101 lines · plain
1; REQUIRES: asserts2; 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-VF44; 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-VF86 7target triple="aarch64-unknown-linux-gnu"8 9; CHECK-VF4: Found an estimated cost of 14 for VF 4 For instruction: %add = fadd float %0, %sum.0710; CHECK-VF8: Found an estimated cost of 28 for VF 8 For instruction: %add = fadd float %0, %sum.0711 12define float @fadd_strict32(ptr noalias nocapture readonly %a, i64 %n) {13entry:14 br label %for.body15 16for.body: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 %iv20 %0 = load float, ptr %arrayidx, align 421 %add = fadd float %0, %sum.0722 %iv.next = add nuw nsw i64 %iv, 123 %exitcond.not = icmp eq i64 %iv.next, %n24 br i1 %exitcond.not, label %for.end, label %for.body25 26for.end:27 ret float %add28}29 30 31; CHECK-VF4: Found an estimated cost of 12 for VF 4 For instruction: %add = fadd double %0, %sum.0732; CHECK-VF8: Found an estimated cost of 24 for VF 8 For instruction: %add = fadd double %0, %sum.0733 34define double @fadd_strict64(ptr noalias nocapture readonly %a, i64 %n) {35entry:36 br label %for.body37 38for.body: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 %iv42 %0 = load double, ptr %arrayidx, align 443 %add = fadd double %0, %sum.0744 %iv.next = add nuw nsw i64 %iv, 145 %exitcond.not = icmp eq i64 %iv.next, %n46 br i1 %exitcond.not, label %for.end, label %for.body47 48for.end:49 ret double %add50}51 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)54 55define float @fmuladd_strict32(ptr %a, ptr %b, i64 %n) {56entry:57 br label %for.body58 59for.body: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 %iv63 %0 = load float, ptr %arrayidx, align 464 %arrayidx2 = getelementptr inbounds float, ptr %b, i64 %iv65 %1 = load float, ptr %arrayidx2, align 466 %muladd = tail call float @llvm.fmuladd.f32(float %0, float %1, float %sum.07)67 %iv.next = add nuw nsw i64 %iv, 168 %exitcond.not = icmp eq i64 %iv.next, %n69 br i1 %exitcond.not, label %for.end, label %for.body70 71for.end:72 ret float %muladd73}74 75declare float @llvm.fmuladd.f32(float, float, float)76 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)79 80define double @fmuladd_strict64(ptr %a, ptr %b, i64 %n) {81entry:82 br label %for.body83 84for.body: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 %iv88 %0 = load double, ptr %arrayidx, align 489 %arrayidx2 = getelementptr inbounds double, ptr %b, i64 %iv90 %1 = load double, ptr %arrayidx2, align 491 %muladd = tail call double @llvm.fmuladd.f64(double %0, double %1, double %sum.07)92 %iv.next = add nuw nsw i64 %iv, 193 %exitcond.not = icmp eq i64 %iv.next, %n94 br i1 %exitcond.not, label %for.end, label %for.body95 96for.end:97 ret double %muladd98}99 100declare double @llvm.fmuladd.f64(double, double, double)101