1 ; RUN: llc -mtriple=arm-eabi -mattr=+v8.2a,+neon,+fullfp16 -float-abi=hard < %s | FileCheck %s
3 %struct.float16x4x2_t = type { [2 x <4 x half>] }
4 %struct.float16x8x2_t = type { [2 x <8 x half>] }
6 define dso_local <4 x half> @test_vabs_f16(<4 x half> %a) {
7 ; CHECKLABEL: test_vabs_f16:
8 ; CHECK: vabs.f16 d0, d0
11 %vabs1.i = tail call <4 x half> @llvm.fabs.v4f16(<4 x half> %a)
12 ret <4 x half> %vabs1.i
15 define dso_local <8 x half> @test_vabsq_f16(<8 x half> %a) {
16 ; CHECKLABEL: test_vabsq_f16:
17 ; CHECK: vabs.f16 q0, q0
20 %vabs1.i = tail call <8 x half> @llvm.fabs.v8f16(<8 x half> %a)
21 ret <8 x half> %vabs1.i
24 define dso_local <4 x i16> @test_vceqz_f16(<4 x half> %a) {
25 ; CHECKLABEL: test_vceqz_f16:
26 ; CHECK: vceq.f16 d0, d0, #0
29 %0 = fcmp oeq <4 x half> %a, zeroinitializer
30 %vceqz.i = sext <4 x i1> %0 to <4 x i16>
31 ret <4 x i16> %vceqz.i
34 define dso_local <8 x i16> @test_vceqzq_f16(<8 x half> %a) {
35 ; CHECKLABEL: test_vceqzq_f16:
36 ; CHECK: vceq.f16 q0, q0, #0
39 %0 = fcmp oeq <8 x half> %a, zeroinitializer
40 %vceqz.i = sext <8 x i1> %0 to <8 x i16>
41 ret <8 x i16> %vceqz.i
44 define dso_local <4 x i16> @test_vcgez_f16(<4 x half> %a) {
45 ; CHECKLABEL: test_vcgez_f16:
46 ; CHECK: vcge.f16 d0, d0, #0
49 %0 = fcmp oge <4 x half> %a, zeroinitializer
50 %vcgez.i = sext <4 x i1> %0 to <4 x i16>
51 ret <4 x i16> %vcgez.i
54 define dso_local <8 x i16> @test_vcgezq_f16(<8 x half> %a) {
55 ; CHECKLABEL: test_vcgezq_f16:
56 ; CHECK: vcge.f16 q0, q0, #0
59 %0 = fcmp oge <8 x half> %a, zeroinitializer
60 %vcgez.i = sext <8 x i1> %0 to <8 x i16>
61 ret <8 x i16> %vcgez.i
64 define dso_local <4 x i16> @test_vcgtz_f16(<4 x half> %a) {
65 ; CHECKLABEL: test_vcgtz_f16:
66 ; CHECK: vcgt.f16 d0, d0, #0
69 %0 = fcmp ogt <4 x half> %a, zeroinitializer
70 %vcgtz.i = sext <4 x i1> %0 to <4 x i16>
71 ret <4 x i16> %vcgtz.i
74 define dso_local <8 x i16> @test_vcgtzq_f16(<8 x half> %a) {
75 ; CHECKLABEL: test_vcgtzq_f16:
76 ; CHECK: vcgt.f16 q0, q0, #0
79 %0 = fcmp ogt <8 x half> %a, zeroinitializer
80 %vcgtz.i = sext <8 x i1> %0 to <8 x i16>
81 ret <8 x i16> %vcgtz.i
84 define dso_local <4 x i16> @test_vclez_f16(<4 x half> %a) {
85 ; CHECKLABEL: test_vclez_f16:
86 ; CHECK: vcle.f16 d0, d0, #0
89 %0 = fcmp ole <4 x half> %a, zeroinitializer
90 %vclez.i = sext <4 x i1> %0 to <4 x i16>
91 ret <4 x i16> %vclez.i
94 define dso_local <8 x i16> @test_vclezq_f16(<8 x half> %a) {
95 ; CHECKLABEL: test_vclezq_f16:
96 ; CHECK: vcle.f16 q0, q0, #0
99 %0 = fcmp ole <8 x half> %a, zeroinitializer
100 %vclez.i = sext <8 x i1> %0 to <8 x i16>
101 ret <8 x i16> %vclez.i
104 define dso_local <4 x i16> @test_vcltz_f16(<4 x half> %a) {
105 ; CHECKLABEL: test_vcltz_f16:
106 ; CHECK: vclt.f16 d0, d0, #0
109 %0 = fcmp olt <4 x half> %a, zeroinitializer
110 %vcltz.i = sext <4 x i1> %0 to <4 x i16>
111 ret <4 x i16> %vcltz.i
114 define dso_local <8 x i16> @test_vcltzq_f16(<8 x half> %a) {
115 ; CHECKLABEL: test_vcltzq_f16:
116 ; CHECK: vclt.f16 q0, q0, #0
119 %0 = fcmp olt <8 x half> %a, zeroinitializer
120 %vcltz.i = sext <8 x i1> %0 to <8 x i16>
121 ret <8 x i16> %vcltz.i
124 define dso_local <4 x half> @test_vcvt_f16_s16(<4 x i16> %a) {
125 ; CHECK-LABEL: test_vcvt_f16_s16:
126 ; CHECK: vcvt.f16.s16 d0, d0
129 %vcvt.i = sitofp <4 x i16> %a to <4 x half>
130 ret <4 x half> %vcvt.i
133 define dso_local <8 x half> @test_vcvtq_f16_s16(<8 x i16> %a) {
134 ; CHECK-LABEL: test_vcvtq_f16_s16:
135 ; CHECK: vcvt.f16.s16 q0, q0
138 %vcvt.i = sitofp <8 x i16> %a to <8 x half>
139 ret <8 x half> %vcvt.i
142 define dso_local <4 x half> @test_vcvt_f16_u16(<4 x i16> %a) {
143 ; CHECK-LABEL: test_vcvt_f16_u16:
144 ; CHECK: vcvt.f16.u16 d0, d0
147 %vcvt.i = uitofp <4 x i16> %a to <4 x half>
148 ret <4 x half> %vcvt.i
151 define dso_local <8 x half> @test_vcvtq_f16_u16(<8 x i16> %a) {
152 ; CHECK-LABEL: test_vcvtq_f16_u16:
153 ; CHECK: vcvt.f16.u16 q0, q0
156 %vcvt.i = uitofp <8 x i16> %a to <8 x half>
157 ret <8 x half> %vcvt.i
160 define dso_local <4 x i16> @test_vcvt_s16_f16(<4 x half> %a) {
161 ; CHECK-LABEL: test_vcvt_s16_f16:
162 ; CHECK: vcvt.s16.f16 d0, d0
165 %vcvt.i = fptosi <4 x half> %a to <4 x i16>
166 ret <4 x i16> %vcvt.i
169 define dso_local <8 x i16> @test_vcvtq_s16_f16(<8 x half> %a) {
170 ; CHECK-LABEL: test_vcvtq_s16_f16:
171 ; CHECK: vcvt.s16.f16 q0, q0
174 %vcvt.i = fptosi <8 x half> %a to <8 x i16>
175 ret <8 x i16> %vcvt.i
178 define dso_local <4 x i16> @test_vcvt_u16_f16(<4 x half> %a) {
179 ; CHECK-LABEL: test_vcvt_u16_f16:
180 ; CHECK: vcvt.u16.f16 d0, d0
183 %vcvt.i = fptoui <4 x half> %a to <4 x i16>
184 ret <4 x i16> %vcvt.i
187 define dso_local <8 x i16> @test_vcvtq_u16_f16(<8 x half> %a) {
188 ; CHECK-LABEL: test_vcvtq_u16_f16:
189 ; CHECK: vcvt.u16.f16 q0, q0
192 %vcvt.i = fptoui <8 x half> %a to <8 x i16>
193 ret <8 x i16> %vcvt.i
196 define dso_local <4 x i16> @test_vcvta_s16_f16(<4 x half> %a) {
197 ; CHECK-LABEL: test_vcvta_s16_f16:
198 ; CHECK: vcvta.s16.f16 d0, d0
201 %vcvta_s16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtas.v4i16.v4f16(<4 x half> %a)
202 ret <4 x i16> %vcvta_s16_v1.i
205 define dso_local <4 x i16> @test_vcvta_u16_f16(<4 x half> %a) {
206 ; CHECK-LABEL: test_vcvta_u16_f16:
207 ; CHECK: vcvta.u16.f16 d0, d0
210 %vcvta_u16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtau.v4i16.v4f16(<4 x half> %a)
211 ret <4 x i16> %vcvta_u16_v1.i
214 define dso_local <8 x i16> @test_vcvtaq_s16_f16(<8 x half> %a) {
215 ; CHECK-LABEL: test_vcvtaq_s16_f16:
216 ; CHECK: vcvta.s16.f16 q0, q0
219 %vcvtaq_s16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtas.v8i16.v8f16(<8 x half> %a)
220 ret <8 x i16> %vcvtaq_s16_v1.i
223 define dso_local <4 x i16> @test_vcvtm_s16_f16(<4 x half> %a) {
224 ; CHECK-LABEL: test_vcvtm_s16_f16:
225 ; CHECK: vcvtm.s16.f16 d0, d0
228 %vcvtm_s16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtms.v4i16.v4f16(<4 x half> %a)
229 ret <4 x i16> %vcvtm_s16_v1.i
232 define dso_local <8 x i16> @test_vcvtmq_s16_f16(<8 x half> %a) {
