| blob | c4de177176e330baf6925db42c298230c0abf804 |
1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
2 ; RUN: llc -mtriple=aarch64 < %s | FileCheck %s
4 define <16 x i8> @lower_trunc_16xi8(i16 %a, i16 %b, i16 %c, i16 %d, i16 %e, i16 %f, i16 %g, i16 %h, i16 %i, i16 %j, i16 %k, i16 %l, i16 %m, i16 %n, i16 %o, i16 %p) {
5 ; CHECK-LABEL: lower_trunc_16xi8:
6 ; CHECK: // %bb.0:
7 ; CHECK-NEXT: fmov s0, w0
8 ; CHECK-NEXT: ldr h1, [sp]
9 ; CHECK-NEXT: add x8, sp, #8
10 ; CHECK-NEXT: ld1 { v1.h }[1], [x8]
11 ; CHECK-NEXT: add x8, sp, #16
12 ; CHECK-NEXT: mov v0.h[1], w1
13 ; CHECK-NEXT: ld1 { v1.h }[2], [x8]
14 ; CHECK-NEXT: add x8, sp, #24
15 ; CHECK-NEXT: mov v0.h[2], w2
16 ; CHECK-NEXT: ld1 { v1.h }[3], [x8]
17 ; CHECK-NEXT: add x8, sp, #32
18 ; CHECK-NEXT: mov v0.h[3], w3
19 ; CHECK-NEXT: ld1 { v1.h }[4], [x8]
20 ; CHECK-NEXT: add x8, sp, #40
21 ; CHECK-NEXT: ld1 { v1.h }[5], [x8]
22 ; CHECK-NEXT: add x8, sp, #48
23 ; CHECK-NEXT: mov v0.h[4], w4
24 ; CHECK-NEXT: ld1 { v1.h }[6], [x8]
25 ; CHECK-NEXT: add x8, sp, #56
26 ; CHECK-NEXT: mov v0.h[5], w5
27 ; CHECK-NEXT: ld1 { v1.h }[7], [x8]
28 ; CHECK-NEXT: mov v0.h[6], w6
29 ; CHECK-NEXT: add v2.8h, v1.8h, v1.8h
30 ; CHECK-NEXT: mov v0.h[7], w7
31 ; CHECK-NEXT: add v3.8h, v0.8h, v0.8h
32 ; CHECK-NEXT: uzp1 v0.16b, v0.16b, v1.16b
33 ; CHECK-NEXT: uzp1 v1.16b, v3.16b, v2.16b
34 ; CHECK-NEXT: eor v0.16b, v0.16b, v1.16b
35 ; CHECK-NEXT: ret
36 %a1 = insertelement <16 x i16> poison, i16 %a, i16 0
37 %b1 = insertelement <16 x i16> %a1, i16 %b, i16 1
38 %c1 = insertelement <16 x i16> %b1, i16 %c, i16 2
39 %d1 = insertelement <16 x i16> %c1, i16 %d, i16 3
40 %e1 = insertelement <16 x i16> %d1, i16 %e, i16 4
41 %f1 = insertelement <16 x i16> %e1, i16 %f, i16 5
42 %g1 = insertelement <16 x i16> %f1, i16 %g, i16 6
43 %h1 = insertelement <16 x i16> %g1, i16 %h, i16 7
44 %i1 = insertelement <16 x i16> %h1, i16 %i, i16 8
45 %j1 = insertelement <16 x i16> %i1, i16 %j, i16 9
46 %k1 = insertelement <16 x i16> %j1, i16 %k, i16 10
47 %l1 = insertelement <16 x i16> %k1, i16 %l, i16 11
48 %m1 = insertelement <16 x i16> %l1, i16 %m, i16 12
49 %n1 = insertelement <16 x i16> %m1, i16 %n, i16 13
50 %o1 = insertelement <16 x i16> %n1, i16 %o, i16 14
51 %p1 = insertelement <16 x i16> %o1, i16 %p, i16 15
52 %t = trunc <16 x i16> %p1 to <16 x i8>
53 %s = add <16 x i16> %p1, %p1
54 %t2 = trunc <16 x i16> %s to <16 x i8>
55 %pro = xor <16 x i8> %t, %t2
56 ret <16 x i8> %pro
57 }
59 define <8 x i16> @lower_trunc_8xi16(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h) {
60 ; CHECK-LABEL: lower_trunc_8xi16:
61 ; CHECK: // %bb.0:
62 ; CHECK-NEXT: fmov s0, w4
63 ; CHECK-NEXT: fmov s1, w0
64 ; CHECK-NEXT: mov v0.s[1], w5
65 ; CHECK-NEXT: mov v1.s[1], w1
66 ; CHECK-NEXT: mov v0.s[2], w6
67 ; CHECK-NEXT: mov v1.s[2], w2
68 ; CHECK-NEXT: mov v0.s[3], w7
69 ; CHECK-NEXT: mov v1.s[3], w3
70 ; CHECK-NEXT: add v2.4s, v0.4s, v0.4s
71 ; CHECK-NEXT: add v3.4s, v1.4s, v1.4s
72 ; CHECK-NEXT: uzp1 v0.8h, v1.8h, v0.8h
73 ; CHECK-NEXT: uzp1 v1.8h, v3.8h, v2.8h
