xref: /aosp_15_r20/external/llvm/test/CodeGen/X86/avx-unpack.ll (revision 9880d6810fe72a1726cb53787c6711e909410d58)
1*9880d681SAndroid Build Coastguard Worker; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
2*9880d681SAndroid Build Coastguard Worker
3*9880d681SAndroid Build Coastguard Worker; CHECK: vunpckhps
4*9880d681SAndroid Build Coastguard Workerdefine <8 x float> @unpackhips(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
5*9880d681SAndroid Build Coastguard Workerentry:
6*9880d681SAndroid Build Coastguard Worker  %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>
7*9880d681SAndroid Build Coastguard Worker  ret <8 x float> %shuffle.i
8*9880d681SAndroid Build Coastguard Worker}
9*9880d681SAndroid Build Coastguard Worker
10*9880d681SAndroid Build Coastguard Worker; CHECK: vunpckhpd
11*9880d681SAndroid Build Coastguard Workerdefine <4 x double> @unpackhipd(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
12*9880d681SAndroid Build Coastguard Workerentry:
13*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
14*9880d681SAndroid Build Coastguard Worker  ret <4 x double> %shuffle.i
15*9880d681SAndroid Build Coastguard Worker}
16*9880d681SAndroid Build Coastguard Worker
17*9880d681SAndroid Build Coastguard Worker; CHECK: vunpcklps
18*9880d681SAndroid Build Coastguard Workerdefine <8 x float> @unpacklops(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
19*9880d681SAndroid Build Coastguard Workerentry:
20*9880d681SAndroid Build Coastguard Worker  %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>
21*9880d681SAndroid Build Coastguard Worker  ret <8 x float> %shuffle.i
22*9880d681SAndroid Build Coastguard Worker}
23*9880d681SAndroid Build Coastguard Worker
24*9880d681SAndroid Build Coastguard Worker; CHECK: vunpcklpd
25*9880d681SAndroid Build Coastguard Workerdefine <4 x double> @unpacklopd(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
26*9880d681SAndroid Build Coastguard Workerentry:
27*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
28*9880d681SAndroid Build Coastguard Worker  ret <4 x double> %shuffle.i
29*9880d681SAndroid Build Coastguard Worker}
30*9880d681SAndroid Build Coastguard Worker
31*9880d681SAndroid Build Coastguard Worker; CHECK-NOT: vunpcklps %ymm
32*9880d681SAndroid Build Coastguard Workerdefine <8 x float> @unpacklops-not(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
33*9880d681SAndroid Build Coastguard Workerentry:
34*9880d681SAndroid Build Coastguard Worker  %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>
35*9880d681SAndroid Build Coastguard Worker  ret <8 x float> %shuffle.i
36*9880d681SAndroid Build Coastguard Worker}
37*9880d681SAndroid Build Coastguard Worker
38*9880d681SAndroid Build Coastguard Worker; CHECK-NOT: vunpcklpd %ymm
39*9880d681SAndroid Build Coastguard Workerdefine <4 x double> @unpacklopd-not(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
40*9880d681SAndroid Build Coastguard Workerentry:
41*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
42*9880d681SAndroid Build Coastguard Worker  ret <4 x double> %shuffle.i
43*9880d681SAndroid Build Coastguard Worker}
44*9880d681SAndroid Build Coastguard Worker
45*9880d681SAndroid Build Coastguard Worker; CHECK-NOT: vunpckhps %ymm
46*9880d681SAndroid Build Coastguard Workerdefine <8 x float> @unpackhips-not(<8 x float> %src1, <8 x float> %src2) nounwind uwtable readnone ssp {
47*9880d681SAndroid Build Coastguard Workerentry:
48*9880d681SAndroid Build Coastguard Worker  %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>
49*9880d681SAndroid Build Coastguard Worker  ret <8 x float> %shuffle.i
50*9880d681SAndroid Build Coastguard Worker}
51*9880d681SAndroid Build Coastguard Worker
52*9880d681SAndroid Build Coastguard Worker; CHECK-NOT: vunpckhpd %ymm
53*9880d681SAndroid Build Coastguard Workerdefine <4 x double> @unpackhipd-not(<4 x double> %src1, <4 x double> %src2) nounwind uwtable readnone ssp {
54*9880d681SAndroid Build Coastguard Workerentry:
