_n_s64
80 ; CHECK: srsra {{d[0-9]+}}, {{d[0-9]+}}, #63
82 %vsrsra = insertelement <1 x i64> undef, i64 %a, i32 0
83 %vsrsra1 = insertelement <1 x i64> undef, i64 %b, i32 0
84 %vsrsra2 = call <1 x i64> @llvm.aarch64.neon.vrsrads.n(<1 x i64> %vsrsra, <1 x i64> %vsrsra1, i32 63)
85 %0 = extractelement <1 x i64> %vsrsra2, i32 0
89 declare <1 x i64> @llvm.aarch64.neon.vrsrads.n(<1 x i64>, <1 x i64>, i32)
91 define i64 @test_vrsrad_n_u64(i64 %a, i64 %b) {
92 ; CHECK: test_vrsrad_n_u64
93 ; CHECK: ursra {{d[0-9]+}}, {{d[0-9]+}}, #63
95 %vursra = insertelement <1 x i64> undef, i64 %a, i32 0
96 %vursra1 = insertelement <1 x i64> undef, i64 %b, i32 0
97 %vursra2 = call <1 x i64> @llvm.aarch64.neon.vrsradu.n(<1 x i64> %vursra, <1 x i64> %vursra1, i32 63)
98 %0 = extractelement <1 x i64> %vursra2, i32 0
102 declare <1 x i64> @llvm.aarch64.neon.vrsradu.n(<1 x i64>, <1 x i64>, i32)
104 define i64 @test_vshld_n_s64(i64 %a) {
105 ; CHECK: test_vshld_n_s64
106 ; CHECK: shl {{d[0-9]+}}, {{d[0-9]+}}, #63
108 %vshl = insertelement <1 x i64> undef, i64 %a, i32 0
109 %vshl1 = call <1 x i64> @llvm.aarch64.neon.vshld.n(<1 x i64> %vshl, i32 63)
110 %0 = extractelement <1 x i64> %vshl1, i32 0
114 declare <1 x i64> @llvm.aarch64.neon.vshld.n(<1 x i64>, i32)
116 define i64 @test_vshld_n_u64(i64 %a) {
117 ; CHECK: test_vshld_n_u64
118 ; CHECK: shl {{d[0-9]+}}, {{d[0-9]+}}, #63
120 %vshl = insertelement <1 x i64> undef, i64 %a, i32 0
121 %vshl1 = call <1 x i64> @llvm.aarch64.neon.vshld.n(<1 x i64> %vshl, i32 63)
122 %0 = extractelement <1 x i64> %vshl1, i32 0
126 define i8 @test_vqshlb_n_s8(i8 %a) {
127 ; CHECK: test_vqshlb_n_s8
128 ; CHECK: sqshl {{b[0-9]+}}, {{b[0-9]+}}, #7
130 %vsqshl = insertelement <1 x i8> undef, i8 %a, i32 0
131 %vsqshl1 = call <1 x i8> @llvm.aarch64.neon.vqshls.n.v1i8(<1 x i8> %vsqshl, i32 7)
132 %0 = extractelement <1 x i8> %vsqshl1, i32 0
136 declare <1 x i8> @llvm.aarch64.neon.vqshls.n.v1i8(<1 x i8>, i32)
138 define i16 @test_vqshlh_n_s16(i16 %a) {
139 ; CHECK: test_vqshlh_n_s16
140 ; CHECK: sqshl {{h[0-9]+}}, {{h[0-9]+}}, #15
142 %vsqshl = insertelement <1 x i16> undef, i16 %a, i32 0
143 %vsqshl1 = call <1 x i16> @llvm.aarch64.neon.vqshls.n.v1i16(<1 x i16> %vsqshl, i32 15)
144 %0 = extractelement <1 x i16> %vsqshl1, i32 0
148 declare <1 x i16> @llvm.aarch64.neon.vqshls.n.v1i16(<1 x i16>, i32)
150 define i32 @test_vqshls_n_s32(i32 %a) {
151 ; CHECK: test_vqshls_n_s32
152 ; CHECK: sqshl {{s[0-9]+}}, {{s[0-9]+}}, #31
154 %vsqshl = insertelement <1 x i32> undef, i32 %a, i32 0
