1 ; RUN: llc -show-mc-encoding -march=arm -mcpu=cortex-a8 -mattr=+neon < %s | FileCheck %s
5 declare <8 x i8> @llvm.arm.neon.vpadd.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
6 declare <4 x i16> @llvm.arm.neon.vpadd.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
7 declare <2 x i32> @llvm.arm.neon.vpadd.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
10 define <8 x i8> @vpadd_8xi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
11 %tmp1 = load <8 x i8>* %A
12 %tmp2 = load <8 x i8>* %B
13 ; CHECK: vpadd.i8 d16, d17, d16 @ encoding: [0xb0,0x0b,0x41,0xf2]
14 %tmp3 = call <8 x i8> @llvm.arm.neon.vpadd.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
19 define <4 x i16> @vpadd_4xi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
20 %tmp1 = load <4 x i16>* %A
21 %tmp2 = load <4 x i16>* %B
22 ; CHECK: vpadd.i16 d16, d17, d16 @ encoding: [0xb0,0x0b,0x51,0xf2]
23 %tmp3 = call <4 x i16> @llvm.arm.neon.vpadd.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
28 define <2 x i32> @vpadd_2xi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
29 %tmp1 = load <2 x i32>* %A
30 %tmp2 = load <2 x i32>* %B
31 ; CHECK: vpadd.i32 d16, d17, d16 @ encoding: [0xb0,0x0b,0x61,0xf2]
32 %tmp3 = call <2 x i32> @llvm.arm.neon.vpadd.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
36 declare <2 x float> @llvm.arm.neon.vpadd.v2f32(<2 x float>, <2 x float>) nounwind readnone
38 ; CHECK: vpadd_2xfloat
39 define <2 x float> @vpadd_2xfloat(<2 x float>* %A, <2 x float>* %B) nounwind {
40 %tmp1 = load <2 x float>* %A
41 %tmp2 = load <2 x float>* %B
42 ; CHECK: vpadd.f32 d16, d16, d17 @ encoding: [0xa1,0x0d,0x40,0xf3]
43 %tmp3 = call <2 x float> @llvm.arm.neon.vpadd.v2f32(<2 x float> %tmp1, <2 x float> %tmp2)
47 declare <4 x i16> @llvm.arm.neon.vpaddls.v4i16.v8i8(<8 x i8>) nounwind readnone
48 declare <2 x i32> @llvm.arm.neon.vpaddls.v2i32.v4i16(<4 x i16>) nounwind readnone
49 declare <1 x i64> @llvm.arm.neon.vpaddls.v1i64.v2i32(<2 x i32>) nounwind readnone
52 define <4 x i16> @vpaddls_8xi8(<8 x i8>* %A) nounwind {
53 %tmp1 = load <8 x i8>* %A
54 ; CHECK: vpaddl.s8 d16, d16 @ encoding: [0x20,0x02,0xf0,0xf3]
55 %tmp2 = call <4 x i16> @llvm.arm.neon.vpaddls.v4i16.v8i8(<8 x i8> %tmp1)
59 ; CHECK: vpaddls_4xi16
60 define <2 x i32> @vpaddls_4xi16(<4 x i16>* %A) nounwind {
61 %tmp1 = load <4 x i16>* %A
62 ; CHECK: vpaddl.s16 d16, d16 @ encoding: [0x20,0x02,0xf4,0xf3]
63 %tmp2 = call <2 x i32> @llvm.arm.neon.vpaddls.v2i32.v4i16(<4 x i16> %tmp1)
67 ; CHECK: vpaddls_2xi32
68 define <1 x i64> @vpaddls_2xi32(<2 x i32>* %A) nounwind {
69 %tmp1 = load <2 x i32>* %A
70 ; CHECK: vpaddl.s32 d16, d16 @ encoding: [0x20,0x02,0xf8,0xf3]
71 %tmp2 = call <1 x i64> @llvm.arm.neon.vpaddls.v1i64.v2i32(<2 x i32> %tmp1)
75 declare <4 x i16> @llvm.arm.neon.vpaddlu.v4i16.v8i8(<8 x i8>) nounwind readnone