55*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x double> %src1, <4 x double> %src2, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
56*9880d681SAndroid Build Coastguard Worker  ret <4 x double> %shuffle.i
57*9880d681SAndroid Build Coastguard Worker}
58*9880d681SAndroid Build Coastguard Worker
59*9880d681SAndroid Build Coastguard Worker;;;;
60*9880d681SAndroid Build Coastguard Worker;;;; Unpack versions using the fp unit for int unpacking
61*9880d681SAndroid Build Coastguard Worker;;;;
62*9880d681SAndroid Build Coastguard Worker
63*9880d681SAndroid Build Coastguard Worker; CHECK: vunpckhps
64*9880d681SAndroid Build Coastguard Workerdefine <8 x i32> @unpackhips1(<8 x i32> %src1, <8 x i32> %src2) nounwind uwtable readnone ssp {
65*9880d681SAndroid Build Coastguard Workerentry:
66*9880d681SAndroid Build Coastguard Worker  %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>
67*9880d681SAndroid Build Coastguard Worker  ret <8 x i32> %shuffle.i
68*9880d681SAndroid Build Coastguard Worker}
69*9880d681SAndroid Build Coastguard Worker
70*9880d681SAndroid Build Coastguard Worker; CHECK: vunpckhps (%
71*9880d681SAndroid Build Coastguard Workerdefine <8 x i32> @unpackhips2(<8 x i32>* %src1, <8 x i32>* %src2) nounwind uwtable readnone ssp {
72*9880d681SAndroid Build Coastguard Workerentry:
73*9880d681SAndroid Build Coastguard Worker  %a = load <8 x i32>, <8 x i32>* %src1
74*9880d681SAndroid Build Coastguard Worker  %b = load <8 x i32>, <8 x i32>* %src2
75*9880d681SAndroid Build Coastguard Worker  %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>
76*9880d681SAndroid Build Coastguard Worker  ret <8 x i32> %shuffle.i
77*9880d681SAndroid Build Coastguard Worker}
78*9880d681SAndroid Build Coastguard Worker
79*9880d681SAndroid Build Coastguard Worker; CHECK: vunpckhpd
80*9880d681SAndroid Build Coastguard Workerdefine <4 x i64> @unpackhipd1(<4 x i64> %src1, <4 x i64> %src2) nounwind uwtable readnone ssp {
81*9880d681SAndroid Build Coastguard Workerentry:
82*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x i64> %src1, <4 x i64> %src2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
83*9880d681SAndroid Build Coastguard Worker  ret <4 x i64> %shuffle.i
84*9880d681SAndroid Build Coastguard Worker}
85*9880d681SAndroid Build Coastguard Worker
86*9880d681SAndroid Build Coastguard Worker; CHECK: vunpckhpd (%
87*9880d681SAndroid Build Coastguard Workerdefine <4 x i64> @unpackhipd2(<4 x i64>* %src1, <4 x i64>* %src2) nounwind uwtable readnone ssp {
88*9880d681SAndroid Build Coastguard Workerentry:
89*9880d681SAndroid Build Coastguard Worker  %a = load <4 x i64>, <4 x i64>* %src1
90*9880d681SAndroid Build Coastguard Worker  %b = load <4 x i64>, <4 x i64>* %src2
91*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
92*9880d681SAndroid Build Coastguard Worker  ret <4 x i64> %shuffle.i
93*9880d681SAndroid Build Coastguard Worker}
94*9880d681SAndroid Build Coastguard Worker
95*9880d681SAndroid Build Coastguard Worker; CHECK: vunpcklps
96*9880d681SAndroid Build Coastguard Workerdefine <8 x i32> @unpacklops1(<8 x i32> %src1, <8 x i32> %src2) nounwind uwtable readnone ssp {
97*9880d681SAndroid Build Coastguard Workerentry:
98*9880d681SAndroid Build Coastguard Worker  %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>
99*9880d681SAndroid Build Coastguard Worker  ret <8 x i32> %shuffle.i
100*9880d681SAndroid Build Coastguard Worker}
101*9880d681SAndroid Build Coastguard Worker
102*9880d681SAndroid Build Coastguard Worker; CHECK: vunpcklps (%
103*9880d681SAndroid Build Coastguard Workerdefine <8 x i32> @unpacklops2(<8 x i32>* %src1, <8 x i32>* %src2) nounwind uwtable readnone ssp {
104*9880d681SAndroid Build Coastguard Workerentry:
105*9880d681SAndroid Build Coastguard Worker  %a = load <8 x i32>, <8 x i32>* %src1
106*9880d681SAndroid Build Coastguard Worker  %b = load <8 x i32>, <8 x i32>* %src2
107*9880d681SAndroid Build Coastguard Worker  %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>
