Skip to content

Commit 56c92bf

Browse files
anton-afanasyevanton-afanasyev
committed
[SLP][Test] Precommit tests for D83779. NFC.
1 parent fc247c8 commit 56c92bf

File tree

1 file changed

+120
-0
lines changed

1 file changed

+120
-0
lines changed

‎llvm/test/Transforms/SLPVectorizer/X86/pr44067.ll

Lines changed: 120 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,120 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2+
; RUN: opt -slp-vectorizer -S -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
3+
4+
; See https://reviews.llvm.org/D83779
5+
6+
define <2 x float> @foo({{float, float}}* %A) {
7+
; CHECK-LABEL: @foo(
8+
; CHECK-NEXT: entry:
9+
; CHECK-NEXT: [[TMP0:%.*]] = bitcast { { float, float } }* [[A:%.*]] to <2 x float>*
10+
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 8
11+
; CHECK-NEXT: [[REORDER_SHUFFLE:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> undef, <2 x i32> <i32 1, i32 0>
12+
; CHECK-NEXT: [[TMP2:%.*]] = fmul <2 x float> [[REORDER_SHUFFLE]], <float 2.000000e+00, float 2.000000e+00>
13+
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
14+
; CHECK-NEXT: [[INS1:%.*]] = insertelement <2 x float> undef, float [[TMP3]], i32 1
15+
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
16+
; CHECK-NEXT: [[INS0:%.*]] = insertelement <2 x float> [[INS1]], float [[TMP4]], i32 0
17+
; CHECK-NEXT: ret <2 x float> [[INS0]]
18+
;
19+
entry:
20+
%0 = bitcast {{float, float}}* %A to <2 x float>*
21+
%1 = load <2 x float>, <2 x float>* %0
22+
%L0 = extractelement <2 x float> %1, i32 0
23+
%L1 = extractelement <2 x float> %1, i32 1
24+
%Mul0 = fmul float %L0, 2.000000e+00
25+
%Mul1 = fmul float %L1, 2.000000e+00
26+
%Ins1 = insertelement <2 x float> undef, float %Mul1, i32 1
27+
%Ins0 = insertelement <2 x float> %Ins1, float %Mul0, i32 0
28+
ret <2 x float> %Ins0
29+
}
30+
31+
32+
%Struct1Ty = type { i16, i16 }
33+
%Struct2Ty = type { %Struct1Ty, %Struct1Ty}
34+
35+
define {%Struct2Ty, %Struct2Ty} @StructOfStructOfStruct(i16 *%Ptr) {
36+
; CHECK-LABEL: @StructOfStructOfStruct(
37+
; CHECK-NEXT: [[GEP0:%.*]] = getelementptr inbounds i16, i16* [[PTR:%.*]], i64 0
38+
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 1
39+
; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 2
40+
; CHECK-NEXT: [[GEP3:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 3
41+
; CHECK-NEXT: [[GEP4:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 4
42+
; CHECK-NEXT: [[GEP5:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 5
43+
; CHECK-NEXT: [[GEP6:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 6
44+
; CHECK-NEXT: [[GEP7:%.*]] = getelementptr inbounds i16, i16* [[PTR]], i64 7
45+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast i16* [[GEP0]] to <8 x i16>*
46+
; CHECK-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* [[TMP1]], align 2
47+
; CHECK-NEXT: [[REORDER_SHUFFLE:%.*]] = shufflevector <8 x i16> [[TMP2]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 7, i32 6, i32 1, i32 0, i32 2, i32 3>
48+
; CHECK-NEXT: [[TMP3:%.*]] = add <8 x i16> [[REORDER_SHUFFLE]], <i16 5, i16 6, i16 8, i16 7, i16 2, i16 1, i16 3, i16 4>
49+
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <8 x i16> [[TMP3]], i32 4
50+
; CHECK-NEXT: [[STRUCTIN0:%.*]] = insertvalue [[STRUCT1TY:%.*]] undef, i16 [[TMP4]], 1
51+
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <8 x i16> [[TMP3]], i32 5
52+
; CHECK-NEXT: [[STRUCTIN1:%.*]] = insertvalue [[STRUCT1TY]] %StructIn0, i16 [[TMP5]], 0
53+
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <8 x i16> [[TMP3]], i32 6
54+
; CHECK-NEXT: [[STRUCTIN2:%.*]] = insertvalue [[STRUCT1TY]] undef, i16 [[TMP6]], 0
