Skip to content

Commit 2251f33

Browse files
RosieSumpterRosieSumpter
committed
[SLP][AArch64] Add SLP vectorizer regression test. NFC
This test is for a missed SLP vectorizer opportunity, reported here https://bugs.llvm.org/show_bug.cgi?id=44593. This is due to a cost modelling issue with vector reduction intrinsics which will be fixed in a future commit (see https://reviews.llvm.org/D104538).
1 parent a427390 commit 2251f33

File tree

1 file changed

+107
-0
lines changed

1 file changed

+107
-0
lines changed

‎llvm/test/Transforms/SLPVectorizer/AArch64/slp-or-reduction.ll

Lines changed: 107 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,107 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2+
; RUN: opt < %s -S -mtriple=aarch64 -slp-vectorizer | FileCheck %s
3+
4+
%struct.buf = type { [8 x i8] }
5+
6+
define i8 @reduce_or(%struct.buf* %a, %struct.buf* %b) {
7+
; CHECK-LABEL: @reduce_or(
8+
; CHECK-NEXT: entry:
9+
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_BUF:%.*]], %struct.buf* [[A:%.*]], i64 0, i32 0, i64 0
10+
; CHECK-NEXT: [[TMP0:%.*]] = load i8, i8* [[ARRAYIDX]], align 1
11+
; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B:%.*]], i64 0, i32 0, i64 0
12+
; CHECK-NEXT: [[TMP1:%.*]] = load i8, i8* [[ARRAYIDX3]], align 1
13+
; CHECK-NEXT: [[XOR12:%.*]] = xor i8 [[TMP1]], [[TMP0]]
14+
; CHECK-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 1
15+
; CHECK-NEXT: [[TMP2:%.*]] = load i8, i8* [[ARRAYIDX_1]], align 1
16+
; CHECK-NEXT: [[ARRAYIDX3_1:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 1
17+
; CHECK-NEXT: [[TMP3:%.*]] = load i8, i8* [[ARRAYIDX3_1]], align 1
18+
; CHECK-NEXT: [[XOR12_1:%.*]] = xor i8 [[TMP3]], [[TMP2]]
19+
; CHECK-NEXT: [[OR13_1:%.*]] = or i8 [[XOR12_1]], [[XOR12]]
20+
; CHECK-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 2
21+
; CHECK-NEXT: [[TMP4:%.*]] = load i8, i8* [[ARRAYIDX_2]], align 1
22+
; CHECK-NEXT: [[ARRAYIDX3_2:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 2
23+
; CHECK-NEXT: [[TMP5:%.*]] = load i8, i8* [[ARRAYIDX3_2]], align 1
24+
; CHECK-NEXT: [[XOR12_2:%.*]] = xor i8 [[TMP5]], [[TMP4]]
25+
; CHECK-NEXT: [[OR13_2:%.*]] = or i8 [[XOR12_2]], [[OR13_1]]
26+
; CHECK-NEXT: [[ARRAYIDX_3:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 3
27+
; CHECK-NEXT: [[TMP6:%.*]] = load i8, i8* [[ARRAYIDX_3]], align 1
28+
; CHECK-NEXT: [[ARRAYIDX3_3:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 3
29+
; CHECK-NEXT: [[TMP7:%.*]] = load i8, i8* [[ARRAYIDX3_3]], align 1
30+
; CHECK-NEXT: [[XOR12_3:%.*]] = xor i8 [[TMP7]], [[TMP6]]
31+
; CHECK-NEXT: [[OR13_3:%.*]] = or i8 [[XOR12_3]], [[OR13_2]]
32+
; CHECK-NEXT: [[ARRAYIDX_4:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 4
33+
; CHECK-NEXT: [[TMP8:%.*]] = load i8, i8* [[ARRAYIDX_4]], align 1
34+
; CHECK-NEXT: [[ARRAYIDX3_4:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 4
35+
; CHECK-NEXT: [[TMP9:%.*]] = load i8, i8* [[ARRAYIDX3_4]], align 1
36+
; CHECK-NEXT: [[XOR12_4:%.*]] = xor i8 [[TMP9]], [[TMP8]]
37+
; CHECK-NEXT: [[OR13_4:%.*]] = or i8 [[XOR12_4]], [[OR13_3]]
38+
; CHECK-NEXT: [[ARRAYIDX_5:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 5
39+
; CHECK-NEXT: [[TMP10:%.*]] = load i8, i8* [[ARRAYIDX_5]], align 1
40+
; CHECK-NEXT: [[ARRAYIDX3_5:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 5
41+
; CHECK-NEXT: [[TMP11:%.*]] = load i8, i8* [[ARRAYIDX3_5]], align 1
42+
; CHECK-NEXT: [[XOR12_5:%.*]] = xor i8 [[TMP11]], [[TMP10]]
43+
; CHECK-NEXT: [[OR13_5:%.*]] = or i8 [[XOR12_5]], [[OR13_4]]