155 %vsqshl1 = call <1 x i32> @llvm.aarch64.neon.vqshls.n.v1i32(<1 x i32> %vsqshl, i32 31)
156 %0 = extractelement <1 x i32> %vsqshl1, i32 0
160 declare <1 x i32> @llvm.aarch64.neon.vqshls.n.v1i32(<1 x i32>, i32)
162 define i64 @test_vqshld_n_s64(i64 %a) {
163 ; CHECK: test_vqshld_n_s64
164 ; CHECK: sqshl {{d[0-9]+}}, {{d[0-9]+}}, #63
166 %vsqshl = insertelement <1 x i64> undef, i64 %a, i32 0
167 %vsqshl1 = call <1 x i64> @llvm.aarch64.neon.vqshls.n.v1i64(<1 x i64> %vsqshl, i32 63)
168 %0 = extractelement <1 x i64> %vsqshl1, i32 0
172 declare <1 x i64> @llvm.aarch64.neon.vqshls.n.v1i64(<1 x i64>, i32)
174 define i8 @test_vqshlb_n_u8(i8 %a) {
175 ; CHECK: test_vqshlb_n_u8
176 ; CHECK: uqshl {{b[0-9]+}}, {{b[0-9]+}}, #7
178 %vuqshl = insertelement <1 x i8> undef, i8 %a, i32 0
179 %vuqshl1 = call <1 x i8> @llvm.aarch64.neon.vqshlu.n.v1i8(<1 x i8> %vuqshl, i32 7)
180 %0 = extractelement <1 x i8> %vuqshl1, i32 0
184 declare <1 x i8> @llvm.aarch64.neon.vqshlu.n.v1i8(<1 x i8>, i32)
186 define i16 @test_vqshlh_n_u16(i16 %a) {
187 ; CHECK: test_vqshlh_n_u16
188 ; CHECK: uqshl {{h[0-9]+}}, {{h[0-9]+}}, #15
190 %vuqshl = insertelement <1 x i16> undef, i16 %a, i32 0
191 %vuqshl1 = call <1 x i16> @llvm.aarch64.neon.vqshlu.n.v1i16(<1 x i16> %vuqshl, i32 15)
192 %0 = extractelement <1 x i16> %vuqshl1, i32 0
196 declare <1 x i16> @llvm.aarch64.neon.vqshlu.n.v1i16(<1 x i16>, i32)
198 define i32 @test_vqshls_n_u32(i32 %a) {
199 ; CHECK: test_vqshls_n_u32
200 ; CHECK: uqshl {{s[0-9]+}}, {{s[0-9]+}}, #31
202 %vuqshl = insertelement <1 x i32> undef, i32 %a, i32 0
203 %vuqshl1 = call <1 x i32> @llvm.aarch64.neon.vqshlu.n.v1i32(<1 x i32> %vuqshl, i32 31)
204 %0 = extractelement <1 x i32> %vuqshl1, i32 0
208 declare <1 x i32> @llvm.aarch64.neon.vqshlu.n.v1i32(<1 x i32>, i32)
210 define i64 @test_vqshld_n_u64(i64 %a) {
211 ; CHECK: test_vqshld_n_u64
212 ; CHECK: uqshl {{d[0-9]+}}, {{d[0-9]+}}, #63
214 %vuqshl = insertelement <1 x i64> undef, i64 %a, i32 0
215 %vuqshl1 = call <1 x i64> @llvm.aarch64.neon.vqshlu.n.v1i64(<1 x i64> %vuqshl, i32 63)
216 %0 = extractelement <1 x i64> %vuqshl1, i32 0
220 declare <1 x i64> @llvm.aarch64.neon.vqshlu.n.v1i64(<1 x i64>, i32)
222 define i8 @test_vqshlub_n_s8(i8 %a) {
223 ; CHECK: test_vqshlub_n_s8
224 ; CHECK: sqshlu {{b[0-9]+}}, {{b[0-9]+}}, #7
226 %vsqshlu = insertelement <1 x i8> undef, i8 %a, i32 0
227 %vsqshlu1 = call <1 x i8> @llvm.aarch64.neon.vsqshlu.v1i8(<1 x i8> %vsqshlu, i32 7)
228 %0 = extractelement <1 x i8> %vsqshlu1, i32 0
232 declare <1 x i8> @llvm.aarch64.neon.vsqshlu.v1i8(<1 x i8>, i32)
234 define i16 @test_vqshluh_n_s16(i16 %a) {
