1 ; RUN: llc -march=mips -mattr=+msa < %s | FileCheck %s
3 define void @and_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
6 %1 = load <16 x i8>* %a
7 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
8 %2 = load <16 x i8>* %b
9 ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
10 %3 = and <16 x i8> %1, %2
11 ; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
12 store <16 x i8> %3, <16 x i8>* %c
13 ; CHECK-DAG: st.b [[R3]], 0($4)
16 ; CHECK: .size and_v16i8
19 define void @and_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
22 %1 = load <8 x i16>* %a
23 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
24 %2 = load <8 x i16>* %b
25 ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
26 %3 = and <8 x i16> %1, %2
27 ; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
28 store <8 x i16> %3, <8 x i16>* %c
29 ; CHECK-DAG: st.h [[R3]], 0($4)
32 ; CHECK: .size and_v8i16
35 define void @and_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
38 %1 = load <4 x i32>* %a
39 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
40 %2 = load <4 x i32>* %b
41 ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
42 %3 = and <4 x i32> %1, %2
43 ; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
44 store <4 x i32> %3, <4 x i32>* %c
45 ; CHECK-DAG: st.w [[R3]], 0($4)
48 ; CHECK: .size and_v4i32
51 define void @and_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
54 %1 = load <2 x i64>* %a
55 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
56 %2 = load <2 x i64>* %b
57 ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
58 %3 = and <2 x i64> %1, %2
59 ; CHECK-DAG: and.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
60 store <2 x i64> %3, <2 x i64>* %c
61 ; CHECK-DAG: st.d [[R3]], 0($4)
64 ; CHECK: .size and_v2i64
67 define void @or_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
70 %1 = load <16 x i8>* %a
71 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
72 %2 = load <16 x i8>* %b
73 ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
74 %3 = or <16 x i8> %1, %2
75 ; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
76 store <16 x i8> %3, <16 x i8>* %c
77 ; CHECK-DAG: st.b [[R3]], 0($4)
80 ; CHECK: .size or_v16i8
83 define void @or_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
86 %1 = load <8 x i16>* %a
87 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
88 %2 = load <8 x i16>* %b
89 ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
90 %3 = or <8 x i16> %1, %2
91 ; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
92 store <8 x i16> %3, <8 x i16>* %c
93 ; CHECK-DAG: st.h [[R3]], 0($4)
96 ; CHECK: .size or_v8i16
99 define void @or_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
102 %1 = load <4 x i32>* %a
103 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
104 %2 = load <4 x i32>* %b
105 ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
106 %3 = or <4 x i32> %1, %2
107 ; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
108 store <4 x i32> %3, <4 x i32>* %c
109 ; CHECK-DAG: st.w [[R3]], 0($4)
112 ; CHECK: .size or_v4i32
115 define void @or_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
118 %1 = load <2 x i64>* %a
119 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
120 %2 = load <2 x i64>* %b
121 ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
122 %3 = or <2 x i64> %1, %2
123 ; CHECK-DAG: or.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
124 store <2 x i64> %3, <2 x i64>* %c
125 ; CHECK-DAG: st.d [[R3]], 0($4)
128 ; CHECK: .size or_v2i64
131 define void @xor_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
134 %1 = load <16 x i8>* %a
135 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
136 %2 = load <16 x i8>* %b
137 ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
138 %3 = xor <16 x i8> %1, %2
139 ; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
140 store <16 x i8> %3, <16 x i8>* %c
141 ; CHECK-DAG: st.b [[R3]], 0($4)
144 ; CHECK: .size xor_v16i8
147 define void @xor_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
150 %1 = load <8 x i16>* %a
151 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
