1 ; RUN: opt -passes=loop-vectorize -force-vector-width=2 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_2
2 ; RUN: opt -passes=loop-vectorize -force-vector-width=4 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_4
3 ; RUN: opt -passes=loop-vectorize -force-vector-width=8 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_8
4 ; RUN: opt -passes=loop-vectorize -force-vector-width=16 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_16
7 target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
8 target triple = "aarch64"
11 define void @i8_factor_2(ptr %data, i64 %n) {
15 ; VF_8-LABEL: Checking a loop in 'i8_factor_2'
16 ; VF_8: Found an estimated cost of 2 for VF 8 For instruction: %tmp2 = load i8, ptr %tmp0, align 1
17 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i8, ptr %tmp1, align 1
18 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: store i8 %tmp2, ptr %tmp0, align 1
19 ; VF_8-NEXT: Found an estimated cost of 2 for VF 8 For instruction: store i8 %tmp3, ptr %tmp1, align 1
20 ; VF_16-LABEL: Checking a loop in 'i8_factor_2'
21 ; VF_16: Found an estimated cost of 2 for VF 16 For instruction: %tmp2 = load i8, ptr %tmp0, align 1
22 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i8, ptr %tmp1, align 1
23 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: store i8 %tmp2, ptr %tmp0, align 1
24 ; VF_16-NEXT: Found an estimated cost of 2 for VF 16 For instruction: store i8 %tmp3, ptr %tmp1, align 1
26 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
27 %tmp0 = getelementptr inbounds %i8.2, ptr %data, i64 %i, i32 0
28 %tmp1 = getelementptr inbounds %i8.2, ptr %data, i64 %i, i32 1
29 %tmp2 = load i8, ptr %tmp0, align 1
30 %tmp3 = load i8, ptr %tmp1, align 1
31 store i8 %tmp2, ptr %tmp0, align 1
32 store i8 %tmp3, ptr %tmp1, align 1
33 %i.next = add nuw nsw i64 %i, 1
34 %cond = icmp slt i64 %i.next, %n
35 br i1 %cond, label %for.body, label %for.end
41 %i16.2 = type {i16, i16}
42 define void @i16_factor_2(ptr %data, i64 %n) {
46 ; VF_4-LABEL: Checking a loop in 'i16_factor_2'
47 ; VF_4: Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load i16, ptr %tmp0, align 2
48 ; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i16, ptr %tmp1, align 2
49 ; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: store i16 %tmp2, ptr %tmp0, align 2
50 ; VF_4-NEXT: Found an estimated cost of 2 for VF 4 For instruction: store i16 %tmp3, ptr %tmp1, align 2
51 ; VF_8-LABEL: Checking a loop in 'i16_factor_2'
52 ; VF_8: Found an estimated cost of 2 for VF 8 For instruction: %tmp2 = load i16, ptr %tmp0, align 2
53 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i16, ptr %tmp1, align 2
54 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: store i16 %tmp2, ptr %tmp0, align 2
55 ; VF_8-NEXT: Found an estimated cost of 2 for VF 8 For instruction: store i16 %tmp3, ptr %tmp1, align 2
56 ; VF_16-LABEL: Checking a loop in 'i16_factor_2'
57 ; VF_16: Found an estimated cost of 4 for VF 16 For instruction: %tmp2 = load i16, ptr %tmp0, align 2
58 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i16, ptr %tmp1, align 2
59 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: store i16 %tmp2, ptr %tmp0, align 2
60 ; VF_16-NEXT: Found an estimated cost of 4 for VF 16 For instruction: store i16 %tmp3, ptr %tmp1, align 2
