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
141
| ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9_10,GFX9 %s
; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9_10,GFX10 %s
; GCN-LABEL: {{^}}test_pk_max_f16_literal_0_1:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, 1.0 op_sel:[0,1] op_sel_hi:[1,0]{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_0_1(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH0000, half 0xH3C00>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_1_0:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, 1.0{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_1_0(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH3C00, half 0xH0000>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_1_1:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, 1.0 op_sel_hi:[1,0]{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_1_1(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH3C00, half 0xH3C00>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_0_m1:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, -1.0 op_sel:[0,1] op_sel_hi:[1,0]{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_0_m1(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH0000, half 0xHBC00>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_m1_0:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, -1.0{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_m1_0(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xHBC00, half 0xH0000>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_m1_m1:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, -1.0 op_sel_hi:[1,0]{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_m1_m1(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xHBC00, half 0xHBC00>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_0_0:
; GFX9_10: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, 0{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_0_0(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH0000, half 0xH0000>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_0_41c8:
; GFX9: s_mov_b32 [[C:s[0-9]+]], 0x41c80000
; GFX9: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, [[C]]{{$}}
; GFX10: v_pk_max_f16 v{{[0-9]+}}, 0x41c8, v{{[0-9]+}} op_sel:[1,0] op_sel_hi:[0,1]{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_0_41c8(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH0000, half 0xH41C8>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_41c8_0:
; GFX9: s_movk_i32 [[C:s[0-9]+]], 0x41c8
; GFX9: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, [[C]]{{$}}
; GFX10: v_pk_max_f16 v{{[0-9]+}}, 0x41c8, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_41c8_0(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH41C8, half 0xH0>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
; GCN-LABEL: {{^}}test_pk_max_f16_literal_42ca_41c8:
; GFX9: s_mov_b32 [[C:s[0-9]+]], 0x41c842ca
; GFX9: v_pk_max_f16 v{{[0-9]+}}, v{{[0-9]+}}, [[C]]{{$}}
; GFX10: v_pk_max_f16 v{{[0-9]+}}, 0x41c842ca, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @test_pk_max_f16_literal_42ca_41c8(<2 x half> addrspace(1)* nocapture %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %arg, i64 %tmp1
%tmp3 = load <2 x half>, <2 x half> addrspace(1)* %tmp2, align 4
%tmp4 = tail call <2 x half> @llvm.maxnum.v2f16(<2 x half> %tmp3, <2 x half> <half 0xH42CA, half 0xH41C8>)
store <2 x half> %tmp4, <2 x half> addrspace(1)* %tmp2, align 4
ret void
}
declare <2 x half> @llvm.maxnum.v2f16(<2 x half>, <2 x half>)
declare i32 @llvm.amdgcn.workitem.id.x()
|