233 ; CHECK-LABEL: test_vcvtmq_s16_f16:
234 ; CHECK: vcvtm.s16.f16 q0, q0
237 %vcvtmq_s16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtms.v8i16.v8f16(<8 x half> %a)
238 ret <8 x i16> %vcvtmq_s16_v1.i
241 define dso_local <4 x i16> @test_vcvtm_u16_f16(<4 x half> %a) {
242 ; CHECK-LABEL: test_vcvtm_u16_f16:
243 ; CHECK: vcvtm.u16.f16 d0, d0
246 %vcvtm_u16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtmu.v4i16.v4f16(<4 x half> %a)
247 ret <4 x i16> %vcvtm_u16_v1.i
250 define dso_local <8 x i16> @test_vcvtmq_u16_f16(<8 x half> %a) {
251 ; CHECK-LABEL: test_vcvtmq_u16_f16:
252 ; CHECK: vcvtm.u16.f16 q0, q0
255 %vcvtmq_u16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtmu.v8i16.v8f16(<8 x half> %a)
256 ret <8 x i16> %vcvtmq_u16_v1.i
259 define dso_local <4 x i16> @test_vcvtn_s16_f16(<4 x half> %a) {
260 ; CHECK-LABEL: test_vcvtn_s16_f16:
261 ; CHECK: vcvtn.s16.f16 d0, d0
264 %vcvtn_s16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtns.v4i16.v4f16(<4 x half> %a)
265 ret <4 x i16> %vcvtn_s16_v1.i
268 define dso_local <8 x i16> @test_vcvtnq_s16_f16(<8 x half> %a) {
269 ; CHECK-LABEL: test_vcvtnq_s16_f16:
270 ; CHECK: vcvtn.s16.f16 q0, q0
273 %vcvtnq_s16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtns.v8i16.v8f16(<8 x half> %a)
274 ret <8 x i16> %vcvtnq_s16_v1.i
277 define dso_local <4 x i16> @test_vcvtn_u16_f16(<4 x half> %a) {
278 ; CHECK-LABEL: test_vcvtn_u16_f16:
279 ; CHECK: vcvtn.u16.f16 d0, d0
282 %vcvtn_u16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtnu.v4i16.v4f16(<4 x half> %a)
283 ret <4 x i16> %vcvtn_u16_v1.i
286 define dso_local <8 x i16> @test_vcvtnq_u16_f16(<8 x half> %a) {
287 ; CHECK-LABEL: test_vcvtnq_u16_f16:
288 ; CHECK: vcvtn.u16.f16 q0, q0
291 %vcvtnq_u16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtnu.v8i16.v8f16(<8 x half> %a)
292 ret <8 x i16> %vcvtnq_u16_v1.i
295 define dso_local <4 x i16> @test_vcvtp_s16_f16(<4 x half> %a) {
296 ; CHECK-LABEL: test_vcvtp_s16_f16:
297 ; CHECK: vcvtp.s16.f16 d0, d0
300 %vcvtp_s16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtps.v4i16.v4f16(<4 x half> %a)
301 ret <4 x i16> %vcvtp_s16_v1.i
304 define dso_local <8 x i16> @test_vcvtpq_s16_f16(<8 x half> %a) {
305 ; CHECK-LABEL: test_vcvtpq_s16_f16:
306 ; CHECK: vcvtp.s16.f16 q0, q0
309 %vcvtpq_s16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtps.v8i16.v8f16(<8 x half> %a)
310 ret <8 x i16> %vcvtpq_s16_v1.i
313 define dso_local <4 x i16> @test_vcvtp_u16_f16(<4 x half> %a) {
314 ; CHECK-LABEL: test_vcvtp_u16_f16:
315 ; CHECK: vcvtp.u16.f16 d0, d0
318 %vcvtp_u16_v1.i = tail call <4 x i16> @llvm.arm.neon.vcvtpu.v4i16.v4f16(<4 x half> %a)
319 ret <4 x i16> %vcvtp_u16_v1.i
322 define dso_local <8 x i16> @test_vcvtpq_u16_f16(<8 x half> %a) {
323 ; CHECK-LABEL: test_vcvtpq_u16_f16:
324 ; CHECK: vcvtp.u16.f16 q0, q0
327 %vcvtpq_u16_v1.i = tail call <8 x i16> @llvm.arm.neon.vcvtpu.v8i16.v8f16(<8 x half> %a)
328 ret <8 x i16> %vcvtpq_u16_v1.i
331 define dso_local <4 x half> @test_vneg_f16(<4 x half> %a) {
332 ; CHECKLABEL: test_vneg_f16:
333 ; CHECK: vneg.f16 d0, d0
336 %sub.i = fsub <4 x half> <half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000>, %a
337 ret <4 x half> %sub.i
340 define dso_local <8 x half> @test_vnegq_f16(<8 x half> %a) {
341 ; CHECKLABEL: test_vnegq_f16:
342 ; CHECK: vneg.f16 q0, q0
345 %sub.i = fsub <8 x half> <half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000>, %a
346 ret <8 x half> %sub.i
349 define dso_local <4 x half> @test_vrecpe_f16(<4 x half> %a) {
350 ; CHECKLABEL: test_vrecpe_f16:
351 ; CHECK: vrecpe.f16 d0, d0
354 %vrecpe_v1.i = tail call <4 x half> @llvm.arm.neon.vrecpe.v4f16(<4 x half> %a)
355 ret <4 x half> %vrecpe_v1.i
358 define dso_local <8 x half> @test_vrecpeq_f16(<8 x half> %a) {
359 ; CHECKLABEL: test_vrecpeq_f16:
360 ; CHECK: vrecpe.f16 q0, q0
363 %vrecpeq_v1.i = tail call <8 x half> @llvm.arm.neon.vrecpe.v8f16(<8 x half> %a)
364 ret <8 x half> %vrecpeq_v1.i
367 define dso_local <4 x half> @test_vrnd_f16(<4 x half> %a) {
368 ; CHECKLABEL: test_vrnd_f16:
369 ; CHECK: vrintz.f16 d0, d0
372 %vrnd_v1.i = tail call <4 x half> @llvm.arm.neon.vrintz.v4f16(<4 x half> %a)
373 ret <4 x half> %vrnd_v1.i
376 define dso_local <8 x half> @test_vrndq_f16(<8 x half> %a) {
377 ; CHECKLABEL: test_vrndq_f16:
378 ; CHECK: vrintz.f16 q0, q0
381 %vrndq_v1.i = tail call <8 x half> @llvm.arm.neon.vrintz.v8f16(<8 x half> %a)
382 ret <8 x half> %vrndq_v1.i
385 define dso_local <4 x half> @test_vrnda_f16(<4 x half> %a) {
386 ; CHECKLABEL: test_vrnda_f16:
387 ; CHECK: vrinta.f16 d0, d0
390 %vrnda_v1.i = tail call <4 x half> @llvm.arm.neon.vrinta.v4f16(<4 x half> %a)
391 ret <4 x half> %vrnda_v1.i
394 define dso_local <8 x half> @test_vrndaq_f16(<8 x half> %a) {
395 ; CHECKLABEL: test_vrndaq_f16:
396 ; CHECK: vrinta.f16 q0, q0
399 %vrndaq_v1.i = tail call <8 x half> @llvm.arm.neon.vrinta.v8f16(<8 x half> %a)
400 ret <8 x half> %vrndaq_v1.i
403 define dso_local <4 x half> @test_vrndm_f16(<4 x half> %a) {
404 ; CHECKLABEL: test_vrndm_f16:
405 ; CHECK: vrintm.f16 d0, d0
408 %vrndm_v1.i = tail call <4 x half> @llvm.arm.neon.vrintm.v4f16(<4 x half> %a)
409 ret <4 x half> %vrndm_v1.i
412 define dso_local <8 x half> @test_vrndmq_f16(<8 x half> %a) {
413 ; CHECKLABEL: test_vrndmq_f16:
414 ; CHECK: vrintm.f16 q0, q0
417 %vrndmq_v1.i = tail call <8 x half> @llvm.arm.neon.vrintm.v8f16(<8 x half> %a)
418 ret <8 x half> %vrndmq_v1.i
421 define dso_local <4 x half> @test_vrndn_f16(<4 x half> %a) {
422 ; CHECKLABEL: test_vrndn_f16:
423 ; CHECK: vrintn.f16 d0, d0
426 %vrndn_v1.i = tail call <4 x half> @llvm.arm.neon.vrintn.v4f16(<4 x half> %a)
427 ret <4 x half> %vrndn_v1.i
430 define dso_local <8 x half> @test_vrndnq_f16(<8 x half> %a) {
431 ; CHECKLABEL: test_vrndnq_f16:
432 ; CHECK: vrintn.f16 q0, q0
435 %vrndnq_v1.i = tail call <8 x half> @llvm.arm.neon.vrintn.v8f16(<8 x half> %a)
436 ret <8 x half> %vrndnq_v1.i
439 define dso_local <4 x half> @test_vrndp_f16(<4 x half> %a) {
440 ; CHECKLABEL: test_vrndp_f16:
441 ; CHECK: vrintp.f16 d0, d0
444 %vrndp_v1.i = tail call <4 x half> @llvm.arm.neon.vrintp.v4f16(<4 x half> %a)
445 ret <4 x half> %vrndp_v1.i
448 define dso_local <8 x half> @test_vrndpq_f16(<8 x half> %a) {
449 ; CHECKLABEL: test_vrndpq_f16:
450 ; CHECK: vrintp.f16 q0, q0
453 %vrndpq_v1.i = tail call <8 x half> @llvm.arm.neon.vrintp.v8f16(<8 x half> %a)
454 ret <8 x half> %vrndpq_v1.i
457 define dso_local <4 x half> @test_vrndx_f16(<4 x half> %a) {
458 ; CHECKLABEL: test_vrndx_f16:
459 ; CHECK: vrintx.f16 d0, d0
462 %vrndx_v1.i = tail call <4 x half> @llvm.arm.neon.vrintx.v4f16(<4 x half> %a)
463 ret <4 x half> %vrndx_v1.i
466 define dso_local <8 x half> @test_vrndxq_f16(<8 x half> %a) {
467 ; CHECKLABEL: test_vrndxq_f16:
468 ; CHECK: vrintx.f16 q0, q0
471 %vrndxq_v1.i = tail call <8 x half> @llvm.arm.neon.vrintx.v8f16(<8 x half> %a)
472 ret <8 x half> %vrndxq_v1.i
475 define dso_local <4 x half> @test_vrsqrte_f16(<4 x half> %a) {
476 ; CHECKLABEL: test_vrsqrte_f16:
477 ; CHECK: vrsqrte.f16 d0, d0
480 %vrsqrte_v1.i = tail call <4 x half> @llvm.arm.neon.vrsqrte.v4f16(<4 x half> %a)
481 ret <4 x half> %vrsqrte_v1.i
484 define dso_local <8 x half> @test_vrsqrteq_f16(<8 x half> %a) {
485 ; CHECKLABEL: test_vrsqrteq_f16:
486 ; CHECK: vrsqrte.f16 q0, q0
489 %vrsqrteq_v1.i = tail call <8 x half> @llvm.arm.neon.vrsqrte.v8f16(<8 x half> %a)
490 ret <8 x half> %vrsqrteq_v1.i
493 define dso_local <4 x half> @test_vadd_f16(<4 x half> %a, <4 x half> %b) {
494 ; CHECKLABEL: test_vadd_f16:
495 ; CHECK: vadd.f16 d0, d0, d1
498 %add.i = fadd <4 x half> %a, %b
499 ret <4 x half> %add.i
502 define dso_local <8 x half> @test_vaddq_f16(<8 x half> %a, <8 x half> %b) {
503 ; CHECKLABEL: test_vaddq_f16:
504 ; CHECK: vadd.f16 q0, q0, q1
507 %add.i = fadd <8 x half> %a, %b
508 ret <8 x half> %add.i
511 define dso_local <4 x half> @test_vabd_f16(<4 x half> %a, <4 x half> %b) {
512 ; CHECKLABEL: test_vabd_f16:
513 ; CHECK: vabd.f16 d0, d0, d1
516 %vabd_v2.i = tail call <4 x half> @llvm.arm.neon.vabds.v4f16(<4 x half> %a, <4 x half> %b)
517 ret <4 x half> %vabd_v2.i
520 define dso_local <8 x half> @test_vabdq_f16(<8 x half> %a, <8 x half> %b) {
521 ; CHECKLABEL: test_vabdq_f16:
522 ; CHECK: vabd.f16 q0, q0, q1
525 %vabdq_v2.i = tail call <8 x half> @llvm.arm.neon.vabds.v8f16(<8 x half> %a, <8 x half> %b)
526 ret <8 x half> %vabdq_v2.i
529 define dso_local <4 x i16> @test_vcage_f16(<4 x half> %a, <4 x half> %b) {
530 ; CHECKLABEL: test_vcage_f16:
531 ; CHECK: vacge.f16 d0, d0, d1
534 %vcage_v2.i = tail call <4 x i16> @llvm.arm.neon.vacge.v4i16.v4f16(<4 x half> %a, <4 x half> %b)
535 ret <4 x i16> %vcage_v2.i
538 define dso_local <8 x i16> @test_vcageq_f16(<8 x half> %a, <8 x half> %b) {
539 ; CHECKLABEL: test_vcageq_f16:
540 ; CHECK: vacge.f16 q0, q0, q1
543 %vcageq_v2.i = tail call <8 x i16> @llvm.arm.neon.vacge.v8i16.v8f16(<8 x half> %a, <8 x half> %b)
544 ret <8 x i16> %vcageq_v2.i
547 define dso_local <4 x i16> @test_vcagt_f16(<4 x half> %a, <4 x half> %b) {
548 ; CHECK-LABEL: test_vcagt_f16:
549 ; CHECK: vacgt.f16 d0, d0, d1
552 %vcagt_v2.i = tail call <4 x i16> @llvm.arm.neon.vacgt.v4i16.v4f16(<4 x half> %a, <4 x half> %b)
553 ret <4 x i16> %vcagt_v2.i
556 define dso_local <8 x i16> @test_vcagtq_f16(<8 x half> %a, <8 x half> %b) {
557 ; CHECK-LABEL: test_vcagtq_f16:
558 ; CHECK: vacgt.f16 q0, q0, q1
561 %vcagtq_v2.i = tail call <8 x i16> @llvm.arm.neon.vacgt.v8i16.v8f16(<8 x half> %a, <8 x half> %b)
562 ret <8 x i16> %vcagtq_v2.i
565 define dso_local <4 x i16> @test_vcale_f16(<4 x half> %a, <4 x half> %b) {
566 ; CHECKLABEL: test_vcale_f16:
567 ; CHECK: vacge.f16 d0, d1, d0
570 %vcale_v2.i = tail call <4 x i16> @llvm.arm.neon.vacge.v4i16.v4f16(<4 x half> %b, <4 x half> %a)
571 ret <4 x i16> %vcale_v2.i
574 define dso_local <8 x i16> @test_vcaleq_f16(<8 x half> %a, <8 x half> %b) {
575 ; CHECKLABEL: test_vcaleq_f16:
576 ; CHECK: vacge.f16 q0, q1, q0
579 %vcaleq_v2.i = tail call <8 x i16> @llvm.arm.neon.vacge.v8i16.v8f16(<8 x half> %b, <8 x half> %a)
580 ret <8 x i16> %vcaleq_v2.i
583 define dso_local <4 x i16> @test_vceq_f16(<4 x half> %a, <4 x half> %b) {
584 ; CHECKLABEL: test_vceq_f16:
585 ; CHECK: vceq.f16 d0, d0, d1
588 %cmp.i = fcmp oeq <4 x half> %a, %b
589 %sext.i = sext <4 x i1> %cmp.i to <4 x i16>
590 ret <4 x i16> %sext.i
593 define dso_local <8 x i16> @test_vceqq_f16(<8 x half> %a, <8 x half> %b) {
594 ; CHECKLABEL: test_vceqq_f16:
595 ; CHECK: vceq.f16 q0, q0, q1
598 %cmp.i = fcmp oeq <8 x half> %a, %b
599 %sext.i = sext <8 x i1> %cmp.i to <8 x i16>
600 ret <8 x i16> %sext.i
603 define dso_local <4 x i16> @test_vcge_f16(<4 x half> %a, <4 x half> %b) {
604 ; CHECKLABEL: test_vcge_f16:
605 ; CHECK: vcge.f16 d0, d0, d1
608 %cmp.i = fcmp oge <4 x half> %a, %b
609 %sext.i = sext <4 x i1> %cmp.i to <4 x i16>
610 ret <4 x i16> %sext.i
613 define dso_local <8 x i16> @test_vcgeq_f16(<8 x half> %a, <8 x half> %b) {
614 ; CHECKLABEL: test_vcgeq_f16:
615 ; CHECK: vcge.f16 q0, q0, q1
618 %cmp.i = fcmp oge <8 x half> %a, %b
619 %sext.i = sext <8 x i1> %cmp.i to <8 x i16>
620 ret <8 x i16> %sext.i
623 define dso_local <4 x i16> @test_vcgt_f16(<4 x half> %a, <4 x half> %b) {
624 ; CHECKLABEL: test_vcgt_f16:
625 ; CHECK: vcgt.f16 d0, d0, d1
628 %cmp.i = fcmp ogt <4 x half> %a, %b
629 %sext.i = sext <4 x i1> %cmp.i to <4 x i16>
630 ret <4 x i16> %sext.i
633 define dso_local <8 x i16> @test_vcgtq_f16(<8 x half> %a, <8 x half> %b) {
634 ; CHECKLABEL: test_vcgtq_f16:
635 ; CHECK: vcgt.f16 q0, q0, q1
638 %cmp.i = fcmp ogt <8 x half> %a, %b
639 %sext.i = sext <8 x i1> %cmp.i to <8 x i16>
640 ret <8 x i16> %sext.i
643 define dso_local <4 x i16> @test_vcle_f16(<4 x half> %a, <4 x half> %b) {
644 ; CHECKLABEL: test_vcle_f16:
645 ; CHECK: vcge.f16 d0, d1, d0
648 %cmp.i = fcmp ole <4 x half> %a, %b
649 %sext.i = sext <4 x i1> %cmp.i to <4 x i16>
650 ret <4 x i16> %sext.i
653 define dso_local <8 x i16> @test_vcleq_f16(<8 x half> %a, <8 x half> %b) {
654 ; CHECKLABEL: test_vcleq_f16:
655 ; CHECK: vcge.f16 q0, q1, q0
