1 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
2 ; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -bb-vectorize-ignore-target-info -instcombine -gvn -S | FileCheck %s
3 ; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -bb-vectorize-aligned-only -bb-vectorize-ignore-target-info -instcombine -gvn -S | FileCheck %s -check-prefix=CHECK-AO
5 ; Simple 3-pair chain with loads and stores
6 define void @test1(double* %a, double* %b, double* %c) nounwind uwtable readonly {
8 %i0 = load double* %a, align 8
9 %i1 = load double* %b, align 8
10 %mul = fmul double %i0, %i1
11 %arrayidx3 = getelementptr inbounds double* %a, i64 1
12 %i3 = load double* %arrayidx3, align 8
13 %arrayidx4 = getelementptr inbounds double* %b, i64 1
14 %i4 = load double* %arrayidx4, align 8
15 %mul5 = fmul double %i3, %i4
16 store double %mul, double* %c, align 8
17 %arrayidx5 = getelementptr inbounds double* %c, i64 1
18 store double %mul5, double* %arrayidx5, align 8
20 ; CHECK-LABEL: @test1(
21 ; CHECK: %i0.v.i0 = bitcast double* %a to <2 x double>*
22 ; CHECK: %i1.v.i0 = bitcast double* %b to <2 x double>*
23 ; CHECK: %i0 = load <2 x double>* %i0.v.i0, align 8
24 ; CHECK: %i1 = load <2 x double>* %i1.v.i0, align 8
25 ; CHECK: %mul = fmul <2 x double> %i0, %i1
26 ; CHECK: %0 = bitcast double* %c to <2 x double>*
27 ; CHECK: store <2 x double> %mul, <2 x double>* %0, align 8
29 ; CHECK-AO-LABEL: @test1(
30 ; CHECK-AO-NOT: <2 x double>
33 ; Simple chain with extending loads and stores
34 define void @test2(float* %a, float* %b, double* %c) nounwind uwtable readonly {
36 %i0f = load float* %a, align 4
37 %i0 = fpext float %i0f to double
38 %i1f = load float* %b, align 4
39 %i1 = fpext float %i1f to double
40 %mul = fmul double %i0, %i1
41 %arrayidx3 = getelementptr inbounds float* %a, i64 1
42 %i3f = load float* %arrayidx3, align 4
43 %i3 = fpext float %i3f to double
44 %arrayidx4 = getelementptr inbounds float* %b, i64 1
45 %i4f = load float* %arrayidx4, align 4
46 %i4 = fpext float %i4f to double
47 %mul5 = fmul double %i3, %i4
48 store double %mul, double* %c, align 8
49 %arrayidx5 = getelementptr inbounds double* %c, i64 1
50 store double %mul5, double* %arrayidx5, align 8
52 ; CHECK-LABEL: @test2(
53 ; CHECK: %i0f.v.i0 = bitcast float* %a to <2 x float>*
54 ; CHECK: %i1f.v.i0 = bitcast float* %b to <2 x float>*
55 ; CHECK: %i0f = load <2 x float>* %i0f.v.i0, align 4
56 ; CHECK: %i0 = fpext <2 x float> %i0f to <2 x double>
57 ; CHECK: %i1f = load <2 x float>* %i1f.v.i0, align 4
58 ; CHECK: %i1 = fpext <2 x float> %i1f to <2 x double>
59 ; CHECK: %mul = fmul <2 x double> %i0, %i1
60 ; CHECK: %0 = bitcast double* %c to <2 x double>*
61 ; CHECK: store <2 x double> %mul, <2 x double>* %0, align 8
63 ; CHECK-AO-LABEL: @test2(
64 ; CHECK-AO-NOT: <2 x double>
67 ; Simple chain with loads and truncating stores
68 define void @test3(double* %a, double* %b, float* %c) nounwind uwtable readonly {
70 %i0 = load double* %a, align 8
71 %i1 = load double* %b, align 8
72 %mul = fmul double %i0, %i1
73 %mulf = fptrunc double %mul to float
74 %arrayidx3 = getelementptr inbounds double* %a, i64 1
75 %i3 = load double* %arrayidx3, align 8
76 %arrayidx4 = getelementptr inbounds double* %b, i64 1
77 %i4 = load double* %arrayidx4, align 8
78 %mul5 = fmul double %i3, %i4
79 %mul5f = fptrunc double %mul5 to float
80 store float %mulf, float* %c, align 8
81 %arrayidx5 = getelementptr inbounds float* %c, i64 1
