xref: /llvm-project/llvm/test/CodeGen/X86/vector-pack-256.ll (revision 34961c600d477386050828de4261d3c3a7cb809c) 
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 {{\.?LCPI[0-9]+_[0-9]+}}(%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 {{\.?LCPI[0-9]+_[0-9]+}}(%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 {{\.?LCPI[0-9]+_[0-9]+}}(%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 {{\.?LCPI[0-9]+_[0-9]+}}(%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:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm1, %ymm1
107; AVX512F-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0
108; AVX512F-NEXT:    retq
109;
110; AVX512BW-LABEL: trunc_concat_packsswb_256:
111; AVX512BW:       # %bb.0:
112; AVX512BW-NEXT:    vpsraw $15, %ymm0, %ymm0
113; AVX512BW-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm1, %ymm1
114; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
115; AVX512BW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
116; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
117; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
118; AVX512BW-NEXT:    retq
119  %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>
120  %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>
121  %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>
122  %4 = trunc <32 x i16> %3 to <32 x i8>
123  ret <32 x i8> %4
124}
125
126define <32 x i8> @trunc_concat_packuswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
127; AVX1-LABEL: trunc_concat_packuswb_256:
128; AVX1:       # %bb.0:
129; AVX1-NEXT:    vpsrlw $15, %xmm0, %xmm2
130; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
131; AVX1-NEXT:    vpsrlw $15, %xmm0, %xmm0
132; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
133; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
134; AVX1-NEXT:    vpackuswb %xmm3, %xmm0, %xmm0
135; AVX1-NEXT:    vpackuswb %xmm1, %xmm2, %xmm1
136; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
137; AVX1-NEXT:    retq
138;
139; AVX2-LABEL: trunc_concat_packuswb_256:
140; AVX2:       # %bb.0:
141; AVX2-NEXT:    vpsrlw $15, %ymm0, %ymm0
142; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
143; AVX2-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0
144; AVX2-NEXT:    retq
145;
146; AVX512F-LABEL: trunc_concat_packuswb_256:
147; AVX512F:       # %bb.0:
148; AVX512F-NEXT:    vpsrlw $15, %ymm0, %ymm0
149; AVX512F-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm1, %ymm1
150; AVX512F-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0
151; AVX512F-NEXT:    retq
152;
153; AVX512BW-LABEL: trunc_concat_packuswb_256:
154; AVX512BW:       # %bb.0:
155; AVX512BW-NEXT:    vpsrlw $15, %ymm0, %ymm0
156; AVX512BW-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm1, %ymm1
157; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
158; AVX512BW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
159; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
160; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
161; AVX512BW-NEXT:    retq
162  %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>
163  %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>
164  %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>
165  %4 = trunc <32 x i16> %3 to <32 x i8>
166  ret <32 x i8> %4
167}
168
169; concat(trunc(x),trunc(y)) -> pack
170
171
172define <16 x i16> @concat_trunc_packssdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
173; AVX1-LABEL: concat_trunc_packssdw_256:
174; AVX1:       # %bb.0:
175; AVX1-NEXT:    vpsrad $17, %xmm0, %xmm2
176; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
177; AVX1-NEXT:    vpsrad $17, %xmm0, %xmm0
178; AVX1-NEXT:    vpsrad $23, %xmm1, %xmm3
179; AVX1-NEXT:    vpackssdw %xmm3, %xmm2, %xmm2
180; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
181; AVX1-NEXT:    vpsrad $23, %xmm1, %xmm1
182; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
183; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
184; AVX1-NEXT:    retq
185;
186; AVX2-LABEL: concat_trunc_packssdw_256:
187; AVX2:       # %bb.0:
188; AVX2-NEXT:    vpsrad $17, %ymm0, %ymm0
189; AVX2-NEXT:    vpsrad $23, %ymm1, %ymm1
190; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
191; AVX2-NEXT:    retq
192;
193; AVX512-LABEL: concat_trunc_packssdw_256:
194; AVX512:       # %bb.0:
195; AVX512-NEXT:    vpsrad $17, %ymm0, %ymm0
196; AVX512-NEXT:    vpsrad $23, %ymm1, %ymm1
197; AVX512-NEXT:    vpmovdw %ymm0, %xmm0
198; AVX512-NEXT:    vpmovdw %ymm1, %xmm1
199; AVX512-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
200; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
201; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
202; AVX512-NEXT:    retq
203  %1 = ashr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
204  %2 = ashr <8 x i32> %a1, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
205  %3 = trunc <8 x i32> %1 to <8 x i16>
206  %4 = trunc <8 x i32> %2 to <8 x i16>
207  %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>
208  ret <16 x i16> %5
209}
210
211define <16 x i16> @concat_trunc_packusdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
212; AVX1-LABEL: concat_trunc_packusdw_256:
213; AVX1:       # %bb.0:
214; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
215; AVX1-NEXT:    vpsrld $17, %xmm2, %xmm2
216; AVX1-NEXT:    vpsrld $17, %xmm0, %xmm0
217; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
218; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
219; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
220; AVX1-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
221; AVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
222; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
223; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
224; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
225; AVX1-NEXT:    retq
226;
227; AVX2-LABEL: concat_trunc_packusdw_256:
228; AVX2:       # %bb.0:
229; AVX2-NEXT:    vpsrld $17, %ymm0, %ymm0
230; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
231; AVX2-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
232; 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]
233; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
234; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
235; AVX2-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
236; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
237; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
238; AVX2-NEXT:    retq
239;
240; AVX512-LABEL: concat_trunc_packusdw_256:
241; AVX512:       # %bb.0:
242; AVX512-NEXT:    vpsrld $17, %ymm0, %ymm0
