xref: /llvm-project/llvm/test/CodeGen/X86/avx-unpack.ll (revision 2f448bf509432c1a19ec46ab8cbc7353c03c6280) 
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
3
4define <8 x float> @unpackhips(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
5; CHECK-LABEL: unpackhips:
6; CHECK:       # %bb.0:
7; CHECK-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
8; CHECK-NEXT:    retq
9  %shuffle.i = shufflevector <8 x float> %src1, <8 x float> %src2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
10  ret <8 x float> %shuffle.i
11}
12
13define <4 x double> @unpackhipd(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
14; CHECK-LABEL: unpackhipd:
15; CHECK:       # %bb.0:
16; CHECK-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
17; CHECK-NEXT:    retq
18  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
19  ret <4 x double> %shuffle.i
20}
21
22define <8 x float> @unpacklops(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
23; CHECK-LABEL: unpacklops:
24; CHECK:       # %bb.0:
25; CHECK-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
26; CHECK-NEXT:    retq
27  %shuffle.i = shufflevector <8 x float> %src1, <8 x float> %src2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
28  ret <8 x float> %shuffle.i
29}
30
31define <4 x double> @unpacklopd(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
32; CHECK-LABEL: unpacklopd:
33; CHECK:       # %bb.0:
34; CHECK-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
35; CHECK-NEXT:    retq
36  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
37  ret <4 x double> %shuffle.i
38}
39
40define <8 x float> @unpacklops_not(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
41; CHECK-LABEL: unpacklops_not:
42; CHECK:       # %bb.0:
43; CHECK-NEXT:    vunpckhps {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
44; CHECK-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
45; CHECK-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
46; CHECK-NEXT:    retq
47  %shuffle.i = shufflevector <8 x float> %src1, <8 x float> %src2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
48  ret <8 x float> %shuffle.i
49}
50
51define <4 x double> @unpacklopd_not(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
52; CHECK-LABEL: unpacklopd_not:
53; CHECK:       # %bb.0:
54; CHECK-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm1[1]
55; CHECK-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
56; CHECK-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
57; CHECK-NEXT:    retq
58  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
59  ret <4 x double> %shuffle.i
60}
61
62define <8 x float> @unpackhips_not(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
63; CHECK-LABEL: unpackhips_not:
64; CHECK:       # %bb.0:
65; CHECK-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[u,2,u,3,u,4,u,5]
66; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,u,3,u,4,u,5,u]
67; CHECK-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
68; CHECK-NEXT:    retq
69  %shuffle.i = shufflevector <8 x float> %src1, <8 x float> %src2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13>
70  ret <8 x float> %shuffle.i
71}
72
73define <4 x double> @unpackhipd_not(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
74; CHECK-LABEL: unpackhipd_not:
75; CHECK:       # %bb.0:
76; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
77; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
78; CHECK-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
79; CHECK-NEXT:    retq
80  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
81  ret <4 x double> %shuffle.i
82}
83
84;;;;
85;;;; Unpack versions using the fp unit for int unpacking
86;;;;
87
88define <8 x i32> @unpackhips1(<8 x i32> %src1, <8 x i32> %src2) nounwind uwtable readnone ssp {
89; CHECK-LABEL: unpackhips1:
90; CHECK:       # %bb.0:
91; CHECK-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
92; CHECK-NEXT:    retq
93  %shuffle.i = shufflevector <8 x i32> %src1, <8 x i32> %src2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
94  ret <8 x i32> %shuffle.i
95}
96
97define <8 x i32> @unpackhips2(ptr %src1, ptr %src2) nounwind uwtable readnone ssp {
98; CHECK-LABEL: unpackhips2:
99; CHECK:       # %bb.0:
100; CHECK-NEXT:    vmovaps (%rdi), %ymm0
101; CHECK-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
102; CHECK-NEXT:    retq
103  %a = load <8 x i32>, ptr %src1
104  %b = load <8 x i32>, ptr %src2
105  %shuffle.i = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
106  ret <8 x i32> %shuffle.i
107}
108
109define <4 x i64> @unpackhipd1(<4 x i64> %src1, <4 x i64> %src2) nounwind uwtable readnone ssp {
110; CHECK-LABEL: unpackhipd1:
111; CHECK:       # %bb.0:
112; CHECK-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