235 ; CHECK: test_vqshluh_n_s16
236 ; CHECK: sqshlu {{h[0-9]+}}, {{h[0-9]+}}, #15
238 %vsqshlu = insertelement <1 x i16> undef, i16 %a, i32 0
239 %vsqshlu1 = call <1 x i16> @llvm.aarch64.neon.vsqshlu.v1i16(<1 x i16> %vsqshlu, i32 15)
240 %0 = extractelement <1 x i16> %vsqshlu1, i32 0
244 declare <1 x i16> @llvm.aarch64.neon.vsqshlu.v1i16(<1 x i16>, i32)
246 define i32 @test_vqshlus_n_s32(i32 %a) {
247 ; CHECK: test_vqshlus_n_s32
248 ; CHECK: sqshlu {{s[0-9]+}}, {{s[0-9]+}}, #31
250 %vsqshlu = insertelement <1 x i32> undef, i32 %a, i32 0
251 %vsqshlu1 = call <1 x i32> @llvm.aarch64.neon.vsqshlu.v1i32(<1 x i32> %vsqshlu, i32 31)
252 %0 = extractelement <1 x i32> %vsqshlu1, i32 0
256 declare <1 x i32> @llvm.aarch64.neon.vsqshlu.v1i32(<1 x i32>, i32)
258 define i64 @test_vqshlud_n_s64(i64 %a) {
259 ; CHECK: test_vqshlud_n_s64
260 ; CHECK: sqshlu {{d[0-9]+}}, {{d[0-9]+}}, #63
262 %vsqshlu = insertelement <1 x i64> undef, i64 %a, i32 0
263 %vsqshlu1 = call <1 x i64> @llvm.aarch64.neon.vsqshlu.v1i64(<1 x i64> %vsqshlu, i32 63)
264 %0 = extractelement <1 x i64> %vsqshlu1, i32 0
268 declare <1 x i64> @llvm.aarch64.neon.vsqshlu.v1i64(<1 x i64>, i32)
270 define i64 @test_vsrid_n_s64(i64 %a, i64 %b) {
271 ; CHECK: test_vsrid_n_s64
272 ; CHECK: sri {{d[0-9]+}}, {{d[0-9]+}}, #63
274 %vsri = insertelement <1 x i64> undef, i64 %a, i32 0
275 %vsri1 = insertelement <1 x i64> undef, i64 %b, i32 0
276 %vsri2 = call <1 x i64> @llvm.aarch64.neon.vsri.v1i64(<1 x i64> %vsri, <1 x i64> %vsri1, i32 63)
277 %0 = extractelement <1 x i64> %vsri2, i32 0
281 declare <1 x i64> @llvm.aarch64.neon.vsri.v1i64(<1 x i64>, <1 x i64>, i32)
283 define i64 @test_vsrid_n_u64(i64 %a, i64 %b) {
284 ; CHECK: test_vsrid_n_u64
285 ; CHECK: sri {{d[0-9]+}}, {{d[0-9]+}}, #63
287 %vsri = insertelement <1 x i64> undef, i64 %a, i32 0
288 %vsri1 = insertelement <1 x i64> undef, i64 %b, i32 0
289 %vsri2 = call <1 x i64> @llvm.aarch64.neon.vsri.v1i64(<1 x i64> %vsri, <1 x i64> %vsri1, i32 63)
290 %0 = extractelement <1 x i64> %vsri2, i32 0
294 define i64 @test_vslid_n_s64(i64 %a, i64 %b) {
295 ; CHECK: test_vslid_n_s64
296 ; CHECK: sli {{d[0-9]+}}, {{d[0-9]+}}, #63
298 %vsli = insertelement <1 x i64> undef, i64 %a, i32 0
299 %vsli1 = insertelement <1 x i64> undef, i64 %b, i32 0
300 %vsli2 = call <1 x i64> @llvm.aarch64.neon.vsli.v1i64(<1 x i64> %vsli, <1 x i64> %vsli1, i32 63)
301 %0 = extractelement <1 x i64> %vsli2, i32 0
305 declare <1 x i64> @llvm.aarch64.neon.vsli.v1i64(<1 x i64>, <1 x i64>, i32)
307 define i64 @test_vslid_n_u64(i64 %a, i64 %b) {