658 %cmp.i = fcmp ole <8 x half> %a, %b
659 %sext.i = sext <8 x i1> %cmp.i to <8 x i16>
660 ret <8 x i16> %sext.i
663 define dso_local <4 x i16> @test_vclt_f16(<4 x half> %a, <4 x half> %b) {
664 ; CHECKLABEL: test_vclt_f16:
665 ; CHECK: vcgt.f16 d0, d1, d0
668 %cmp.i = fcmp olt <4 x half> %a, %b
669 %sext.i = sext <4 x i1> %cmp.i to <4 x i16>
670 ret <4 x i16> %sext.i
673 define dso_local <8 x i16> @test_vcltq_f16(<8 x half> %a, <8 x half> %b) {
674 ; CHECKLABEL: test_vcltq_f16:
675 ; CHECK: vcgt.f16 q0, q1, q0
678 %cmp.i = fcmp olt <8 x half> %a, %b
679 %sext.i = sext <8 x i1> %cmp.i to <8 x i16>
680 ret <8 x i16> %sext.i
683 define dso_local <4 x half> @test_vcvt_n_f16_s16(<4 x i16> %a) {
684 ; CHECKLABEL: test_vcvt_n_f16_s16:
685 ; CHECK: vcvt.f16.s16 d0, d0, #2
688 %vcvt_n1 = tail call <4 x half> @llvm.arm.neon.vcvtfxs2fp.v4f16.v4i16(<4 x i16> %a, i32 2)
689 ret <4 x half> %vcvt_n1
692 declare <4 x half> @llvm.arm.neon.vcvtfxs2fp.v4f16.v4i16(<4 x i16>, i32) #2
694 define dso_local <8 x half> @test_vcvtq_n_f16_s16(<8 x i16> %a) {
695 ; CHECKLABEL: test_vcvtq_n_f16_s16:
696 ; CHECK: vcvt.f16.s16 q0, q0, #2
699 %vcvt_n1 = tail call <8 x half> @llvm.arm.neon.vcvtfxs2fp.v8f16.v8i16(<8 x i16> %a, i32 2)
700 ret <8 x half> %vcvt_n1
703 declare <8 x half> @llvm.arm.neon.vcvtfxs2fp.v8f16.v8i16(<8 x i16>, i32) #2
705 define dso_local <4 x half> @test_vcvt_n_f16_u16(<4 x i16> %a) {
706 ; CHECKLABEL: test_vcvt_n_f16_u16:
707 ; CHECK: vcvt.f16.u16 d0, d0, #2
710 %vcvt_n1 = tail call <4 x half> @llvm.arm.neon.vcvtfxu2fp.v4f16.v4i16(<4 x i16> %a, i32 2)
711 ret <4 x half> %vcvt_n1
714 declare <4 x half> @llvm.arm.neon.vcvtfxu2fp.v4f16.v4i16(<4 x i16>, i32) #2
716 define dso_local <8 x half> @test_vcvtq_n_f16_u16(<8 x i16> %a) {
717 ; CHECKLABEL: test_vcvtq_n_f16_u16:
718 ; CHECK: vcvt.f16.u16 q0, q0, #2
721 %vcvt_n1 = tail call <8 x half> @llvm.arm.neon.vcvtfxu2fp.v8f16.v8i16(<8 x i16> %a, i32 2)
722 ret <8 x half> %vcvt_n1
725 declare <8 x half> @llvm.arm.neon.vcvtfxu2fp.v8f16.v8i16(<8 x i16>, i32) #2
727 define dso_local <4 x i16> @test_vcvt_n_s16_f16(<4 x half> %a) {
728 ; CHECKLABEL: test_vcvt_n_s16_f16:
729 ; CHECK: vcvt.s16.f16 d0, d0, #2
732 %vcvt_n1 = tail call <4 x i16> @llvm.arm.neon.vcvtfp2fxs.v4i16.v4f16(<4 x half> %a, i32 2)
733 ret <4 x i16> %vcvt_n1
736 declare <4 x i16> @llvm.arm.neon.vcvtfp2fxs.v4i16.v4f16(<4 x half>, i32) #2
738 define dso_local <8 x i16> @test_vcvtq_n_s16_f16(<8 x half> %a) {
739 ; CHECKLABEL: test_vcvtq_n_s16_f16:
740 ; CHECK: vcvt.s16.f16 q0, q0, #2
743 %vcvt_n1 = tail call <8 x i16> @llvm.arm.neon.vcvtfp2fxs.v8i16.v8f16(<8 x half> %a, i32 2)
744 ret <8 x i16> %vcvt_n1
747 declare <8 x i16> @llvm.arm.neon.vcvtfp2fxs.v8i16.v8f16(<8 x half>, i32) #2
749 define dso_local <4 x i16> @test_vcvt_n_u16_f16(<4 x half> %a) {
750 ; CHECKLABEL: test_vcvt_n_u16_f16:
751 ; CHECK: vcvt.u16.f16 d0, d0, #2
754 %vcvt_n1 = tail call <4 x i16> @llvm.arm.neon.vcvtfp2fxu.v4i16.v4f16(<4 x half> %a, i32 2)
755 ret <4 x i16> %vcvt_n1
758 declare <4 x i16> @llvm.arm.neon.vcvtfp2fxu.v4i16.v4f16(<4 x half>, i32) #2
760 define dso_local <8 x i16> @test_vcvtq_n_u16_f16(<8 x half> %a) {
761 ; CHECKLABEL: test_vcvtq_n_u16_f16:
762 ; CHECK: vcvt.u16.f16 q0, q0, #2
765 %vcvt_n1 = tail call <8 x i16> @llvm.arm.neon.vcvtfp2fxu.v8i16.v8f16(<8 x half> %a, i32 2)
766 ret <8 x i16> %vcvt_n1
769 declare <8 x i16> @llvm.arm.neon.vcvtfp2fxu.v8i16.v8f16(<8 x half>, i32) #2
771 define dso_local <4 x half> @test_vmax_f16(<4 x half> %a, <4 x half> %b) {
772 ; CHECKLABEL: test_vmax_f16:
773 ; CHECK: vmax.f16 d0, d0, d1
776 %vmax_v2.i = tail call <4 x half> @llvm.arm.neon.vmaxs.v4f16(<4 x half> %a, <4 x half> %b)
777 ret <4 x half> %vmax_v2.i
780 define dso_local <8 x half> @test_vmaxq_f16(<8 x half> %a, <8 x half> %b) {
781 ; CHECKLABEL: test_vmaxq_f16:
782 ; CHECK: vmax.f16 q0, q0, q1
785 %vmaxq_v2.i = tail call <8 x half> @llvm.arm.neon.vmaxs.v8f16(<8 x half> %a, <8 x half> %b)
786 ret <8 x half> %vmaxq_v2.i
789 define dso_local <4 x half> @test_vmaxnm_f16(<4 x half> %a, <4 x half> %b) {
790 ; CHECK-LABEL: test_vmaxnm_f16:
791 ; CHECK: vmaxnm.f16 d0, d0, d1
794 %vmaxnm_v2.i = tail call <4 x half> @llvm.arm.neon.vmaxnm.v4f16(<4 x half> %a, <4 x half> %b)
795 ret <4 x half> %vmaxnm_v2.i
798 define dso_local <8 x half> @test_vmaxnmq_f16(<8 x half> %a, <8 x half> %b) {
799 ; CHECK-LABEL: test_vmaxnmq_f16:
800 ; CHECK: vmaxnm.f16 q0, q0, q1
803 %vmaxnmq_v2.i = tail call <8 x half> @llvm.arm.neon.vmaxnm.v8f16(<8 x half> %a, <8 x half> %b)
804 ret <8 x half> %vmaxnmq_v2.i
807 define dso_local <4 x half> @test_vmin_f16(<4 x half> %a, <4 x half> %b) {
808 ; CHECK-LABEL: test_vmin_f16:
809 ; CHECK: vmin.f16 d0, d0, d1
812 %vmin_v2.i = tail call <4 x half> @llvm.arm.neon.vmins.v4f16(<4 x half> %a, <4 x half> %b)
813 ret <4 x half> %vmin_v2.i
816 define dso_local <8 x half> @test_vminq_f16(<8 x half> %a, <8 x half> %b) {
817 ; CHECK-LABEL: test_vminq_f16:
818 ; CHECK: vmin.f16 q0, q0, q1
821 %vminq_v2.i = tail call <8 x half> @llvm.arm.neon.vmins.v8f16(<8 x half> %a, <8 x half> %b)
822 ret <8 x half> %vminq_v2.i
825 define dso_local <4 x half> @test_vminnm_f16(<4 x half> %a, <4 x half> %b) {
826 ; CHECK-LABEL: test_vminnm_f16:
827 ; CHECK: vminnm.f16 d0, d0, d1
830 %vminnm_v2.i = tail call <4 x half> @llvm.arm.neon.vminnm.v4f16(<4 x half> %a, <4 x half> %b)
831 ret <4 x half> %vminnm_v2.i
834 define dso_local <8 x half> @test_vminnmq_f16(<8 x half> %a, <8 x half> %b) {
835 ; CHECK-LABEL: test_vminnmq_f16:
836 ; CHECK: vminnm.f16 q0, q0, q1
839 %vminnmq_v2.i = tail call <8 x half> @llvm.arm.neon.vminnm.v8f16(<8 x half> %a, <8 x half> %b)
840 ret <8 x half> %vminnmq_v2.i
843 define dso_local <4 x half> @test_vmul_f16(<4 x half> %a, <4 x half> %b) {
844 ; CHECKLABEL: test_vmul_f16:
845 ; CHECK: vmul.f16 d0, d0, d1
848 %mul.i = fmul <4 x half> %a, %b
849 ret <4 x half> %mul.i
852 define dso_local <8 x half> @test_vmulq_f16(<8 x half> %a, <8 x half> %b) {
853 ; CHECKLABEL: test_vmulq_f16:
854 ; CHECK: vmul.f16 q0, q0, q1
857 %mul.i = fmul <8 x half> %a, %b
858 ret <8 x half> %mul.i
861 define dso_local <4 x half> @test_vpadd_f16(<4 x half> %a, <4 x half> %b) {
862 ; CHECKLABEL: test_vpadd_f16:
863 ; CHECK: vpadd.f16 d0, d0, d1
866 %vpadd_v2.i = tail call <4 x half> @llvm.arm.neon.vpadd.v4f16(<4 x half> %a, <4 x half> %b)
867 ret <4 x half> %vpadd_v2.i
870 define dso_local <4 x half> @test_vpmax_f16(<4 x half> %a, <4 x half> %b) {
871 ; CHECKLABEL: test_vpmax_f16:
872 ; CHECK: vpmax.f16 d0, d0, d1
875 %vpmax_v2.i = tail call <4 x half> @llvm.arm.neon.vpmaxs.v4f16(<4 x half> %a, <4 x half> %b)
876 ret <4 x half> %vpmax_v2.i
879 define dso_local <4 x half> @test_vpmin_f16(<4 x half> %a, <4 x half> %b) {
880 ; CHECKLABEL: test_vpmin_f16:
881 ; CHECK: vpmin.f16 d0, d0, d1
884 %vpmin_v2.i = tail call <4 x half> @llvm.arm.neon.vpmins.v4f16(<4 x half> %a, <4 x half> %b)
885 ret <4 x half> %vpmin_v2.i
888 define dso_local <4 x half> @test_vrecps_f16(<4 x half> %a, <4 x half> %b) {
889 ; CHECKLABEL: test_vrecps_f16:
890 ; CHECK: vrecps.f16 d0, d0, d1
893 %vrecps_v2.i = tail call <4 x half> @llvm.arm.neon.vrecps.v4f16(<4 x half> %a, <4 x half> %b)
894 ret <4 x half> %vrecps_v2.i
897 define dso_local <8 x half> @test_vrecpsq_f16(<8 x half> %a, <8 x half> %b) {
898 ; CHECKLABEL: test_vrecpsq_f16:
899 ; CHECK: vrecps.f16 q0, q0, q1
902 %vrecpsq_v2.i = tail call <8 x half> @llvm.arm.neon.vrecps.v8f16(<8 x half> %a, <8 x half> %b)
903 ret <8 x half> %vrecpsq_v2.i
906 define dso_local <4 x half> @test_vrsqrts_f16(<4 x half> %a, <4 x half> %b) {
907 ; CHECKLABEL: test_vrsqrts_f16:
908 ; CHECK: vrsqrts.f16 d0, d0, d1
911 %vrsqrts_v2.i = tail call <4 x half> @llvm.arm.neon.vrsqrts.v4f16(<4 x half> %a, <4 x half> %b)
912 ret <4 x half> %vrsqrts_v2.i
915 define dso_local <8 x half> @test_vrsqrtsq_f16(<8 x half> %a, <8 x half> %b) {
916 ; CHECKLABEL: test_vrsqrtsq_f16:
917 ; CHECK: vrsqrts.f16 q0, q0, q1
920 %vrsqrtsq_v2.i = tail call <8 x half> @llvm.arm.neon.vrsqrts.v8f16(<8 x half> %a, <8 x half> %b)
921 ret <8 x half> %vrsqrtsq_v2.i
924 define dso_local <4 x half> @test_vsub_f16(<4 x half> %a, <4 x half> %b) {
925 ; CHECKLABEL: test_vsub_f16:
926 ; CHECK: vsub.f16 d0, d0, d1
929 %sub.i = fsub <4 x half> %a, %b
930 ret <4 x half> %sub.i
933 define dso_local <8 x half> @test_vsubq_f16(<8 x half> %a, <8 x half> %b) {
934 ; CHECKLABEL: test_vsubq_f16:
935 ; CHECK: vsub.f16 q0, q0, q1
938 %sub.i = fsub <8 x half> %a, %b
939 ret <8 x half> %sub.i
942 define dso_local <4 x half> @test_vfma_f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
943 ; CHECK-LABEL: test_vfma_f16:
944 ; CHECK: vfma.f16 d0, d1, d2
947 %0 = tail call <4 x half> @llvm.fma.v4f16(<4 x half> %b, <4 x half> %c, <4 x half> %a)
951 define dso_local <8 x half> @test_vfmaq_f16(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
952 ; CHECK-LABEL: test_vfmaq_f16:
953 ; CHECK: vfma.f16 q0, q1, q2
956 %0 = tail call <8 x half> @llvm.fma.v8f16(<8 x half> %b, <8 x half> %c, <8 x half> %a)
960 define dso_local <4 x half> @test_vfms_f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
961 ; CHECK-LABEL: test_vfms_f16:
962 ; CHECK: vneg.f16 [[D16:d[0-9]+]], d1
963 ; CHECK-NEXT: vfma.f16 d0, [[D16]], d2
966 %sub.i = fsub <4 x half> <half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000>, %b
967 %0 = tail call <4 x half> @llvm.fma.v4f16(<4 x half> %sub.i, <4 x half> %c, <4 x half> %a)
971 define dso_local <8 x half> @test_vfmsq_f16(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
972 ; CHECK-LABEL: test_vfmsq_f16:
973 ; CHECK: vneg.f16 [[Q8:q[0-9]+]], q1
974 ; CHECK-NEXT: vfma.f16 q0, [[Q8]], q2
977 %sub.i = fsub <8 x half> <half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000, half 0xH8000>, %b
978 %0 = tail call <8 x half> @llvm.fma.v8f16(<8 x half> %sub.i, <8 x half> %c, <8 x half> %a)
982 define dso_local <4 x half> @test_vmul_lane_f16(<4 x half> %a, <4 x half> %b) {
983 ; CHECK-LABEL: test_vmul_lane_f16:
984 ; CHECK: vmul.f16 d0, d0, d1[3]
987 %shuffle = shufflevector <4 x half> %b, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
988 %mul = fmul <4 x half> %shuffle, %a
992 define dso_local <8 x half> @test_vmulq_lane_f16(<8 x half> %a, <4 x half> %b) {
993 ; CHECK-LABEL: test_vmulq_lane_f16:
994 ; CHECK: vmul.f16 q0, q0, d2[3]
997 %shuffle = shufflevector <4 x half> %b, <4 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
998 %mul = fmul <8 x half> %shuffle, %a
1002 define dso_local <4 x half> @test_vmul_n_f16(<4 x half> %a, float %b.coerce) {
1003 ; CHECK-LABEL: test_vmul_n_f16:
1004 ; CHECK: vmul.f16 d0, d0, d1[0]
1007 %0 = bitcast float %b.coerce to i32
1008 %tmp.0.extract.trunc = trunc i32 %0 to i16
1009 %1 = bitcast i16 %tmp.0.extract.trunc to half
1010 %vecinit = insertelement <4 x half> undef, half %1, i32 0
1011 %vecinit4 = shufflevector <4 x half> %vecinit, <4 x half> undef, <4 x i32> zeroinitializer
1012 %mul = fmul <4 x half> %vecinit4, %a
1016 define dso_local <8 x half> @test_vmulq_n_f16(<8 x half> %a, float %b.coerce) {
1017 ; CHECK-LABEL: test_vmulq_n_f16:
1018 ; CHECK: vmul.f16 q0, q0, d2[0]
1021 %0 = bitcast float %b.coerce to i32
1022 %tmp.0.extract.trunc = trunc i32 %0 to i16
1023 %1 = bitcast i16 %tmp.0.extract.trunc to half
1024 %vecinit = insertelement <8 x half> undef, half %1, i32 0
1025 %vecinit8 = shufflevector <8 x half> %vecinit, <8 x half> undef, <8 x i32> zeroinitializer
1026 %mul = fmul <8 x half> %vecinit8, %a
1030 define dso_local <4 x half> @test_vbsl_f16(<4 x i16> %a, <4 x half> %b, <4 x half> %c) {
1031 ; CHECKLABEL: test_vbsl_f16:
1032 ; CHECK: vbsl d0, d1, d2
1035 %0 = bitcast <4 x i16> %a to <8 x i8>
1036 %1 = bitcast <4 x half> %b to <8 x i8>
1037 %2 = bitcast <4 x half> %c to <8 x i8>
1038 %vbsl_v.i = tail call <8 x i8> @llvm.arm.neon.vbsl.v8i8(<8 x i8> %0, <8 x i8> %1, <8 x i8> %2)
1039 %3 = bitcast <8 x i8> %vbsl_v.i to <4 x half>
1043 define dso_local <8 x half> @test_vbslq_f16(<8 x i16> %a, <8 x half> %b, <8 x half> %c) {
1044 ; CHECKLABEL: test_vbslq_f16:
1045 ; CHECK: vbsl q0, q1, q2
1048 %0 = bitcast <8 x i16> %a to <16 x i8>
1049 %1 = bitcast <8 x half> %b to <16 x i8>
1050 %2 = bitcast <8 x half> %c to <16 x i8>
1051 %vbslq_v.i = tail call <16 x i8> @llvm.arm.neon.vbsl.v16i8(<16 x i8> %0, <16 x i8> %1, <16 x i8> %2)
1052 %3 = bitcast <16 x i8> %vbslq_v.i to <8 x half>
1056 define dso_local %struct.float16x4x2_t @test_vzip_f16(<4 x half> %a, <4 x half> %b) {
1057 ; CHECK-LABEL: test_vzip_f16:
1058 ; CHECK: vzip.16 d0, d1
1061 %vzip.i = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
1062 %vzip1.i = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
1063 %.fca.0.0.insert = insertvalue %struct.float16x4x2_t undef, <4 x half> %vzip.i, 0, 0
1064 %.fca.0.1.insert = insertvalue %struct.float16x4x2_t %.fca.0.0.insert, <4 x half> %vzip1.i, 0, 1
1065 ret %struct.float16x4x2_t %.fca.0.1.insert
1068 define dso_local %struct.float16x8x2_t @test_vzipq_f16(<8 x half> %a, <8 x half> %b) {
1069 ; CHECK-LABEL: test_vzipq_f16:
1070 ; CHECK: vzip.16 q0, q1
1073 %vzip.i = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
1074 %vzip1.i = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
1075 %.fca.0.0.insert = insertvalue %struct.float16x8x2_t undef, <8 x half> %vzip.i, 0, 0
1076 %.fca.0.1.insert = insertvalue %struct.float16x8x2_t %.fca.0.0.insert, <8 x half> %vzip1.i, 0, 1
1077 ret %struct.float16x8x2_t %.fca.0.1.insert
1080 define dso_local %struct.float16x4x2_t @test_vuzp_f16(<4 x half> %a, <4 x half> %b) {
1081 ; CHECK-LABEL: test_vuzp_f16:
1082 ; CHECK: vuzp.16 d0, d1
1085 %vuzp.i = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
1086 %vuzp1.i = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
1087 %.fca.0.0.insert = insertvalue %struct.float16x4x2_t undef, <4 x half> %vuzp.i, 0, 0
1088 %.fca.0.1.insert = insertvalue %struct.float16x4x2_t %.fca.0.0.insert, <4 x half> %vuzp1.i, 0, 1
1089 ret %struct.float16x4x2_t %.fca.0.1.insert
1092 define dso_local %struct.float16x8x2_t @test_vuzpq_f16(<8 x half> %a, <8 x half> %b) {
1093 ; CHECK-LABEL: test_vuzpq_f16:
1094 ; CHECK: vuzp.16 q0, q1
1097 %vuzp.i = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
1098 %vuzp1.i = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
1099 %.fca.0.0.insert = insertvalue %struct.float16x8x2_t undef, <8 x half> %vuzp.i, 0, 0
1100 %.fca.0.1.insert = insertvalue %struct.float16x8x2_t %.fca.0.0.insert, <8 x half> %vuzp1.i, 0, 1
1101 ret %struct.float16x8x2_t %.fca.0.1.insert
1104 define dso_local %struct.float16x4x2_t @test_vtrn_f16(<4 x half> %a, <4 x half> %b) {
1105 ; CHECK-LABEL: test_vtrn_f16:
1106 ; CHECK: vtrn.16 d0, d1
1109 %vtrn.i = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
1110 %vtrn1.i = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
1111 %.fca.0.0.insert = insertvalue %struct.float16x4x2_t undef, <4 x half> %vtrn.i, 0, 0
1112 %.fca.0.1.insert = insertvalue %struct.float16x4x2_t %.fca.0.0.insert, <4 x half> %vtrn1.i, 0, 1
1113 ret %struct.float16x4x2_t %.fca.0.1.insert
1116 define dso_local %struct.float16x8x2_t @test_vtrnq_f16(<8 x half> %a, <8 x half> %b) {
1117 ; CHECK-LABEL: test_vtrnq_f16:
1118 ; CHECK: vtrn.16 q0, q1
1121 %vtrn.i = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 0, i32 8, i32 2, i32 10, i32 4, i32 12, i32 6, i32 14>
1122 %vtrn1.i = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 1, i32 9, i32 3, i32 11, i32 5, i32 13, i32 7, i32 15>
1123 %.fca.0.0.insert = insertvalue %struct.float16x8x2_t undef, <8 x half> %vtrn.i, 0, 0
1124 %.fca.0.1.insert = insertvalue %struct.float16x8x2_t %.fca.0.0.insert, <8 x half> %vtrn1.i, 0, 1
1125 ret %struct.float16x8x2_t %.fca.0.1.insert
1128 define dso_local <4 x half> @test_vmov_n_f16(float %a.coerce) {
1129 ; CHECK-LABEL: test_vmov_n_f16:
1130 ; CHECK: vdup.16 d0, d0[0]
1133 %0 = bitcast float %a.coerce to i32
1134 %tmp.0.extract.trunc = trunc i32 %0 to i16
1135 %1 = bitcast i16 %tmp.0.extract.trunc to half
1136 %vecinit = insertelement <4 x half> undef, half %1, i32 0
1137 %vecinit4 = shufflevector <4 x half> %vecinit, <4 x half> undef, <4 x i32> zeroinitializer
1138 ret <4 x half> %vecinit4
1141 define dso_local <8 x half> @test_vmovq_n_f16(float %a.coerce) {
1142 ; CHECK-LABEL: test_vmovq_n_f16:
1143 ; CHECK: vdup.16 q0, d0[0]
1146 %0 = bitcast float %a.coerce to i32
1147 %tmp.0.extract.trunc = trunc i32 %0 to i16
1148 %1 = bitcast i16 %tmp.0.extract.trunc to half
1149 %vecinit = insertelement <8 x half> undef, half %1, i32 0
1150 %vecinit8 = shufflevector <8 x half> %vecinit, <8 x half> undef, <8 x i32> zeroinitializer
1151 ret <8 x half> %vecinit8
1154 define dso_local <4 x half> @test_vdup_n_f16(float %a.coerce) {
1155 ; CHECK-LABEL: test_vdup_n_f16:
1156 ; CHECK: vdup.16 d0, d0[0]
1159 %0 = bitcast float %a.coerce to i32
1160 %tmp.0.extract.trunc = trunc i32 %0 to i16
1161 %1 = bitcast i16 %tmp.0.extract.trunc to half
1162 %vecinit = insertelement <4 x half> undef, half %1, i32 0
1163 %vecinit4 = shufflevector <4 x half> %vecinit, <4 x half> undef, <4 x i32> zeroinitializer
1164 ret <4 x half> %vecinit4
1167 define dso_local <8 x half> @test_vdupq_n_f16(float %a.coerce) {
1168 ; CHECK-LABEL: test_vdupq_n_f16:
1169 ; CHECK: vdup.16 q0, d0[0]
1172 %0 = bitcast float %a.coerce to i32
1173 %tmp.0.extract.trunc = trunc i32 %0 to i16
1174 %1 = bitcast i16 %tmp.0.extract.trunc to half
1175 %vecinit = insertelement <8 x half> undef, half %1, i32 0
1176 %vecinit8 = shufflevector <8 x half> %vecinit, <8 x half> undef, <8 x i32> zeroinitializer
1177 ret <8 x half> %vecinit8
1180 define dso_local <4 x half> @test_vdup_lane_f16(<4 x half> %a) {
1181 ; CHECK-LABEL: test_vdup_lane_f16:
1182 ; CHECK: vdup.32 d0, d0[3]
1185 %shuffle = shufflevector <4 x half> %a, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
1186 ret <4 x half> %shuffle
1189 define dso_local <8 x half> @test_vdupq_lane_f16(<4 x half> %a) {
1190 ; CHECK-LABEL: test_vdupq_lane_f16:
1191 ; CHECK: vdup.16 q0, d0[3]
1194 %shuffle = shufflevector <4 x half> %a, <4 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
