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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc --mtriple=loongarch64 --mattr=+lasx,-frecipe < %s | FileCheck %s --check-prefix=FAULT
; RUN: llc --mtriple=loongarch64 --mattr=+lasx,+frecipe < %s | FileCheck %s
define void @fdiv_v8f32(ptr %res, ptr %a0, ptr %a1) nounwind {
; FAULT-LABEL: fdiv_v8f32:
; FAULT: # %bb.0:
; FAULT-NEXT: xvld $xr0, $a1, 0
; FAULT-NEXT: xvld $xr1, $a2, 0
; FAULT-NEXT: xvfdiv.s $xr0, $xr0, $xr1
; FAULT-NEXT: xvst $xr0, $a0, 0
; FAULT-NEXT: ret
;
; CHECK-LABEL: fdiv_v8f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xvld $xr0, $a2, 0
; CHECK-NEXT: xvld $xr1, $a1, 0
; CHECK-NEXT: xvfrecipe.s $xr2, $xr0
; CHECK-NEXT: xvfmul.s $xr3, $xr1, $xr2
; CHECK-NEXT: xvfnmsub.s $xr0, $xr0, $xr3, $xr1
; CHECK-NEXT: xvfmadd.s $xr0, $xr2, $xr0, $xr3
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <8 x float>, ptr %a0
%v1 = load <8 x float>, ptr %a1
%v2 = fdiv fast <8 x float> %v0, %v1
store <8 x float> %v2, ptr %res
ret void
}
define void @fdiv_v4f64(ptr %res, ptr %a0, ptr %a1) nounwind {
; FAULT-LABEL: fdiv_v4f64:
; FAULT: # %bb.0:
; FAULT-NEXT: xvld $xr0, $a1, 0
; FAULT-NEXT: xvld $xr1, $a2, 0
; FAULT-NEXT: xvfdiv.d $xr0, $xr0, $xr1
; FAULT-NEXT: xvst $xr0, $a0, 0
; FAULT-NEXT: ret
;
; CHECK-LABEL: fdiv_v4f64:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xvld $xr0, $a2, 0
; CHECK-NEXT: xvld $xr1, $a1, 0
; CHECK-NEXT: lu52i.d $a1, $zero, -1025
; CHECK-NEXT: xvreplgr2vr.d $xr2, $a1
; CHECK-NEXT: xvfrecipe.d $xr3, $xr0
; CHECK-NEXT: xvfmadd.d $xr2, $xr0, $xr3, $xr2
; CHECK-NEXT: xvfnmsub.d $xr2, $xr2, $xr3, $xr3
; CHECK-NEXT: xvfmul.d $xr3, $xr1, $xr2
; CHECK-NEXT: xvfnmsub.d $xr0, $xr0, $xr3, $xr1
; CHECK-NEXT: xvfmadd.d $xr0, $xr2, $xr0, $xr3
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <4 x double>, ptr %a0
%v1 = load <4 x double>, ptr %a1
%v2 = fdiv fast <4 x double> %v0, %v1
store <4 x double> %v2, ptr %res
ret void
}
;; 1.0 / vec
define void @one_fdiv_v8f32(ptr %res, ptr %a0) nounwind {
; FAULT-LABEL: one_fdiv_v8f32:
; FAULT: # %bb.0:
; FAULT-NEXT: xvld $xr0, $a1, 0
; FAULT-NEXT: xvfrecip.s $xr0, $xr0
; FAULT-NEXT: xvst $xr0, $a0, 0
; FAULT-NEXT: ret
;
; CHECK-LABEL: one_fdiv_v8f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xvld $xr0, $a1, 0
; CHECK-NEXT: xvfrecipe.s $xr1, $xr0
; CHECK-NEXT: lu12i.w $a1, -264192
; CHECK-NEXT: xvreplgr2vr.w $xr2, $a1
; CHECK-NEXT: xvfmadd.s $xr0, $xr0, $xr1, $xr2
; CHECK-NEXT: xvfnmsub.s $xr0, $xr0, $xr1, $xr1
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <8 x float>, ptr %a0
%div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %v0
store <8 x float> %div, ptr %res
ret void
}
define void @one_fdiv_v4f64(ptr %res, ptr %a0) nounwind {
; FAULT-LABEL: one_fdiv_v4f64:
; FAULT: # %bb.0:
; FAULT-NEXT: xvld $xr0, $a1, 0
; FAULT-NEXT: xvfrecip.d $xr0, $xr0
; FAULT-NEXT: xvst $xr0, $a0, 0
; FAULT-NEXT: ret
;
; CHECK-LABEL: one_fdiv_v4f64:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xvld $xr0, $a1, 0
; CHECK-NEXT: xvfrecipe.d $xr1, $xr0
; CHECK-NEXT: lu52i.d $a1, $zero, 1023
; CHECK-NEXT: xvreplgr2vr.d $xr2, $a1
; CHECK-NEXT: xvfnmsub.d $xr3, $xr0, $xr1, $xr2
; CHECK-NEXT: xvfmadd.d $xr1, $xr1, $xr3, $xr1
; CHECK-NEXT: xvfnmsub.d $xr0, $xr0, $xr1, $xr2
; CHECK-NEXT: xvfmadd.d $xr0, $xr1, $xr0, $xr1
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <4 x double>, ptr %a0
%div = fdiv fast <4 x double> <double 1.0, double 1.0, double 1.0, double 1.0>, %v0
store <4 x double> %div, ptr %res
ret void
}
|