308 ; CHECK: test_vslid_n_u64
309 ; CHECK: sli {{d[0-9]+}}, {{d[0-9]+}}, #63
311 %vsli = insertelement <1 x i64> undef, i64 %a, i32 0
312 %vsli1 = insertelement <1 x i64> undef, i64 %b, i32 0
313 %vsli2 = call <1 x i64> @llvm.aarch64.neon.vsli.v1i64(<1 x i64> %vsli, <1 x i64> %vsli1, i32 63)
314 %0 = extractelement <1 x i64> %vsli2, i32 0
318 define i8 @test_vqshrnh_n_s16(i16 %a) {
319 ; CHECK: test_vqshrnh_n_s16
320 ; CHECK: sqshrn {{b[0-9]+}}, {{h[0-9]+}}, #8
322 %vsqshrn = insertelement <1 x i16> undef, i16 %a, i32 0
323 %vsqshrn1 = call <1 x i8> @llvm.aarch64.neon.vsqshrn.v1i8(<1 x i16> %vsqshrn, i32 8)
324 %0 = extractelement <1 x i8> %vsqshrn1, i32 0
328 declare <1 x i8> @llvm.aarch64.neon.vsqshrn.v1i8(<1 x i16>, i32)
330 define i16 @test_vqshrns_n_s32(i32 %a) {
331 ; CHECK: test_vqshrns_n_s32
332 ; CHECK: sqshrn {{h[0-9]+}}, {{s[0-9]+}}, #16
334 %vsqshrn = insertelement <1 x i32> undef, i32 %a, i32 0
335 %vsqshrn1 = call <1 x i16> @llvm.aarch64.neon.vsqshrn.v1i16(<1 x i32> %vsqshrn, i32 16)
336 %0 = extractelement <1 x i16> %vsqshrn1, i32 0
340 declare <1 x i16> @llvm.aarch64.neon.vsqshrn.v1i16(<1 x i32>, i32)
342 define i32 @test_vqshrnd_n_s64(i64 %a) {
343 ; CHECK: test_vqshrnd_n_s64
344 ; CHECK: sqshrn {{s[0-9]+}}, {{d[0-9]+}}, #32
346 %vsqshrn = insertelement <1 x i64> undef, i64 %a, i32 0
347 %vsqshrn1 = call <1 x i32> @llvm.aarch64.neon.vsqshrn.v1i32(<1 x i64> %vsqshrn, i32 32)
348 %0 = extractelement <1 x i32> %vsqshrn1, i32 0
352 declare <1 x i32> @llvm.aarch64.neon.vsqshrn.v1i32(<1 x i64>, i32)
354 define i8 @test_vqshrnh_n_u16(i16 %a) {
355 ; CHECK: test_vqshrnh_n_u16
356 ; CHECK: uqshrn {{b[0-9]+}}, {{h[0-9]+}}, #8
358 %vuqshrn = insertelement <1 x i16> undef, i16 %a, i32 0
359 %vuqshrn1 = call <1 x i8> @llvm.aarch64.neon.vuqshrn.v1i8(<1 x i16> %vuqshrn, i32 8)
360 %0 = extractelement <1 x i8> %vuqshrn1, i32 0
364 declare <1 x i8> @llvm.aarch64.neon.vuqshrn.v1i8(<1 x i16>, i32)
366 define i16 @test_vqshrns_n_u32(i32 %a) {
367 ; CHECK: test_vqshrns_n_u32
368 ; CHECK: uqshrn {{h[0-9]+}}, {{s[0-9]+}}, #16
370 %vuqshrn = insertelement <1 x i32> undef, i32 %a, i32 0
371 %vuqshrn1 = call <1 x i16> @llvm.aarch64.neon.vuqshrn.v1i16(<1 x i32> %vuqshrn, i32 16)
372 %0 = extractelement <1 x i16> %vuqshrn1, i32 0
376 declare <1 x i16> @llvm.aarch64.neon.vuqshrn.v1i16(<1 x i32>, i32)
378 define i32 @test_vqshrnd_n_u64(i64 %a) {
379 ; CHECK: test_vqshrnd_n_u64
380 ; CHECK: uqshrn {{s[0-9]+}}, {{d[0-9]+}}, #32
382 %vuqshrn = insertelement <1 x i64> undef, i64 %a, i32 0
