2 ; RUN: opt < %s -passes=loop-vectorize -debug -disable-output -force-ordered-reductions=true -hints-allow-reordering=false \
3 ; RUN: -force-vector-width=4 -force-vector-interleave=1 -S 2>&1 | FileCheck %s --check-prefix=CHECK-VF4
4 ; RUN: opt < %s -passes=loop-vectorize -debug -disable-output -force-ordered-reductions=true -hints-allow-reordering=false \
5 ; RUN: -force-vector-width=8 -force-vector-interleave=1 -S 2>&1 | FileCheck %s --check-prefix=CHECK-VF8
7 target triple="aarch64-unknown-linux-gnu"
9 ; CHECK-VF4: Found an estimated cost of 14 for VF 4 For instruction: %add = fadd float %0, %sum.07
10 ; CHECK-VF8: Found an estimated cost of 28 for VF 8 For instruction: %add = fadd float %0, %sum.07
12 define float @fadd_strict32(ptr noalias nocapture readonly %a, i64 %n) {
17 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
18 %sum.07 = phi float [ 0.000000e+00, %entry ], [ %add, %for.body ]
19 %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
20 %0 = load float, ptr %arrayidx, align 4
21 %add = fadd float %0, %sum.07
22 %iv.next = add nuw nsw i64 %iv, 1
23 %exitcond.not = icmp eq i64 %iv.next, %n
24 br i1 %exitcond.not, label %for.end, label %for.body
31 ; CHECK-VF4: Found an estimated cost of 12 for VF 4 For instruction: %add = fadd double %0, %sum.07
32 ; CHECK-VF8: Found an estimated cost of 24 for VF 8 For instruction: %add = fadd double %0, %sum.07
34 define double @fadd_strict64(ptr noalias nocapture readonly %a, i64 %n) {
39 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
40 %sum.07 = phi double [ 0.000000e+00, %entry ], [ %add, %for.body ]
41 %arrayidx = getelementptr inbounds double, ptr %a, i64 %iv
42 %0 = load double, ptr %arrayidx, align 4
43 %add = fadd double %0, %sum.07
44 %iv.next = add nuw nsw i64 %iv, 1
45 %exitcond.not = icmp eq i64 %iv.next, %n
46 br i1 %exitcond.not, label %for.end, label %for.body
52 ; CHECK-VF4: Found an estimated cost of 16 for VF 4 For instruction: %muladd = tail call float @llvm.fmuladd.f32(float %0, float %1, float %sum.07)
53 ; CHECK-VF8: Found an estimated cost of 32 for VF 8 For instruction: %muladd = tail call float @llvm.fmuladd.f32(float %0, float %1, float %sum.07)
55 define float @fmuladd_strict32(ptr %a, ptr %b, i64 %n) {
60 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
61 %sum.07 = phi float [ 0.000000e+00, %entry ], [ %muladd, %for.body ]
62 %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
63 %0 = load float, ptr %arrayidx, align 4
64 %arrayidx2 = getelementptr inbounds float, ptr %b, i64 %iv
65 %1 = load float, ptr %arrayidx2, align 4
66 %muladd = tail call float @llvm.fmuladd.f32(float %0, float %1, float %sum.07)
67 %iv.next = add nuw nsw i64 %iv, 1
68 %exitcond.not = icmp eq i64 %iv.next, %n
69 br i1 %exitcond.not, label %for.end, label %for.body
75 declare float @llvm.fmuladd.f32(float, float, float)
77 ; CHECK-VF4: Found an estimated cost of 16 for VF 4 For instruction: %muladd = tail call double @llvm.fmuladd.f64(double %0, double %1, double %sum.07)
78 ; CHECK-VF8: Found an estimated cost of 32 for VF 8 For instruction: %muladd = tail call double @llvm.fmuladd.f64(double %0, double %1, double %sum.07)
80 define double @fmuladd_strict64(ptr %a, ptr %b, i64 %n) {
85 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
86 %sum.07 = phi double [ 0.000000e+00, %entry ], [ %muladd, %for.body ]
87 %arrayidx = getelementptr inbounds double, ptr %a, i64 %iv
88 %0 = load double, ptr %arrayidx, align 4
89 %arrayidx2 = getelementptr inbounds double, ptr %b, i64 %iv
90 %1 = load double, ptr %arrayidx2, align 4
91 %muladd = tail call double @llvm.fmuladd.f64(double %0, double %1, double %sum.07)
92 %iv.next = add nuw nsw i64 %iv, 1
93 %exitcond.not = icmp eq i64 %iv.next, %n
94 br i1 %exitcond.not, label %for.end, label %for.body
100 declare double @llvm.fmuladd.f64(double, double, double)