108*9880d681SAndroid Build Coastguard Worker  ret <8 x i32> %shuffle.i
109*9880d681SAndroid Build Coastguard Worker}
110*9880d681SAndroid Build Coastguard Worker
111*9880d681SAndroid Build Coastguard Worker; CHECK: vunpcklpd
112*9880d681SAndroid Build Coastguard Workerdefine <4 x i64> @unpacklopd1(<4 x i64> %src1, <4 x i64> %src2) nounwind uwtable readnone ssp {
113*9880d681SAndroid Build Coastguard Workerentry:
114*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x i64> %src1, <4 x i64> %src2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
115*9880d681SAndroid Build Coastguard Worker  ret <4 x i64> %shuffle.i
116*9880d681SAndroid Build Coastguard Worker}
117*9880d681SAndroid Build Coastguard Worker
118*9880d681SAndroid Build Coastguard Worker; CHECK: vunpcklpd (%
119*9880d681SAndroid Build Coastguard Workerdefine <4 x i64> @unpacklopd2(<4 x i64>* %src1, <4 x i64>* %src2) nounwind uwtable readnone ssp {
120*9880d681SAndroid Build Coastguard Workerentry:
121*9880d681SAndroid Build Coastguard Worker  %a = load <4 x i64>, <4 x i64>* %src1
122*9880d681SAndroid Build Coastguard Worker  %b = load <4 x i64>, <4 x i64>* %src2
123*9880d681SAndroid Build Coastguard Worker  %shuffle.i = shufflevector <4 x i64> %a, <4 x i64> %b, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
124*9880d681SAndroid Build Coastguard Worker  ret <4 x i64> %shuffle.i
125*9880d681SAndroid Build Coastguard Worker}
126*9880d681SAndroid Build Coastguard Worker
127*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpckhwd
128*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpckhwd
129*9880d681SAndroid Build Coastguard Worker; CHECK: vinsertf128
130*9880d681SAndroid Build Coastguard Workerdefine <16 x i16> @unpackhwd_undef(<16 x i16> %src1) nounwind uwtable readnone ssp {
131*9880d681SAndroid Build Coastguard Workerentry:
132*9880d681SAndroid Build Coastguard Worker  %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>
133*9880d681SAndroid Build Coastguard Worker  ret <16 x i16> %shuffle.i
134*9880d681SAndroid Build Coastguard Worker}
135*9880d681SAndroid Build Coastguard Worker
136*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpcklwd
137*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpcklwd
138*9880d681SAndroid Build Coastguard Worker; CHECK: vinsertf128
139*9880d681SAndroid Build Coastguard Workerdefine <16 x i16> @unpacklwd_undef(<16 x i16> %src1) nounwind uwtable readnone ssp {
140*9880d681SAndroid Build Coastguard Workerentry:
141*9880d681SAndroid Build Coastguard Worker  %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>
142*9880d681SAndroid Build Coastguard Worker  ret <16 x i16> %shuffle.i
143*9880d681SAndroid Build Coastguard Worker}
144*9880d681SAndroid Build Coastguard Worker
145*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpckhbw
146*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpckhbw
147*9880d681SAndroid Build Coastguard Worker; CHECK: vinsertf128
148*9880d681SAndroid Build Coastguard Workerdefine <32 x i8> @unpackhbw_undef(<32 x i8> %src1, <32 x i8> %src2) nounwind uwtable readnone ssp {
149*9880d681SAndroid Build Coastguard Workerentry:
150*9880d681SAndroid Build Coastguard Worker  %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>
151*9880d681SAndroid Build Coastguard Worker  ret <32 x i8> %shuffle.i
152*9880d681SAndroid Build Coastguard Worker}
153*9880d681SAndroid Build Coastguard Worker
154*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpcklbw
155*9880d681SAndroid Build Coastguard Worker; CHECK: vpunpcklbw
156*9880d681SAndroid Build Coastguard Worker; CHECK: vinsertf128
157*9880d681SAndroid Build Coastguard Workerdefine <32 x i8> @unpacklbw_undef(<32 x i8> %src1) nounwind uwtable readnone ssp {
158*9880d681SAndroid Build Coastguard Workerentry:
159*9880d681SAndroid Build Coastguard Worker  %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>
160*9880d681SAndroid Build Coastguard Worker  ret <32 x i8> %shuffle.i
161*9880d681SAndroid Build Coastguard Worker}
162