1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=armv8 -mattr=+v8.1a | FileCheck %s
4 ;-----------------------------------------------------------------------------
7 declare <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16>, <4 x i16>)
8 declare <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16>, <8 x i16>)
9 declare <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32>, <2 x i32>)
10 declare <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32>, <4 x i32>)
12 declare <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16>, <4 x i16>)
13 declare <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16>, <8 x i16>)
14 declare <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32>, <2 x i32>)
15 declare <4 x i32> @llvm.sadd.sat.v4i32(<4 x i32>, <4 x i32>)
17 declare <4 x i16> @llvm.ssub.sat.v4i16(<4 x i16>, <4 x i16>)
18 declare <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16>, <8 x i16>)
19 declare <2 x i32> @llvm.ssub.sat.v2i32(<2 x i32>, <2 x i32>)
20 declare <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32>, <4 x i32>)
22 declare <4 x i16> @llvm.arm.neon.vqrdmlah.v4i16(<4 x i16>, <4 x i16>, <4 x i16>)
23 declare <2 x i32> @llvm.arm.neon.vqrdmlah.v2i32(<2 x i32>, <2 x i32>, <2 x i32>)
24 declare <8 x i16> @llvm.arm.neon.vqrdmlah.v8i16(<8 x i16>, <8 x i16>, <8 x i16>)
25 declare <4 x i32> @llvm.arm.neon.vqrdmlah.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
26 declare <4 x i16> @llvm.arm.neon.vqrdmlsh.v4i16(<4 x i16>, <4 x i16>, <4 x i16>)
27 declare <2 x i32> @llvm.arm.neon.vqrdmlsh.v2i32(<2 x i32>, <2 x i32>, <2 x i32>)
28 declare <8 x i16> @llvm.arm.neon.vqrdmlsh.v8i16(<8 x i16>, <8 x i16>, <8 x i16>)
29 declare <4 x i32> @llvm.arm.neon.vqrdmlsh.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
31 ; The sadd intrinsics in this file previously transformed into sqrdmlah where they
32 ; shouldn't. They should produce vqrdmulh and vadd.
34 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmulah_v4i16(<4 x i16> %acc, <4 x i16> %mhs, <4 x i16> %rhs) {
35 ; CHECK-LABEL: test_vqrdmulah_v4i16:
37 ; CHECK-NEXT: vqrdmulh.s16 d16, d1, d2
38 ; CHECK-NEXT: vqadd.s16 d0, d0, d16
40 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %mhs, <4 x i16> %rhs)
41 %retval = call <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16> %acc, <4 x i16> %prod)
45 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmulah_v8i16(<8 x i16> %acc, <8 x i16> %mhs, <8 x i16> %rhs) {
46 ; CHECK-LABEL: test_vqrdmulah_v8i16:
48 ; CHECK-NEXT: vqrdmulh.s16 q8, q1, q2
49 ; CHECK-NEXT: vqadd.s16 q0, q0, q8
51 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %mhs, <8 x i16> %rhs)
52 %retval = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %acc, <8 x i16> %prod)
56 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmulah_v2i32(<2 x i32> %acc, <2 x i32> %mhs, <2 x i32> %rhs) {
57 ; CHECK-LABEL: test_vqrdmulah_v2i32:
59 ; CHECK-NEXT: vqrdmulh.s32 d16, d1, d2
60 ; CHECK-NEXT: vqadd.s32 d0, d0, d16
62 %prod = call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %mhs, <2 x i32> %rhs)
63 %retval = call <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32> %acc, <2 x i32> %prod)
67 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmulah_v4i32(<4 x i32> %acc, <4 x i32> %mhs, <4 x i32> %rhs) {
68 ; CHECK-LABEL: test_vqrdmulah_v4i32:
70 ; CHECK-NEXT: vqrdmulh.s32 q8, q1, q2
71 ; CHECK-NEXT: vqadd.s32 q0, q0, q8
73 %prod = call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %mhs, <4 x i32> %rhs)
