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
171
172
173
174
175
176
177
178
179
180
|
; RUN: llc -mtriple=aarch64 %s -o - | FileCheck %s
; RUN: llc -mtriple=aarch64 -global-isel=true -global-isel-abort=2 %s -o - | FileCheck %s
; Div whose result is unused should be removed unless we have strict exceptions
; CHECK-LABEL: unused_div:
; CHECK-NOT: fdiv
; CHECK: ret
define void @unused_div(float %x, float %y) #0 {
entry:
%add = fdiv float %x, %y
ret void
}
; CHECK-LABEL: unused_div_fpexcept_strict:
; CHECK: fdiv s0, s0, s1
; CHECK-NEXT: ret
define void @unused_div_fpexcept_strict(float %x, float %y) #0 {
entry:
%add = call float @llvm.experimental.constrained.fdiv.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
ret void
}
; CHECK-LABEL: unused_div_round_dynamic:
; CHECK-NOT: fdiv
; CHECK: ret
define void @unused_div_round_dynamic(float %x, float %y) #0 {
entry:
%add = call float @llvm.experimental.constrained.fdiv.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
ret void
}
; Machine CSE should eliminate the second add unless we have strict exceptions
; CHECK-LABEL: add_twice:
; CHECK: fadd [[ADD:s[0-9]+]], s0, s1
; CHECK-NEXT: cmp w0, #0
; CHECK-NEXT: fmul [[MUL:s[0-9]+]], [[ADD]], [[ADD]]
; CHECK-NEXT: fcsel s0, [[ADD]], [[MUL]], eq
; CHECK-NEXT: ret
define float @add_twice(float %x, float %y, i32 %n) #0 {
entry:
%add = fadd float %x, %y
%tobool.not = icmp eq i32 %n, 0
br i1 %tobool.not, label %if.end, label %if.then
if.then:
%add1 = fadd float %x, %y
%mul = fmul float %add, %add1
br label %if.end
if.end:
%a.0 = phi float [ %mul, %if.then ], [ %add, %entry ]
ret float %a.0
}
; CHECK-LABEL: add_twice_fpexcept_strict:
; CHECK: fmov [[X:s[0-9]+]], s0
; CHECK-NEXT: fadd s0, s0, s1
; CHECK-NEXT: cbz w0, [[LABEL:.LBB[0-9_]+]]
; CHECK: fadd [[ADD:s[0-9]+]], [[X]], s1
; CHECK-NEXT: fmul s0, s0, [[ADD]]
; CHECK: [[LABEL]]:
; CHECK-NEXT: ret
define float @add_twice_fpexcept_strict(float %x, float %y, i32 %n) #0 {
entry:
%add = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
%tobool.not = icmp eq i32 %n, 0
br i1 %tobool.not, label %if.end, label %if.then
if.then:
%add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
%mul = call float @llvm.experimental.constrained.fmul.f32(float %add, float %add1, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
br label %if.end
if.end:
%a.0 = phi float [ %mul, %if.then ], [ %add, %entry ]
ret float %a.0
}
; CHECK-LABEL: add_twice_round_dynamic:
; CHECK: fadd s0, s0, s1
; CHECK-NEXT: cbz w0, [[LABEL:.LBB[0-9_]+]]
; CHECK-NOT: fadd
; CHECK: fmul s0, s0, s0
; CHECK: [[LABEL]]:
; CHECK-NEXT: ret
define float @add_twice_round_dynamic(float %x, float %y, i32 %n) #0 {
entry:
%add = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
%tobool.not = icmp eq i32 %n, 0
br i1 %tobool.not, label %if.end, label %if.then
if.then:
%add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
%mul = call float @llvm.experimental.constrained.fmul.f32(float %add, float %add1, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
br label %if.end
if.end:
%a.0 = phi float [ %mul, %if.then ], [ %add, %entry ]
ret float %a.0
}
; Two adds separated by llvm.set.rounding should be preserved when rounding is
; dynamic (as they may give different results) or when we have strict exceptions
; (the llvm.set.rounding is irrelevant, but both could trap).
