1; RUN: opt < %s -instcombine -S | FileCheck %s
2
3define <4 x float> @test1(<4 x float> %v1) {
4; CHECK-LABEL: @test1(
5; CHECK: ret <4 x float> %v1
6  %v2 = shufflevector <4 x float> %v1, <4 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
7  ret <4 x float> %v2
8}
9
10define <4 x float> @test2(<4 x float> %v1) {
11; CHECK-LABEL: @test2(
12; CHECK: ret <4 x float> %v1
13  %v2 = shufflevector <4 x float> %v1, <4 x float> %v1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
14  ret <4 x float> %v2
15}
16
17define float @test3(<4 x float> %A, <4 x float> %B, float %f) {
18; CHECK-LABEL: @test3(
19; CHECK: ret float %f
20        %C = insertelement <4 x float> %A, float %f, i32 0
21        %D = shufflevector <4 x float> %C, <4 x float> %B, <4 x i32> <i32 5, i32 0, i32 2, i32 7>
22        %E = extractelement <4 x float> %D, i32 1
23        ret float %E
24}
25
26define i32 @test4(<4 x i32> %X) {
27; CHECK-LABEL: @test4(
28; CHECK-NEXT: extractelement
29; CHECK-NEXT: ret
30        %tmp152.i53899.i = shufflevector <4 x i32> %X, <4 x i32> undef, <4 x i32> zeroinitializer
31        %tmp34 = extractelement <4 x i32> %tmp152.i53899.i, i32 0
32        ret i32 %tmp34
33}
34
35define i32 @test5(<4 x i32> %X) {
36; CHECK-LABEL: @test5(
37; CHECK-NEXT: extractelement
38; CHECK-NEXT: ret
39        %tmp152.i53899.i = shufflevector <4 x i32> %X, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 undef, i32 undef>
40        %tmp34 = extractelement <4 x i32> %tmp152.i53899.i, i32 0
41        ret i32 %tmp34
42}
43
44define float @test6(<4 x float> %X) {
45; CHECK-LABEL: @test6(
46; CHECK-NEXT: extractelement
47; CHECK-NEXT: ret
48        %X1 = bitcast <4 x float> %X to <4 x i32>
49        %tmp152.i53899.i = shufflevector <4 x i32> %X1, <4 x i32> undef, <4 x i32> zeroinitializer
50        %tmp152.i53900.i = bitcast <4 x i32> %tmp152.i53899.i to <4 x float>
51        %tmp34 = extractelement <4 x float> %tmp152.i53900.i, i32 0
52        ret float %tmp34
53}
54
55define <4 x float> @test7(<4 x float> %tmp45.i) {
56; CHECK-LABEL: @test7(
57; CHECK-NEXT: ret <4 x float> %tmp45.i
58        %tmp1642.i = shufflevector <4 x float> %tmp45.i, <4 x float> undef, <4 x i32> < i32 0, i32 1, i32 6, i32 7 >
59        ret <4 x float> %tmp1642.i
60}
61
62; This should turn into a single shuffle.
63define <4 x float> @test8(<4 x float> %tmp, <4 x float> %tmp1) {
64; CHECK-LABEL: @test8(
65; CHECK-NEXT: shufflevector
66; CHECK-NEXT: ret
67        %tmp4 = extractelement <4 x float> %tmp, i32 1
68        %tmp2 = extractelement <4 x float> %tmp, i32 3
69        %tmp1.upgrd.1 = extractelement <4 x float> %tmp1, i32 0
70        %tmp128 = insertelement <4 x float> undef, float %tmp4, i32 0
71        %tmp130 = insertelement <4 x float> %tmp128, float undef, i32 1
72        %tmp132 = insertelement <4 x float> %tmp130, float %tmp2, i32 2
73        %tmp134 = insertelement <4 x float> %tmp132, float %tmp1.upgrd.1, i32 3
74        ret <4 x float> %tmp134
75}
76
77; Test fold of two shuffles where the first shuffle vectors inputs are a
78; different length then the second.
79define <4 x i8> @test9(<16 x i8> %tmp6) nounwind {
80; CHECK-LABEL: @test9(
81; CHECK-NEXT: shufflevector
82; CHECK-NEXT: ret
83	%tmp7 = shufflevector <16 x i8> %tmp6, <16 x i8> undef, <4 x i32> < i32 13, i32 9, i32 4, i32 13 >		; <<4 x i8>> [#uses=1]
84	%tmp9 = shufflevector <4 x i8> %tmp7, <4 x i8> undef, <4 x i32> < i32 3, i32 1, i32 2, i32 0 >		; <<4 x i8>> [#uses=1]
85	ret <4 x i8> %tmp9
86}
87
88; Same as test9, but make sure that "undef" mask values are not confused with
89; mask values of 2*N, where N is the mask length.  These shuffles should not
90; be folded (because [8,9,4,8] may not be a mask supported by the target).