74 %retval = call <4 x i32> @llvm.sadd.sat.v4i32(<4 x i32> %acc, <4 x i32> %prod)
78 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmulsh_v4i16(<4 x i16> %acc, <4 x i16> %mhs, <4 x i16> %rhs) {
79 ; CHECK-LABEL: test_vqrdmulsh_v4i16:
81 ; CHECK-NEXT: vqrdmulh.s16 d16, d1, d2
82 ; CHECK-NEXT: vqsub.s16 d0, d0, d16
84 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %mhs, <4 x i16> %rhs)
85 %retval = call <4 x i16> @llvm.ssub.sat.v4i16(<4 x i16> %acc, <4 x i16> %prod)
89 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmulsh_v8i16(<8 x i16> %acc, <8 x i16> %mhs, <8 x i16> %rhs) {
90 ; CHECK-LABEL: test_vqrdmulsh_v8i16:
92 ; CHECK-NEXT: vqrdmulh.s16 q8, q1, q2
93 ; CHECK-NEXT: vqsub.s16 q0, q0, q8
95 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %mhs, <8 x i16> %rhs)
96 %retval = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %acc, <8 x i16> %prod)
100 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmulsh_v2i32(<2 x i32> %acc, <2 x i32> %mhs, <2 x i32> %rhs) {
101 ; CHECK-LABEL: test_vqrdmulsh_v2i32:
103 ; CHECK-NEXT: vqrdmulh.s32 d16, d1, d2
104 ; CHECK-NEXT: vqsub.s32 d0, d0, d16
106 %prod = call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %mhs, <2 x i32> %rhs)
107 %retval = call <2 x i32> @llvm.ssub.sat.v2i32(<2 x i32> %acc, <2 x i32> %prod)
108 ret <2 x i32> %retval
111 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmulsh_v4i32(<4 x i32> %acc, <4 x i32> %mhs, <4 x i32> %rhs) {
112 ; CHECK-LABEL: test_vqrdmulsh_v4i32:
114 ; CHECK-NEXT: vqrdmulh.s32 q8, q1, q2
115 ; CHECK-NEXT: vqsub.s32 q0, q0, q8
117 %prod = call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %mhs, <4 x i32> %rhs)
118 %retval = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> %acc, <4 x i32> %prod)
119 ret <4 x i32> %retval
122 ;-----------------------------------------------------------------------------
125 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmulah_lane_s16(<4 x i16> %acc, <4 x i16> %x, <4 x i16> %v) {
126 ; CHECK-LABEL: test_vqrdmulah_lane_s16:
127 ; CHECK: @ %bb.0: @ %entry
128 ; CHECK-NEXT: vqrdmulh.s16 d16, d1, d2[3]
129 ; CHECK-NEXT: vqadd.s16 d0, d0, d16
132 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
133 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %x, <4 x i16> %shuffle)
134 %retval = call <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16> %acc, <4 x i16> %prod)
135 ret <4 x i16> %retval
138 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmulahq_lane_s16(<8 x i16> %acc, <8 x i16> %x, <4 x i16> %v) {
139 ; CHECK-LABEL: test_vqrdmulahq_lane_s16:
140 ; CHECK: @ %bb.0: @ %entry
141 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
142 ; CHECK-NEXT: vqrdmulh.s16 q8, q1, d4[2]
143 ; CHECK-NEXT: vqadd.s16 q0, q0, q8
146 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
147 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %x, <8 x i16> %shuffle)
148 %retval = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %acc, <8 x i16> %prod)
149 ret <8 x i16> %retval
152 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmulah_lane_s32(<2 x i32> %acc, <2 x i32> %x, <2 x i32> %v) {
153 ; CHECK-LABEL: test_vqrdmulah_lane_s32:
154 ; CHECK: @ %bb.0: @ %entry
155 ; CHECK-NEXT: vqrdmulh.s32 d16, d1, d2[1]
156 ; CHECK-NEXT: vqadd.s32 d0, d0, d16
159 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
160 %prod = tail call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %x, <2 x i32> %shuffle)