74 ; CHECK-NEXT: eor v0.16b, v0.16b, v1.16b
75 ; CHECK-NEXT: ret
76 %a1 = insertelement <8 x i32> poison, i32 %a, i32 0
77 %b1 = insertelement <8 x i32> %a1, i32 %b, i32 1
78 %c1 = insertelement <8 x i32> %b1, i32 %c, i32 2
79 %d1 = insertelement <8 x i32> %c1, i32 %d, i32 3
80 %e1 = insertelement <8 x i32> %d1, i32 %e, i32 4
81 %f1 = insertelement <8 x i32> %e1, i32 %f, i32 5
82 %g1 = insertelement <8 x i32> %f1, i32 %g, i32 6
83 %h1 = insertelement <8 x i32> %g1, i32 %h, i32 7
84 %t = trunc <8 x i32> %h1 to <8 x i16>
85 %s = add <8 x i32> %h1, %h1
86 %t2 = trunc <8 x i32> %s to <8 x i16>
87 %o = xor <8 x i16> %t, %t2
88 ret <8 x i16> %o
89 }
91 define <4 x i32> @lower_trunc_4xi32(i64 %a, i64 %b, i64 %c, i64 %d) {
92 ; CHECK-LABEL: lower_trunc_4xi32:
93 ; CHECK: // %bb.0:
94 ; CHECK-NEXT: fmov d0, x2
95 ; CHECK-NEXT: fmov d1, x0
96 ; CHECK-NEXT: mov v0.d[1], x3
97 ; CHECK-NEXT: mov v1.d[1], x1
98 ; CHECK-NEXT: add v2.2d, v0.2d, v0.2d
99 ; CHECK-NEXT: add v3.2d, v1.2d, v1.2d
100 ; CHECK-NEXT: uzp1 v0.4s, v1.4s, v0.4s
101 ; CHECK-NEXT: uzp1 v1.4s, v3.4s, v2.4s
102 ; CHECK-NEXT: eor v0.16b, v0.16b, v1.16b
103 ; CHECK-NEXT: ret
104 %a1 = insertelement <4 x i64> poison, i64 %a, i64 0
105 %b1 = insertelement <4 x i64> %a1, i64 %b, i64 1
106 %c1 = insertelement <4 x i64> %b1, i64 %c, i64 2
107 %d1 = insertelement <4 x i64> %c1, i64 %d, i64 3
108 %t = trunc <4 x i64> %d1 to <4 x i32>
109 %s = add <4 x i64> %d1, %d1
110 %t2 = trunc <4 x i64> %s to <4 x i32>
111 %o = xor <4 x i32> %t, %t2
112 ret <4 x i32> %o
113 }
115 define <8 x i32> @lower_trunc_8xi32(i64 %a, i64 %b, i64 %c, i64 %d, i64 %e, i64 %f, i64 %g, i64 %h) {
116 ; CHECK-LABEL: lower_trunc_8xi32:
117 ; CHECK: // %bb.0:
118 ; CHECK-NEXT: fmov d0, x2
119 ; CHECK-NEXT: fmov d1, x0
120 ; CHECK-NEXT: fmov d2, x6
121 ; CHECK-NEXT: fmov d3, x4
122 ; CHECK-NEXT: mov v0.d[1], x3
123 ; CHECK-NEXT: mov v1.d[1], x1
124 ; CHECK-NEXT: mov v2.d[1], x7
125 ; CHECK-NEXT: mov v3.d[1], x5
126 ; CHECK-NEXT: add v4.2d, v0.2d, v0.2d
127 ; CHECK-NEXT: add v5.2d, v1.2d, v1.2d
128 ; CHECK-NEXT: add v6.2d, v2.2d, v2.2d
129 ; CHECK-NEXT: add v7.2d, v3.2d, v3.2d
130 ; CHECK-NEXT: uzp1 v2.4s, v3.4s, v2.4s
131 ; CHECK-NEXT: uzp1 v0.4s, v1.4s, v0.4s
132 ; CHECK-NEXT: uzp1 v3.4s, v5.4s, v4.4s
133 ; CHECK-NEXT: uzp1 v1.4s, v7.4s, v6.4s
134 ; CHECK-NEXT: eor v0.16b, v0.16b, v3.16b
135 ; CHECK-NEXT: eor v1.16b, v2.16b, v1.16b
136 ; CHECK-NEXT: ret
137 %a1 = insertelement <8 x i64> poison, i64 %a, i64 0
138 %b1 = insertelement <8 x i64> %a1, i64 %b, i64 1
139 %c1 = insertelement <8 x i64> %b1, i64 %c, i64 2
140 %d1 = insertelement <8 x i64> %c1, i64 %d, i64 3
141 %e1 = insertelement <8 x i64> %d1, i64 %e, i64 4
142 %f1 = insertelement <8 x i64> %e1, i64 %f, i64 5
143 %g1 = insertelement <8 x i64> %f1, i64 %g, i64 6
144 %h1 = insertelement <8 x i64> %g1, i64 %h, i64 7
145 %t = trunc <8 x i64> %h1 to <8 x i32>
146 %s = add <8 x i64> %h1, %h1
147 %t2 = trunc <8 x i64> %s to <8 x i32>
148 %o = xor <8 x i32> %t, %t2
149 ret <8 x i32> %o
150 }