76 declare <2 x i32> @llvm.arm.neon.vpaddlu.v2i32.v4i16(<4 x i16>) nounwind readnone
77 declare <1 x i64> @llvm.arm.neon.vpaddlu.v1i64.v2i32(<2 x i32>) nounwind readnone
80 define <4 x i16> @vpaddlu_8xi8(<8 x i8>* %A) nounwind {
81 %tmp1 = load <8 x i8>* %A
82 ; CHECK: vpaddl.u8 d16, d16 @ encoding: [0xa0,0x02,0xf0,0xf3]
83 %tmp2 = call <4 x i16> @llvm.arm.neon.vpaddlu.v4i16.v8i8(<8 x i8> %tmp1)
87 ; CHECK: vpaddlu_4xi16
88 define <2 x i32> @vpaddlu_4xi16(<4 x i16>* %A) nounwind {
89 %tmp1 = load <4 x i16>* %A
90 ; CHECK: vpaddl.u16 d16, d16 @ encoding: [0xa0,0x02,0xf4,0xf3]
91 %tmp2 = call <2 x i32> @llvm.arm.neon.vpaddlu.v2i32.v4i16(<4 x i16> %tmp1)
95 ; CHECK: vpaddlu_2xi32
96 define <1 x i64> @vpaddlu_2xi32(<2 x i32>* %A) nounwind {
97 %tmp1 = load <2 x i32>* %A
98 ; CHECK: vpaddl.u32 d16, d16 @ encoding: [0xa0,0x02,0xf8,0xf3]
99 %tmp2 = call <1 x i64> @llvm.arm.neon.vpaddlu.v1i64.v2i32(<2 x i32> %tmp1)
103 declare <8 x i16> @llvm.arm.neon.vpaddls.v8i16.v16i8(<16 x i8>) nounwind readnone
104 declare <4 x i32> @llvm.arm.neon.vpaddls.v4i32.v8i16(<8 x i16>) nounwind readnone
105 declare <2 x i64> @llvm.arm.neon.vpaddls.v2i64.v4i32(<4 x i32>) nounwind readnone
107 ; CHECK: vpaddls_16xi8
108 define <8 x i16> @vpaddls_16xi8(<16 x i8>* %A) nounwind {
109 %tmp1 = load <16 x i8>* %A
110 ; CHECK: vpaddl.s8 q8, q8 @ encoding: [0x60,0x02,0xf0,0xf3]
111 %tmp2 = call <8 x i16> @llvm.arm.neon.vpaddls.v8i16.v16i8(<16 x i8> %tmp1)
115 ; CHECK: vpaddls_8xi16
116 define <4 x i32> @vpaddls_8xi16(<8 x i16>* %A) nounwind {
117 %tmp1 = load <8 x i16>* %A
118 ; CHECK: vpaddl.s16 q8, q8 @ encoding: [0x60,0x02,0xf4,0xf3]
119 %tmp2 = call <4 x i32> @llvm.arm.neon.vpaddls.v4i32.v8i16(<8 x i16> %tmp1)
123 ; CHECK: vpaddls_4xi32
124 define <2 x i64> @vpaddls_4xi32(<4 x i32>* %A) nounwind {
125 %tmp1 = load <4 x i32>* %A
126 ; CHECK: vpaddl.s32 q8, q8 @ encoding: [0x60,0x02,0xf8,0xf3]
127 %tmp2 = call <2 x i64> @llvm.arm.neon.vpaddls.v2i64.v4i32(<4 x i32> %tmp1)
131 declare <8 x i16> @llvm.arm.neon.vpaddlu.v8i16.v16i8(<16 x i8>) nounwind readnone
132 declare <4 x i32> @llvm.arm.neon.vpaddlu.v4i32.v8i16(<8 x i16>) nounwind readnone
133 declare <2 x i64> @llvm.arm.neon.vpaddlu.v2i64.v4i32(<4 x i32>) nounwind readnone
135 ; CHECK: vpaddlu_16xi8
136 define <8 x i16> @vpaddlu_16xi8(<16 x i8>* %A) nounwind {
137 %tmp1 = load <16 x i8>* %A
138 ; CHECK: vpaddl.u8 q8, q8 @ encoding: [0xe0,0x02,0xf0,0xf3]
139 %tmp2 = call <8 x i16> @llvm.arm.neon.vpaddlu.v8i16.v16i8(<16 x i8> %tmp1)
143 ; CHECK: vpaddlu_8xi16
144 define <4 x i32> @vpaddlu_8xi16(<8 x i16>* %A) nounwind {
145 %tmp1 = load <8 x i16>* %A
146 ; CHECK: vpaddl.u16 q8, q8 @ encoding: [0xe0,0x02,0xf4,0xf3]
147 %tmp2 = call <4 x i32> @llvm.arm.neon.vpaddlu.v4i32.v8i16(<8 x i16> %tmp1)
151 ; CHECK: vpaddlu_4xi32
152 define <2 x i64> @vpaddlu_4xi32(<4 x i32>* %A) nounwind {
