Skip to content

Commit b3f4535

Browse files
alexey-bataevalexey-bataev
committed
[SLP][NFC]Add a test for bottom to top reordering.
1 parent d785168 commit b3f4535

File tree

1 file changed

+98
-0
lines changed

1 file changed

+98
-0
lines changed

‎llvm/test/Transforms/SLPVectorizer/X86/bottom-to-top-reorder.ll

Lines changed: 98 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,98 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2+
; RUN: opt -slp-vectorizer -S -mtriple=x86_64-unknown-linux-gnu -mattr=+avx < %s | FileCheck %s
3+
4+
define void @test(i32* %0, i32* %1, i32* %2) {
5+
; CHECK-LABEL: @test(
6+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP0:%.*]], i64 4
7+
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP1:%.*]], i64 1
8+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 1
9+
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 5
10+
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 2
11+
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
12+
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 6
13+
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 3
14+
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 3
15+
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 7
16+
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP2:%.*]], i64 2
17+
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 1
18+
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 3
19+
; CHECK-NEXT: [[TMP17:%.*]] = bitcast i32* [[TMP1]] to <4 x i32>*
20+
; CHECK-NEXT: [[TMP18:%.*]] = load <4 x i32>, <4 x i32>* [[TMP17]], align 4
21+
; CHECK-NEXT: [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[TMP18]], <4 x i32> poison, <4 x i32> <i32 2, i32 0, i32 1, i32 3>
22+
; CHECK-NEXT: [[TMP19:%.*]] = bitcast i32* [[TMP0]] to <4 x i32>*
23+
; CHECK-NEXT: [[TMP20:%.*]] = load <4 x i32>, <4 x i32>* [[TMP19]], align 4
24+
; CHECK-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP4]] to <4 x i32>*
25+
; CHECK-NEXT: [[TMP22:%.*]] = load <4 x i32>, <4 x i32>* [[TMP21]], align 4
26+
; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <4 x i32> [[TMP22]], <4 x i32> poison, <4 x i32> <i32 2, i32 0, i32 1, i32 3>
27+
; CHECK-NEXT: [[TMP23:%.*]] = sub <4 x i32> <i32 0, i32 0, i32 undef, i32 0>, [[TMP20]]
28+
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP23]], <4 x i32> poison, <4 x i32> <i32 2, i32 0, i32 1, i32 3>
29+
; CHECK-NEXT: [[TMP24:%.*]] = sub <4 x i32> [[SHUFFLE]], [[SHUFFLE1]]
30+
; CHECK-NEXT: [[TMP25:%.*]] = add <4 x i32> [[TMP24]], [[SHUFFLE2]]
31+
; CHECK-NEXT: [[TMP26:%.*]] = add <4 x i32> [[TMP25]], <i32 1, i32 0, i32 0, i32 0>
32+
; CHECK-NEXT: [[TMP27:%.*]] = sub <4 x i32> [[TMP25]], <i32 1, i32 0, i32 0, i32 0>
33+
; CHECK-NEXT: [[TMP28:%.*]] = shufflevector <4 x i32> [[TMP26]], <4 x i32> [[TMP27]], <4 x i32> <i32 0, i32 1, i32 2, i32 7>
34+
; CHECK-NEXT: [[TMP29:%.*]] = add <4 x i32> [[TMP28]], zeroinitializer
35+
; CHECK-NEXT: [[TMP30:%.*]] = sub <4 x i32> [[TMP28]], zeroinitializer
36+
; CHECK-NEXT: [[TMP31:%.*]] = shufflevector <4 x i32> [[TMP29]], <4 x i32> [[TMP30]], <4 x i32> <i32 0, i32 5, i32 6, i32 7>
37+
; CHECK-NEXT: [[TMP32:%.*]] = add <4 x i32> [[TMP31]], zeroinitializer
38+
; CHECK-NEXT: [[TMP33:%.*]] = sub <4 x i32> [[TMP31]], zeroinitializer
39+
; CHECK-NEXT: [[TMP34:%.*]] = shufflevector <4 x i32> [[TMP32]], <4 x i32> [[TMP33]], <4 x i32> <i32 0, i32 5, i32 6, i32 3>
40+
; CHECK-NEXT: [[TMP35:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
41+
; CHECK-NEXT: store <4 x i32> [[TMP34]], <4 x i32>* [[TMP35]], align 4
42+
; CHECK-NEXT: ret void
43+
;
44+
%4 = load i32, i32* %1, align 4
45+
%5 = load i32, i32* %0, align 4
46+
%6 = getelementptr inbounds i32, i32* %0, i64 4
47+
%7 = load i32, i32* %6, align 4
48+
%8 = getelementptr inbounds i32, i32* %1, i64 1
49+
%9 = load i32, i32* %8, align 4
50+
%10 = getelementptr inbounds i32, i32* %0, i64 1
51+
%11 = load i32, i32* %10, align 4
52+
%12 = getelementptr inbounds i32, i32* %0, i64 5
53+
%13 = load i32, i32* %12, align 4
54+
%14 = getelementptr inbounds i32, i32* %1, i64 2
55+
%15 = load i32, i32* %14, align 4
56+
%16 = getelementptr inbounds i32, i32* %0, i64 2
57+
%17 = load i32, i32* %16, align 4
58+
%18 = getelementptr inbounds i32, i32* %0, i64 6
59+
%19 = load i32, i32* %18, align 4
60+
%20 = getelementptr inbounds i32, i32* %1, i64 3
61+
%21 = load i32, i32* %20, align 4
62+
%22 = getelementptr inbounds i32, i32* %0, i64 3
63+
%23 = load i32, i32* %22, align 4
64+
%24 = getelementptr inbounds i32, i32* %0, i64 7
65+
%25 = load i32, i32* %24, align 4
66+
%26 = sub i32 0, %23
67+
%27 = sub i32 %26, %25
68+
%28 = add i32 %27, %21
69+
%29 = sub i32 undef, %17
70+
%30 = sub i32 %29, %19
71+
%31 = add i32 %30, %15
72+
%32 = sub i32 0, %11
73+
%33 = sub i32 %32, %13
74+
%34 = add i32 %33, %9
75+
%35 = sub i32 0, %5
76+
%36 = sub i32 %35, %7
77+
%37 = add i32 %36, %4
78+
%38 = add i32 %31, 1
79+
%39 = add i32 %38, 0
80+
%40 = add i32 %39, 0
81+
store i32 %40, i32* %2, align 4
82+
%41 = getelementptr inbounds i32, i32* %2, i64 2
83+
%42 = add i32 0, %34
84+
%43 = sub i32 %42, 0
85+
%44 = sub i32 %43, 0
86+
store i32 %44, i32* %41, align 4
87+
%45 = getelementptr inbounds i32, i32* %2, i64 1
88+
%46 = add i32 %37, 0
89+
%47 = sub i32 %46, 0
90+
%48 = sub i32 %47, 0
91+
store i32 %48, i32* %45, align 4
92+
%49 = getelementptr inbounds i32, i32* %2, i64 3
93+
%50 = sub i32 %28, 0
94+
%51 = sub i32 %50, 0
95+
%52 = add i32 %51, 0
96+
store i32 %52, i32* %49, align 4
97+
ret void
98+
}

0 commit comments

Comments
 (0)