1; RUN: llc < %s -mcpu=x86-64 -mattr=-sse2 -x86-experimental-vector-shuffle-lowering | FileCheck %s --check-prefix=SSE1
2
3target triple = "x86_64-unknown-unknown"
4
5define <4 x float> @shuffle_v4f32_0001(<4 x float> %a, <4 x float> %b) {
6; SSE1-LABEL: shuffle_v4f32_0001:
7; SSE1:       # BB#0:
8; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,0,1]
9; SSE1-NEXT:    retq
10  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
11  ret <4 x float> %shuffle
12}
13define <4 x float> @shuffle_v4f32_0020(<4 x float> %a, <4 x float> %b) {
14; SSE1-LABEL: shuffle_v4f32_0020:
15; SSE1:       # BB#0:
16; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,2,0]
17; SSE1-NEXT:    retq
18  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 0>
19  ret <4 x float> %shuffle
20}
21define <4 x float> @shuffle_v4f32_0300(<4 x float> %a, <4 x float> %b) {
22; SSE1-LABEL: shuffle_v4f32_0300:
23; SSE1:       # BB#0:
24; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3,0,0]
25; SSE1-NEXT:    retq
26  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 3, i32 0, i32 0>
27  ret <4 x float> %shuffle
28}
29define <4 x float> @shuffle_v4f32_1000(<4 x float> %a, <4 x float> %b) {
30; SSE1-LABEL: shuffle_v4f32_1000:
31; SSE1:       # BB#0:
32; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0,0,0]
33; SSE1-NEXT:    retq
34  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 1, i32 0, i32 0, i32 0>
35  ret <4 x float> %shuffle
36}
37define <4 x float> @shuffle_v4f32_2200(<4 x float> %a, <4 x float> %b) {
38; SSE1-LABEL: shuffle_v4f32_2200:
39; SSE1:       # BB#0:
40; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2,0,0]
41; SSE1-NEXT:    retq
42  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 2, i32 2, i32 0, i32 0>
43  ret <4 x float> %shuffle
44}
45define <4 x float> @shuffle_v4f32_3330(<4 x float> %a, <4 x float> %b) {
46; SSE1-LABEL: shuffle_v4f32_3330:
47; SSE1:       # BB#0:
48; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,0]
49; SSE1-NEXT:    retq
50  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 3, i32 3, i32 3, i32 0>
51  ret <4 x float> %shuffle
52}
53define <4 x float> @shuffle_v4f32_3210(<4 x float> %a, <4 x float> %b) {
54; SSE1-LABEL: shuffle_v4f32_3210:
55; SSE1:       # BB#0:
56; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
57; SSE1-NEXT:    retq
58  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
59  ret <4 x float> %shuffle
60}
61define <4 x float> @shuffle_v4f32_0011(<4 x float> %a, <4 x float> %b) {
62; SSE1-LABEL: shuffle_v4f32_0011:
63; SSE1:       # BB#0:
64; SSE1-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0,0,1,1]
65; SSE1-NEXT:    retq
66  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
67  ret <4 x float> %shuffle
68}
69define <4 x float> @shuffle_v4f32_2233(<4 x float> %a, <4 x float> %b) {
70; SSE1-LABEL: shuffle_v4f32_2233:
71; SSE1:       # BB#0:
72; SSE1-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2,2,3,3]
73; SSE1-NEXT:    retq
74  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
75  ret <4 x float> %shuffle
76}
77define <4 x float> @shuffle_v4f32_0022(<4 x float> %a, <4 x float> %b) {
78; SSE1-LABEL: shuffle_v4f32_0022:
79; SSE1:       # BB#0:
80; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,2,2]
81; SSE1-NEXT:    retq
82  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
83  ret <4 x float> %shuffle
84}
85define <4 x float> @shuffle_v4f32_1133(<4 x float> %a, <4 x float> %b) {
86; SSE1-LABEL: shuffle_v4f32_1133:
87; SSE1:       # BB#0:
88; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,3,3]
89; SSE1-NEXT:    retq
90  %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
91  ret <4 x float> %shuffle
92}
93
94define <4 x float> @shuffle_v4f32_4zzz(<4 x float> %a) {
95; SSE1-LABEL: shuffle_v4f32_4zzz:
96; SSE1:       # BB#0:
97; SSE1-NEXT:    xorps %xmm1, %xmm1
98; SSE1-NEXT:    movss %xmm0, %xmm1
99; SSE1-NEXT:    movaps %xmm1, %xmm0
100; SSE1-NEXT:    retq
101  %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 4, i32 1, i32 2, i32 3>
102  ret <4 x float> %shuffle
103}
104
105define <4 x float> @shuffle_v4f32_z4zz(<4 x float> %a) {
106; SSE1-LABEL: shuffle_v4f32_z4zz:
107; SSE1:       # BB#0:
108; SSE1-NEXT:    xorps %xmm1, %xmm1
109; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
110; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[3,0]
111; SSE1-NEXT:    retq
112  %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 2, i32 4, i32 3, i32 0>
113  ret <4 x float> %shuffle
114}
115
116define <4 x float> @shuffle_v4f32_zz4z(<4 x float> %a) {
117; SSE1-LABEL: shuffle_v4f32_zz4z:
118; SSE1:       # BB#0:
119; SSE1-NEXT:    xorps %xmm1, %xmm1
120; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
