aboutsummaryrefslogtreecommitdiff
path: root/llvm/test/CodeGen/AMDGPU/fabs-r600.ll
blob: 7e1aa99c3ec40f97ceacdb6db45ff27844087a57 (plain)
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
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 %s


; DAGCombiner will transform:
; (fabsf (f32 bitcast (i32 a))) => (f32 bitcast (and (i32 a), 0x7FFFFFFF))
; unless isFabsFree returns true
define amdgpu_kernel void @s_fabsf_fn_free(ptr addrspace(1) %out, i32 %in) {
; R600-LABEL: s_fabsf_fn_free:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     MOV * T0.W, KC0[2].Z,
; R600-NEXT:     MOV T0.X, |PV.W|,
; R600-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %bc= bitcast i32 %in to float
  %fabs = call float @fabsf(float %bc)
  store float %fabs, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @s_fabsf_free(ptr addrspace(1) %out, i32 %in) {
; R600-LABEL: s_fabsf_free:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     MOV * T0.W, KC0[2].Z,
; R600-NEXT:     MOV T0.X, |PV.W|,
; R600-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %bc= bitcast i32 %in to float
  %fabs = call float @llvm.fabs.f32(float %bc)
  store float %fabs, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @s_fabsf_f32(ptr addrspace(1) %out, float %in) {
; R600-LABEL: s_fabsf_f32:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     MOV * T0.W, KC0[2].Z,
; R600-NEXT:     MOV T0.X, |PV.W|,
; R600-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %fabs = call float @llvm.fabs.f32(float %in)
  store float %fabs, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @fabs_v2f32(ptr addrspace(1) %out, <2 x float> %in) {
; R600-LABEL: fabs_v2f32:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 5, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     MOV * T0.W, KC0[3].X,
; R600-NEXT:     MOV T0.Y, |PV.W|,
; R600-NEXT:     MOV * T0.W, KC0[2].W,
; R600-NEXT:     MOV T0.X, |PV.W|,
; R600-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %fabs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %in)
  store <2 x float> %fabs, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @fabsf_v4f32(ptr addrspace(1) %out, <4 x float> %in) {
; R600-LABEL: fabsf_v4f32:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 9, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     MOV T0.W, KC0[4].X,
; R600-NEXT:     MOV * T1.W, KC0[3].W,
; R600-NEXT:     MOV * T0.W, |PV.W|,
; R600-NEXT:     MOV T0.Z, |T1.W|,
; R600-NEXT:     MOV * T1.W, KC0[3].Z,
; R600-NEXT:     MOV T0.Y, |PV.W|,
; R600-NEXT:     MOV * T1.W, KC0[3].Y,
; R600-NEXT:     MOV T0.X, |PV.W|,
; R600-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %fabs = call <4 x float> @llvm.fabs.v4f32(<4 x float> %in)
  store <4 x float> %fabs, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @fabsf_fn_fold(ptr addrspace(1) %out, float %in0, float %in1) {
; R600-LABEL: fabsf_fn_fold:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     LSHR T0.X, KC0[2].Y, literal.x,
; R600-NEXT:     MUL_IEEE * T1.X, |KC0[2].Z|, KC0[2].W,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %fabs = call float @fabsf(float %in0)
  %fmul = fmul float %fabs, %in1
  store float %fmul, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @fabs_fold(ptr addrspace(1) %out, float %in0, float %in1) {
; R600-LABEL: fabs_fold:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     LSHR T0.X, KC0[2].Y, literal.x,
; R600-NEXT:     MUL_IEEE * T1.X, |KC0[2].Z|, KC0[2].W,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %fabs = call float @llvm.fabs.f32(float %in0)
  %fmul = fmul float %fabs, %in1
  store float %fmul, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @bitpreserve_fabsf_f32(ptr addrspace(1) %out, float %in) {
; R600-LABEL: bitpreserve_fabsf_f32:
; R600:       ; %bb.0:
; R600-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; R600-NEXT:    CF_END
; R600-NEXT:    PAD
; R600-NEXT:    ALU clause starting at 4:
; R600-NEXT:     LSHR T0.X, KC0[2].Y, literal.x,
; R600-NEXT:     ADD * T1.X, |KC0[2].Z|, 1.0,
; R600-NEXT:    2(2.802597e-45), 0(0.000000e+00)
  %in.bc = bitcast float %in to i32
  %int.abs = and i32 %in.bc, 2147483647
  %bc = bitcast i32 %int.abs to float
  %fadd = fadd float %bc, 1.0
  store float %fadd, ptr addrspace(1) %out
  ret void
}

declare float @fabsf(float) readnone
declare float @llvm.fabs.f32(float) readnone
declare <2 x float> @llvm.fabs.v2f32(<2 x float>) readnone
declare <4 x float> @llvm.fabs.v4f32(<4 x float>) readnone