153 %tmp1 = load <4 x i32>* %A
154 ; CHECK: vpaddl.u32 q8, q8 @ encoding: [0xe0,0x02,0xf8,0xf3]
155 %tmp2 = call <2 x i64> @llvm.arm.neon.vpaddlu.v2i64.v4i32(<4 x i32> %tmp1)
159 declare <4 x i16> @llvm.arm.neon.vpadals.v4i16.v8i8(<4 x i16>, <8 x i8>) nounwind readnone
160 declare <2 x i32> @llvm.arm.neon.vpadals.v2i32.v4i16(<2 x i32>, <4 x i16>) nounwind readnone
161 declare <1 x i64> @llvm.arm.neon.vpadals.v1i64.v2i32(<1 x i64>, <2 x i32>) nounwind readnone
163 ; CHECK: vpadals_8xi8
164 define <4 x i16> @vpadals_8xi8(<4 x i16>* %A, <8 x i8>* %B) nounwind {
165 %tmp1 = load <4 x i16>* %A
166 %tmp2 = load <8 x i8>* %B
167 ; CHECK: vpadal.s8 d16, d17 @ encoding: [0x21,0x06,0xf0,0xf3]
168 %tmp3 = call <4 x i16> @llvm.arm.neon.vpadals.v4i16.v8i8(<4 x i16> %tmp1, <8 x i8> %tmp2)
172 ; CHECK: vpadals_4xi16
173 define <2 x i32> @vpadals_4xi16(<2 x i32>* %A, <4 x i16>* %B) nounwind {
174 %tmp1 = load <2 x i32>* %A
175 %tmp2 = load <4 x i16>* %B
176 ; CHECK: vpadal.s16 d16, d17 @ encoding: [0x21,0x06,0xf4,0xf3]
177 %tmp3 = call <2 x i32> @llvm.arm.neon.vpadals.v2i32.v4i16(<2 x i32> %tmp1, <4 x i16> %tmp2)
181 ; CHECK: vpadals_2xi32
182 define <1 x i64> @vpadals_2xi32(<1 x i64>* %A, <2 x i32>* %B) nounwind {
183 %tmp1 = load <1 x i64>* %A
184 %tmp2 = load <2 x i32>* %B
185 ; CHECK: vpadal.s32 d16, d17 @ encoding: [0x21,0x06,0xf8,0xf3]
186 %tmp3 = call <1 x i64> @llvm.arm.neon.vpadals.v1i64.v2i32(<1 x i64> %tmp1, <2 x i32> %tmp2)
190 declare <4 x i16> @llvm.arm.neon.vpadalu.v4i16.v8i8(<4 x i16>, <8 x i8>) nounwind readnone
191 declare <2 x i32> @llvm.arm.neon.vpadalu.v2i32.v4i16(<2 x i32>, <4 x i16>) nounwind readnone
192 declare <1 x i64> @llvm.arm.neon.vpadalu.v1i64.v2i32(<1 x i64>, <2 x i32>) nounwind readnone
194 ; CHECK: vpadalu_8xi8
195 define <4 x i16> @vpadalu_8xi8(<4 x i16>* %A, <8 x i8>* %B) nounwind {
196 %tmp1 = load <4 x i16>* %A
197 %tmp2 = load <8 x i8>* %B
198 ; CHECK: vpadal.u8 d16, d17 @ encoding: [0xa1,0x06,0xf0,0xf3]
199 %tmp3 = call <4 x i16> @llvm.arm.neon.vpadalu.v4i16.v8i8(<4 x i16> %tmp1, <8 x i8> %tmp2)
203 ; CHECK: vpadalu_4xi16
204 define <2 x i32> @vpadalu_4xi16(<2 x i32>* %A, <4 x i16>* %B) nounwind {
205 %tmp1 = load <2 x i32>* %A
206 %tmp2 = load <4 x i16>* %B
207 ; CHECK: vpadal.u16 d16, d17 @ encoding: [0xa1,0x06,0xf4,0xf3]
208 %tmp3 = call <2 x i32> @llvm.arm.neon.vpadalu.v2i32.v4i16(<2 x i32> %tmp1, <4 x i16> %tmp2)
212 ; CHECK: vpadalu_2xi32
213 define <1 x i64> @vpadalu_2xi32(<1 x i64>* %A, <2 x i32>* %B) nounwind {
214 %tmp1 = load <1 x i64>* %A
215 %tmp2 = load <2 x i32>* %B
216 ; CHECK: vpadal.u32 d16, d17 @ encoding: [0xa1,0x06,0xf8,0xf3]
217 %tmp3 = call <1 x i64> @llvm.arm.neon.vpadalu.v1i64.v2i32(<1 x i64> %tmp1, <2 x i32> %tmp2)
221 declare <8 x i16> @llvm.arm.neon.vpadals.v8i16.v16i8(<8 x i16>, <16 x i8>) nounwind readnone
222 declare <4 x i32> @llvm.arm.neon.vpadals.v4i32.v8i16(<4 x i32>, <8 x i16>) nounwind readnone