82 store float %mul5f, float* %arrayidx5, align 4
84 ; CHECK-LABEL: @test3(
85 ; CHECK: %i0.v.i0 = bitcast double* %a to <2 x double>*
86 ; CHECK: %i1.v.i0 = bitcast double* %b to <2 x double>*
87 ; CHECK: %i0 = load <2 x double>* %i0.v.i0, align 8
88 ; CHECK: %i1 = load <2 x double>* %i1.v.i0, align 8
89 ; CHECK: %mul = fmul <2 x double> %i0, %i1
90 ; CHECK: %mulf = fptrunc <2 x double> %mul to <2 x float>
91 ; CHECK: %0 = bitcast float* %c to <2 x float>*
92 ; CHECK: store <2 x float> %mulf, <2 x float>* %0, align 8
94 ; CHECK-AO-LABEL: @test3(
95 ; CHECK-AO: %i0 = load double* %a, align 8
96 ; CHECK-AO: %i1 = load double* %b, align 8
97 ; CHECK-AO: %arrayidx3 = getelementptr inbounds double* %a, i64 1
98 ; CHECK-AO: %i3 = load double* %arrayidx3, align 8
99 ; CHECK-AO: %arrayidx4 = getelementptr inbounds double* %b, i64 1
100 ; CHECK-AO: %i4 = load double* %arrayidx4, align 8
101 ; CHECK-AO: %mul.v.i1.1 = insertelement <2 x double> undef, double %i1, i32 0
102 ; CHECK-AO: %mul.v.i1.2 = insertelement <2 x double> %mul.v.i1.1, double %i4, i32 1
103 ; CHECK-AO: %mul.v.i0.1 = insertelement <2 x double> undef, double %i0, i32 0
104 ; CHECK-AO: %mul.v.i0.2 = insertelement <2 x double> %mul.v.i0.1, double %i3, i32 1
105 ; CHECK-AO: %mul = fmul <2 x double> %mul.v.i0.2, %mul.v.i1.2
106 ; CHECK-AO: %mulf = fptrunc <2 x double> %mul to <2 x float>
107 ; CHECK-AO: %0 = bitcast float* %c to <2 x float>*
108 ; CHECK-AO: store <2 x float> %mulf, <2 x float>* %0, align 8
112 ; Simple 3-pair chain with loads and stores (unreachable)
113 define void @test4(i1 %bool, double* %a, double* %b, double* %c) nounwind uwtable readonly {
115 br i1 %bool, label %if.then1, label %if.end
122 %i0 = load double* %a, align 8
123 %i1 = load double* %b, align 8
124 %mul = fmul double %i0, %i1
125 %arrayidx3 = getelementptr inbounds double* %a, i64 1
126 %i3 = load double* %arrayidx3, align 8
127 %arrayidx4 = getelementptr inbounds double* %b, i64 1
128 %i4 = load double* %arrayidx4, align 8
129 %mul5 = fmul double %i3, %i4
130 store double %mul, double* %c, align 8
131 %arrayidx5 = getelementptr inbounds double* %c, i64 1
132 store double %mul5, double* %arrayidx5, align 8
137 ; CHECK-LABEL: @test4(
138 ; CHECK-NOT: <2 x double>
139 ; CHECK-AO-LABEL: @test4(
140 ; CHECK-AO-NOT: <2 x double>
143 ; Simple 3-pair chain with loads and stores
144 define void @test5(double* %a, double* %b, double* %c) nounwind uwtable readonly {
146 %i0 = load double* %a, align 8
147 %i1 = load double* %b, align 8
148 %mul = fmul double %i0, %i1
149 %arrayidx3 = getelementptr inbounds double* %a, i64 1
150 %i3 = load double* %arrayidx3, align 8
151 %arrayidx4 = getelementptr inbounds double* %b, i64 1
152 %i4 = load double* %arrayidx4, align 8
153 %mul5 = fmul double %i3, %i4
154 %arrayidx5 = getelementptr inbounds double* %c, i64 1
155 store double %mul5, double* %arrayidx5, align 8
156 store double %mul, double* %c, align 4
158 ; CHECK-LABEL: @test5(
159 ; CHECK: %i0.v.i0 = bitcast double* %a to <2 x double>*
160 ; CHECK: %i1.v.i0 = bitcast double* %b to <2 x double>*
161 ; CHECK: %i0 = load <2 x double>* %i0.v.i0, align 8
162 ; CHECK: %i1 = load <2 x double>* %i1.v.i0, align 8
163 ; CHECK: %mul = fmul <2 x double> %i0, %i1
164 ; CHECK: %0 = bitcast double* %c to <2 x double>*
165 ; CHECK: store <2 x double> %mul, <2 x double>* %0, align 4
167 ; CHECK-AO-LABEL: @test5(
168 ; CHECK-AO-NOT: <2 x double>