1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck %s
; extract element 0 as shift
define i32 @cast_v4i32_to_i128_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%trunc = trunc i128 %bigint to i32
ret i32 %trunc
}
; extract element 1 as shift
define i32 @cast_v4i32_to_i128_lshr_32_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_32_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_mov_b32_e32 v0, v1
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 32
%trunc = trunc i128 %srl to i32
ret i32 %trunc
}
; extract element 2 as shift
define i32 @cast_v4i32_to_i128_lshr_64_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_64_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_mov_b32_e32 v0, v2
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 64
%trunc = trunc i128 %srl to i32
ret i32 %trunc
}
; extract element 3 as shift
define i32 @cast_v4i32_to_i128_lshr_96_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_96_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_mov_b32_e32 v0, v3
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 96
%trunc = trunc i128 %srl to i32
ret i32 %trunc
}
; Shift not aligned to element, not a simple extract
define i32 @cast_v4i32_to_i128_lshr_33_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_33_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_alignbit_b32 v0, v2, v1, 1
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 33
%trunc = trunc i128 %srl to i32
ret i32 %trunc
}
; extract misaligned element
define i32 @cast_v4i32_to_i128_lshr_31_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_31_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_alignbit_b32 v0, v1, v0, 31
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 31
%trunc = trunc i128 %srl to i32
ret i32 %trunc
}
; extract misaligned element
define i32 @cast_v4i32_to_i128_lshr_48_trunc_i32(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_48_trunc_i32:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: s_mov_b32 s4, 0x1000706
; CHECK-NEXT: v_perm_b32 v0, v1, v2, s4
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 48
%trunc = trunc i128 %srl to i32
ret i32 %trunc
}
; extract elements 1 and 2 with shift
define i64 @cast_v4i32_to_i128_lshr_32_trunc_i64(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_32_trunc_i64:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_mov_b32_e32 v0, v1
; CHECK-NEXT: v_mov_b32_e32 v1, v2
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 32
%trunc = trunc i128 %srl to i64
ret i64 %trunc
}
; extract elements 2 and 3 with shift
define i64 @cast_v4i32_to_i128_lshr_64_trunc_i64(<4 x i32> %arg) {
; CHECK-LABEL: cast_v4i32_to_i128_lshr_64_trunc_i64:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_mov_b32_e32 v1, v3
; CHECK-NEXT: v_mov_b32_e32 v0, v2
; CHECK-NEXT: s_setpc_b64 s[30:31]
%bigint = bitcast <4 x i32> %arg to i128
%srl = lshr i128 %bigint, 64
%trunc = trunc i128 %srl to i64
ret i64 %trunc
}
; FIXME: We don't process this case because we see multiple bitcasts
; before a 32-bit build_vector
define i32 @build_vector_i16_to_shift(i16 %arg0, i16 %arg1, i16 %arg2, i16 %arg3) {
; CHECK-LABEL: build_vector_i16_to_shift:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: s_mov_b32 s4, 0x5040100
; CHECK-NEXT: v_perm_b32 v0, v3, v2, s4
; CHECK-NEXT: s_setpc_b64 s[30:31]
%ins.0 = insertelement <4 x i16> poison, i16 %arg0, i32 0
%ins.1 = insertelement <4 x i16> %ins.0, i16 %arg1, i32 1
%ins.2 = insertelement <4 x i16> %ins.1, i16 %arg2, i32 2
%ins.3 = insertelement <4 x i16> %ins.2, i16 %arg3, i32 3
%cast = bitcast <4 x i16> %ins.3 to i64
%srl = lshr i64 %cast, 32
%trunc = trunc i64 %srl to i32
ret i32 %trunc
}
|