44+
; CHECK-NEXT: [[ARRAYIDX_6:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 6
45+
; CHECK-NEXT: [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX_6]], align 1
46+
; CHECK-NEXT: [[ARRAYIDX3_6:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 6
47+
; CHECK-NEXT: [[TMP13:%.*]] = load i8, i8* [[ARRAYIDX3_6]], align 1
48+
; CHECK-NEXT: [[XOR12_6:%.*]] = xor i8 [[TMP13]], [[TMP12]]
49+
; CHECK-NEXT: [[OR13_6:%.*]] = or i8 [[XOR12_6]], [[OR13_5]]
50+
; CHECK-NEXT: [[ARRAYIDX_7:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[A]], i64 0, i32 0, i64 7
51+
; CHECK-NEXT: [[TMP14:%.*]] = load i8, i8* [[ARRAYIDX_7]], align 1
52+
; CHECK-NEXT: [[ARRAYIDX3_7:%.*]] = getelementptr inbounds [[STRUCT_BUF]], %struct.buf* [[B]], i64 0, i32 0, i64 7
53+
; CHECK-NEXT: [[TMP15:%.*]] = load i8, i8* [[ARRAYIDX3_7]], align 1
54+
; CHECK-NEXT: [[XOR12_7:%.*]] = xor i8 [[TMP15]], [[TMP14]]
55+
; CHECK-NEXT: [[OR13_7:%.*]] = or i8 [[XOR12_7]], [[OR13_6]]
56+
; CHECK-NEXT: ret i8 [[OR13_7]]
57+
;
58+
entry:
59+
%arrayidx = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 0
60+
%0 = load i8, i8* %arrayidx, align 1
61+
%arrayidx3 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 0
62+
%1 = load i8, i8* %arrayidx3, align 1
63+
%xor12 = xor i8 %1, %0
64+
%arrayidx.1 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 1
65+
%2 = load i8, i8* %arrayidx.1, align 1
66+
%arrayidx3.1 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 1
67+
%3 = load i8, i8* %arrayidx3.1, align 1
68+
%xor12.1 = xor i8 %3, %2
69+
%or13.1 = or i8 %xor12.1, %xor12
70+
%arrayidx.2 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 2
71+
%4 = load i8, i8* %arrayidx.2, align 1
72+
%arrayidx3.2 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 2
73+
%5 = load i8, i8* %arrayidx3.2, align 1
74+
%xor12.2 = xor i8 %5, %4
75+
%or13.2 = or i8 %xor12.2, %or13.1
76+
%arrayidx.3 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 3
77+
%6 = load i8, i8* %arrayidx.3, align 1
78+
%arrayidx3.3 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 3
79+
%7 = load i8, i8* %arrayidx3.3, align 1
80+
%xor12.3 = xor i8 %7, %6
81+
%or13.3 = or i8 %xor12.3, %or13.2
82+
%arrayidx.4 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 4
83+
%8 = load i8, i8* %arrayidx.4, align 1
84+
%arrayidx3.4 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 4
85+
%9 = load i8, i8* %arrayidx3.4, align 1
86+
%xor12.4 = xor i8 %9, %8
87+
%or13.4 = or i8 %xor12.4, %or13.3
88+
%arrayidx.5 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 5
89+
%10 = load i8, i8* %arrayidx.5, align 1
90+
%arrayidx3.5 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 5
91+
%11 = load i8, i8* %arrayidx3.5, align 1
92+
%xor12.5 = xor i8 %11, %10
93+
%or13.5 = or i8 %xor12.5, %or13.4
94+
%arrayidx.6 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 6
95+
%12 = load i8, i8* %arrayidx.6, align 1
96+
%arrayidx3.6 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 6
97+
%13 = load i8, i8* %arrayidx3.6, align 1
98+
%xor12.6 = xor i8 %13, %12
99+
%or13.6 = or i8 %xor12.6, %or13.5
100+
%arrayidx.7 = getelementptr inbounds %struct.buf, %struct.buf* %a, i64 0, i32 0, i64 7
101+
%14 = load i8, i8* %arrayidx.7, align 1
102+
%arrayidx3.7 = getelementptr inbounds %struct.buf, %struct.buf* %b, i64 0, i32 0, i64 7
103+
%15 = load i8, i8* %arrayidx3.7, align 1
104+
%xor12.7 = xor i8 %15, %14
105+
%or13.7 = or i8 %xor12.7, %or13.6
106+
ret i8 %or13.7
107+
}

0 commit comments

Comments
 (0)