161 %retval = call <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32> %acc, <2 x i32> %prod)
162 ret <2 x i32> %retval
165 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmulahq_lane_s32(<4 x i32> %acc,<4 x i32> %x, <2 x i32> %v) {
166 ; CHECK-LABEL: test_vqrdmulahq_lane_s32:
167 ; CHECK: @ %bb.0: @ %entry
168 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
169 ; CHECK-NEXT: vqrdmulh.s32 q8, q1, d4[0]
170 ; CHECK-NEXT: vqadd.s32 q0, q0, q8
173 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <4 x i32> zeroinitializer
174 %prod = tail call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %x, <4 x i32> %shuffle)
175 %retval = call <4 x i32> @llvm.sadd.sat.v4i32(<4 x i32> %acc, <4 x i32> %prod)
176 ret <4 x i32> %retval
179 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmulsh_lane_s16(<4 x i16> %acc, <4 x i16> %x, <4 x i16> %v) {
180 ; CHECK-LABEL: test_vqrdmulsh_lane_s16:
181 ; CHECK: @ %bb.0: @ %entry
182 ; CHECK-NEXT: vqrdmulh.s16 d16, d1, d2[3]
183 ; CHECK-NEXT: vqsub.s16 d0, d0, d16
186 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
187 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %x, <4 x i16> %shuffle)
188 %retval = call <4 x i16> @llvm.ssub.sat.v4i16(<4 x i16> %acc, <4 x i16> %prod)
189 ret <4 x i16> %retval
192 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmulshq_lane_s16(<8 x i16> %acc, <8 x i16> %x, <4 x i16> %v) {
193 ; CHECK-LABEL: test_vqrdmulshq_lane_s16:
194 ; CHECK: @ %bb.0: @ %entry
195 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
196 ; CHECK-NEXT: vqrdmulh.s16 q8, q1, d4[2]
197 ; CHECK-NEXT: vqsub.s16 q0, q0, q8
200 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
201 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %x, <8 x i16> %shuffle)
202 %retval = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %acc, <8 x i16> %prod)
203 ret <8 x i16> %retval
206 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmulsh_lane_s32(<2 x i32> %acc, <2 x i32> %x, <2 x i32> %v) {
207 ; CHECK-LABEL: test_vqrdmulsh_lane_s32:
208 ; CHECK: @ %bb.0: @ %entry
209 ; CHECK-NEXT: vqrdmulh.s32 d16, d1, d2[1]
210 ; CHECK-NEXT: vqsub.s32 d0, d0, d16
213 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
214 %prod = tail call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %x, <2 x i32> %shuffle)
215 %retval = call <2 x i32> @llvm.ssub.sat.v2i32(<2 x i32> %acc, <2 x i32> %prod)
216 ret <2 x i32> %retval
219 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmulshq_lane_s32(<4 x i32> %acc,<4 x i32> %x, <2 x i32> %v) {
220 ; CHECK-LABEL: test_vqrdmulshq_lane_s32:
221 ; CHECK: @ %bb.0: @ %entry
222 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
223 ; CHECK-NEXT: vqrdmulh.s32 q8, q1, d4[0]
224 ; CHECK-NEXT: vqsub.s32 q0, q0, q8
227 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <4 x i32> zeroinitializer
228 %prod = tail call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %x, <4 x i32> %shuffle)
229 %retval = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> %acc, <4 x i32> %prod)
230 ret <4 x i32> %retval
235 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmlah_s16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) {
236 ; CHECK-LABEL: test_vqrdmlah_s16:
237 ; CHECK: @ %bb.0: @ %entry
238 ; CHECK-NEXT: vqrdmlah.s16 d0, d1, d2
241 %vqrdmlah_v3.i = tail call <4 x i16> @llvm.arm.neon.vqrdmlah.v4i16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) #3
242 ret <4 x i16> %vqrdmlah_v3.i