62 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
63 %tmp0 = getelementptr inbounds %i16.2, ptr %data, i64 %i, i32 0
64 %tmp1 = getelementptr inbounds %i16.2, ptr %data, i64 %i, i32 1
65 %tmp2 = load i16, ptr %tmp0, align 2
66 %tmp3 = load i16, ptr %tmp1, align 2
67 store i16 %tmp2, ptr %tmp0, align 2
68 store i16 %tmp3, ptr %tmp1, align 2
69 %i.next = add nuw nsw i64 %i, 1
70 %cond = icmp slt i64 %i.next, %n
71 br i1 %cond, label %for.body, label %for.end
77 %i32.2 = type {i32, i32}
78 define void @i32_factor_2(ptr %data, i64 %n) {
82 ; VF_2-LABEL: Checking a loop in 'i32_factor_2'
83 ; VF_2: Found an estimated cost of 2 for VF 2 For instruction: %tmp2 = load i32, ptr %tmp0, align 4
84 ; VF_2-NEXT: Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load i32, ptr %tmp1, align 4
85 ; VF_2-NEXT: Found an estimated cost of 0 for VF 2 For instruction: store i32 %tmp2, ptr %tmp0, align 4
86 ; VF_2-NEXT: Found an estimated cost of 2 for VF 2 For instruction: store i32 %tmp3, ptr %tmp1, align 4
87 ; VF_4-LABEL: Checking a loop in 'i32_factor_2'
88 ; VF_4: Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load i32, ptr %tmp0, align 4
89 ; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i32, ptr %tmp1, align 4
90 ; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: store i32 %tmp2, ptr %tmp0, align 4
91 ; VF_4-NEXT: Found an estimated cost of 2 for VF 4 For instruction: store i32 %tmp3, ptr %tmp1, align 4
92 ; VF_8-LABEL: Checking a loop in 'i32_factor_2'
93 ; VF_8: Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load i32, ptr %tmp0, align 4
94 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i32, ptr %tmp1, align 4
95 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: store i32 %tmp2, ptr %tmp0, align 4
96 ; VF_8-NEXT: Found an estimated cost of 4 for VF 8 For instruction: store i32 %tmp3, ptr %tmp1, align 4
97 ; VF_16-LABEL: Checking a loop in 'i32_factor_2'
98 ; VF_16: Found an estimated cost of 8 for VF 16 For instruction: %tmp2 = load i32, ptr %tmp0, align 4
99 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i32, ptr %tmp1, align 4
100 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: store i32 %tmp2, ptr %tmp0, align 4
101 ; VF_16-NEXT: Found an estimated cost of 8 for VF 16 For instruction: store i32 %tmp3, ptr %tmp1, align 4
103 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
104 %tmp0 = getelementptr inbounds %i32.2, ptr %data, i64 %i, i32 0
105 %tmp1 = getelementptr inbounds %i32.2, ptr %data, i64 %i, i32 1
106 %tmp2 = load i32, ptr %tmp0, align 4
107 %tmp3 = load i32, ptr %tmp1, align 4
108 store i32 %tmp2, ptr %tmp0, align 4
109 store i32 %tmp3, ptr %tmp1, align 4
110 %i.next = add nuw nsw i64 %i, 1
111 %cond = icmp slt i64 %i.next, %n
112 br i1 %cond, label %for.body, label %for.end
118 %i64.2 = type {i64, i64}
119 define void @i64_factor_2(ptr %data, i64 %n) {
123 ; VF_2-LABEL: Checking a loop in 'i64_factor_2'
124 ; VF_2: Found an estimated cost of 2 for VF 2 For instruction: %tmp2 = load i64, ptr %tmp0, align 8
125 ; VF_2-NEXT: Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load i64, ptr %tmp1, align 8
126 ; VF_2-NEXT: Found an estimated cost of 0 for VF 2 For instruction: store i64 %tmp2, ptr %tmp0, align 8