121; SSE1-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,2]
122; SSE1-NEXT:    movaps %xmm1, %xmm0
123; SSE1-NEXT:    retq
124  %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 0, i32 0, i32 4, i32 0>
125  ret <4 x float> %shuffle
126}
127
128define <4 x float> @shuffle_v4f32_zuu4(<4 x float> %a) {
129; SSE1-LABEL: shuffle_v4f32_zuu4:
130; SSE1:       # BB#0:
131; SSE1-NEXT:    xorps %xmm1, %xmm1
132; SSE1-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
133; SSE1-NEXT:    movaps %xmm1, %xmm0
134; SSE1-NEXT:    retq
135  %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 0, i32 undef, i32 undef, i32 4>
136  ret <4 x float> %shuffle
137}
138
139define <4 x float> @shuffle_v4f32_zzz7(<4 x float> %a) {
140; SSE1-LABEL: shuffle_v4f32_zzz7:
141; SSE1:       # BB#0:
142; SSE1-NEXT:    xorps %xmm1, %xmm1
143; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
144; SSE1-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
145; SSE1-NEXT:    movaps %xmm1, %xmm0
146; SSE1-NEXT:    retq
147  %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 0, i32 1, i32 2, i32 7>
148  ret <4 x float> %shuffle
149}
150
151define <4 x float> @shuffle_v4f32_z6zz(<4 x float> %a) {
152; SSE1-LABEL: shuffle_v4f32_z6zz:
153; SSE1:       # BB#0:
154; SSE1-NEXT:    xorps %xmm1, %xmm1
155; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
156; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
157; SSE1-NEXT:    retq
158  %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 0, i32 6, i32 2, i32 3>
159  ret <4 x float> %shuffle
160}
161
162define <4 x float> @insert_reg_and_zero_v4f32(float %a) {
163; SSE1-LABEL: insert_reg_and_zero_v4f32:
164; SSE1:       # BB#0:
165; SSE1-NEXT:    xorps %xmm1, %xmm1
166; SSE1-NEXT:    movss %xmm0, %xmm1
167; SSE1-NEXT:    movaps %xmm1, %xmm0
168; SSE1-NEXT:    retq
169  %v = insertelement <4 x float> undef, float %a, i32 0
170  %shuffle = shufflevector <4 x float> %v, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
171  ret <4 x float> %shuffle
172}
173
174define <4 x float> @insert_mem_and_zero_v4f32(float* %ptr) {
175; SSE1-LABEL: insert_mem_and_zero_v4f32:
176; SSE1:       # BB#0:
177; SSE1-NEXT:    movss (%rdi), %xmm0
178; SSE1-NEXT:    retq
179  %a = load float* %ptr
180  %v = insertelement <4 x float> undef, float %a, i32 0
181  %shuffle = shufflevector <4 x float> %v, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
182  ret <4 x float> %shuffle
183}
184
185define <4 x float> @insert_mem_lo_v4f32(<2 x float>* %ptr, <4 x float> %b) {
186; SSE1-LABEL: insert_mem_lo_v4f32:
187; SSE1:       # BB#0:
188; SSE1-NEXT:    movq (%rdi), %rax
189; SSE1-NEXT:    movl %eax, -{{[0-9]+}}(%rsp)
190; SSE1-NEXT:    shrq $32, %rax
191; SSE1-NEXT:    movl %eax, -{{[0-9]+}}(%rsp)
192; SSE1-NEXT:    movss -{{[0-9]+}}(%rsp), %xmm1
193; SSE1-NEXT:    movss -{{[0-9]+}}(%rsp), %xmm2
194; SSE1-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
195; SSE1-NEXT:    xorps %xmm2, %xmm2
196; SSE1-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,1]
197; SSE1-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
198; SSE1-NEXT:    movaps %xmm1, %xmm0
199; SSE1-NEXT:    retq
200  %a = load <2 x float>* %ptr
201  %v = shufflevector <2 x float> %a, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
202  %shuffle = shufflevector <4 x float> %v, <4 x float> %b, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
203  ret <4 x float> %shuffle
204}
205
206define <4 x float> @insert_mem_hi_v4f32(<2 x float>* %ptr, <4 x float> %b) {
207; SSE1-LABEL: insert_mem_hi_v4f32:
208; SSE1:       # BB#0:
209; SSE1-NEXT:    movq (%rdi), %rax
210; SSE1-NEXT:    movl %eax, -{{[0-9]+}}(%rsp)
211; SSE1-NEXT:    shrq $32, %rax
212; SSE1-NEXT:    movl %eax, -{{[0-9]+}}(%rsp)
213; SSE1-NEXT:    movss -{{[0-9]+}}(%rsp), %xmm1
214; SSE1-NEXT:    movss -{{[0-9]+}}(%rsp), %xmm2
215; SSE1-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
216; SSE1-NEXT:    xorps %xmm2, %xmm2
217; SSE1-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,1]
218; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,1]
219; SSE1-NEXT:    retq
220  %a = load <2 x float>* %ptr
221  %v = shufflevector <2 x float> %a, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
222  %shuffle = shufflevector <4 x float> %v, <4 x float> %b, <4 x i32> <i32 4, i32 5, i32 0, i32 1>
223  ret <4 x float> %shuffle
224}
225
226define <4 x float> @shuffle_mem_v4f32_3210(<4 x float>* %ptr) {
227; SSE1-LABEL: shuffle_mem_v4f32_3210:
228; SSE1:       # BB#0:
229; SSE1-NEXT:    movaps (%rdi), %xmm0
230; SSE1-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,2,1,0]
231; SSE1-NEXT:    retq
232  %a = load <4 x float>* %ptr
233  %shuffle = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
234  ret <4 x float> %shuffle
235}
236