1; RUN: opt %loadNPMPolly -polly-stmt-granularity=bb -polly-optree-normalize-phi=true '-passes=print<polly-optree>' -disable-output < %s | FileCheck %s -match-full-lines 2 3target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 4 5define internal fastcc void @kernel_atax(ptr nocapture readonly %A, ptr nocapture readonly %x, ptr nocapture %y, ptr nocapture %tmp) unnamed_addr #0 { 6entry: 7 br label %entry.split 8 9entry.split: ; preds = %entry 10 call void @llvm.memset.p0.i64(ptr %y, i8 0, i64 16800, i32 8, i1 false) 11 br label %for.body3 12 13for.body3: ; preds = %for.inc40, %entry.split 14 %indvars.iv8 = phi i64 [ 0, %entry.split ], [ %indvars.iv.next9, %for.inc40 ] 15 %arrayidx5 = getelementptr inbounds double, ptr %tmp, i64 %indvars.iv8 16 store double 0.000000e+00, ptr %arrayidx5, align 8, !tbaa !6 17 br label %for.body8 18 19for.body8: ; preds = %for.body8, %for.body3 20 %0 = phi double [ 0.000000e+00, %for.body3 ], [ %add, %for.body8 ] 21 %indvars.iv = phi i64 [ 0, %for.body3 ], [ %indvars.iv.next, %for.body8 ] 22 %arrayidx14 = getelementptr inbounds [2100 x double], ptr %A, i64 %indvars.iv8, i64 %indvars.iv 23 %1 = load double, ptr %arrayidx14, align 8, !tbaa !6 24 %arrayidx16 = getelementptr inbounds double, ptr %x, i64 %indvars.iv 25 %2 = load double, ptr %arrayidx16, align 8, !tbaa !6 26 %mul = fmul double %1, %2 27 %add = fadd double %0, %mul 28 store double %add, ptr %arrayidx5, align 8, !tbaa !6 29 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 30 %exitcond = icmp eq i64 %indvars.iv.next, 2 31 br i1 %exitcond, label %for.end21, label %for.body8 32 33for.end21: ; preds = %for.body8 34 br label %for.body24 35 36for.body24: ; preds = %for.body24.for.body24_crit_edge, %for.end21 37 %3 = phi double [ %add, %for.end21 ], [ %.pre, %for.body24.for.body24_crit_edge ] 38 %indvars.iv5 = phi i64 [ 0, %for.end21 ], [ %indvars.iv.next6, %for.body24.for.body24_crit_edge ] 39 %arrayidx26 = getelementptr inbounds double, ptr %y, i64 %indvars.iv5 40 %4 = load double, ptr %arrayidx26, align 8, !tbaa !6 41 %arrayidx30 = getelementptr inbounds [2100 x double], ptr %A, i64 %indvars.iv8, i64 %indvars.iv5 42 %5 = load double, ptr %arrayidx30, align 8, !tbaa !6 43 %mul33 = fmul double %5, %3 44 %add34 = fadd double %4, %mul33 45 store double %add34, ptr %arrayidx26, align 8, !tbaa !6 46 %indvars.iv.next6 = add nuw nsw i64 %indvars.iv5, 1 47 %exitcond7 = icmp eq i64 %indvars.iv.next6, 2 48 br i1 %exitcond7, label %for.inc40, label %for.body24.for.body24_crit_edge 49 50for.body24.for.body24_crit_edge: ; preds = %for.body24 51 %.pre = load double, ptr %arrayidx5, align 8, !tbaa !6 52 br label %for.body24 53 54for.inc40: ; preds = %for.body24 55 %indvars.iv.next9 = add nuw nsw i64 %indvars.iv8, 1 56 %exitcond10 = icmp eq i64 %indvars.iv.next9, 2 57 br i1 %exitcond10, label %for.end42, label %for.body3 58 59for.end42: ; preds = %for.inc40 60 ret void 61} 62 63; Function Attrs: argmemonly nounwind 64declare void @llvm.memset.p0.i64(ptr nocapture writeonly, i8, i64, i32, i1) #1 65 66attributes #0 = { noinline norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" } 67attributes #1 = { argmemonly nounwind } 68 69!llvm.module.flags = !{!0} 70!llvm.ident = !{!1} 71 72!0 = !{i32 1, !"wchar_size", i32 4} 73!1 = !{!"clang version 6.0.0 (trunk 312565) (llvm/trunk 312564)"} 74!2 = !{!3, !3, i64 0} 75!3 = !{!"any pointer", !4, i64 0} 76!4 = !{!"omnipotent char", !5, i64 0} 77!5 = !{!"Simple C/C++ TBAA"} 78!6 = !{!7, !7, i64 0} 79!7 = !