245 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmlah_s32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) {
246 ; CHECK-LABEL: test_vqrdmlah_s32:
247 ; CHECK: @ %bb.0: @ %entry
248 ; CHECK-NEXT: vqrdmlah.s32 d0, d1, d2
251 %vqrdmlah_v3.i = tail call <2 x i32> @llvm.arm.neon.vqrdmlah.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) #3
252 ret <2 x i32> %vqrdmlah_v3.i
255 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmlahq_s16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) {
256 ; CHECK-LABEL: test_vqrdmlahq_s16:
257 ; CHECK: @ %bb.0: @ %entry
258 ; CHECK-NEXT: vqrdmlah.s16 q0, q1, q2
261 %vqrdmlahq_v3.i = tail call <8 x i16> @llvm.arm.neon.vqrdmlah.v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) #3
262 ret <8 x i16> %vqrdmlahq_v3.i
265 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmlahq_s32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
266 ; CHECK-LABEL: test_vqrdmlahq_s32:
267 ; CHECK: @ %bb.0: @ %entry
268 ; CHECK-NEXT: vqrdmlah.s32 q0, q1, q2
271 %vqrdmlahq_v3.i = tail call <4 x i32> @llvm.arm.neon.vqrdmlah.v4i32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) #3
272 ret <4 x i32> %vqrdmlahq_v3.i
275 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmlah_lane_s16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) {
276 ; CHECK-LABEL: test_vqrdmlah_lane_s16:
277 ; CHECK: @ %bb.0: @ %entry
278 ; CHECK-NEXT: vqrdmlah.s16 d0, d1, d2[3]
281 %lane = shufflevector <4 x i16> %c, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
282 %vqrdmlah_v3.i = tail call <4 x i16> @llvm.arm.neon.vqrdmlah.v4i16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %lane) #3
283 ret <4 x i16> %vqrdmlah_v3.i
286 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmlah_lane_s32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) {
287 ; CHECK-LABEL: test_vqrdmlah_lane_s32:
288 ; CHECK: @ %bb.0: @ %entry
289 ; CHECK-NEXT: vqrdmlah.s32 d0, d1, d2[1]
292 %lane = shufflevector <2 x i32> %c, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
293 %vqrdmlah_v3.i = tail call <2 x i32> @llvm.arm.neon.vqrdmlah.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %lane) #3
294 ret <2 x i32> %vqrdmlah_v3.i
297 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmlahq_lane_s16(<8 x i16> %a, <8 x i16> %b, <4 x i16> %c) {
298 ; CHECK-LABEL: test_vqrdmlahq_lane_s16:
299 ; CHECK: @ %bb.0: @ %entry
300 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
301 ; CHECK-NEXT: vqrdmlah.s16 q0, q1, d4[3]
304 %lane = shufflevector <4 x i16> %c, <4 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
305 %vqrdmlahq_v3.i = tail call <8 x i16> @llvm.arm.neon.vqrdmlah.v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %lane) #3
306 ret <8 x i16> %vqrdmlahq_v3.i
309 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmlahq_lane_s32(<4 x i32> %a, <4 x i32> %b, <2 x i32> %c) {
310 ; CHECK-LABEL: test_vqrdmlahq_lane_s32:
311 ; CHECK: @ %bb.0: @ %entry
312 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
313 ; CHECK-NEXT: vqrdmlah.s32 q0, q1, d4[1]
316 %lane = shufflevector <2 x i32> %c, <2 x i32> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
317 %vqrdmlahq_v3.i = tail call <4 x i32> @llvm.arm.neon.vqrdmlah.v4i32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %lane) #3
318 ret <4 x i32> %vqrdmlahq_v3.i
321 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmlsh_s16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) {
322 ; CHECK-LABEL: test_vqrdmlsh_s16:
323 ; CHECK: @ %bb.0: @ %entry
324 ; CHECK-NEXT: vqrdmlsh.s16 d0, d1, d2
327 %vqrdmlsh_v3.i = tail call <4 x i16> @llvm.arm.neon.vqrdmlsh.v4i16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) #3
328 ret <4 x i16> %vqrdmlsh_v3.i
331 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmlsh_s32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) {
332 ; CHECK-LABEL: test_vqrdmlsh_s32:
333 ; CHECK: @ %bb.0: @ %entry
334 ; CHECK-NEXT: vqrdmlsh.s32 d0, d1, d2
337 %vqrdmlsh_v3.i = tail call <2 x i32> @llvm.arm.neon.vqrdmlsh.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) #3
338 ret <2 x i32> %vqrdmlsh_v3.i
341 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmlshq_s16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) {
342 ; CHECK-LABEL: test_vqrdmlshq_s16:
343 ; CHECK: @ %bb.0: @ %entry
344 ; CHECK-NEXT: vqrdmlsh.s16 q0, q1, q2
347 %vqrdmlshq_v3.i = tail call <8 x i16> @llvm.arm.neon.vqrdmlsh.v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) #3
348 ret <8 x i16> %vqrdmlshq_v3.i
351 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmlshq_s32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
352 ; CHECK-LABEL: test_vqrdmlshq_s32:
353 ; CHECK: @ %bb.0: @ %entry
354 ; CHECK-NEXT: vqrdmlsh.s32 q0, q1, q2
357 %vqrdmlshq_v3.i = tail call <4 x i32> @llvm.arm.neon.vqrdmlsh.v4i32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) #3
358 ret <4 x i32> %vqrdmlshq_v3.i
361 define arm_aapcs_vfpcc <4 x i16> @test_vqrdmlsh_lane_s16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %c) {
362 ; CHECK-LABEL: test_vqrdmlsh_lane_s16:
363 ; CHECK: @ %bb.0: @ %entry
364 ; CHECK-NEXT: vqrdmlsh.s16 d0, d1, d2[3]
367 %lane = shufflevector <4 x i16> %c, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
368 %vqrdmlsh_v3.i = tail call <4 x i16> @llvm.arm.neon.vqrdmlsh.v4i16(<4 x i16> %a, <4 x i16> %b, <4 x i16> %lane) #3
369 ret <4 x i16> %vqrdmlsh_v3.i
372 define arm_aapcs_vfpcc <2 x i32> @test_vqrdmlsh_lane_s32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) {
373 ; CHECK-LABEL: test_vqrdmlsh_lane_s32:
374 ; CHECK: @ %bb.0: @ %entry
375 ; CHECK-NEXT: vqrdmlsh.s32 d0, d1, d2[1]
378 %lane = shufflevector <2 x i32> %c, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
379 %vqrdmlsh_v3.i = tail call <2 x i32> @llvm.arm.neon.vqrdmlsh.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %lane) #3
380 ret <2 x i32> %vqrdmlsh_v3.i
383 define arm_aapcs_vfpcc <8 x i16> @test_vqrdmlshq_lane_s16(<8 x i16> %a, <8 x i16> %b, <4 x i16> %c) {
384 ; CHECK-LABEL: test_vqrdmlshq_lane_s16:
385 ; CHECK: @ %bb.0: @ %entry
386 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
387 ; CHECK-NEXT: vqrdmlsh.s16 q0, q1, d4[3]
390 %lane = shufflevector <4 x i16> %c, <4 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
391 %vqrdmlshq_v3.i = tail call <8 x i16> @llvm.arm.neon.vqrdmlsh.v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %lane) #3
392 ret <8 x i16> %vqrdmlshq_v3.i
395 define arm_aapcs_vfpcc <4 x i32> @test_vqrdmlshq_lane_s32(<4 x i32> %a, <4 x i32> %b, <2 x i32> %c) {
396 ; CHECK-LABEL: test_vqrdmlshq_lane_s32:
397 ; CHECK: @ %bb.0: @ %entry
398 ; CHECK-NEXT: @ kill: def $d4 killed $d4 def $q2
399 ; CHECK-NEXT: vqrdmlsh.s32 q0, q1, d4[1]
402 %lane = shufflevector <2 x i32> %c, <2 x i32> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
403 %vqrdmlshq_v3.i = tail call <4 x i32> @llvm.arm.neon.vqrdmlsh.v4i32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %lane) #3
404 ret <4 x i32> %vqrdmlshq_v3.i