152 %2 = load <8 x i16>* %b
153 ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
154 %3 = xor <8 x i16> %1, %2
155 ; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
156 store <8 x i16> %3, <8 x i16>* %c
157 ; CHECK-DAG: st.h [[R3]], 0($4)
160 ; CHECK: .size xor_v8i16
163 define void @xor_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
166 %1 = load <4 x i32>* %a
167 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
168 %2 = load <4 x i32>* %b
169 ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
170 %3 = xor <4 x i32> %1, %2
171 ; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
172 store <4 x i32> %3, <4 x i32>* %c
173 ; CHECK-DAG: st.w [[R3]], 0($4)
176 ; CHECK: .size xor_v4i32
179 define void @xor_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
182 %1 = load <2 x i64>* %a
183 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
184 %2 = load <2 x i64>* %b
185 ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
186 %3 = xor <2 x i64> %1, %2
187 ; CHECK-DAG: xor.v [[R3:\$w[0-9]+]], [[R1]], [[R2]]
188 store <2 x i64> %3, <2 x i64>* %c
189 ; CHECK-DAG: st.d [[R3]], 0($4)
192 ; CHECK: .size xor_v2i64
195 define void @sll_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
198 %1 = load <16 x i8>* %a
199 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
200 %2 = load <16 x i8>* %b
201 ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
202 %3 = shl <16 x i8> %1, %2
203 ; CHECK-DAG: sll.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
204 store <16 x i8> %3, <16 x i8>* %c
205 ; CHECK-DAG: st.b [[R3]], 0($4)
208 ; CHECK: .size sll_v16i8
211 define void @sll_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
214 %1 = load <8 x i16>* %a
215 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
216 %2 = load <8 x i16>* %b
217 ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
218 %3 = shl <8 x i16> %1, %2
219 ; CHECK-DAG: sll.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
220 store <8 x i16> %3, <8 x i16>* %c
221 ; CHECK-DAG: st.h [[R3]], 0($4)
224 ; CHECK: .size sll_v8i16
227 define void @sll_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
230 %1 = load <4 x i32>* %a
231 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
232 %2 = load <4 x i32>* %b
233 ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
234 %3 = shl <4 x i32> %1, %2
235 ; CHECK-DAG: sll.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
236 store <4 x i32> %3, <4 x i32>* %c
237 ; CHECK-DAG: st.w [[R3]], 0($4)
240 ; CHECK: .size sll_v4i32
243 define void @sll_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
246 %1 = load <2 x i64>* %a
247 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
248 %2 = load <2 x i64>* %b
249 ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
250 %3 = shl <2 x i64> %1, %2
251 ; CHECK-DAG: sll.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
252 store <2 x i64> %3, <2 x i64>* %c
253 ; CHECK-DAG: st.d [[R3]], 0($4)
256 ; CHECK: .size sll_v2i64
259 define void @sra_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
262 %1 = load <16 x i8>* %a
263 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
264 %2 = load <16 x i8>* %b
265 ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
266 %3 = ashr <16 x i8> %1, %2
267 ; CHECK-DAG: sra.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
268 store <16 x i8> %3, <16 x i8>* %c
269 ; CHECK-DAG: st.b [[R3]], 0($4)
272 ; CHECK: .size sra_v16i8
275 define void @sra_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
278 %1 = load <8 x i16>* %a
279 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
280 %2 = load <8 x i16>* %b
281 ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
282 %3 = ashr <8 x i16> %1, %2
283 ; CHECK-DAG: sra.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
284 store <8 x i16> %3, <8 x i16>* %c
285 ; CHECK-DAG: st.h [[R3]], 0($4)
288 ; CHECK: .size sra_v8i16
291 define void @sra_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
294 %1 = load <4 x i32>* %a
295 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
296 %2 = load <4 x i32>* %b
297 ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
