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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc --mtriple=loongarch64 --mattr=+lasx,-frecipe < %s | FileCheck %s --check-prefix=FAULT
; RUN: llc --mtriple=loongarch64 --mattr=+lasx,+frecipe < %s | FileCheck %s
;; 1.0 / (fsqrt vec)
define void @one_div_sqrt_v8f32(ptr %res, ptr %a0) nounwind {
; FAULT-LABEL: one_div_sqrt_v8f32:
; FAULT: # %bb.0: # %entry
; FAULT-NEXT: xvld $xr0, $a1, 0
; FAULT-NEXT: xvfrsqrt.s $xr0, $xr0
; FAULT-NEXT: xvst $xr0, $a0, 0
; FAULT-NEXT: ret
;
; CHECK-LABEL: one_div_sqrt_v8f32:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xvld $xr0, $a1, 0
; CHECK-NEXT: xvfrsqrte.s $xr1, $xr0
; CHECK-NEXT: xvfmul.s $xr1, $xr0, $xr1
; CHECK-NEXT: xvfmul.s $xr0, $xr0, $xr1
; CHECK-NEXT: lu12i.w $a1, -261120
; CHECK-NEXT: xvreplgr2vr.w $xr2, $a1
; CHECK-NEXT: xvfmadd.s $xr0, $xr0, $xr1, $xr2
; CHECK-NEXT: lu12i.w $a1, -266240
; CHECK-NEXT: xvreplgr2vr.w $xr2, $a1
; CHECK-NEXT: xvfmul.s $xr1, $xr1, $xr2
; CHECK-NEXT: xvfmul.s $xr0, $xr1, $xr0
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <8 x float>, ptr %a0, align 16
%sqrt = call fast <8 x float> @llvm.sqrt.v8f32 (<8 x float> %v0)
%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>, %sqrt
store <8 x float> %div, ptr %res, align 16
ret void
}
define void @one_div_sqrt_v4f64(ptr %res, ptr %a0) nounwind {
; FAULT-LABEL: one_div_sqrt_v4f64:
; FAULT: # %bb.0: # %entry
; FAULT-NEXT: xvld $xr0, $a1, 0
; FAULT-NEXT: xvfrsqrt.d $xr0, $xr0
; FAULT-NEXT: xvst $xr0, $a0, 0
; FAULT-NEXT: ret
;
; CHECK-LABEL: one_div_sqrt_v4f64:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: xvld $xr0, $a1, 0
; CHECK-NEXT: xvfrsqrte.d $xr1, $xr0
; CHECK-NEXT: xvfmul.d $xr1, $xr0, $xr1
; CHECK-NEXT: xvfmul.d $xr2, $xr0, $xr1
; CHECK-NEXT: ori $a1, $zero, 0
; CHECK-NEXT: lu32i.d $a1, -524288
; CHECK-NEXT: lu52i.d $a1, $a1, -1024
; CHECK-NEXT: xvreplgr2vr.d $xr3, $a1
; CHECK-NEXT: xvfmadd.d $xr2, $xr2, $xr1, $xr3
; CHECK-NEXT: lu52i.d $a1, $zero, -1026
; CHECK-NEXT: xvreplgr2vr.d $xr4, $a1
; CHECK-NEXT: xvfmul.d $xr1, $xr1, $xr4
; CHECK-NEXT: xvfmul.d $xr1, $xr1, $xr2
; CHECK-NEXT: xvfmul.d $xr0, $xr0, $xr1
; CHECK-NEXT: xvfmadd.d $xr0, $xr0, $xr1, $xr3
; CHECK-NEXT: xvfmul.d $xr1, $xr1, $xr4
; CHECK-NEXT: xvfmul.d $xr0, $xr1, $xr0
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <4 x double>, ptr %a0, align 16
%sqrt = call fast <4 x double> @llvm.sqrt.v4f64 (<4 x double> %v0)
%div = fdiv fast <4 x double> <double 1.0, double 1.0, double 1.0, double 1.0>, %sqrt
store <4 x double> %div, ptr %res, align 16
ret void
}
declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)
declare <4 x double> @llvm.sqrt.v4f64(<4 x double>)
|
