1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx  | FileCheck %s --check-prefix=AVX1
3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2
4; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512f  | FileCheck %s --check-prefixes=AVX512,AVX512F
5; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX512,AVX512BW
6
7; trunc(concat(x,y)) -> pack
8
9define <16 x i16> @trunc_concat_packssdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
10; AVX1-LABEL: trunc_concat_packssdw_256:
11; AVX1:       # %bb.0:
12; AVX1-NEXT:    vpsrad $17, %xmm0, %xmm2
13; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
14; AVX1-NEXT:    vpsrad $17, %xmm0, %xmm0
15; AVX1-NEXT:    vpsrad $23, %xmm1, %xmm3
16; AVX1-NEXT:    vpackssdw %xmm3, %xmm2, %xmm2
17; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
18; AVX1-NEXT:    vpsrad $23, %xmm1, %xmm1
19; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
20; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
21; AVX1-NEXT:    retq
22;
23; AVX2-LABEL: trunc_concat_packssdw_256:
24; AVX2:       # %bb.0:
25; AVX2-NEXT:    vpsrad $17, %ymm0, %ymm0
26; AVX2-NEXT:    vpsrad $23, %ymm1, %ymm1
27; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
28; AVX2-NEXT:    retq
29;
30; AVX512-LABEL: trunc_concat_packssdw_256:
31; AVX512:       # %bb.0:
32; AVX512-NEXT:    vpsrad $17, %ymm0, %ymm0
33; AVX512-NEXT:    vpsrad $23, %ymm1, %ymm1
34; AVX512-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
35; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
36; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
37; AVX512-NEXT:    vpmovdw %zmm0, %ymm0
38; AVX512-NEXT:    retq
39  %1 = ashr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
40  %2 = ashr <8 x i32> %a1, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
41  %3 = shufflevector <8 x i32> %1, <8 x i32> %2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15>
42  %4 = trunc <16 x i32> %3 to <16 x i16>
43  ret <16 x i16> %4
44}
45
46define <16 x i16> @trunc_concat_packusdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
47; AVX1-LABEL: trunc_concat_packusdw_256:
48; AVX1:       # %bb.0:
49; AVX1-NEXT:    vpsrld $17, %xmm0, %xmm2
50; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
51; AVX1-NEXT:    vpsrld $17, %xmm0, %xmm0
52; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm1, %ymm1
53; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
54; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
55; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
56; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
57; AVX1-NEXT:    retq
58;
59; AVX2-LABEL: trunc_concat_packusdw_256:
60; AVX2:       # %bb.0:
61; AVX2-NEXT:    vpsrld $17, %ymm0, %ymm0
62; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15]
63; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
64; AVX2-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0
65; AVX2-NEXT:    retq
66;
67; AVX512-LABEL: trunc_concat_packusdw_256:
68; AVX512:       # %bb.0:
69; AVX512-NEXT:    vpsrld $17, %ymm0, %ymm0
70; AVX512-NEXT:    vpandd {{.*}}(%rip){1to8}, %ymm1, %ymm1
71; AVX512-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
72; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
73; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
74; AVX512-NEXT:    vpmovdw %zmm0, %ymm0
75; AVX512-NEXT:    retq
76  %1 = lshr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
77  %2 = and  <8 x i32> %a1, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
78  %3 = shufflevector <8 x i32> %1, <8 x i32> %2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15>
79  %4 = trunc <16 x i32> %3 to <16 x i16>
80  ret <16 x i16> %4
81}
82
83define <32 x i8> @trunc_concat_packsswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
84; AVX1-LABEL: trunc_concat_packsswb_256:
85; AVX1:       # %bb.0:
86; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm2
87; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
88; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm0
89; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm1, %ymm1
90; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
91; AVX1-NEXT:    vpacksswb %xmm3, %xmm0, %xmm0
92; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
93; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
94; AVX1-NEXT:    retq
95;
96; AVX2-LABEL: trunc_concat_packsswb_256:
97; AVX2:       # %bb.0:
98; AVX2-NEXT:    vpsraw $15, %ymm0, %ymm0
99; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
100; AVX2-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0
101; AVX2-NEXT:    retq
102;
103; AVX512F-LABEL: trunc_concat_packsswb_256:
104; AVX512F:       # %bb.0:
105; AVX512F-NEXT:    vpsraw $15, %ymm0, %ymm0
106; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
107; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
108; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
109; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
110; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
111; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
112; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
113; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
114; AVX512F-NEXT:    retq
115;
116; AVX512BW-LABEL: trunc_concat_packsswb_256:
117; AVX512BW:       # %bb.0:
118; AVX512BW-NEXT:    vpsraw $15, %ymm0, %ymm0
119; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
120; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
121; AVX512BW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
122; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
123; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
124; AVX512BW-NEXT:    retq
125  %1 = ashr <16 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
126  %2 = and  <16 x i16> %a1, <i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1>
127  %3 = shufflevector <16 x i16> %1, <16 x i16> %2, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
128  %4 = trunc <32 x i16> %3 to <32 x i8>
129  ret <32 x i8> %4
130}
131
132define <32 x i8> @trunc_concat_packuswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
133; AVX1-LABEL: trunc_concat_packuswb_256:
134; AVX1:       # %bb.0:
135; AVX1-NEXT:    vpsrlw $15, %xmm0, %xmm2
136; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
137; AVX1-NEXT:    vpsrlw $15, %xmm0, %xmm0
138; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm1, %ymm1
139; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
140; AVX1-NEXT:    vpackuswb %xmm3, %xmm0, %xmm0
141; AVX1-NEXT:    vpackuswb %xmm1, %xmm2, %xmm1
142; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
143; AVX1-NEXT:    retq
144;
145; AVX2-LABEL: trunc_concat_packuswb_256:
146; AVX2:       # %bb.0:
147; AVX2-NEXT:    vpsrlw $15, %ymm0, %ymm0
148; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
149; AVX2-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0
150; AVX2-NEXT:    retq
151;
152; AVX512F-LABEL: trunc_concat_packuswb_256:
153; AVX512F:       # %bb.0:
154; AVX512F-NEXT:    vpsrlw $15, %ymm0, %ymm0
155; AVX512F-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
156; AVX512F-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
157; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
158; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
159; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
160; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
161; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
162; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
163; AVX512F-NEXT:    retq
164;
165; AVX512BW-LABEL: trunc_concat_packuswb_256:
166; AVX512BW:       # %bb.0:
167; AVX512BW-NEXT:    vpsrlw $15, %ymm0, %ymm0
168; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
169; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
170; AVX512BW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
171; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
172; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
173; AVX512BW-NEXT:    retq
174  %1 = lshr <16 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
175  %2 = and  <16 x i16> %a1, <i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1>
176  %3 = shufflevector <16 x i16> %1, <16 x i16> %2, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
177  %4 = trunc <32 x i16> %3 to <32 x i8>
178  ret <32 x i8> %4
179}
180
181; concat(trunc(x),trunc(y)) -> pack
182
183
184define <16 x i16> @concat_trunc_packssdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
185; AVX1-LABEL: concat_trunc_packssdw_256:
186; AVX1:       # %bb.0:
187; AVX1-NEXT:    vpsrad $17, %xmm0, %xmm2
188; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
189; AVX1-NEXT:    vpsrad $17, %xmm0, %xmm0
190; AVX1-NEXT:    vpsrad $23, %xmm1, %xmm3
191; AVX1-NEXT:    vpackssdw %xmm3, %xmm2, %xmm2
192; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
193; AVX1-NEXT:    vpsrad $23, %xmm1, %xmm1
194; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
195; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
196; AVX1-NEXT:    retq
197;
198; AVX2-LABEL: concat_trunc_packssdw_256:
199; AVX2:       # %bb.0:
200; AVX2-NEXT:    vpsrad $17, %ymm0, %ymm0
201; AVX2-NEXT:    vpsrad $23, %ymm1, %ymm1
202; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
203; AVX2-NEXT:    retq
204;
205; AVX512-LABEL: concat_trunc_packssdw_256:
206; AVX512:       # %bb.0:
207; AVX512-NEXT:    vpsrad $17, %ymm0, %ymm0
208; AVX512-NEXT:    vpsrad $23, %ymm1, %ymm1
209; AVX512-NEXT:    vpmovdw %ymm0, %xmm0
210; AVX512-NEXT:    vpmovdw %ymm1, %xmm1
211; AVX512-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
212; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
213; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
214; AVX512-NEXT:    retq
215  %1 = ashr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
216  %2 = ashr <8 x i32> %a1, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
217  %3 = trunc <8 x i32> %1 to <8 x i16>
218  %4 = trunc <8 x i32> %2 to <8 x i16>
219  %5 = shufflevector <8 x i16> %3, <8 x i16> %4, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15>
220  ret <16 x i16> %5
221}
222
223define <16 x i16> @concat_trunc_packusdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
224; AVX1-LABEL: concat_trunc_packusdw_256:
225; AVX1:       # %bb.0:
226; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
227; AVX1-NEXT:    vpsrld $17, %xmm2, %xmm2
228; AVX1-NEXT:    vpsrld $17, %xmm0, %xmm0
229; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
230; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
231; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
232; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
233; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
234; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
235; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
236; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
237; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
238; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
239; AVX1-NEXT:    retq
240;
241; AVX2-LABEL: concat_trunc_packusdw_256:
242; AVX2:       # %bb.0:
243; AVX2-NEXT:    vpsrld $17, %ymm0, %ymm0
244; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
245; AVX2-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
246; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u]
247; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
248; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
249; AVX2-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
250; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
251; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
252; AVX2-NEXT:    retq
253;
254; AVX512-LABEL: concat_trunc_packusdw_256:
255; AVX512:       # %bb.0:
256; AVX512-NEXT:    vpsrld $17, %ymm0, %ymm0
257; AVX512-NEXT:    vpmovdw %ymm0, %xmm0
258; AVX512-NEXT:    vpmovdw %ymm1, %xmm1
259; AVX512-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
260; AVX512-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
261; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
262; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
263; AVX512-NEXT:    retq
264  %1 = lshr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
265  %2 = and  <8 x i32> %a1, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
266  %3 = trunc <8 x i32> %1 to <8 x i16>
267  %4 = trunc <8 x i32> %2 to <8 x i16>