113; CHECK-NEXT:    retq
114  %shuffle.i = shufflevector <4 x i64> %src1, <4 x i64> %src2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
115  ret <4 x i64> %shuffle.i
116}
117
118define <4 x i64> @unpackhipd2(ptr %src1, ptr %src2) nounwind uwtable readnone ssp {
119; CHECK-LABEL: unpackhipd2:
120; CHECK:       # %bb.0:
121; CHECK-NEXT:    vmovaps (%rdi), %ymm0
122; CHECK-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
123; CHECK-NEXT:    retq
124  %a = load <4 x i64>, ptr %src1
125  %b = load <4 x i64>, ptr %src2
126  %shuffle.i = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
127  ret <4 x i64> %shuffle.i
128}
129
130define <8 x i32> @unpacklops1(<8 x i32> %src1, <8 x i32> %src2) nounwind uwtable readnone ssp {
131; CHECK-LABEL: unpacklops1:
132; CHECK:       # %bb.0:
133; CHECK-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
134; CHECK-NEXT:    retq
135  %shuffle.i = shufflevector <8 x i32> %src1, <8 x i32> %src2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
136  ret <8 x i32> %shuffle.i
137}
138
139define <8 x i32> @unpacklops2(ptr %src1, ptr %src2) nounwind uwtable readnone ssp {
140; CHECK-LABEL: unpacklops2:
141; CHECK:       # %bb.0:
142; CHECK-NEXT:    vmovaps (%rdi), %ymm0
143; CHECK-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
144; CHECK-NEXT:    retq
145  %a = load <8 x i32>, ptr %src1
146  %b = load <8 x i32>, ptr %src2
147  %shuffle.i = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
148  ret <8 x i32> %shuffle.i
149}
150
151define <4 x i64> @unpacklopd1(<4 x i64> %src1, <4 x i64> %src2) nounwind uwtable readnone ssp {
152; CHECK-LABEL: unpacklopd1:
153; CHECK:       # %bb.0:
154; CHECK-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
155; CHECK-NEXT:    retq
156  %shuffle.i = shufflevector <4 x i64> %src1, <4 x i64> %src2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
157  ret <4 x i64> %shuffle.i
158}
159
160define <4 x i64> @unpacklopd2(ptr %src1, ptr %src2) nounwind uwtable readnone ssp {
161; CHECK-LABEL: unpacklopd2:
162; CHECK:       # %bb.0:
163; CHECK-NEXT:    vmovaps (%rdi), %ymm0
164; CHECK-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
165; CHECK-NEXT:    retq
166  %a = load <4 x i64>, ptr %src1
167  %b = load <4 x i64>, ptr %src2
168  %shuffle.i = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
169  ret <4 x i64> %shuffle.i
170}
171
172define <16 x i16> @unpackhwd_undef(<16 x i16> %src1) nounwind uwtable readnone ssp {
173; CHECK-LABEL: unpackhwd_undef:
174; CHECK:       # %bb.0:
175; CHECK-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm0[4,4,5,5,6,6,7,7]
176; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
177; CHECK-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
178; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
179; CHECK-NEXT:    retq
180  %shuffle.i = shufflevector <16 x i16> %src1, <16 x i16> %src1, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
181  ret <16 x i16> %shuffle.i
182}
183
184define <16 x i16> @unpacklwd_undef(<16 x i16> %src1) nounwind uwtable readnone ssp {
185; CHECK-LABEL: unpacklwd_undef:
186; CHECK:       # %bb.0:
187; CHECK-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0,0,1,1,2,2,3,3]
188; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
189; CHECK-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
190; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
191; CHECK-NEXT:    retq
192  %shuffle.i = shufflevector <16 x i16> %src1, <16 x i16> %src1, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27>
193  ret <16 x i16> %shuffle.i
194}
195
196define <32 x i8> @unpackhbw_undef(<32 x i8> %src1, <32 x i8> %src2) nounwind uwtable readnone ssp {
197; CHECK-LABEL: unpackhbw_undef:
198; CHECK:       # %bb.0:
199; CHECK-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
200; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
201; CHECK-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
202; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
203; CHECK-NEXT:    retq
204  %shuffle.i = shufflevector <32 x i8> %src1, <32 x i8> %src1, <32 x i32> <i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63>
205  ret <32 x i8> %shuffle.i
206}
207
208define <32 x i8> @unpacklbw_undef(<32 x i8> %src1) nounwind uwtable readnone ssp {
209; CHECK-LABEL: unpacklbw_undef:
210; CHECK:       # %bb.0:
211; CHECK-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
212; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
213; CHECK-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
214; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
215; CHECK-NEXT:    retq
216  %shuffle.i = shufflevector <32 x i8> %src1, <32 x i8> %src1, <32 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55>
217  ret <32 x i8> %shuffle.i
218}
219
220