1195 ret <8 x half> %shuffle
1198 define dso_local <4 x half> @test_vext_f16(<4 x half> %a, <4 x half> %b) {
1199 ; CHECK-LABEL: test_vext_f16:
1200 ; CHECK: vext.16 d0, d0, d1, #2
1203 %vext = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
1204 ret <4 x half> %vext
1207 define dso_local <8 x half> @test_vextq_f16(<8 x half> %a, <8 x half> %b) {
1208 ; CHECK-LABEL: test_vextq_f16:
1209 ; CHECK: vext.16 q0, q0, q1, #5
1212 %vext = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12>
1213 ret <8 x half> %vext
1216 define dso_local <4 x half> @test_vrev64_f16(<4 x half> %a) {
1218 %shuffle.i = shufflevector <4 x half> %a, <4 x half> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
1219 ret <4 x half> %shuffle.i
1222 define dso_local <8 x half> @test_vrev64q_f16(<8 x half> %a) {
1224 %shuffle.i = shufflevector <8 x half> %a, <8 x half> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
1225 ret <8 x half> %shuffle.i
1228 define <4 x half> @test_vld_dup1_4xhalf(half* %b) {
1229 ; CHECK-LABEL: test_vld_dup1_4xhalf:
1230 ; CHECK: vld1.16 {d0[]}, [r0:16]
1234 %b1 = load half, half* %b, align 2
1235 %vecinit = insertelement <4 x half> undef, half %b1, i32 0
1236 %vecinit2 = insertelement <4 x half> %vecinit, half %b1, i32 1
1237 %vecinit3 = insertelement <4 x half> %vecinit2, half %b1, i32 2
1238 %vecinit4 = insertelement <4 x half> %vecinit3, half %b1, i32 3
1239 ret <4 x half> %vecinit4
1242 define <8 x half> @test_vld_dup1_8xhalf(half* %b) local_unnamed_addr {
1243 ; CHECK-LABEL: test_vld_dup1_8xhalf:
1244 ; CHECK: vld1.16 {d0[], d1[]}, [r0:16]
1248 %b1 = load half, half* %b, align 2
1249 %vecinit = insertelement <8 x half> undef, half %b1, i32 0
1250 %vecinit8 = shufflevector <8 x half> %vecinit, <8 x half> undef, <8 x i32> zeroinitializer
1251 ret <8 x half> %vecinit8
1254 define <8 x half> @test_shufflevector8xhalf(<4 x half> %a) {
1255 ; CHECK-LABEL: test_shufflevector8xhalf:
1256 ; CHECK: vmov.f64 d1, d0
1260 %r = shufflevector <4 x half> %a, <4 x half> %a, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
1264 declare <4 x half> @llvm.fabs.v4f16(<4 x half>)
1265 declare <8 x half> @llvm.fabs.v8f16(<8 x half>)
1266 declare <4 x i16> @llvm.arm.neon.vcvtas.v4i16.v4f16(<4 x half>)
1267 declare <4 x i16> @llvm.arm.neon.vcvtau.v4i16.v4f16(<4 x half>)
1268 declare <8 x i16> @llvm.arm.neon.vcvtas.v8i16.v8f16(<8 x half>)
1269 declare <4 x i16> @llvm.arm.neon.vcvtms.v4i16.v4f16(<4 x half>)
1270 declare <8 x i16> @llvm.arm.neon.vcvtms.v8i16.v8f16(<8 x half>)
1271 declare <4 x i16> @llvm.arm.neon.vcvtmu.v4i16.v4f16(<4 x half>)
1272 declare <8 x i16> @llvm.arm.neon.vcvtmu.v8i16.v8f16(<8 x half>)
1273 declare <4 x i16> @llvm.arm.neon.vcvtns.v4i16.v4f16(<4 x half>)
1274 declare <8 x i16> @llvm.arm.neon.vcvtns.v8i16.v8f16(<8 x half>)
1275 declare <4 x i16> @llvm.arm.neon.vcvtnu.v4i16.v4f16(<4 x half>)
1276 declare <8 x i16> @llvm.arm.neon.vcvtnu.v8i16.v8f16(<8 x half>)
1277 declare <4 x i16> @llvm.arm.neon.vcvtps.v4i16.v4f16(<4 x half>)
1278 declare <8 x i16> @llvm.arm.neon.vcvtps.v8i16.v8f16(<8 x half>)
1279 declare <4 x i16> @llvm.arm.neon.vcvtpu.v4i16.v4f16(<4 x half>)
1280 declare <8 x i16> @llvm.arm.neon.vcvtpu.v8i16.v8f16(<8 x half>)
1281 declare <4 x half> @llvm.arm.neon.vrecpe.v4f16(<4 x half>)
1282 declare <8 x half> @llvm.arm.neon.vrecpe.v8f16(<8 x half>)
1283 declare <4 x half> @llvm.arm.neon.vrintz.v4f16(<4 x half>)
1284 declare <8 x half> @llvm.arm.neon.vrintz.v8f16(<8 x half>)
1285 declare <4 x half> @llvm.arm.neon.vrinta.v4f16(<4 x half>)
1286 declare <8 x half> @llvm.arm.neon.vrinta.v8f16(<8 x half>)
1287 declare <4 x half> @llvm.arm.neon.vrintm.v4f16(<4 x half>)
1288 declare <8 x half> @llvm.arm.neon.vrintm.v8f16(<8 x half>)
1289 declare <4 x half> @llvm.arm.neon.vrintn.v4f16(<4 x half>)
1290 declare <8 x half> @llvm.arm.neon.vrintn.v8f16(<8 x half>)
1291 declare <4 x half> @llvm.arm.neon.vrintp.v4f16(<4 x half>)
1292 declare <8 x half> @llvm.arm.neon.vrintp.v8f16(<8 x half>)
1293 declare <4 x half> @llvm.arm.neon.vrintx.v4f16(<4 x half>)
1294 declare <8 x half> @llvm.arm.neon.vrintx.v8f16(<8 x half>)
1295 declare <4 x half> @llvm.arm.neon.vrsqrte.v4f16(<4 x half>)
1296 declare <8 x half> @llvm.arm.neon.vrsqrte.v8f16(<8 x half>)
1297 declare <4 x half> @llvm.arm.neon.vabds.v4f16(<4 x half>, <4 x half>)
1298 declare <8 x half> @llvm.arm.neon.vabds.v8f16(<8 x half>, <8 x half>)
1299 declare <4 x i16> @llvm.arm.neon.vacge.v4i16.v4f16(<4 x half>, <4 x half>)
1300 declare <8 x i16> @llvm.arm.neon.vacge.v8i16.v8f16(<8 x half>, <8 x half>)
1301 declare <4 x i16> @llvm.arm.neon.vacgt.v4i16.v4f16(<4 x half>, <4 x half>)
1302 declare <8 x i16> @llvm.arm.neon.vacgt.v8i16.v8f16(<8 x half>, <8 x half>)
1303 declare <4 x half> @llvm.arm.neon.vmaxs.v4f16(<4 x half>, <4 x half>)
1304 declare <8 x half> @llvm.arm.neon.vmaxs.v8f16(<8 x half>, <8 x half>)
1305 declare <4 x half> @llvm.arm.neon.vmaxnm.v4f16(<4 x half>, <4 x half>)
1306 declare <8 x half> @llvm.arm.neon.vmaxnm.v8f16(<8 x half>, <8 x half>)
1307 declare <4 x half> @llvm.arm.neon.vmins.v4f16(<4 x half>, <4 x half>)
1308 declare <8 x half> @llvm.arm.neon.vmins.v8f16(<8 x half>, <8 x half>)
1309 declare <4 x half> @llvm.arm.neon.vminnm.v4f16(<4 x half>, <4 x half>)
1310 declare <8 x half> @llvm.arm.neon.vminnm.v8f16(<8 x half>, <8 x half>)
1311 declare <4 x half> @llvm.arm.neon.vpadd.v4f16(<4 x half>, <4 x half>)
1312 declare <4 x half> @llvm.arm.neon.vpmaxs.v4f16(<4 x half>, <4 x half>)
1313 declare <4 x half> @llvm.arm.neon.vpmins.v4f16(<4 x half>, <4 x half>)
1314 declare <4 x half> @llvm.arm.neon.vrecps.v4f16(<4 x half>, <4 x half>)
1315 declare <8 x half> @llvm.arm.neon.vrecps.v8f16(<8 x half>, <8 x half>)
1316 declare <4 x half> @llvm.arm.neon.vrsqrts.v4f16(<4 x half>, <4 x half>)
1317 declare <8 x half> @llvm.arm.neon.vrsqrts.v8f16(<8 x half>, <8 x half>)
1318 declare <4 x half> @llvm.fma.v4f16(<4 x half>, <4 x half>, <4 x half>)
1319 declare <8 x half> @llvm.fma.v8f16(<8 x half>, <8 x half>, <8 x half>)