223 declare <2 x i64> @llvm.arm.neon.vpadals.v2i64.v4i32(<2 x i64>, <4 x i32>) nounwind readnone
225 ; CHECK: vpadals_16xi8
226 define <8 x i16> @vpadals_16xi8(<8 x i16>* %A, <16 x i8>* %B) nounwind {
227 %tmp1 = load <8 x i16>* %A
228 %tmp2 = load <16 x i8>* %B
229 ; CHECK: vpadal.s8 q9, q8 @ encoding: [0x60,0x26,0xf0,0xf3]
230 %tmp3 = call <8 x i16> @llvm.arm.neon.vpadals.v8i16.v16i8(<8 x i16> %tmp1, <16 x i8> %tmp2)
234 ; CHECK: vpadals_8xi16
235 define <4 x i32> @vpadals_8xi16(<4 x i32>* %A, <8 x i16>* %B) nounwind {
236 %tmp1 = load <4 x i32>* %A
237 %tmp2 = load <8 x i16>* %B
238 ; CHECK: vpadal.s16 q9, q8 @ encoding: [0x60,0x26,0xf4,0xf3]
239 %tmp3 = call <4 x i32> @llvm.arm.neon.vpadals.v4i32.v8i16(<4 x i32> %tmp1, <8 x i16> %tmp2)
243 ; CHECK: vpadals_4xi32
244 define <2 x i64> @vpadals_4xi32(<2 x i64>* %A, <4 x i32>* %B) nounwind {
245 %tmp1 = load <2 x i64>* %A
246 %tmp2 = load <4 x i32>* %B
247 ; CHECK: vpadal.s32 q9, q8 @ encoding: [0x60,0x26,0xf8,0xf3]
248 %tmp3 = call <2 x i64> @llvm.arm.neon.vpadals.v2i64.v4i32(<2 x i64> %tmp1, <4 x i32> %tmp2)
252 declare <8 x i16> @llvm.arm.neon.vpadalu.v8i16.v16i8(<8 x i16>, <16 x i8>) nounwind readnone
253 declare <4 x i32> @llvm.arm.neon.vpadalu.v4i32.v8i16(<4 x i32>, <8 x i16>) nounwind readnone
254 declare <2 x i64> @llvm.arm.neon.vpadalu.v2i64.v4i32(<2 x i64>, <4 x i32>) nounwind readnone
256 ; CHECK: vpadalu_16xi8
257 define <8 x i16> @vpadalu_16xi8(<8 x i16>* %A, <16 x i8>* %B) nounwind {
258 %tmp1 = load <8 x i16>* %A
259 %tmp2 = load <16 x i8>* %B
260 ; CHECK: vpadal.u8 q9, q8 @ encoding: [0xe0,0x26,0xf0,0xf3]
261 %tmp3 = call <8 x i16> @llvm.arm.neon.vpadalu.v8i16.v16i8(<8 x i16> %tmp1, <16 x i8> %tmp2)
265 ; CHECK: vpadalu_8xi16
266 define <4 x i32> @vpadalu_8xi16(<4 x i32>* %A, <8 x i16>* %B) nounwind {
267 %tmp1 = load <4 x i32>* %A
268 %tmp2 = load <8 x i16>* %B
269 ; CHECK: vpadal.u16 q9, q8 @ encoding: [0xe0,0x26,0xf4,0xf3]
270 %tmp3 = call <4 x i32> @llvm.arm.neon.vpadalu.v4i32.v8i16(<4 x i32> %tmp1, <8 x i16> %tmp2)
274 ; CHECK: vpadalu_4xi32
275 define <2 x i64> @vpadalu_4xi32(<2 x i64>* %A, <4 x i32>* %B) nounwind {
276 %tmp1 = load <2 x i64>* %A
277 %tmp2 = load <4 x i32>* %B
278 ; CHECK: vpadal.u32 q9, q8 @ encoding: [0xe0,0x26,0xf8,0xf3]
279 %tmp3 = call <2 x i64> @llvm.arm.neon.vpadalu.v2i64.v4i32(<2 x i64> %tmp1, <4 x i32> %tmp2)
283 declare <8 x i8> @llvm.arm.neon.vpmins.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
284 declare <4 x i16> @llvm.arm.neon.vpmins.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
285 declare <2 x i32> @llvm.arm.neon.vpmins.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
288 define <8 x i8> @vpmins_8xi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
289 %tmp1 = load <8 x i8>* %A
290 %tmp2 = load <8 x i8>* %B
