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
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn -mcpu=fiji < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GCN-SAFE,VI,VI-SAFE %s
; RUN: llc -enable-no-signed-zeros-fp-math -mtriple=amdgcn -mcpu=fiji < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GCN-NSZ,VI,VI-NSZ %s
; --------------------------------------------------------------------------------
; rcp tests
; --------------------------------------------------------------------------------
define half @v_fneg_rcp_f16(half %a) #0 {
; GCN-LABEL: v_fneg_rcp_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_rcp_f16_e64 v0, -v0
; GCN-NEXT: s_setpc_b64 s[30:31]
%rcp = call half @llvm.amdgcn.rcp.f16(half %a)
%fneg = fneg half %rcp
ret half %fneg
}
define half @v_fneg_rcp_fneg_f16(half %a) #0 {
; GCN-LABEL: v_fneg_rcp_fneg_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_rcp_f16_e32 v0, v0
; GCN-NEXT: s_setpc_b64 s[30:31]
%fneg.a = fneg half %a
%rcp = call half @llvm.amdgcn.rcp.f16(half %fneg.a)
%fneg = fneg half %rcp
ret half %fneg
}
define { half, half } @v_fneg_rcp_store_use_fneg_f16(half %a) #0 {
; GCN-LABEL: v_fneg_rcp_store_use_fneg_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_rcp_f16_e32 v2, v0
; GCN-NEXT: v_xor_b32_e32 v1, 0x8000, v0
; GCN-NEXT: v_mov_b32_e32 v0, v2
; GCN-NEXT: s_setpc_b64 s[30:31]
%fneg.a = fneg half %a
%rcp = call half @llvm.amdgcn.rcp.f16(half %fneg.a)
%fneg = fneg half %rcp
%insert.0 = insertvalue { half, half } poison, half %fneg, 0
%insert.1 = insertvalue { half, half } %insert.0, half %fneg.a, 1
ret { half, half } %insert.1
}
define { half, half } @v_fneg_rcp_multi_use_fneg_f16(half %a, half %c) #0 {
; GCN-LABEL: v_fneg_rcp_multi_use_fneg_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_rcp_f16_e32 v2, v0
; GCN-NEXT: v_mul_f16_e64 v1, -v0, v1
; GCN-NEXT: v_mov_b32_e32 v0, v2
; GCN-NEXT: s_setpc_b64 s[30:31]
%fneg.a = fneg half %a
%rcp = call half @llvm.amdgcn.rcp.f16(half %fneg.a)
%fneg = fneg half %rcp
%use1 = fmul half %fneg.a, %c
%insert.0 = insertvalue { half, half } poison, half %fneg, 0
%insert.1 = insertvalue { half, half } %insert.0, half %use1, 1
ret { half, half } %insert.1
}
; --------------------------------------------------------------------------------
; sin tests
; --------------------------------------------------------------------------------
define half @v_fneg_amdgcn_sin_f16(half %a) #0 {
; GCN-LABEL: v_fneg_amdgcn_sin_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_sin_f16_e64 v0, -v0
; GCN-NEXT: s_setpc_b64 s[30:31]
%sin = call half @llvm.amdgcn.sin.f16(half %a)
%fneg = fneg half %sin
ret half %fneg
}
; --------------------------------------------------------------------------------
; vintrp tests
; --------------------------------------------------------------------------------
define { float, float } @v_fneg_interp_p1_f16(float %a, float %b) #0 {
; GCN-LABEL: v_fneg_interp_p1_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_mul_f32_e64 v1, v0, -v1
; GCN-NEXT: s_mov_b32 m0, 0
; GCN-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 2, 2), 3
; GCN-NEXT: v_interp_p1ll_f16 v0, v1, attr0.x
; GCN-NEXT: v_interp_p1ll_f16 v1, v1, attr0.y
; GCN-NEXT: s_setpc_b64 s[30:31]
%mul = fmul float %a, %b
%fneg = fneg float %mul
%intrp0 = call float @llvm.amdgcn.interp.p1.f16(float %fneg, i32 0, i32 0, i1 false, i32 0)
%intrp1 = call float @llvm.amdgcn.interp.p1.f16(float %fneg, i32 1, i32 0, i1 false, i32 0)
%insert.0 = insertvalue { float, float } poison, float %intrp0, 0
%insert.1 = insertvalue { float, float } %insert.0, float %intrp1, 1
ret { float, float } %insert.1
}
define { half, half } @v_fneg_interp_p2_f16(float %a, float %b) #0 {
; GCN-LABEL: v_fneg_interp_p2_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_mul_f32_e64 v1, v0, -v1
; GCN-NEXT: v_mov_b32_e32 v2, 4.0
; GCN-NEXT: s_mov_b32 m0, 0
; GCN-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 2, 2), 3
; GCN-NEXT: v_interp_p2_f16 v0, v1, attr0.x, v2
; GCN-NEXT: v_interp_p2_f16 v1, v1, attr0.y, v2
; GCN-NEXT: s_setpc_b64 s[30:31]
%mul = fmul float %a, %b
%fneg = fneg float %mul
%intrp0 = call half @llvm.amdgcn.interp.p2.f16(float 4.0, float %fneg, i32 0, i32 0, i1 false, i32 0)
%intrp1 = call half @llvm.amdgcn.interp.p2.f16(float 4.0, float %fneg, i32 1, i32 0, i1 false, i32 0)
%insert.0 = insertvalue { half, half } poison, half %intrp0, 0
%insert.1 = insertvalue { half, half } %insert.0, half %intrp1, 1
ret { half, half } %insert.1
}
; --------------------------------------------------------------------------------
; arithmetic.fence tests
; --------------------------------------------------------------------------------
; FIXME: Legalization/promote is broken
define half @v_fneg_arithmetic_fence_f16(half %a) #0 {
; GCN-LABEL: v_fneg_arithmetic_fence_f16:
; GCN: ; %bb.0:
; GCN-NEXT: ;ARITH_FENCE
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_xor_b32_e32 v0, 0x8000, v0
; GCN-NEXT: s_setpc_b64 s[30:31]
%fence = call half @llvm.arithmetic.fence.f16(half %a)
%fneg = fneg half %fence
ret half %fneg
}
define half @v_fneg_arithmetic_fence_fmul_f16(half %a, half %b) #0 {
; GCN-LABEL: v_fneg_arithmetic_fence_fmul_f16:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: v_mul_f16_e32 v0, v0, v1
; GCN-NEXT: ;ARITH_FENCE
; GCN-NEXT: v_xor_b32_e32 v0, 0x8000, v0
; GCN-NEXT: s_setpc_b64 s[30:31]
%mul = fmul half %a, %b
%fence = call half @llvm.arithmetic.fence.f16(half %mul)
%fneg = fneg half %fence
ret half %fneg
}
declare half @llvm.amdgcn.rcp.f16(half) #1
declare half @llvm.amdgcn.sin.f16(half) #1
declare half @llvm.arithmetic.fence.f16(half) #1
declare float @llvm.amdgcn.interp.p1.f16(float, i32, i32, i1, i32) #0
declare half @llvm.amdgcn.interp.p2.f16(float, float, i32, i32, i1, i32) #0
attributes #0 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind readnone }
attributes #2 = { nounwind "unsafe-fp-math"="true" }
attributes #3 = { nounwind "no-signed-zeros-fp-math"="true" }
attributes #4 = { nounwind "amdgpu-ieee"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GCN-NSZ: {{.*}}
; GCN-SAFE: {{.*}}
; VI: {{.*}}
; VI-NSZ: {{.*}}
; VI-SAFE: {{.*}}
|