268  %5 = shufflevector <8 x i16> %3, <8 x i16> %4, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15>
269  ret <16 x i16> %5
270}
271
272define <32 x i8> @concat_trunc_packsswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
273; AVX1-LABEL: concat_trunc_packsswb_256:
274; AVX1:       # %bb.0:
275; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
276; AVX1-NEXT:    vpsraw $15, %xmm2, %xmm2
277; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm0
278; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
279; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm1, %ymm1
280; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
281; AVX1-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
282; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
283; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
284; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
285; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
286; AVX1-NEXT:    retq
287;
288; AVX2-LABEL: concat_trunc_packsswb_256:
289; AVX2:       # %bb.0:
290; AVX2-NEXT:    vpsraw $15, %ymm0, %ymm0
291; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
292; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
293; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
294; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
295; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
296; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
297; AVX2-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
298; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
299; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
300; AVX2-NEXT:    retq
301;
302; AVX512F-LABEL: concat_trunc_packsswb_256:
303; AVX512F:       # %bb.0:
304; AVX512F-NEXT:    vpsraw $15, %ymm0, %ymm0
305; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
306; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
307; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
308; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
309; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
310; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
311; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
312; AVX512F-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
313; AVX512F-NEXT:    retq
314;
315; AVX512BW-LABEL: concat_trunc_packsswb_256:
316; AVX512BW:       # %bb.0:
317; AVX512BW-NEXT:    vpsraw $15, %ymm0, %ymm0
318; AVX512BW-NEXT:    vpmovwb %ymm0, %xmm0
319; AVX512BW-NEXT:    vpmovwb %ymm1, %xmm1
320; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
321; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
322; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
323; AVX512BW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
324; AVX512BW-NEXT:    retq
325  %1 = ashr <16 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
326  %2 = and  <16 x i16> %a1, <i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1>
327  %3 = trunc <16 x i16> %1 to <16 x i8>
328  %4 = trunc <16 x i16> %2 to <16 x i8>
329  %5 = shufflevector <16 x i8> %3, <16 x i8> %4, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
330  ret <32 x i8> %5
331}
332
333define <32 x i8> @concat_trunc_packuswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
334; AVX1-LABEL: concat_trunc_packuswb_256:
335; AVX1:       # %bb.0:
336; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
337; AVX1-NEXT:    vpsrlw $15, %xmm2, %xmm2
338; AVX1-NEXT:    vpsrlw $15, %xmm0, %xmm0
339; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
340; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm1, %ymm1
341; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
342; AVX1-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
343; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
344; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
345; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
346; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
347; AVX1-NEXT:    retq
348;
349; AVX2-LABEL: concat_trunc_packuswb_256:
350; AVX2:       # %bb.0:
351; AVX2-NEXT:    vpsrlw $15, %ymm0, %ymm0
352; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
353; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
354; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
355; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
356; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
357; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
358; AVX2-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
359; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
360; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
361; AVX2-NEXT:    retq
362;
363; AVX512F-LABEL: concat_trunc_packuswb_256:
364; AVX512F:       # %bb.0:
365; AVX512F-NEXT:    vpsrlw $15, %ymm0, %ymm0
366; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
367; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
368; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
369; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
370; AVX512F-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
371; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
372; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
373; AVX512F-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
374; AVX512F-NEXT:    retq
375;
376; AVX512BW-LABEL: concat_trunc_packuswb_256:
377; AVX512BW:       # %bb.0:
378; AVX512BW-NEXT:    vpsrlw $15, %ymm0, %ymm0
379; AVX512BW-NEXT:    vpmovwb %ymm0, %xmm0
380; AVX512BW-NEXT:    vpmovwb %ymm1, %xmm1
381; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
382; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
383; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
384; AVX512BW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
385; AVX512BW-NEXT:    retq
386  %1 = lshr <16 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
387  %2 = and  <16 x i16> %a1, <i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1, i16  1>
388  %3 = trunc <16 x i16> %1 to <16 x i8>
389  %4 = trunc <16 x i16> %2 to <16 x i8>
390  %5 = shufflevector <16 x i8> %3, <16 x i8> %4, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
391  ret <32 x i8> %5
392}
393