291 ; CHECK: vpmin.s8 d16, d16, d17 @ encoding: [0xb1,0x0a,0x40,0xf2]
292 %tmp3 = call <8 x i8> @llvm.arm.neon.vpmins.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
296 ; CHECK: vpmins_4xi16
297 define <4 x i16> @vpmins_4xi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
298 %tmp1 = load <4 x i16>* %A
299 %tmp2 = load <4 x i16>* %B
300 ; CHECK: vpmin.s16 d16, d16, d17 @ encoding: [0xb1,0x0a,0x50,0xf2]
301 %tmp3 = call <4 x i16> @llvm.arm.neon.vpmins.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
305 ; CHECK: vpmins_2xi32
306 define <2 x i32> @vpmins_2xi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
307 %tmp1 = load <2 x i32>* %A
308 %tmp2 = load <2 x i32>* %B
309 ; CHECK: vpmin.s32 d16, d16, d17 @ encoding: [0xb1,0x0a,0x60,0xf2]
310 %tmp3 = call <2 x i32> @llvm.arm.neon.vpmins.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
314 declare <8 x i8> @llvm.arm.neon.vpminu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
315 declare <4 x i16> @llvm.arm.neon.vpminu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
316 declare <2 x i32> @llvm.arm.neon.vpminu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
319 define <8 x i8> @vpminu_8xi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
320 %tmp1 = load <8 x i8>* %A
321 %tmp2 = load <8 x i8>* %B
322 ; CHECK: vpmin.u8 d16, d16, d17 @ encoding: [0xb1,0x0a,0x40,0xf3]
323 %tmp3 = call <8 x i8> @llvm.arm.neon.vpminu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
327 ; CHECK: vpminu_4xi16
328 define <4 x i16> @vpminu_4xi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
329 %tmp1 = load <4 x i16>* %A
330 %tmp2 = load <4 x i16>* %B
331 ; CHECK: vpmin.u16 d16, d16, d17 @ encoding: [0xb1,0x0a,0x50,0xf3]
332 %tmp3 = call <4 x i16> @llvm.arm.neon.vpminu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
336 ; CHECK: vpminu_2xi32
337 define <2 x i32> @vpminu_2xi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
338 %tmp1 = load <2 x i32>* %A
339 %tmp2 = load <2 x i32>* %B
340 ; CHECK: vpmin.u32 d16, d16, d17 @ encoding: [0xb1,0x0a,0x60,0xf3]
341 %tmp3 = call <2 x i32> @llvm.arm.neon.vpminu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
345 declare <2 x float> @llvm.arm.neon.vpmins.v2f32(<2 x float>, <2 x float>) nounwind readnone
347 ; CHECK: vpmin_2xfloat
348 define <2 x float> @vpmin_2xfloat(<2 x float>* %A, <2 x float>* %B) nounwind {
349 %tmp1 = load <2 x float>* %A
350 %tmp2 = load <2 x float>* %B
351 ; CHECK: vpmin.f32 d16, d16, d17 @ encoding: [0xa1,0x0f,0x60,0xf3]
352 %tmp3 = call <2 x float> @llvm.arm.neon.vpmins.v2f32(<2 x float> %tmp1, <2 x float> %tmp2)
353 ret <2 x float> %tmp3
356 declare <8 x i8> @llvm.arm.neon.vpmaxs.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
357 declare <4 x i16> @llvm.arm.neon.vpmaxs.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