1320 declare <8 x i8> @llvm.arm.neon.vbsl.v8i8(<8 x i8>, <8 x i8>, <8 x i8>)
1321 declare <16 x i8> @llvm.arm.neon.vbsl.v16i8(<16 x i8>, <16 x i8>, <16 x i8>)
1322 declare { <8 x half>, <8 x half> } @llvm.arm.neon.vld2lane.v8f16.p0i8(i8*, <8 x half>, <8 x half>, i32, i32)
1323 declare { <4 x half>, <4 x half> } @llvm.arm.neon.vld2lane.v4f16.p0i8(i8*, <4 x half>, <4 x half>, i32, i32)
1324 declare { <8 x half>, <8 x half>, <8 x half> } @llvm.arm.neon.vld3lane.v8f16.p0i8(i8*, <8 x half>, <8 x half>, <8 x half>, i32, i32)
1325 declare { <4 x half>, <4 x half>, <4 x half> } @llvm.arm.neon.vld3lane.v4f16.p0i8(i8*, <4 x half>, <4 x half>, <4 x half>, i32, i32)
1326 declare { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.arm.neon.vld4lane.v8f16.p0i8(i8*, <8 x half>, <8 x half>, <8 x half>, <8 x half>, i32, i32)
1327 declare { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.arm.neon.vld4lane.v4f16.p0i8(i8*, <4 x half>, <4 x half>, <4 x half>, <4 x half>, i32, i32)
1328 declare void @llvm.arm.neon.vst2lane.p0i8.v8f16(i8*, <8 x half>, <8 x half>, i32, i32)
1329 declare void @llvm.arm.neon.vst2lane.p0i8.v4f16(i8*, <4 x half>, <4 x half>, i32, i32)
1330 declare void @llvm.arm.neon.vst3lane.p0i8.v8f16(i8*, <8 x half>, <8 x half>, <8 x half>, i32, i32)
1331 declare void @llvm.arm.neon.vst3lane.p0i8.v4f16(i8*, <4 x half>, <4 x half>, <4 x half>, i32, i32)
1332 declare void @llvm.arm.neon.vst4lane.p0i8.v8f16(i8*, <8 x half>, <8 x half>, <8 x half>, <8 x half>, i32, i32)
1333 declare void @llvm.arm.neon.vst4lane.p0i8.v4f16(i8*, <4 x half>, <4 x half>, <4 x half>, <4 x half>, i32, i32)
1335 define { <8 x half>, <8 x half> } @test_vld2q_lane_f16(i8*, <8 x half>, <8 x half>) {
1336 ; CHECK-LABEL: test_vld2q_lane_f16:
1337 ; CHECK: vld2.16 {d1[3], d3[3]}, [r0]
1340 %3 = tail call { <8 x half>, <8 x half> } @llvm.arm.neon.vld2lane.v8f16.p0i8(i8* %0, <8 x half> %1, <8 x half> %2, i32 7, i32 2)
1341 ret { <8 x half>, <8 x half> } %3
1344 define { <4 x half>, <4 x half> } @test_vld2_lane_f16(i8*, <4 x half>, <4 x half>) {
1345 ; CHECK-LABEL: test_vld2_lane_f16:
1346 ; CHECK: vld2.16 {d0[3], d1[3]}, [r0]
1349 %3 = tail call { <4 x half>, <4 x half> } @llvm.arm.neon.vld2lane.v4f16.p0i8(i8* %0, <4 x half> %1, <4 x half> %2, i32 3, i32 2)
1350 ret { <4 x half>, <4 x half> } %3
1353 define { <8 x half>, <8 x half>, <8 x half> } @test_vld3q_lane_f16(i8*, <8 x half>, <8 x half>, <8 x half>) {
1354 ; CHECK-LABEL: test_vld3q_lane_f16:
1355 ; CHECK: vld3.16 {d1[3], d3[3], d5[3]}, [r0]
1358 %4 = tail call { <8 x half>, <8 x half>, <8 x half> } @llvm.arm.neon.vld3lane.v8f16.p0i8(i8* %0, <8 x half> %1, <8 x half> %2, <8 x half> %3, i32 7, i32 2)
1359 ret { <8 x half>, <8 x half>, <8 x half> } %4
1362 define { <4 x half>, <4 x half>, <4 x half> } @test_vld3_lane_f16(i8*, <4 x half>, <4 x half>, <4 x half>) {
1363 ; CHECK-LABEL: test_vld3_lane_f16:
1364 ; CHECK: vld3.16 {d0[3], d1[3], d2[3]}, [r0]
1367 %4 = tail call { <4 x half>, <4 x half>, <4 x half> } @llvm.arm.neon.vld3lane.v4f16.p0i8(i8* %0, <4 x half> %1, <4 x half> %2, <4 x half> %3, i32 3, i32 2)
1368 ret { <4 x half>, <4 x half>, <4 x half> } %4
1370 define { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @test_vld4lane_v8f16_p0i8(i8*, <8 x half>, <8 x half>, <8 x half>, <8 x half>) {
1371 ; CHECK-LABEL: test_vld4lane_v8f16_p0i8:
1372 ; CHECK: vld4.16 {d1[3], d3[3], d5[3], d7[3]}, [r0]
1375 %5 = tail call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.arm.neon.vld4lane.v8f16.p0i8(i8* %0, <8 x half> %1, <8 x half> %2, <8 x half> %3, <8 x half> %4, i32 7, i32 2)
1376 ret { <8 x half>, <8 x half>, <8 x half>, <8 x half> } %5
1378 define { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @test_vld4lane_v4f16_p0i8(i8*, <4 x half>, <4 x half>, <4 x half>, <4 x half>) {
1379 ; CHECK-LABEL: test_vld4lane_v4f16_p0i8:
1380 ; CHECK: vld4.16 {d0[3], d1[3], d2[3], d3[3]}, [r0]
1383 %5 = tail call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.arm.neon.vld4lane.v4f16.p0i8(i8* %0, <4 x half> %1, <4 x half> %2, <4 x half> %3, <4 x half> %4, i32 3, i32 2)
1384 ret { <4 x half>, <4 x half>, <4 x half>, <4 x half> } %5
1386 define void @test_vst2lane_p0i8_v8f16(i8*, <8 x half>, <8 x half>) {
1387 ; CHECK-LABEL: test_vst2lane_p0i8_v8f16:
1388 ; CHECK: vst2.16 {d0[0], d2[0]}, [r0]
1391 tail call void @llvm.arm.neon.vst2lane.p0i8.v8f16(i8* %0, <8 x half> %1, <8 x half> %2, i32 0, i32 1)
1394 define void @test_vst2lane_p0i8_v4f16(i8*, <4 x half>, <4 x half>) {
1395 ; CHECK-LABEL: test_vst2lane_p0i8_v4f16:
1396 ; CHECK: vst2.16 {d0[0], d1[0]}, [r0:32]
1399 tail call void @llvm.arm.neon.vst2lane.p0i8.v4f16(i8* %0, <4 x half> %1, <4 x half> %2, i32 0, i32 0)
1402 define void @test_vst3lane_p0i8_v8f16(i8*, <8 x half>, <8 x half>, <8 x half>) {
1403 ; CHECK-LABEL: test_vst3lane_p0i8_v8f16:
1404 ; CHECK: vst3.16 {d0[0], d2[0], d4[0]}, [r0]
1407 tail call void @llvm.arm.neon.vst3lane.p0i8.v8f16(i8* %0, <8 x half> %1, <8 x half> %2, <8 x half> %3, i32 0, i32 0)
1410 define void @test_vst3lane_p0i8_v4f16(i8*, <4 x half>, <4 x half>, <4 x half>) {
1411 ; CHECK-LABEL: test_vst3lane_p0i8_v4f16:
1412 ; CHECK: vst3.16 {d0[0], d1[0], d2[0]}, [r0]
1415 tail call void @llvm.arm.neon.vst3lane.p0i8.v4f16(i8* %0, <4 x half> %1, <4 x half> %2, <4 x half> %3, i32 0, i32 0)
1418 define void @test_vst4lane_p0i8_v8f16(i8*, <8 x half>, <8 x half>, <8 x half>, <8 x half>) {
1419 ; CHECK-LABEL: test_vst4lane_p0i8_v8f16:
1420 ; CHECK: vst4.16 {d0[0], d2[0], d4[0], d6[0]}, [r0:64]
1423 tail call void @llvm.arm.neon.vst4lane.p0i8.v8f16(i8* %0, <8 x half> %1, <8 x half> %2, <8 x half> %3, <8 x half> %4, i32 0, i32 0)
1426 define void @test_vst4lane_p0i8_v4f16(i8*, <4 x half>, <4 x half>, <4 x half>, <4 x half>) {
1427 ; CHECK-LABEL: test_vst4lane_p0i8_v4f16:
1428 ; CHECK: vst4.16 {d0[0], d1[0], d2[0], d3[0]}, [r0:64]
1431 tail call void @llvm.arm.neon.vst4lane.p0i8.v4f16(i8* %0, <4 x half> %1, <4 x half> %2, <4 x half> %3, <4 x half> %4, i32 0, i32 0)