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
|
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -passes=load-store-vectorizer -S -o - %s | FileCheck %s
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -aa-pipeline=basic-aa -passes='function(load-store-vectorizer)' -S -o - %s | FileCheck %s
target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
; Vectorized subsets of the load/store chains in the presence of
; interleaved loads/stores
; CHECK-LABEL: @interleave_2L_2S(
; CHECK: load <2 x i32>
; CHECK: load i32
; CHECK: store <2 x i32>
; CHECK: load i32
define void @interleave_2L_2S(ptr noalias %ptr) {
%next.gep1 = getelementptr i32, ptr %ptr, i64 1
%next.gep2 = getelementptr i32, ptr %ptr, i64 2
%l1 = load i32, ptr %next.gep1, align 4
%l2 = load i32, ptr %ptr, align 4
store i32 0, ptr %next.gep1, align 4
store i32 0, ptr %ptr, align 4
%l3 = load i32, ptr %next.gep1, align 4
%l4 = load i32, ptr %next.gep2, align 4
ret void
}
; CHECK-LABEL: @interleave_3L_2S_1L(
; CHECK: load <3 x i32>
; CHECK: store <2 x i32>
; CHECK: load i32
define void @interleave_3L_2S_1L(ptr noalias %ptr) {
%next.gep1 = getelementptr i32, ptr %ptr, i64 1
%next.gep2 = getelementptr i32, ptr %ptr, i64 2
%l2 = load i32, ptr %ptr, align 4
%l1 = load i32, ptr %next.gep1, align 4
store i32 0, ptr %next.gep1, align 4
store i32 0, ptr %ptr, align 4
%l3 = load i32, ptr %next.gep1, align 4
%l4 = load i32, ptr %next.gep2, align 4
ret void
}
; CHECK-LABEL: @chain_suffix(
; CHECK: load i32
; CHECK: store <2 x i32>
; CHECK: load <2 x i32>
define void @chain_suffix(ptr noalias %ptr) {
%next.gep1 = getelementptr i32, ptr %ptr, i64 1
%next.gep2 = getelementptr i32, ptr %ptr, i64 2
%l2 = load i32, ptr %ptr, align 4
store i32 0, ptr %next.gep1, align 4
store i32 0, ptr %ptr, align 4
%l3 = load i32, ptr %next.gep1, align 4
%l4 = load i32, ptr %next.gep2, align 4
ret void
}
; CHECK-LABEL: @chain_prefix_suffix(
; CHECK: load <2 x i32>
; CHECK: store <2 x i32>
; CHECK: load <3 x i32>
define void @chain_prefix_suffix(ptr noalias %ptr) {
%next.gep1 = getelementptr i32, ptr %ptr, i64 1
%next.gep2 = getelementptr i32, ptr %ptr, i64 2
%next.gep3 = getelementptr i32, ptr %ptr, i64 3
%l1 = load i32, ptr %ptr, align 4
%l2 = load i32, ptr %next.gep1, align 4
store i32 0, ptr %next.gep1, align 4
store i32 0, ptr %next.gep2, align 4
%l3 = load i32, ptr %next.gep1, align 4
%l4 = load i32, ptr %next.gep2, align 4
%l5 = load i32, ptr %next.gep3, align 4
ret void
}
; FIXME: If the chain is too long and TLI says misaligned is not fast,
; then LSV fails to vectorize anything in that chain.
; To reproduce below, add a tmp5 (ptr+4) and load tmp5 into l6 and l7.
; CHECK-LABEL: @interleave_get_longest
; CHECK: load <3 x i32>
; CHECK: load i32
; CHECK: store <2 x i32> zeroinitializer
; CHECK: load i32
; CHECK: load i32
; CHECK: load i32
define void @interleave_get_longest(ptr noalias %ptr) {
%tmp2 = getelementptr i32, ptr %ptr, i64 1
%tmp3 = getelementptr i32, ptr %ptr, i64 2
%tmp4 = getelementptr i32, ptr %ptr, i64 3
%l1 = load i32, ptr %tmp2, align 4
%l2 = load i32, ptr %ptr, align 4
store i32 0, ptr %tmp2, align 4
store i32 0, ptr %ptr, align 4
%l3 = load i32, ptr %tmp2, align 4
%l4 = load i32, ptr %tmp3, align 4
%l5 = load i32, ptr %tmp4, align 4
%l6 = load i32, ptr %tmp4, align 4
%l7 = load i32, ptr %tmp4, align 4
ret void
}
|
