128 lines · plain
1; RUN: opt < %s -passes=loop-interchange -cache-line-size=64 -pass-remarks-missed='loop-interchange' -pass-remarks-output=%t -S \2; RUN: -verify-dom-info -verify-loop-info -verify-loop-lcssa 2>&1 | FileCheck -check-prefix=IR %s3; RUN: FileCheck --input-file=%t %s4 5; Both tests should be rejected as interchange candidates. For now, they are6; rejected for dependence analysis reasons, but that's because support for 'S'7; scalar dependencies was removed. When that is properly, the inner loop only8; reductions should still not be supported currently, see discussion at D530279; for more information on the required checks.10 11@A = common global [500 x [500 x i32]] zeroinitializer12@X = common global i32 013@B = common global [500 x [500 x i32]] zeroinitializer14@Y = common global i32 015 16;; global X17 18;; for( int i=1;i<N;i++)19;; for( int j=1;j<N;j++)20;; X+=A[j][i];21 22; CHECK: --- !Missed23; CHECK-NEXT: Pass: loop-interchange24; CHECK-NEXT: Name: Dependence25; CHECK-NEXT: Function: reduction_0126 27; IR-LABEL: @reduction_01(28; IR-NOT: split29 30define void @reduction_01(i32 %N) {31entry:32 %cmp16 = icmp sgt i32 %N, 133 br i1 %cmp16, label %for.body3.lr.ph, label %for.end834 35for.body3.lr.ph: ; preds = %for.cond1.for.inc6_crit_edge, %entry36 %indvars.iv18 = phi i64 [ %indvars.iv.next19, %for.cond1.for.inc6_crit_edge ], [ 1, %entry ]37 %X.promoted = load i32, ptr @X38 br label %for.body339 40for.body3: ; preds = %for.body3, %for.body3.lr.ph41 %indvars.iv = phi i64 [ 1, %for.body3.lr.ph ], [ %indvars.iv.next, %for.body3 ]42 %add15 = phi i32 [ %X.promoted, %for.body3.lr.ph ], [ %add, %for.body3 ]43 %arrayidx5 = getelementptr inbounds [500 x [500 x i32]], ptr @A, i64 0, i64 %indvars.iv, i64 %indvars.iv1844 %0 = load i32, ptr %arrayidx545 %add = add nsw i32 %add15, %046 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 147 %lftr.wideiv = trunc i64 %indvars.iv.next to i3248 %exitcond = icmp eq i32 %lftr.wideiv, %N49 br i1 %exitcond, label %for.cond1.for.inc6_crit_edge, label %for.body350 51for.cond1.for.inc6_crit_edge: ; preds = %for.body352 %add.lcssa = phi i32 [ %add, %for.body3 ]53 store i32 %add.lcssa, ptr @X54 %indvars.iv.next19 = add nuw nsw i64 %indvars.iv18, 155 %lftr.wideiv20 = trunc i64 %indvars.iv.next19 to i3256 %exitcond21 = icmp eq i32 %lftr.wideiv20, %N57 br i1 %exitcond21, label %for.end8, label %for.body3.lr.ph58 59for.end8: ; preds = %for.cond1.for.inc6_crit_edge, %entry60 ret void61}62 63;; Not tightly nested. Do not interchange.64;; for( int i=1;i<N;i++)65;; for( int j=1;j<N;j++) {66;; for( int k=1;k<N;k++) {67;; X+=A[k][j];68;; }69;; Y+=B[j][i];70;; }71 72;; Not tightly nested. Do not interchange.73;; Not interchanged hence the phi's in the inner loop will not be split.74 75; CHECK: --- !Missed76; CHECK-NEXT: Pass: loop-interchange77; CHECK-NEXT: Name: UnsupportedPHIOuter78; CHECK-NEXT: Function: reduction_0379 80; IR-LABEL: @reduction_03(81; IR-NOT: split82 83define void @reduction_03(i32 %N) {84entry:85 %cmp35 = icmp sgt i32 %N, 186 br i1 %cmp35, label %for.cond4.preheader.lr.ph, label %for.end1987 88for.cond4.preheader.lr.ph: ; preds = %for.cond1.for.inc17_crit_edge, %entry89 %indvars.iv41 = phi i64 [ %indvars.iv.next42, %for.cond1.for.inc17_crit_edge ], [ 1, %entry ]90 %Y.promoted = load i32, ptr @Y91 br label %for.body6.lr.ph92 93for.body6.lr.ph: ; preds = %for.cond4.for.end_crit_edge, %for.cond4.preheader.lr.ph94 %indvars.iv37 = phi i64 [ 1, %for.cond4.preheader.lr.ph ], [ %indvars.iv.next38, %for.cond4.for.end_crit_edge ]95 %add1334 = phi i32 [ %Y.promoted, %for.cond4.preheader.lr.ph ], [ %add13, %for.cond4.for.end_crit_edge ]96 %X.promoted = load i32, ptr @X97 br label %for.body698 99for.body6: ; preds = %for.body6, %for.body6.lr.ph100 %indvars.iv = phi i64 [ 1, %for.body6.lr.ph ], [ %indvars.iv.next, %for.body6 ]101 %arrayidx8 = getelementptr inbounds [500 x [500 x i32]], ptr @A, i64 0, i64 %indvars.iv, i64 %indvars.iv37102 %0 = load i32, ptr %arrayidx8103 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1104 %lftr.wideiv = trunc i64 %indvars.iv.next to i32105 %exitcond = icmp eq i32 %lftr.wideiv, %N106 br i1 %exitcond, label %for.cond4.for.end_crit_edge, label %for.body6107 108for.cond4.for.end_crit_edge: ; preds = %for.body6109 %arrayidx12 = getelementptr inbounds [500 x [500 x i32]], ptr @B, i64 0, i64 %indvars.iv37, i64 %indvars.iv41110 %1 = load i32, ptr %arrayidx12111 %add13 = add nsw i32 %add1334, %1112 %indvars.iv.next38 = add nuw nsw i64 %indvars.iv37, 1113 %lftr.wideiv39 = trunc i64 %indvars.iv.next38 to i32114 %exitcond40 = icmp eq i32 %lftr.wideiv39, %N115 br i1 %exitcond40, label %for.cond1.for.inc17_crit_edge, label %for.body6.lr.ph116 117for.cond1.for.inc17_crit_edge: ; preds = %for.cond4.for.end_crit_edge118 %add13.lcssa = phi i32 [ %add13, %for.cond4.for.end_crit_edge ]119 store i32 %add13.lcssa, ptr @Y120 %indvars.iv.next42 = add nuw nsw i64 %indvars.iv41, 1121 %lftr.wideiv43 = trunc i64 %indvars.iv.next42 to i32122 %exitcond44 = icmp eq i32 %lftr.wideiv43, %N123 br i1 %exitcond44, label %for.end19, label %for.cond4.preheader.lr.ph124 125for.end19: ; preds = %for.cond1.for.inc17_crit_edge, %entry126 ret void127}128