91define <4 x i8> @test9a(<16 x i8> %tmp6) nounwind {
92; CHECK-LABEL: @test9a(
93; CHECK-NEXT: shufflevector
94; CHECK-NEXT: shufflevector
95; CHECK-NEXT: ret
96	%tmp7 = shufflevector <16 x i8> %tmp6, <16 x i8> undef, <4 x i32> < i32 undef, i32 9, i32 4, i32 8 >		; <<4 x i8>> [#uses=1]
97	%tmp9 = shufflevector <4 x i8> %tmp7, <4 x i8> undef, <4 x i32> < i32 3, i32 1, i32 2, i32 0 >		; <<4 x i8>> [#uses=1]
98	ret <4 x i8> %tmp9
99}
100
101; Test fold of two shuffles where the first shuffle vectors inputs are a
102; different length then the second.
103define <4 x i8> @test9b(<4 x i8> %tmp6, <4 x i8> %tmp7) nounwind {
104; CHECK-LABEL: @test9b(
105; CHECK-NEXT: shufflevector
106; CHECK-NEXT: ret
107  %tmp1 = shufflevector <4 x i8> %tmp6, <4 x i8> %tmp7, <8 x i32> <i32 0, i32 1, i32 4, i32 5, i32 4, i32 5, i32 2, i32 3>		; <<4 x i8>> [#uses=1]
108  %tmp9 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 4, i32 5>		; <<4 x i8>> [#uses=1]
109  ret <4 x i8> %tmp9
110}
111
112; Redundant vector splats should be removed.  Radar 8597790.
113define <4 x i32> @test10(<4 x i32> %tmp5) nounwind {
114; CHECK-LABEL: @test10(
115; CHECK-NEXT: shufflevector
116; CHECK-NEXT: ret
117  %tmp6 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
118  %tmp7 = shufflevector <4 x i32> %tmp6, <4 x i32> undef, <4 x i32> zeroinitializer
119  ret <4 x i32> %tmp7
120}
121
122; Test fold of two shuffles where the two shufflevector inputs's op1 are
123; the same
124define <8 x i8> @test11(<16 x i8> %tmp6) nounwind {
125; CHECK-LABEL: @test11(
126; CHECK-NEXT: shufflevector <16 x i8> %tmp6, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
127; CHECK-NEXT: ret
128  %tmp1 = shufflevector <16 x i8> %tmp6, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>		; <<4 x i8>> [#uses=1]
129  %tmp2 = shufflevector <16 x i8> %tmp6, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>		; <<4 x i8>> [#uses=1]
130  %tmp3 = shufflevector <4 x i8> %tmp1, <4 x i8> %tmp2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>		; <<8 x i8>> [#uses=1]
131  ret <8 x i8> %tmp3
132}
133
134; Test fold of two shuffles where the first shufflevector's inputs are
135; the same as the second
136define <8 x i8> @test12(<8 x i8> %tmp6, <8 x i8> %tmp2) nounwind {
137; CHECK-LABEL: @test12(
138; CHECK-NEXT: shufflevector <8 x i8> %tmp6, <8 x i8> %tmp2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 9, i32 8, i32 11, i32 12>
139; CHECK-NEXT: ret
140  %tmp1 = shufflevector <8 x i8> %tmp6, <8 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 undef, i32 7>	; <<8 x i8>> [#uses=1]
141  %tmp3 = shufflevector <8 x i8> %tmp1, <8 x i8> %tmp2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 9, i32 8, i32 11, i32 12>		; <<8 x i8>> [#uses=1]
