1; RUN: llc -verify-machineinstrs < %s -enable-misched -mcpu=a2 -mtriple=powerpc64-unknown-linux-gnu | FileCheck %s 2; RUN: llc -verify-machineinstrs < %s -enable-misched -enable-aa-sched-mi -mcpu=a2 -mtriple=powerpc64-unknown-linux-gnu | FileCheck %s 3target 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-f128:128:128-v128:128:128-n32:64" 4 5@aa = external global [256 x [256 x double]], align 32 6@bb = external global [256 x [256 x double]], align 32 7@cc = external global [256 x [256 x double]], align 32 8@.str1 = external hidden unnamed_addr constant [6 x i8], align 1 9@X = external global [16000 x double], align 32 10@Y = external global [16000 x double], align 32 11@Z = external global [16000 x double], align 32 12@U = external global [16000 x double], align 32 13@V = external global [16000 x double], align 32 14@.str137 = external hidden unnamed_addr constant [14 x i8], align 1 15 16declare void @check(i32 signext) nounwind 17 18declare signext i32 @printf(i8* nocapture, ...) nounwind 19 20declare signext i32 @init(i8*) nounwind 21 22define signext i32 @s000() nounwind { 23entry: 24 %call = tail call signext i32 @init(i8* getelementptr inbounds ([6 x i8], [6 x i8]* @.str1, i64 0, i64 0)) 25 %call1 = tail call i64 @clock() nounwind 26 br label %for.cond2.preheader 27 28; CHECK: @s000 29 30for.cond2.preheader: ; preds = %for.end, %entry 31 %nl.018 = phi i32 [ 0, %entry ], [ %inc9, %for.end ] 32 br label %for.body4 33 34for.body4: ; preds = %for.body4, %for.cond2.preheader 35 %indvars.iv = phi i64 [ 0, %for.cond2.preheader ], [ %indvars.iv.next.15, %for.body4 ] 36 %arrayidx = getelementptr inbounds [16000 x double], [16000 x double]* @Y, i64 0, i64 %indvars.iv 37 %arrayidx6 = getelementptr inbounds [16000 x double], [16000 x double]* @X, i64 0, i64 %indvars.iv 38 %0 = bitcast double* %arrayidx to <1 x double>* 39 %1 = load <1 x double>, <1 x double>* %0, align 32 40 %add = fadd <1 x double> %1, <double 1.000000e+00> 41 %2 = bitcast double* %arrayidx6 to <1 x double>* 42 store <1 x double> %add, <1 x double>* %2, align 32 43 %indvars.iv.next.322 = or i64 %indvars.iv, 4 44 %arrayidx.4 = getelementptr inbounds [16000 x double], [16000 x double]* @Y, i64 0, i64 %indvars.iv.next.322 45 %arrayidx6.4 = getelementptr inbounds [16000 x double], [16000 x double]* @X, i64 0, i64 %indvars.iv.next.322 46 %3 = bitcast double* %arrayidx.4 to <1 x double>* 47 %4 = load <1 x double>, <1 x double>* %3, align 32 48 %add.4 = fadd <1 x double> %4, <double 1.000000e+00> 49 %5 = bitcast double* %arrayidx6.4 to <1 x double>* 50 store <1 x double> %add.4, <1 x double>* %5, align 32 51 %indvars.iv.next.726 = or i64 %indvars.iv, 8 52 %arrayidx.8 = getelementptr inbounds [16000 x double], [16000 x double]* @Y, i64 0, i64 %indvars.iv.next.726 53 %arrayidx6.8 = getelementptr inbounds [16000 x double], [16000 x double]* @X, i64 0, i64 %indvars.iv.next.726 54 %6 = bitcast double* %arrayidx.8 to <1 x double>* 55 %7 = load <1 x double>, <1 x double>* %6, align 32 56 %add.8 = fadd <1 x double> %7, <double 1.000000e+00> 57 %8 = bitcast double* %arrayidx6.8 to <1 x double>* 58 store <1 x double> %add.8, <1 x double>* %8, align 32 59 %indvars.iv.next.1130 = or i64 %indvars.iv, 12 60 %arrayidx.12 = getelementptr inbounds [16000 x double], [16000 x double]* @Y, i64 0, i64 %indvars.iv.next.1130 61 %arrayidx6.12 = getelementptr inbounds [16000 x double], [16000 x double]* @X, i64 0, i64 %indvars.iv.next.1130 62 %9 = bitcast double* %arrayidx.12 to <1 x double>* 63 %10 = load <1 x double>, <1 x double>* %9, align 32 64 %add.12 = fadd <1 x double> %10, <double 1.000000e+00> 65 %11 = bitcast double* %arrayidx6.12 to <1 x double>* 66 store <1 x double> %add.12, <1 x double>* %11, align 32 67 %indvars.iv.next.15 = add i64 %indvars.iv, 16 68 %lftr.wideiv.15 = trunc i64 %indvars.iv.next.15 to i32 69 %exitcond.15 = icmp eq i32 %lftr.wideiv.15, 16000 70 br i1 %exitcond.15, label %for.end, label %for.body4 71 72; All of the loads should come before all of the stores. 73; CHECK: mtctr 74; CHECK: stfd 75; CHECK-NOT: lfd 76; CHECK: bdnz 77 78for.end: ; preds = %for.body4 79 %call7 = tail call signext i32 @dummy(double* getelementptr inbounds ([16000 x double], [16000 x double]* @X, i64 0, i64 0), double* getelementptr inbounds ([16000 x double], [16000 x double]* @Y, i64 0, i64 0), double* getelementptr inbounds ([16000 x double], [16000 x double]* @Z, i64 0, i64 0), double* getelementptr inbounds ([16000 x double], [16000 x double]* @U, i64 0, i64 0), double* getelementptr inbounds ([16000 x double], [16000 x double]* @V, i64 0, i64 0), [256 x double]* getelementptr inbounds ([256 x [256 x double]], [256 x [256 x double]]* @aa, i64 0, i64 0), [256 x double]* getelementptr inbounds ([256 x [256 x double]], [256 x [256 x double]]* @bb, i64 0, i64 0), [256 x double]* getelementptr inbounds ([256 x [256 x double]], [256 x [256 x double]]* @cc, i64 0, i64 0), double 0.000000e+00) nounwind 80 %inc9 = add nsw i32 %nl.018, 1 81 %exitcond = icmp eq i32 %inc9, 400000 82 br i1 %exitcond, label %for.end10, label %for.cond2.preheader 83 84for.end10: ; preds = %for.end 85 %call11 = tail call i64 @clock() nounwind 86 %sub = sub nsw i64 %call11, %call1 87 %conv = sitofp i64 %sub to double 88 %div = fdiv double %conv, 1.000000e+06 89 %call12 = tail call signext i32 (i8*, ...) @printf(i8* getelementptr inbounds ([14 x i8], [14 x i8]* @.str137, i64 0, i64 0), double %div) nounwind 90 tail call void @check(i32 signext 1) 91 ret i32 0 92} 93 94declare i64 @clock() nounwind 95 96declare signext i32 @dummy(double*, double*, double*, double*, double*, [256 x double]*, [256 x double]*, [256 x double]*, double) 97