127 ; VF_2-NEXT: Found an estimated cost of 2 for VF 2 For instruction: store i64 %tmp3, ptr %tmp1, align 8
128 ; VF_4-LABEL: Checking a loop in 'i64_factor_2'
129 ; VF_4: Found an estimated cost of 4 for VF 4 For instruction: %tmp2 = load i64, ptr %tmp0, align 8
130 ; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i64, ptr %tmp1, align 8
131 ; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: store i64 %tmp2, ptr %tmp0, align 8
132 ; VF_4-NEXT: Found an estimated cost of 4 for VF 4 For instruction: store i64 %tmp3, ptr %tmp1, align 8
133 ; VF_8-LABEL: Checking a loop in 'i64_factor_2'
134 ; VF_8: Found an estimated cost of 8 for VF 8 For instruction: %tmp2 = load i64, ptr %tmp0, align 8
135 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i64, ptr %tmp1, align 8
136 ; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: store i64 %tmp2, ptr %tmp0, align 8
137 ; VF_8-NEXT: Found an estimated cost of 8 for VF 8 For instruction: store i64 %tmp3, ptr %tmp1, align 8
138 ; VF_16-LABEL: Checking a loop in 'i64_factor_2'
139 ; VF_16: Found an estimated cost of 16 for VF 16 For instruction: %tmp2 = load i64, ptr %tmp0, align 8
140 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i64, ptr %tmp1, align 8
141 ; VF_16-NEXT: Found an estimated cost of 0 for VF 16 For instruction: store i64 %tmp2, ptr %tmp0, align 8
142 ; VF_16-NEXT: Found an estimated cost of 16 for VF 16 For instruction: store i64 %tmp3, ptr %tmp1, align 8
144 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
145 %tmp0 = getelementptr inbounds %i64.2, ptr %data, i64 %i, i32 0
146 %tmp1 = getelementptr inbounds %i64.2, ptr %data, i64 %i, i32 1
147 %tmp2 = load i64, ptr %tmp0, align 8
148 %tmp3 = load i64, ptr %tmp1, align 8
149 store i64 %tmp2, ptr %tmp0, align 8
150 store i64 %tmp3, ptr %tmp1, align 8
151 %i.next = add nuw nsw i64 %i, 1
152 %cond = icmp slt i64 %i.next, %n
153 br i1 %cond, label %for.body, label %for.end
159 %i64.8 = type {i64, i64, i64, i64, i64, i64, i64, i64}
160 define void @i64_factor_8(ptr %data, i64 %n) {
164 ; The interleave factor in this test is 8, which is greater than the maximum
165 ; allowed factor for AArch64 (4). Thus, we will fall back to the basic TTI
166 ; implementation for determining the cost of the interleaved load group. The
167 ; stores do not form a legal interleaved group because the group would contain
170 ; VF_2-LABEL: Checking a loop in 'i64_factor_8'
171 ; VF_2: Found an estimated cost of 8 for VF 2 For instruction: %tmp2 = load i64, ptr %tmp0, align 8
172 ; VF_2-NEXT: Found an estimated cost of 8 for VF 2 For instruction: %tmp3 = load i64, ptr %tmp1, align 8
173 ; VF_2-NEXT: Found an estimated cost of 12 for VF 2 For instruction: store i64 %tmp2, ptr %tmp0, align 8
174 ; VF_2-NEXT: Found an estimated cost of 12 for VF 2 For instruction: store i64 %tmp3, ptr %tmp1, align 8
176 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
177 %tmp0 = getelementptr inbounds %i64.8, ptr %data, i64 %i, i32 2
178 %tmp1 = getelementptr inbounds %i64.8, ptr %data, i64 %i, i32 6
179 %tmp2 = load i64, ptr %tmp0, align 8
180 %tmp3 = load i64, ptr %tmp1, align 8
181 store i64 %tmp2, ptr %tmp0, align 8
182 store i64 %tmp3, ptr %tmp1, align 8
183 %i.next = add nuw nsw i64 %i, 1
184 %cond = icmp slt i64 %i.next, %n
185 br i1 %cond, label %for.body, label %for.end