243; AVX512-NEXT:    vpmovdw %ymm0, %xmm0
244; AVX512-NEXT:    vpmovdw %ymm1, %xmm1
245; AVX512-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm1, %xmm1
246; AVX512-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
247; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
248; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
249; AVX512-NEXT:    retq
250  %1 = lshr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
251  %2 = and  <8 x i32> %a1, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
252  %3 = trunc <8 x i32> %1 to <8 x i16>
253  %4 = trunc <8 x i32> %2 to <8 x i16>
254  %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>
255  ret <16 x i16> %5
256}
257
258define <32 x i8> @concat_trunc_packsswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
259; AVX1-LABEL: concat_trunc_packsswb_256:
260; AVX1:       # %bb.0:
261; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
262; AVX1-NEXT:    vpsraw $15, %xmm2, %xmm2
263; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm0
264; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
265; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
266; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
267; AVX1-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
268; AVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
269; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
270; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
271; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
272; AVX1-NEXT:    retq
273;
274; AVX2-LABEL: concat_trunc_packsswb_256:
275; AVX2:       # %bb.0:
276; AVX2-NEXT:    vpsraw $15, %ymm0, %ymm0
277; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
278; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
279; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
280; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
281; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
282; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
283; AVX2-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
284; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
285; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
286; AVX2-NEXT:    retq
287;
288; AVX512F-LABEL: concat_trunc_packsswb_256:
289; AVX512F:       # %bb.0:
290; AVX512F-NEXT:    vpsraw $15, %ymm0, %ymm0
291; 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
292; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
293; 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
294; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
295; AVX512F-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm1, %xmm1
296; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
297; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
298; AVX512F-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
299; AVX512F-NEXT:    retq
300;
301; AVX512BW-LABEL: concat_trunc_packsswb_256:
302; AVX512BW:       # %bb.0:
303; AVX512BW-NEXT:    vpsraw $15, %ymm0, %ymm0
304; AVX512BW-NEXT:    vpmovwb %ymm0, %xmm0
305; AVX512BW-NEXT:    vpmovwb %ymm1, %xmm1
306; AVX512BW-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm1, %xmm1
307; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
308; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
309; AVX512BW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
310; AVX512BW-NEXT:    retq
311  %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>
312  %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>
313  %3 = trunc <16 x i16> %1 to <16 x i8>
314  %4 = trunc <16 x i16> %2 to <16 x i8>
315  %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>
316  ret <32 x i8> %5
317}
318
319define <32 x i8> @concat_trunc_packuswb_256(<16 x i16> %a0, <16 x i16> %a1) nounwind {
320; AVX1-LABEL: concat_trunc_packuswb_256:
321; AVX1:       # %bb.0:
322; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
323; AVX1-NEXT:    vpsrlw $15, %xmm2, %xmm2
324; AVX1-NEXT:    vpsrlw $15, %xmm0, %xmm0
325; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
326; AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
327; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
328; AVX1-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
329; AVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
330; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
331; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
332; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
333; AVX1-NEXT:    retq
334;
335; AVX2-LABEL: concat_trunc_packuswb_256:
336; AVX2:       # %bb.0:
337; AVX2-NEXT:    vpsrlw $15, %ymm0, %ymm0
338; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
339; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
340; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
341; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
342; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
343; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
344; AVX2-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
345; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
346; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
347; AVX2-NEXT:    retq
348;
349; AVX512F-LABEL: concat_trunc_packuswb_256:
350; AVX512F:       # %bb.0:
351; AVX512F-NEXT:    vpsrlw $15, %ymm0, %ymm0
352; 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
353; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
354; 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
355; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
356; AVX512F-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm1, %xmm1
357; AVX512F-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
358; AVX512F-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
359; AVX512F-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
360; AVX512F-NEXT:    retq
361;
362; AVX512BW-LABEL: concat_trunc_packuswb_256:
363; AVX512BW:       # %bb.0:
364; AVX512BW-NEXT:    vpsrlw $15, %ymm0, %ymm0
365; AVX512BW-NEXT:    vpmovwb %ymm0, %xmm0
366; AVX512BW-NEXT:    vpmovwb %ymm1, %xmm1
367; AVX512BW-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm1, %xmm1
368; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
369; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
370; AVX512BW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
371; AVX512BW-NEXT:    retq
372  %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>
373  %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>
374  %3 = trunc <16 x i16> %1 to <16 x i8>
375  %4 = trunc <16 x i16> %2 to <16 x i8>
376  %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>
377  ret <32 x i8> %5
378}
379