142  ret <8 x i8> %tmp3
143}
144
145; Test fold of two shuffles where the first shufflevector's inputs are
146; the same as the second
147define <8 x i8> @test12a(<8 x i8> %tmp6, <8 x i8> %tmp2) nounwind {
148; CHECK-LABEL: @test12a(
149; CHECK-NEXT: shufflevector <8 x i8> %tmp2, <8 x i8> %tmp6, <8 x i32> <i32 0, i32 3, i32 1, i32 4, i32 8, i32 9, i32 10, i32 11>
150; CHECK-NEXT: ret
151  %tmp1 = shufflevector <8 x i8> %tmp6, <8 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 undef, i32 7>	; <<8 x i8>> [#uses=1]
152  %tmp3 = shufflevector <8 x i8> %tmp2, <8 x i8> %tmp1, <8 x i32> <i32 0, i32 3, i32 1, i32 4, i32 8, i32 9, i32 10, i32 11>		; <<8 x i8>> [#uses=1]
153  ret <8 x i8> %tmp3
154}
155
156define <2 x i8> @test13a(i8 %x1, i8 %x2) {
157; CHECK-LABEL: @test13a(
158; CHECK-NEXT: insertelement {{.*}} undef, i8 %x1, i32 1
159; CHECK-NEXT: insertelement {{.*}} i8 %x2, i32 0
160; CHECK-NEXT: add {{.*}} <i8 7, i8 5>
161; CHECK-NEXT: ret
162  %A = insertelement <2 x i8> undef, i8 %x1, i32 0
163  %B = insertelement <2 x i8> %A, i8 %x2, i32 1
164  %C = add <2 x i8> %B, <i8 5, i8 7>
165  %D = shufflevector <2 x i8> %C, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
166  ret <2 x i8> %D
167}
168
169define <2 x i8> @test13b(i8 %x) {
170; CHECK-LABEL: @test13b(
171; CHECK-NEXT: insertelement <2 x i8> undef, i8 %x, i32 1
172; CHECK-NEXT: ret
173  %A = insertelement <2 x i8> undef, i8 %x, i32 0
174  %B = shufflevector <2 x i8> %A, <2 x i8> undef, <2 x i32> <i32 undef, i32 0>
175  ret <2 x i8> %B
176}
177
178define <2 x i8> @test13c(i8 %x1, i8 %x2) {
179; CHECK-LABEL: @test13c(
180; CHECK-NEXT: insertelement <2 x i8> {{.*}}, i32 0
181; CHECK-NEXT: insertelement <2 x i8> {{.*}}, i32 1
182; CHECK-NEXT: ret
183  %A = insertelement <4 x i8> undef, i8 %x1, i32 0
184  %B = insertelement <4 x i8> %A, i8 %x2, i32 2
185  %C = shufflevector <4 x i8> %B, <4 x i8> undef, <2 x i32> <i32 0, i32 2>
186  ret <2 x i8> %C
187}
188
189define void @test14(i16 %conv10) {
190  %tmp = alloca <4 x i16>, align 8
191  %vecinit6 = insertelement <4 x i16> undef, i16 23, i32 3
192  store <4 x i16> %vecinit6, <4 x i16>* undef
193  %tmp1 = load <4 x i16>, <4 x i16>* undef
194  %vecinit11 = insertelement <4 x i16> undef, i16 %conv10, i32 3
195  %div = udiv <4 x i16> %tmp1, %vecinit11
196  store <4 x i16> %div, <4 x i16>* %tmp
197  %tmp4 = load <4 x i16>, <4 x i16>* %tmp
198  %tmp5 = shufflevector <4 x i16> %tmp4, <4 x i16> undef, <2 x i32> <i32 2, i32 0>
199  %cmp = icmp ule <2 x i16> %tmp5, undef
200  %sext = sext <2 x i1> %cmp to <2 x i16>
201  ret void
202}
203
204; Check that sequences of insert/extract element are
205; collapsed into valid shuffle instruction with correct shuffle indexes.