298 %3 = ashr <4 x i32> %1, %2
299 ; CHECK-DAG: sra.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
300 store <4 x i32> %3, <4 x i32>* %c
301 ; CHECK-DAG: st.w [[R3]], 0($4)
304 ; CHECK: .size sra_v4i32
307 define void @sra_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
310 %1 = load <2 x i64>* %a
311 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
312 %2 = load <2 x i64>* %b
313 ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
314 %3 = ashr <2 x i64> %1, %2
315 ; CHECK-DAG: sra.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
316 store <2 x i64> %3, <2 x i64>* %c
317 ; CHECK-DAG: st.d [[R3]], 0($4)
320 ; CHECK: .size sra_v2i64
323 define void @srl_v16i8(<16 x i8>* %c, <16 x i8>* %a, <16 x i8>* %b) nounwind {
326 %1 = load <16 x i8>* %a
327 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
328 %2 = load <16 x i8>* %b
329 ; CHECK-DAG: ld.b [[R2:\$w[0-9]+]], 0($6)
330 %3 = lshr <16 x i8> %1, %2
331 ; CHECK-DAG: srl.b [[R3:\$w[0-9]+]], [[R1]], [[R2]]
332 store <16 x i8> %3, <16 x i8>* %c
333 ; CHECK-DAG: st.b [[R3]], 0($4)
336 ; CHECK: .size srl_v16i8
339 define void @srl_v8i16(<8 x i16>* %c, <8 x i16>* %a, <8 x i16>* %b) nounwind {
342 %1 = load <8 x i16>* %a
343 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
344 %2 = load <8 x i16>* %b
345 ; CHECK-DAG: ld.h [[R2:\$w[0-9]+]], 0($6)
346 %3 = lshr <8 x i16> %1, %2
347 ; CHECK-DAG: srl.h [[R3:\$w[0-9]+]], [[R1]], [[R2]]
348 store <8 x i16> %3, <8 x i16>* %c
349 ; CHECK-DAG: st.h [[R3]], 0($4)
352 ; CHECK: .size srl_v8i16
355 define void @srl_v4i32(<4 x i32>* %c, <4 x i32>* %a, <4 x i32>* %b) nounwind {
358 %1 = load <4 x i32>* %a
359 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
360 %2 = load <4 x i32>* %b
361 ; CHECK-DAG: ld.w [[R2:\$w[0-9]+]], 0($6)
362 %3 = lshr <4 x i32> %1, %2
363 ; CHECK-DAG: srl.w [[R3:\$w[0-9]+]], [[R1]], [[R2]]
364 store <4 x i32> %3, <4 x i32>* %c
365 ; CHECK-DAG: st.w [[R3]], 0($4)
368 ; CHECK: .size srl_v4i32
371 define void @srl_v2i64(<2 x i64>* %c, <2 x i64>* %a, <2 x i64>* %b) nounwind {
374 %1 = load <2 x i64>* %a
375 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
376 %2 = load <2 x i64>* %b
377 ; CHECK-DAG: ld.d [[R2:\$w[0-9]+]], 0($6)
378 %3 = lshr <2 x i64> %1, %2
379 ; CHECK-DAG: srl.d [[R3:\$w[0-9]+]], [[R1]], [[R2]]
380 store <2 x i64> %3, <2 x i64>* %c
381 ; CHECK-DAG: st.d [[R3]], 0($4)
384 ; CHECK: .size srl_v2i64
387 define void @ctlz_v16i8(<16 x i8>* %c, <16 x i8>* %a) nounwind {
390 %1 = load <16 x i8>* %a
391 ; CHECK-DAG: ld.b [[R1:\$w[0-9]+]], 0($5)
392 %2 = tail call <16 x i8> @llvm.ctlz.v16i8 (<16 x i8> %1)
393 ; CHECK-DAG: nlzc.b [[R3:\$w[0-9]+]], [[R1]]
394 store <16 x i8> %2, <16 x i8>* %c
395 ; CHECK-DAG: st.b [[R3]], 0($4)
398 ; CHECK: .size ctlz_v16i8
401 define void @ctlz_v8i16(<8 x i16>* %c, <8 x i16>* %a) nounwind {
404 %1 = load <8 x i16>* %a
405 ; CHECK-DAG: ld.h [[R1:\$w[0-9]+]], 0($5)
406 %2 = tail call <8 x i16> @llvm.ctlz.v8i16 (<8 x i16> %1)
407 ; CHECK-DAG: nlzc.h [[R3:\$w[0-9]+]], [[R1]]
408 store <8 x i16> %2, <8 x i16>* %c
409 ; CHECK-DAG: st.h [[R3]], 0($4)
412 ; CHECK: .size ctlz_v8i16
415 define void @ctlz_v4i32(<4 x i32>* %c, <4 x i32>* %a) nounwind {
418 %1 = load <4 x i32>* %a
419 ; CHECK-DAG: ld.w [[R1:\$w[0-9]+]], 0($5)
420 %2 = tail call <4 x i32> @llvm.ctlz.v4i32 (<4 x i32> %1)
421 ; CHECK-DAG: nlzc.w [[R3:\$w[0-9]+]], [[R1]]
422 store <4 x i32> %2, <4 x i32>* %c
423 ; CHECK-DAG: st.w [[R3]], 0($4)
426 ; CHECK: .size ctlz_v4i32
429 define void @ctlz_v2i64(<2 x i64>* %c, <2 x i64>* %a) nounwind {
432 %1 = load <2 x i64>* %a
433 ; CHECK-DAG: ld.d [[R1:\$w[0-9]+]], 0($5)
434 %2 = tail call <2 x i64> @llvm.ctlz.v2i64 (<2 x i64> %1)
435 ; CHECK-DAG: nlzc.d [[R3:\$w[0-9]+]], [[R1]]
436 store <2 x i64> %2, <2 x i64>* %c
437 ; CHECK-DAG: st.d [[R3]], 0($4)
440 ; CHECK: .size ctlz_v2i64
443 declare <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %val)
444 declare <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %val)
445 declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val)
446 declare <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %val)