{!"double", !4, i64 0} 80 81 82; CHECK: Statistics { 83; CHECK: Operand trees forwarded: 2 84; CHECK: Statements with forwarded operand trees: 2 85; CHECK: } 86 87; CHECK-NEXT: After statements { 88; CHECK-NEXT: Stmt_for_body3 89; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 0] 90; CHECK-NEXT: { Stmt_for_body3[i0] -> MemRef_tmp[i0] }; 91; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 1] 92; CHECK-NEXT: { Stmt_for_body3[i0] -> MemRef1__phi[] }; 93; CHECK-NEXT: Instructions { 94; CHECK-NEXT: store double 0.000000e+00, ptr %arrayidx5, align 8, !tbaa !2 95; CHECK-NEXT: } 96; CHECK-NEXT: Stmt_for_body8 97; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 1] 98; CHECK-NEXT: { Stmt_for_body8[i0, i1] -> MemRef1__phi[] }; 99; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 1] 100; CHECK-NEXT: { Stmt_for_body8[i0, i1] -> MemRef1__phi[] }; 101; CHECK-NEXT: new: { Stmt_for_body8[i0, i1] -> MemRef_tmp[i0] }; 102; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0] 103; CHECK-NEXT: { Stmt_for_body8[i0, i1] -> MemRef_A[i0, i1] }; 104; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0] 105; CHECK-NEXT: { Stmt_for_body8[i0, i1] -> MemRef_x[i1] }; 106; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 0] 107; CHECK-NEXT: { Stmt_for_body8[i0, i1] -> MemRef_tmp[i0] }; 108; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 1] 109; CHECK-NEXT: { Stmt_for_body8[i0, i1] -> MemRef_add[] }; 110; CHECK-NEXT: Instructions { 111; CHECK-NEXT: %0 = phi double [ 0.000000e+00, %for.body3 ], [ %add, %for.body8 ] 112; CHECK-NEXT: %1 = load double, ptr %arrayidx14, align 8, !tbaa !2 113; CHECK-NEXT: %2 = load double, ptr %arrayidx16, align 8, !tbaa !2 114; CHECK-NEXT: %mul = fmul double %1, %2 115; CHECK-NEXT: %add = fadd double %0, %mul 116; CHECK-NEXT: store double %add, ptr %arrayidx5, align 8, !tbaa !2 117; CHECK-NEXT: %exitcond = icmp eq i64 %indvars.iv.next, 2 118; CHECK-NEXT: } 119; CHECK-NEXT: Stmt_for_end21 120; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 1] 121; CHECK-NEXT: { Stmt_for_end21[i0] -> MemRef_add[] }; 122; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 1] 123; CHECK-NEXT: { Stmt_for_end21[i0] -> MemRef5__phi[] }; 124; CHECK-NEXT: Instructions { 125; CHECK-NEXT: } 126; CHECK-NEXT: Stmt_for_body24 127; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 1] 128; CHECK-NEXT: { Stmt_for_body24[i0, i1] -> MemRef5__phi[] }; 129; CHECK-NEXT: new: { Stmt_for_body24[i0, i1] -> MemRef_tmp[i0] }; 130; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0] 131; CHECK-NEXT: { Stmt_for_body24[i0, i1] -> MemRef_y[i1] }; 132; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0] 133; CHECK-NEXT: { Stmt_for_body24[i0, i1] -> MemRef_A[i0, i1] }; 134; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 0] 135; CHECK-NEXT: { Stmt_for_body24[i0, i1] -> MemRef_y[i1] }; 136; CHECK-NEXT: Instructions { 137; CHECK-NEXT: %3 = phi double [ %add, %for.end21 ], [ %.pre, %for.body24.for.body24_crit_edge ] 138; CHECK-NEXT: %4 = load double, ptr %arrayidx26, align 8, !tbaa !2 139; CHECK-NEXT: %5 = load double, ptr %arrayidx30, align 8, !tbaa !2 140; CHECK-NEXT: %mul33 = fmul double %5, %3 141; CHECK-NEXT: %add34 = fadd double %4, %mul33 142; CHECK-NEXT: store double %add34, ptr %arrayidx26, align 8, !tbaa !2 143; CHECK-NEXT: %exitcond7 = icmp eq i64 %indvars.iv.next6, 2 144; CHECK-NEXT: } 145; CHECK-NEXT: Stmt_for_body24_for_body24_crit_edge 146; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 1] 147; CHECK-NEXT: { Stmt_for_body24_for_body24_crit_edge[i0, i1] -> MemRef5__phi[] }; 148; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0] 149; CHECK-NEXT: { Stmt_for_body24_for_body24_crit_edge[i0, i1] -> MemRef_tmp[i0] }; 150; CHECK-NEXT: Instructions { 151; CHECK-NEXT: %.pre = load double, ptr %arrayidx5, align 8, !tbaa !2 152; CHECK-NEXT: } 153; CHECK-NEXT: } 154