206
207define <4 x float> @test15a(<4 x float> %LHS, <4 x float> %RHS) {
208; CHECK-LABEL: @test15a
209; CHECK-NEXT: shufflevector <4 x float> %LHS, <4 x float> %RHS, <4 x i32> <i32 4, i32 0, i32 6, i32 6>
210; CHECK-NEXT: ret <4 x float> %tmp4
211  %tmp1 = extractelement <4 x float> %LHS, i32 0
212  %tmp2 = insertelement <4 x float> %RHS, float %tmp1, i32 1
213  %tmp3 = extractelement <4 x float> %RHS, i32 2
214  %tmp4 = insertelement <4 x float> %tmp2, float %tmp3, i32 3
215  ret <4 x float> %tmp4
216}
217
218define <4 x float> @test15b(<4 x float> %LHS, <4 x float> %RHS) {
219; CHECK-LABEL: @test15b
220; CHECK-NEXT: shufflevector <4 x float> %LHS, <4 x float> %RHS, <4 x i32> <i32 4, i32 3, i32 6, i32 6>
221; CHECK-NEXT: ret <4 x float> %tmp5
222  %tmp0 = extractelement <4 x float> %LHS, i32 3
223  %tmp1 = insertelement <4 x float> %RHS, float %tmp0, i32 0
224  %tmp2 = extractelement <4 x float> %tmp1, i32 0
225  %tmp3 = insertelement <4 x float> %RHS, float %tmp2, i32 1
226  %tmp4 = extractelement <4 x float> %RHS, i32 2
227  %tmp5 = insertelement <4 x float> %tmp3, float %tmp4, i32 3
228  ret <4 x float> %tmp5
229}
230
231define <1 x i32> @test16a(i32 %ele) {
232; CHECK-LABEL: @test16a(
233; CHECK-NEXT: ret <1 x i32> <i32 2>
234  %tmp0 = insertelement <2 x i32> <i32 1, i32 undef>, i32 %ele, i32 1
235  %tmp1 = shl <2 x i32> %tmp0, <i32 1, i32 1>
236  %tmp2 = shufflevector <2 x i32> %tmp1, <2 x i32> undef, <1 x i32> <i32 0>
237  ret <1 x i32> %tmp2
238}
239
240define <4 x i8> @test16b(i8 %ele) {
241; CHECK-LABEL: @test16b(
242; CHECK-NEXT: ret <4 x i8> <i8 2, i8 2, i8 2, i8 2>
243  %tmp0 = insertelement <8 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 undef, i8 1>, i8 %ele, i32 6
244  %tmp1 = shl <8 x i8> %tmp0, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
245  %tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 4>
246  ret <4 x i8> %tmp2
247}
248
249; If composition of two shuffles is identity, shuffles can be removed.
250define <4 x i32> @shuffle_17ident(<4 x i32> %v) nounwind uwtable {
251; CHECK-LABEL: @shuffle_17ident(
252; CHECK-NOT: shufflevector
253  %shuffle = shufflevector <4 x i32> %v, <4 x i32> zeroinitializer,
254                           <4 x i32> <i32 1, i32 2, i32 3, i32 0>
255  %shuffle2 = shufflevector <4 x i32> %shuffle, <4 x i32> zeroinitializer,
256                            <4 x i32> <i32 3, i32 0, i32 1, i32 2>
257  ret <4 x i32> %shuffle2
258}
259
260; swizzle can be put after operation
261define <4 x i32> @shuffle_17and(<4 x i32> %v1, <4 x i32> %v2) nounwind uwtable {
262; CHECK-LABEL: @shuffle_17and(
263; CHECK-NOT: shufflevector
264; CHECK: and <4 x i32> %v1, %v2
265; CHECK: shufflevector
266  %t1 = shufflevector <4 x i32> %v1, <4 x i32> zeroinitializer,
267                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
268  %t2 = shufflevector <4 x i32> %v2, <4 x i32> zeroinitializer,
269                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
270  %r = and <4 x i32> %t1, %t2
271  ret <4 x i32> %r
272}
273
274define <4 x i32> @shuffle_17add(<4 x i32> %v1, <4 x i32> %v2) nounwind uwtable {
275; CHECK-LABEL: @shuffle_17add(
276; CHECK-NOT: shufflevector
277; CHECK: add <4 x i32> %v1, %v2
278; CHECK: shufflevector
279  %t1 = shufflevector <4 x i32> %v1, <4 x i32> zeroinitializer,
280                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
281  %t2 = shufflevector <4 x i32> %v2, <4 x i32> zeroinitializer,
282                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
283  %r = add <4 x i32> %t1, %t2
284  ret <4 x i32> %r
285}
286