55+
; CHECK-NEXT: [[TMP7:%.*]] = extractelement <8 x i16> [[TMP3]], i32 7
56+
; CHECK-NEXT: [[STRUCTIN3:%.*]] = insertvalue [[STRUCT1TY]] %StructIn2, i16 [[TMP7]], 1
57+
; CHECK-NEXT: [[TMP8:%.*]] = extractelement <8 x i16> [[TMP3]], i32 0
58+
; CHECK-NEXT: [[STRUCTIN4:%.*]] = insertvalue [[STRUCT1TY]] undef, i16 [[TMP8]], 0
59+
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <8 x i16> [[TMP3]], i32 1
60+
; CHECK-NEXT: [[STRUCTIN5:%.*]] = insertvalue [[STRUCT1TY]] %StructIn4, i16 [[TMP9]], 1
61+
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <8 x i16> [[TMP3]], i32 2
62+
; CHECK-NEXT: [[STRUCTIN6:%.*]] = insertvalue [[STRUCT1TY]] undef, i16 [[TMP10]], 1
63+
; CHECK-NEXT: [[TMP11:%.*]] = extractelement <8 x i16> [[TMP3]], i32 3
64+
; CHECK-NEXT: [[STRUCTIN7:%.*]] = insertvalue [[STRUCT1TY]] %StructIn6, i16 [[TMP11]], 0
65+
; CHECK-NEXT: [[STRUCT2IN0:%.*]] = insertvalue [[STRUCT2TY:%.*]] undef, [[STRUCT1TY]] %StructIn1, 0
66+
; CHECK-NEXT: [[STRUCT2IN1:%.*]] = insertvalue [[STRUCT2TY]] %Struct2In0, [[STRUCT1TY]] %StructIn3, 1
67+
; CHECK-NEXT: [[STRUCT2IN2:%.*]] = insertvalue [[STRUCT2TY]] undef, [[STRUCT1TY]] %StructIn5, 0
68+
; CHECK-NEXT: [[STRUCT2IN3:%.*]] = insertvalue [[STRUCT2TY]] %Struct2In2, [[STRUCT1TY]] %StructIn7, 1
69+
; CHECK-NEXT: [[RET0:%.*]] = insertvalue { [[STRUCT2TY]], [[STRUCT2TY]] } undef, [[STRUCT2TY]] %Struct2In3, 1
70+
; CHECK-NEXT: [[RET1:%.*]] = insertvalue { [[STRUCT2TY]], [[STRUCT2TY]] } [[RET0]], [[STRUCT2TY]] %Struct2In1, 0
71+
; CHECK-NEXT: ret { [[STRUCT2TY]], [[STRUCT2TY]] } [[RET1]]
72+
;
73+
%GEP0 = getelementptr inbounds i16, i16* %Ptr, i64 0
74+
%L0 = load i16, i16 * %GEP0
75+
%GEP1 = getelementptr inbounds i16, i16* %Ptr, i64 1
76+
%L1 = load i16, i16 * %GEP1
77+
%GEP2 = getelementptr inbounds i16, i16* %Ptr, i64 2
78+
%L2 = load i16, i16 * %GEP2
79+
%GEP3 = getelementptr inbounds i16, i16* %Ptr, i64 3
80+
%L3 = load i16, i16 * %GEP3
81+
%GEP4 = getelementptr inbounds i16, i16* %Ptr, i64 4
82+
%L4 = load i16, i16 * %GEP4
83+
%GEP5 = getelementptr inbounds i16, i16* %Ptr, i64 5
84+
%L5 = load i16, i16 * %GEP5
85+
%GEP6 = getelementptr inbounds i16, i16* %Ptr, i64 6
86+
%L6 = load i16, i16 * %GEP6
87+
%GEP7 = getelementptr inbounds i16, i16* %Ptr, i64 7
88+
%L7 = load i16, i16 * %GEP7
89+
90+
%Fadd0 = add i16 %L0, 1
91+
%Fadd1 = add i16 %L1, 2
92+
%Fadd2 = add i16 %L2, 3
93+
%Fadd3 = add i16 %L3, 4
94+
%Fadd4 = add i16 %L4, 5
95+
%Fadd5 = add i16 %L5, 6
96+
%Fadd6 = add i16 %L6, 7
97+
%Fadd7 = add i16 %L7, 8
98+
99+
%StructIn0 = insertvalue %Struct1Ty undef, i16 %Fadd1, 1
100+
%StructIn1 = insertvalue %Struct1Ty %StructIn0, i16 %Fadd0, 0
101+
102+
%StructIn2 = insertvalue %Struct1Ty undef, i16 %Fadd2, 0
103+
%StructIn3 = insertvalue %Struct1Ty %StructIn2, i16 %Fadd3, 1
104+
105+
%StructIn4 = insertvalue %Struct1Ty undef, i16 %Fadd4, 0
106+
%StructIn5 = insertvalue %Struct1Ty %StructIn4, i16 %Fadd5, 1
107+
108+
%StructIn6 = insertvalue %Struct1Ty undef, i16 %Fadd7, 1
109+
%StructIn7 = insertvalue %Struct1Ty %StructIn6, i16 %Fadd6, 0
110+
111+
%Struct2In0 = insertvalue %Struct2Ty undef, %Struct1Ty %StructIn1, 0
112+
%Struct2In1 = insertvalue %Struct2Ty %Struct2In0, %Struct1Ty %StructIn3, 1
113+
114+
%Struct2In2 = insertvalue %Struct2Ty undef, %Struct1Ty %StructIn5, 0
115+
%Struct2In3 = insertvalue %Struct2Ty %Struct2In2, %Struct1Ty %StructIn7, 1
116+
117+
%Ret0 = insertvalue {%Struct2Ty, %Struct2Ty} undef, %Struct2Ty %Struct2In3, 1
118+
%Ret1 = insertvalue {%Struct2Ty, %Struct2Ty} %Ret0, %Struct2Ty %Struct2In1, 0
119+
ret {%Struct2Ty, %Struct2Ty} %Ret1
120+
}

0 commit comments

Comments
 (0)