383 %vuqshrn1 = call <1 x i32> @llvm.aarch64.neon.vuqshrn.v1i32(<1 x i64> %vuqshrn, i32 32)
384 %0 = extractelement <1 x i32> %vuqshrn1, i32 0
388 declare <1 x i32> @llvm.aarch64.neon.vuqshrn.v1i32(<1 x i64>, i32)
390 define i8 @test_vqrshrnh_n_s16(i16 %a) {
391 ; CHECK: test_vqrshrnh_n_s16
392 ; CHECK: sqrshrn {{b[0-9]+}}, {{h[0-9]+}}, #8
394 %vsqrshrn = insertelement <1 x i16> undef, i16 %a, i32 0
395 %vsqrshrn1 = call <1 x i8> @llvm.aarch64.neon.vsqrshrn.v1i8(<1 x i16> %vsqrshrn, i32 8)
396 %0 = extractelement <1 x i8> %vsqrshrn1, i32 0
400 declare <1 x i8> @llvm.aarch64.neon.vsqrshrn.v1i8(<1 x i16>, i32)
402 define i16 @test_vqrshrns_n_s32(i32 %a) {
403 ; CHECK: test_vqrshrns_n_s32
404 ; CHECK: sqrshrn {{h[0-9]+}}, {{s[0-9]+}}, #16
406 %vsqrshrn = insertelement <1 x i32> undef, i32 %a, i32 0
407 %vsqrshrn1 = call <1 x i16> @llvm.aarch64.neon.vsqrshrn.v1i16(<1 x i32> %vsqrshrn, i32 16)
408 %0 = extractelement <1 x i16> %vsqrshrn1, i32 0
412 declare <1 x i16> @llvm.aarch64.neon.vsqrshrn.v1i16(<1 x i32>, i32)
414 define i32 @test_vqrshrnd_n_s64(i64 %a) {
415 ; CHECK: test_vqrshrnd_n_s64
416 ; CHECK: sqrshrn {{s[0-9]+}}, {{d[0-9]+}}, #32
418 %vsqrshrn = insertelement <1 x i64> undef, i64 %a, i32 0
419 %vsqrshrn1 = call <1 x i32> @llvm.aarch64.neon.vsqrshrn.v1i32(<1 x i64> %vsqrshrn, i32 32)
420 %0 = extractelement <1 x i32> %vsqrshrn1, i32 0
424 declare <1 x i32> @llvm.aarch64.neon.vsqrshrn.v1i32(<1 x i64>, i32)
426 define i8 @test_vqrshrnh_n_u16(i16 %a) {
427 ; CHECK: test_vqrshrnh_n_u16
428 ; CHECK: uqrshrn {{b[0-9]+}}, {{h[0-9]+}}, #8
430 %vuqrshrn = insertelement <1 x i16> undef, i16 %a, i32 0
431 %vuqrshrn1 = call <1 x i8> @llvm.aarch64.neon.vuqrshrn.v1i8(<1 x i16> %vuqrshrn, i32 8)
432 %0 = extractelement <1 x i8> %vuqrshrn1, i32 0
436 declare <1 x i8> @llvm.aarch64.neon.vuqrshrn.v1i8(<1 x i16>, i32)
438 define i16 @test_vqrshrns_n_u32(i32 %a) {
439 ; CHECK: test_vqrshrns_n_u32
440 ; CHECK: uqrshrn {{h[0-9]+}}, {{s[0-9]+}}, #16
442 %vuqrshrn = insertelement <1 x i32> undef, i32 %a, i32 0
443 %vuqrshrn1 = call <1 x i16> @llvm.aarch64.neon.vuqrshrn.v1i16(<1 x i32> %vuqrshrn, i32 16)
444 %0 = extractelement <1 x i16> %vuqrshrn1, i32 0
448 declare <1 x i16> @llvm.aarch64.neon.vuqrshrn.v1i16(<1 x i32>, i32)
450 define i32 @test_vqrshrnd_n_u64(i64 %a) {
451 ; CHECK: test_vqrshrnd_n_u64
452 ; CHECK: uqrshrn {{s[0-9]+}}, {{d[0-9]+}}, #32
454 %vuqrshrn = insertelement <1 x i64> undef, i64 %a, i32 0
455 %vuqrshrn1 = call <1 x i32> @llvm.aarch64.neon.vuqrshrn.v1i32(<1 x i64> %vuqrshrn, i32 32)
456 %0 = extractelement <1 x i32> %vuqrshrn1, i32 0