358 declare <2 x i32> @llvm.arm.neon.vpmaxs.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
361 define <8 x i8> @vpmaxs_8xi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
362 %tmp1 = load <8 x i8>* %A
363 %tmp2 = load <8 x i8>* %B
364 ; CHECK: vpmax.s8 d16, d16, d17 @ encoding: [0xa1,0x0a,0x40,0xf2]
365 %tmp3 = call <8 x i8> @llvm.arm.neon.vpmaxs.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
369 ; CHECK: vpmaxs_4xi16
370 define <4 x i16> @vpmaxs_4xi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
371 %tmp1 = load <4 x i16>* %A
372 %tmp2 = load <4 x i16>* %B
373 ; CHECK: vpmax.s16 d16, d16, d17 @ encoding: [0xa1,0x0a,0x50,0xf2]
374 %tmp3 = call <4 x i16> @llvm.arm.neon.vpmaxs.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
378 ; CHECK: vpmaxs_2xi32
379 define <2 x i32> @vpmaxs_2xi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
380 %tmp1 = load <2 x i32>* %A
381 %tmp2 = load <2 x i32>* %B
382 ; CHECK: vpmax.s32 d16, d16, d17 @ encoding: [0xa1,0x0a,0x60,0xf2]
383 %tmp3 = call <2 x i32> @llvm.arm.neon.vpmaxs.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
387 declare <8 x i8> @llvm.arm.neon.vpmaxu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
388 declare <4 x i16> @llvm.arm.neon.vpmaxu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
389 declare <2 x i32> @llvm.arm.neon.vpmaxu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
392 define <8 x i8> @vpmaxu_8xi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
393 %tmp1 = load <8 x i8>* %A
394 %tmp2 = load <8 x i8>* %B
395 ; CHECK: vpmax.u8 d16, d16, d17 @ encoding: [0xa1,0x0a,0x40,0xf3]
396 %tmp3 = call <8 x i8> @llvm.arm.neon.vpmaxu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
400 ; CHECK: vpmaxu_4xi16
401 define <4 x i16> @vpmaxu_4xi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
402 %tmp1 = load <4 x i16>* %A
403 %tmp2 = load <4 x i16>* %B
404 ; CHECK: vpmax.u16 d16, d16, d17 @ encoding: [0xa1,0x0a,0x50,0xf3]
405 %tmp3 = call <4 x i16> @llvm.arm.neon.vpmaxu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
409 ; CHECK: vpmaxu_2xi32
410 define <2 x i32> @vpmaxu_2xi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
411 %tmp1 = load <2 x i32>* %A
412 %tmp2 = load <2 x i32>* %B
413 ; CHECK: vpmax.u32 d16, d16, d17 @ encoding: [0xa1,0x0a,0x60,0xf3]
414 %tmp3 = call <2 x i32> @llvm.arm.neon.vpmaxu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
418 declare <2 x float> @llvm.arm.neon.vpmaxs.v2f32(<2 x float>, <2 x float>) nounwind readnone
420 ; CHECK: vpmax_2xfloat
421 define <2 x float> @vpmax_2xfloat(<2 x float>* %A, <2 x float>* %B) nounwind {
422 %tmp1 = load <2 x float>* %A
423 %tmp2 = load <2 x float>* %B
424 ; CHECK: vpmax.f32 d16, d16, d17 @ encoding: [0xa1,0x0f,0x40,0xf3]
425 %tmp3 = call <2 x float> @llvm.arm.neon.vpmaxs.v2f32(<2 x float> %tmp1, <2 x float> %tmp2)
426 ret <2 x float> %tmp3