1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-- -mattr=+avx  | FileCheck %s
3; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s
4
5define float @matching_scalar(<4 x float>* dereferenceable(16) %p) {
6; CHECK-LABEL: @matching_scalar(
7; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to float*
8; CHECK-NEXT:    [[R:%.*]] = load float, float* [[BC]], align 16
9; CHECK-NEXT:    ret float [[R]]
10;
11  %bc = bitcast <4 x float>* %p to float*
12  %r = load float, float* %bc, align 16
13  ret float %r
14}
15
16define i32 @nonmatching_scalar(<4 x float>* dereferenceable(16) %p) {
17; CHECK-LABEL: @nonmatching_scalar(
18; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to i32*
19; CHECK-NEXT:    [[R:%.*]] = load i32, i32* [[BC]], align 16
20; CHECK-NEXT:    ret i32 [[R]]
21;
22  %bc = bitcast <4 x float>* %p to i32*
23  %r = load i32, i32* %bc, align 16
24  ret i32 %r
25}
26
27define i64 @larger_scalar(<4 x float>* dereferenceable(16) %p) {
28; CHECK-LABEL: @larger_scalar(
29; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to i64*
30; CHECK-NEXT:    [[R:%.*]] = load i64, i64* [[BC]], align 16
31; CHECK-NEXT:    ret i64 [[R]]
32;
33  %bc = bitcast <4 x float>* %p to i64*
34  %r = load i64, i64* %bc, align 16
35  ret i64 %r
36}
37
38define i8 @smaller_scalar(<4 x float>* dereferenceable(16) %p) {
39; CHECK-LABEL: @smaller_scalar(
40; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to i8*
41; CHECK-NEXT:    [[R:%.*]] = load i8, i8* [[BC]], align 16
42; CHECK-NEXT:    ret i8 [[R]]
43;
44  %bc = bitcast <4 x float>* %p to i8*
45  %r = load i8, i8* %bc, align 16
46  ret i8 %r
47}
48
49define i8 @smaller_scalar_256bit_vec(<8 x float>* dereferenceable(32) %p) {
50; CHECK-LABEL: @smaller_scalar_256bit_vec(
51; CHECK-NEXT:    [[BC:%.*]] = bitcast <8 x float>* [[P:%.*]] to i8*
52; CHECK-NEXT:    [[R:%.*]] = load i8, i8* [[BC]], align 32
53; CHECK-NEXT:    ret i8 [[R]]
54;
55  %bc = bitcast <8 x float>* %p to i8*
56  %r = load i8, i8* %bc, align 32
57  ret i8 %r
58}
59
60define i8 @smaller_scalar_less_aligned(<4 x float>* dereferenceable(16) %p) {
61; CHECK-LABEL: @smaller_scalar_less_aligned(
62; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to i8*
63; CHECK-NEXT:    [[R:%.*]] = load i8, i8* [[BC]], align 4
64; CHECK-NEXT:    ret i8 [[R]]
65;
66  %bc = bitcast <4 x float>* %p to i8*
67  %r = load i8, i8* %bc, align 4
68  ret i8 %r
69}
70
71define float @matching_scalar_small_deref(<4 x float>* dereferenceable(15) %p) {
72; CHECK-LABEL: @matching_scalar_small_deref(
73; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to float*
74; CHECK-NEXT:    [[R:%.*]] = load float, float* [[BC]], align 16
75; CHECK-NEXT:    ret float [[R]]
76;
77  %bc = bitcast <4 x float>* %p to float*
78  %r = load float, float* %bc, align 16
79  ret float %r
80}
81
82define float @matching_scalar_volatile(<4 x float>* dereferenceable(16) %p) {
83; CHECK-LABEL: @matching_scalar_volatile(
84; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float>* [[P:%.*]] to float*
85; CHECK-NEXT:    [[R:%.*]] = load volatile float, float* [[BC]], align 16
86; CHECK-NEXT:    ret float [[R]]
87;
88  %bc = bitcast <4 x float>* %p to float*
89  %r = load volatile float, float* %bc, align 16
90  ret float %r
91}
92
93define float @nonvector(double* dereferenceable(16) %p) {
94; CHECK-LABEL: @nonvector(
95; CHECK-NEXT:    [[BC:%.*]] = bitcast double* [[P:%.*]] to float*
96; CHECK-NEXT:    [[R:%.*]] = load float, float* [[BC]], align 16
97; CHECK-NEXT:    ret float [[R]]
98;
99  %bc = bitcast double* %p to float*
100  %r = load float, float* %bc, align 16
101  ret float %r
102}
103