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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -verify-machineinstrs -mcpu=pwr8 -mtriple=powerpc64le < %s | FileCheck %s
define i64 @test1(i64* %a, i64* %b) {
; CHECK-LABEL: test1:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: mr 5, 3
; CHECK-NEXT: ld 3, 0(3)
; CHECK-NEXT: ld 4, 0(4)
; CHECK-NEXT: mtvsrd 34, 3
; CHECK-NEXT: add 3, 3, 4
; CHECK-NEXT: mtvsrd 35, 4
; CHECK-NEXT: vavgsb 2, 2, 3
; CHECK-NEXT: stxsdx 34, 0, 5
; CHECK-NEXT: blr
entry:
%lhs = load i64, i64* %a, align 8
%rhs = load i64, i64* %b, align 8
%sum = add i64 %lhs, %rhs
%lv = insertelement <2 x i64> undef, i64 %lhs, i32 0
%rv = insertelement <2 x i64> undef, i64 %rhs, i32 0
%lhc = bitcast <2 x i64> %lv to <16 x i8>
%rhc = bitcast <2 x i64> %rv to <16 x i8>
%add = call <16 x i8> @llvm.ppc.altivec.vavgsb(<16 x i8> %lhc, <16 x i8> %rhc)
%cb = bitcast <16 x i8> %add to <2 x i64>
%fv = extractelement <2 x i64> %cb, i32 0
store i64 %fv, i64* %a, align 8
ret i64 %sum
}
define i64 @test2(i64* %a, i64* %b) {
; CHECK-LABEL: test2:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: mr 5, 3
; CHECK-NEXT: ld 3, 0(3)
; CHECK-NEXT: ld 4, 0(4)
; CHECK-NEXT: mtvsrd 34, 3
; CHECK-NEXT: add 3, 3, 4
; CHECK-NEXT: mtvsrd 35, 4
; CHECK-NEXT: vadduhm 2, 2, 3
; CHECK-NEXT: stxsdx 34, 0, 5
; CHECK-NEXT: blr
entry:
%lhs = load i64, i64* %a, align 8
%rhs = load i64, i64* %b, align 8
%sum = add i64 %lhs, %rhs
%lv = insertelement <2 x i64> undef, i64 %lhs, i32 0
%rv = insertelement <2 x i64> undef, i64 %rhs, i32 0
%lhc = bitcast <2 x i64> %lv to <8 x i16>
%rhc = bitcast <2 x i64> %rv to <8 x i16>
%add = add <8 x i16> %lhc, %rhc
%cb = bitcast <8 x i16> %add to <2 x i64>
%fv = extractelement <2 x i64> %cb, i32 0
store i64 %fv, i64* %a, align 8
ret i64 %sum
}
; Ensure that vec-ops with multiple uses aren't simplified.
define signext i16 @vecop_uses(i16* %addr) {
; CHECK-LABEL: vecop_uses:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: li 4, 16
; CHECK-NEXT: lxvd2x 1, 0, 3
; CHECK-NEXT: lxvd2x 0, 3, 4
; CHECK-NEXT: xxswapd 35, 1
; CHECK-NEXT: xxswapd 34, 0
; CHECK-NEXT: vminsh 2, 3, 2
; CHECK-NEXT: xxswapd 35, 34
; CHECK-NEXT: vminsh 2, 2, 3
; CHECK-NEXT: xxspltw 35, 34, 2
; CHECK-NEXT: vminsh 2, 2, 3
; CHECK-NEXT: vsplth 3, 2, 6
; CHECK-NEXT: vminsh 2, 2, 3
; CHECK-NEXT: xxswapd 0, 34
; CHECK-NEXT: mffprd 3, 0
; CHECK-NEXT: clrldi 3, 3, 48
; CHECK-NEXT: extsh 3, 3
; CHECK-NEXT: blr
entry:
%0 = bitcast i16* %addr to <16 x i16>*
%1 = load <16 x i16>, <16 x i16>* %0, align 2
%2 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> %1)
ret i16 %2
}
define signext i32 @vecop_uses2([4 x i32]* %a, [4 x i32]* %b, [4 x i32]* %c) {
; CHECK-LABEL: vecop_uses2:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: lxvd2x 0, 0, 3
; CHECK-NEXT: lxvd2x 1, 0, 4
; CHECK-NEXT: xxswapd 34, 0
; CHECK-NEXT: xxswapd 35, 1
; CHECK-NEXT: xxsldwi 0, 34, 34, 3
; CHECK-NEXT: vmuluwm 2, 3, 2
; CHECK-NEXT: mffprwz 3, 0
; CHECK-NEXT: xxswapd 0, 34
; CHECK-NEXT: extsw 3, 3
; CHECK-NEXT: stxvd2x 0, 0, 5
; CHECK-NEXT: blr
entry:
%0 = bitcast [4 x i32]* %a to <4 x i32>*
%1 = load <4 x i32>, <4 x i32>* %0, align 4
%2 = bitcast [4 x i32]* %b to <4 x i32>*
%3 = load <4 x i32>, <4 x i32>* %2, align 4
%4 = mul <4 x i32> %3, %1
%5 = bitcast [4 x i32]* %c to <4 x i32>*
store <4 x i32> %4, <4 x i32>* %5, align 4
%6 = extractelement <4 x i32> %1, i32 3
ret i32 %6
}
declare <16 x i8> @llvm.ppc.altivec.vavgsb(<16 x i8>, <16 x i8>)
declare i16 @llvm.vector.reduce.smin.v16i16(<16 x i16>)
|