460 declare <1 x i32> @llvm.aarch64.neon.vuqrshrn.v1i32(<1 x i64>, i32)
462 define i8 @test_vqshrunh_n_s16(i16 %a) {
463 ; CHECK: test_vqshrunh_n_s16
464 ; CHECK: sqshrun {{b[0-9]+}}, {{h[0-9]+}}, #8
466 %vsqshrun = insertelement <1 x i16> undef, i16 %a, i32 0
467 %vsqshrun1 = call <1 x i8> @llvm.aarch64.neon.vsqshrun.v1i8(<1 x i16> %vsqshrun, i32 8)
468 %0 = extractelement <1 x i8> %vsqshrun1, i32 0
472 declare <1 x i8> @llvm.aarch64.neon.vsqshrun.v1i8(<1 x i16>, i32)
474 define i16 @test_vqshruns_n_s32(i32 %a) {
475 ; CHECK: test_vqshruns_n_s32
476 ; CHECK: sqshrun {{h[0-9]+}}, {{s[0-9]+}}, #16
478 %vsqshrun = insertelement <1 x i32> undef, i32 %a, i32 0
479 %vsqshrun1 = call <1 x i16> @llvm.aarch64.neon.vsqshrun.v1i16(<1 x i32> %vsqshrun, i32 16)
480 %0 = extractelement <1 x i16> %vsqshrun1, i32 0
484 declare <1 x i16> @llvm.aarch64.neon.vsqshrun.v1i16(<1 x i32>, i32)
486 define i32 @test_vqshrund_n_s64(i64 %a) {
487 ; CHECK: test_vqshrund_n_s64
488 ; CHECK: sqshrun {{s[0-9]+}}, {{d[0-9]+}}, #32
490 %vsqshrun = insertelement <1 x i64> undef, i64 %a, i32 0
491 %vsqshrun1 = call <1 x i32> @llvm.aarch64.neon.vsqshrun.v1i32(<1 x i64> %vsqshrun, i32 32)
492 %0 = extractelement <1 x i32> %vsqshrun1, i32 0
496 declare <1 x i32> @llvm.aarch64.neon.vsqshrun.v1i32(<1 x i64>, i32)
498 define i8 @test_vqrshrunh_n_s16(i16 %a) {
499 ; CHECK: test_vqrshrunh_n_s16
500 ; CHECK: sqrshrun {{b[0-9]+}}, {{h[0-9]+}}, #8
502 %vsqrshrun = insertelement <1 x i16> undef, i16 %a, i32 0
503 %vsqrshrun1 = call <1 x i8> @llvm.aarch64.neon.vsqrshrun.v1i8(<1 x i16> %vsqrshrun, i32 8)
504 %0 = extractelement <1 x i8> %vsqrshrun1, i32 0
508 declare <1 x i8> @llvm.aarch64.neon.vsqrshrun.v1i8(<1 x i16>, i32)
510 define i16 @test_vqrshruns_n_s32(i32 %a) {
511 ; CHECK: test_vqrshruns_n_s32
512 ; CHECK: sqrshrun {{h[0-9]+}}, {{s[0-9]+}}, #16
514 %vsqrshrun = insertelement <1 x i32> undef, i32 %a, i32 0
515 %vsqrshrun1 = call <1 x i16> @llvm.aarch64.neon.vsqrshrun.v1i16(<1 x i32> %vsqrshrun, i32 16)
516 %0 = extractelement <1 x i16> %vsqrshrun1, i32 0
520 declare <1 x i16> @llvm.aarch64.neon.vsqrshrun.v1i16(<1 x i32>, i32)
522 define i32 @test_vqrshrund_n_s64(i64 %a) {
523 ; CHECK: test_vqrshrund_n_s64
524 ; CHECK: sqrshrun {{s[0-9]+}}, {{d[0-9]+}}, #32
526 %vsqrshrun = insertelement <1 x i64> undef, i64 %a, i32 0
527 %vsqrshrun1 = call <1 x i32> @llvm.aarch64.neon.vsqrshrun.v1i32(<1 x i64> %vsqrshrun, i32 32)
528 %0 = extractelement <1 x i32> %vsqrshrun1, i32 0
532 declare <1 x i32> @llvm.aarch64.neon.vsqrshrun.v1i32(<1 x i64>, i32)