287define <4 x i32> @shuffle_17addnsw(<4 x i32> %v1, <4 x i32> %v2) nounwind uwtable {
288; CHECK-LABEL: @shuffle_17addnsw(
289; CHECK-NOT: shufflevector
290; CHECK: add nsw <4 x i32> %v1, %v2
291; CHECK: shufflevector
292  %t1 = shufflevector <4 x i32> %v1, <4 x i32> zeroinitializer,
293                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
294  %t2 = shufflevector <4 x i32> %v2, <4 x i32> zeroinitializer,
295                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
296  %r = add nsw <4 x i32> %t1, %t2
297  ret <4 x i32> %r
298}
299
300define <4 x i32> @shuffle_17addnuw(<4 x i32> %v1, <4 x i32> %v2) nounwind uwtable {
301; CHECK-LABEL: @shuffle_17addnuw(
302; CHECK-NOT: shufflevector
303; CHECK: add nuw <4 x i32> %v1, %v2
304; CHECK: shufflevector
305  %t1 = shufflevector <4 x i32> %v1, <4 x i32> zeroinitializer,
306                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
307  %t2 = shufflevector <4 x i32> %v2, <4 x i32> zeroinitializer,
308                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
309  %r = add nuw <4 x i32> %t1, %t2
310  ret <4 x i32> %r
311}
312
313define <4 x float> @shuffle_17fsub_fast(<4 x float> %v1, <4 x float> %v2) nounwind uwtable {
314; CHECK-LABEL: @shuffle_17fsub_fast(
315; CHECK-NEXT: [[VAR1:%[a-zA-Z0-9.]+]] = fsub fast <4 x float> %v1, %v2
316; CHECK-NEXT: shufflevector <4 x float> [[VAR1]], <4 x float> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
317; CHECK-NEXT: ret <4 x float>
318  %t1 = shufflevector <4 x float> %v1, <4 x float> zeroinitializer,
319                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
320  %t2 = shufflevector <4 x float> %v2, <4 x float> zeroinitializer,
321                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
322  %r = fsub fast <4 x float> %t1, %t2
323  ret <4 x float> %r
324}
325
326define <4 x i32> @shuffle_17addconst(<4 x i32> %v1, <4 x i32> %v2) {
327; CHECK-LABEL: @shuffle_17addconst(
328; CHECK-NOT: shufflevector
329; CHECK: [[VAR1:%[a-zA-Z0-9.]+]] = add <4 x i32> %v1, <i32 4, i32 1, i32 2, i32 3>
330; CHECK: [[VAR2:%[a-zA-Z0-9.]+]] = shufflevector <4 x i32> [[VAR1]], <4 x i32> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
331; CHECK: ret <4 x i32> [[VAR2]]
332  %t1 = shufflevector <4 x i32> %v1, <4 x i32> zeroinitializer,
333                      <4 x i32> <i32 1, i32 2, i32 3, i32 0>
334  %r = add <4 x i32> %t1, <i32 1, i32 2, i32 3, i32 4>
335  ret <4 x i32> %r
336}
337
338define <4 x i32> @shuffle_17add2(<4 x i32> %v) {
339; CHECK-LABEL: @shuffle_17add2(
340; CHECK-NOT: shufflevector
341; CHECK: [[VAR:%[a-zA-Z0-9.]+]] = shl <4 x i32> %v, <i32 1, i32 1, i32 1, i32 1>
342; CHECK: ret <4 x i32> [[VAR]]
343  %t1 = shufflevector <4 x i32> %v, <4 x i32> zeroinitializer,
344                      <4 x i32> <i32 3, i32 2, i32 1, i32 0>
345  %t2 = add <4 x i32> %t1, %t1
346  %r = shufflevector <4 x i32> %t2, <4 x i32> zeroinitializer,
347                     <4 x i32> <i32 3, i32 2, i32 1, i32 0>
348  ret <4 x i32> %r
349}
350
351define <4 x i32> @shuffle_17mulsplat(<4 x i32> %v) {
352; CHECK-LABEL: @shuffle_17mulsplat(
353; CHECK-NOT: shufflevector
354; CHECK: [[VAR1:%[a-zA-Z0-9.]+]] = mul <4 x i32> %v, %v
355; CHECK: [[VAR2:%[a-zA-Z0-9.]+]] = shufflevector <4 x i32> [[VAR1]], <4 x i32> undef, <4 x i32> zeroinitializer
356; CHECK: ret <4 x i32> [[VAR2]]
357  %s1 = shufflevector <4 x i32> %v,
358                      <4 x i32> zeroinitializer,
359                      <4 x i32> zeroinitializer
360  %m1 = mul <4 x i32> %s1, %s1
361  %s2 = shufflevector <4 x i32> %m1,
362                      <4 x i32> zeroinitializer,
