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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=fiji -o - < %s | FileCheck --check-prefixes=GCN %s
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 -o - < %s | FileCheck --check-prefixes=GCN %s
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 -o - < %s | FileCheck --check-prefixes=GCN %s
; Test vector bitfield extract.
define i32 @v_srl_mask_i32(i32 %value) {
; GCN-LABEL: v_srl_mask_i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_bfe_u32 v0, v0, 8, 5
; GCN-NEXT: s_setpc_b64 s[30:31]
%1 = lshr i32 %value, 8
%2 = and i32 %1, 31
ret i32 %2
}
; Test scalar bitfield extract.
define amdgpu_ps i32 @s_srl_mask_i32(i32 inreg %value) {
; GCN-LABEL: s_srl_mask_i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_bfe_u32 s0, s0, 0x50008
; GCN-NEXT: ; return to shader part epilog
%1 = lshr i32 %value, 8
%2 = and i32 %1, 31
ret i32 %2
}
; Don't generate G_UBFX if the offset + width is too big.
define amdgpu_ps i32 @s_srl_big_mask_i32(i32 inreg %value) {
; GCN-LABEL: s_srl_big_mask_i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_lshr_b32 s0, s0, 30
; GCN-NEXT: ; return to shader part epilog
%1 = lshr i32 %value, 30
%2 = and i32 %1, 31
ret i32 %2
}
; Test vector bitfield extract.
define i32 @v_mask_srl_i32(i32 %value) {
; GCN-LABEL: v_mask_srl_i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_bfe_u32 v0, v0, 8, 5
; GCN-NEXT: s_setpc_b64 s[30:31]
%1 = and i32 %value, 7936 ; 31 << 8
%2 = lshr i32 %1, 8
ret i32 %2
}
; Test scalar bitfield extract.
define amdgpu_ps i32 @s_mask_srl_i32(i32 inreg %value) {
; GCN-LABEL: s_mask_srl_i32:
; GCN: ; %bb.0:
; GCN-NEXT: s_bfe_u32 s0, s0, 0x50008
; GCN-NEXT: ; return to shader part epilog
%1 = and i32 %value, 7936 ; 31 << 8
%2 = lshr i32 %1, 8
ret i32 %2
}
; Test vector bitfield extract for 64-bits.
define i64 @v_srl_mask_i64(i64 %value) {
; GCN-LABEL: v_srl_mask_i64:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_lshrrev_b64 v[0:1], 25, v[0:1]
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: v_bfe_u32 v0, v0, 0, 10
; GCN-NEXT: s_setpc_b64 s[30:31]
%1 = lshr i64 %value, 25
%2 = and i64 %1, 1023
ret i64 %2
}
; Test scalar bitfield extract for 64-bits.
define amdgpu_ps i64 @s_srl_mask_i64(i64 inreg %value) {
; GCN-LABEL: s_srl_mask_i64:
; GCN: ; %bb.0:
; GCN-NEXT: s_bfe_u64 s[0:1], s[0:1], 0xa0019
; GCN-NEXT: ; return to shader part epilog
%1 = lshr i64 %value, 25
%2 = and i64 %1, 1023
ret i64 %2
}
; Don't generate G_UBFX if the offset + width is too big.
define amdgpu_ps i64 @s_srl_big_mask_i64(i64 inreg %value) {
; GCN-LABEL: s_srl_big_mask_i64:
; GCN: ; %bb.0:
; GCN-NEXT: s_lshr_b32 s0, s1, 28
; GCN-NEXT: s_mov_b32 s1, 0
; GCN-NEXT: ; return to shader part epilog
%1 = lshr i64 %value, 60
%2 = and i64 %1, 63
ret i64 %2
}
; Test vector bitfield extract for 64-bits.
; TODO: No need for a 64-bit shift instruction when the extracted value is
; entirely contained within the upper or lower half.
define i64 @v_mask_srl_i64(i64 %value) {
; GCN-LABEL: v_mask_srl_i64:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_lshrrev_b64 v[0:1], 25, v[0:1]
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: v_bfe_u32 v0, v0, 0, 10
; GCN-NEXT: s_setpc_b64 s[30:31]
%1 = and i64 %value, 34326183936 ; 1023 << 25
%2 = lshr i64 %1, 25
ret i64 %2
}
; Test scalar bitfield extract for 64-bits.
define amdgpu_ps i64 @s_mask_srl_i64(i64 inreg %value) {
; GCN-LABEL: s_mask_srl_i64:
; GCN: ; %bb.0:
; GCN-NEXT: s_bfe_u64 s[0:1], s[0:1], 0xa0019
; GCN-NEXT: ; return to shader part epilog
%1 = and i64 %value, 34326183936 ; 1023 << 25
%2 = lshr i64 %1, 25
ret i64 %2
}
|