; CHECK-LABEL: set_rounding:
; CHECK-DAG: fadd [[SREG:s[0-9]+]], s0, s1
; CHECK-DAG: mrs [[XREG1:x[0-9]+]], FPCR
; CHECK-DAG: orr [[XREG2:x[0-9]+]], [[XREG1]], #0xc00000
; CHECK: msr FPCR, [[XREG2]]
; CHECK-NEXT: mrs [[XREG3:x[0-9]+]], FPCR
; CHECK-NEXT: and [[XREG4:x[0-9]+]], [[XREG3]], #0xffffffffff3fffff
; CHECK-NEXT: msr FPCR, [[XREG4]]
; CHECK-NEXT: fsub s0, [[SREG]], [[SREG]]
; CHECK-NEXT: ret
define float @set_rounding(float %x, float %y) #0 {
entry:
%add1 = fadd float %x, %y
call void @llvm.set.rounding(i32 0)
%add2 = fadd float %x, %y
call void @llvm.set.rounding(i32 1)
%sub = fsub float %add1, %add2
ret float %sub
}
; CHECK-LABEL: set_rounding_fpexcept_strict:
; CHECK-DAG: fadd [[SREG1:s[0-9]+]], s0, s1
; CHECK-DAG: mrs [[XREG1:x[0-9]+]], FPCR
; CHECK-DAG: orr [[XREG2:x[0-9]+]], [[XREG1]], #0xc00000
; CHECK: msr FPCR, [[XREG2]]
; CHECK-DAG: fadd [[SREG2:s[0-9]+]], s0, s1
; CHECK-DAG: mrs [[XREG3:x[0-9]+]], FPCR
; CHECK-DAG: and [[XREG4:x[0-9]+]], [[XREG3]], #0xffffffffff3fffff
; CHECK-NEXT: msr FPCR, [[XREG4]]
; CHECK-NEXT: fsub s0, [[SREG1]], [[SREG2]]
; CHECK-NEXT: ret
define float @set_rounding_fpexcept_strict(float %x, float %y) #0 {
entry:
%add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
call void @llvm.set.rounding(i32 0)
%add2 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
call void @llvm.set.rounding(i32 1)
%sub = call float @llvm.experimental.constrained.fsub.f32(float %add1, float %add2, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
ret float %sub
}
; CHECK-LABEL: set_rounding_round_dynamic:
; CHECK-DAG: fadd [[SREG1:s[0-9]+]], s0, s1
; CHECK-DAG: mrs [[XREG1:x[0-9]+]], FPCR
; CHECK-DAG: orr [[XREG2:x[0-9]+]], [[XREG1]], #0xc00000
; CHECK: msr FPCR, [[XREG2]]
; CHECK-DAG: fadd [[SREG2:s[0-9]+]], s0, s1
; CHECK-DAG: mrs [[XREG3:x[0-9]+]], FPCR
; CHECK-DAG: and [[XREG4:x[0-9]+]], [[XREG3]], #0xffffffffff3fffff
; CHECK-NEXT: msr FPCR, [[XREG4]]
; CHECK-NEXT: fsub s0, [[SREG1]], [[SREG2]]
; CHECK-NEXT: ret
define float @set_rounding_round_dynamic(float %x, float %y) #0 {
entry:
%add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
call void @llvm.set.rounding(i32 0)
%add2 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
call void @llvm.set.rounding(i32 1)
%sub = call float @llvm.experimental.constrained.fsub.f32(float %add1, float %add2, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
ret float %sub
}
declare float @llvm.experimental.constrained.fadd.f32(float, float, metadata, metadata) #0
declare float @llvm.experimental.constrained.fsub.f32(float, float, metadata, metadata) #0
declare float @llvm.experimental.constrained.fmul.f32(float, float, metadata, metadata) #0
declare float @llvm.experimental.constrained.fdiv.f32(float, float, metadata, metadata) #0
declare i32 @llvm.get.rounding()
declare void @llvm.set.rounding(i32)
attributes #0 = { strictfp }
|