363                      <4 x i32> <i32 1, i32 1, i32 1, i32 1>
364  ret <4 x i32> %s2
365}
366
367; Do not reorder shuffle and binop if LHS of shuffles are of different size
368define <2 x i32> @pr19717(<4 x i32> %in0, <2 x i32> %in1) {
369; CHECK-LABEL: @pr19717(
370; CHECK: shufflevector
371; CHECK: shufflevector
372; CHECK: mul
373  %shuffle = shufflevector <4 x i32> %in0, <4 x i32> %in0, <2 x i32> zeroinitializer
374  %shuffle4 = shufflevector <2 x i32> %in1, <2 x i32> %in1, <2 x i32> zeroinitializer
375  %mul = mul <2 x i32> %shuffle, %shuffle4
376  ret <2 x i32> %mul
377}
378
379define <4 x i16> @pr19717a(<8 x i16> %in0, <8 x i16> %in1) {
380; CHECK-LABEL: @pr19717a(
381; CHECK: [[VAR1:%[a-zA-Z0-9.]+]] = mul <8 x i16> %in0, %in1
382; CHECK: [[VAR2:%[a-zA-Z0-9.]+]] = shufflevector <8 x i16> [[VAR1]], <8 x i16> undef, <4 x i32> <i32 5, i32 5, i32 5, i32 5>
383; CHECK: ret <4 x i16> [[VAR2]]
384  %shuffle = shufflevector <8 x i16> %in0, <8 x i16> %in0, <4 x i32> <i32 5, i32 5, i32 5, i32 5>
385  %shuffle1 = shufflevector <8 x i16> %in1, <8 x i16> %in1, <4 x i32> <i32 5, i32 5, i32 5, i32 5>
386  %mul = mul <4 x i16> %shuffle, %shuffle1
387  ret <4 x i16> %mul
388}
389
390define <8 x i8> @pr19730(<16 x i8> %in0) {
391; CHECK-LABEL: @pr19730(
392; CHECK: shufflevector
393  %shuffle = shufflevector <16 x i8> %in0, <16 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
394  %shuffle1 = shufflevector <8 x i8> %shuffle, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
395  ret <8 x i8> %shuffle1
396}
397
398define i32 @pr19737(<4 x i32> %in0) {
399; CHECK-LABEL: @pr19737(
400; CHECK: [[VAR:%[a-zA-Z0-9.]+]] = extractelement <4 x i32> %in0, i32 0
401; CHECK: ret i32 [[VAR]]
402  %shuffle.i = shufflevector <4 x i32> zeroinitializer, <4 x i32> %in0, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
403  %neg.i = xor <4 x i32> %shuffle.i, <i32 -1, i32 -1, i32 -1, i32 -1>
404  %and.i = and <4 x i32> %in0, %neg.i
405  %rv = extractelement <4 x i32> %and.i, i32 0
406  ret i32 %rv
407}
408
409; In PR20059 ( http://llvm.org/pr20059 ), shufflevector operations are reordered/removed
410; for an srem operation. This is not a valid optimization because it may cause a trap
411; on div-by-zero.
412
413define <4 x i32> @pr20059(<4 x i32> %p1, <4 x i32> %p2) {
414; CHECK-LABEL: @pr20059(
415; CHECK-NEXT: %splat1 = shufflevector <4 x i32> %p1, <4 x i32> undef, <4 x i32> zeroinitializer
416; CHECK-NEXT: %splat2 = shufflevector <4 x i32> %p2, <4 x i32> undef, <4 x i32> zeroinitializer
417; CHECK-NEXT: %retval = srem <4 x i32> %splat1, %splat2
418  %splat1 = shufflevector <4 x i32> %p1, <4 x i32> undef, <4 x i32> zeroinitializer
419  %splat2 = shufflevector <4 x i32> %p2, <4 x i32> undef, <4 x i32> zeroinitializer
420  %retval = srem <4 x i32> %splat1, %splat2
421  ret <4 x i32> %retval
422}
423
424define <4 x i32> @pr20114(<4 x i32> %__mask) {
425; CHECK-LABEL: @pr20114
426; CHECK: shufflevector
427; CHECK: and
428  %mask01.i = shufflevector <4 x i32> %__mask, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
429  %masked_new.i.i.i = and <4 x i32> bitcast (<2 x i64> <i64 ptrtoint (<4 x i32> (<4 x i32>)* @pr20114 to i64), i64 ptrtoint (<4 x i32> (<4 x i32>)* @pr20114 to i64)> to <4 x i32>), %mask01.i
430  ret <4 x i32> %masked_new.i.i.i
431}
432
433define <2 x i32*> @pr23113(<4 x i32*> %A) {
434; CHECK-LABEL: @pr23113
435; CHECK: %[[V:.*]] = shufflevector <4 x i32*> %A, <4 x i32*> undef, <2 x i32> <i32 0, i32 1>
436; CHECK-NEXT: ret <2 x i32*> %[[V]]
437  %1 = shufflevector <4 x i32*> %A, <4 x i32*> undef, <2 x i32> <i32 0, i32 1>
438  ret <2 x i32*> %1
439}
440