1; RUN: opt < %s -passes=loop-interchange -cache-line-size=64 -pass-remarks-missed='loop-interchange' \ 2; RUN: -pass-remarks-output=%t -verify-loop-info -verify-dom-info -S | FileCheck -check-prefix=IR %s 3; RUN: FileCheck --input-file=%t %s 4 5; RUN: opt < %s -passes=loop-interchange -cache-line-size=64 -pass-remarks-missed='loop-interchange' \ 6; RUN: -da-disable-delinearization-checks -pass-remarks-output=%t \ 7; RUN: -verify-loop-info -verify-dom-info -S | FileCheck -check-prefix=IR %s 8; RUN: FileCheck --check-prefix=DELIN --input-file=%t %s 9 10target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 11 12@A = common global [100 x [100 x i32]] zeroinitializer 13@B = common global [100 x [100 x [100 x i32]]] zeroinitializer 14@C = common global [100 x [100 x i64]] zeroinitializer 15 16;;--------------------------------------Test case 01------------------------------------ 17;; This loop can be interchanged with -da-disable-delinearization-checks, otherwise it cannot 18;; be interchanged due to dependence. 19;; for(int i=0;i<N-1;i++) 20;; for(int j=1;j<N-1;j++) 21;; A[j+1][i+1] = A[j+1][i+1] + k; 22 23; CHECK: Name: Dependence 24; CHECK-NEXT: Function: interchange_01 25 26; DELIN: Name: Interchanged 27; DELIN-NEXT: Function: interchange_01 28define void @interchange_01(i32 %k, i32 %N) { 29 entry: 30 %sub = add nsw i32 %N, -1 31 %cmp26 = icmp sgt i32 %N, 1 32 br i1 %cmp26, label %for.cond1.preheader.lr.ph, label %for.end17 33 34 for.cond1.preheader.lr.ph: 35 %cmp324 = icmp sgt i32 %sub, 1 36 %0 = add i32 %N, -2 37 %1 = sext i32 %sub to i64 38 br label %for.cond1.preheader 39 40 for.cond.loopexit: 41 %cmp = icmp slt i64 %indvars.iv.next29, %1 42 br i1 %cmp, label %for.cond1.preheader, label %for.end17 43 44 for.cond1.preheader: 45 %indvars.iv28 = phi i64 [ 0, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next29, %for.cond.loopexit ] 46 %indvars.iv.next29 = add nuw nsw i64 %indvars.iv28, 1 47 br i1 %cmp324, label %for.body4, label %for.cond.loopexit 48 49 for.body4: 50 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body4 ], [ 1, %for.cond1.preheader ] 51 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 52 %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], ptr @A, i64 0, i64 %indvars.iv.next, i64 %indvars.iv.next29 53 %2 = load i32, ptr %arrayidx7 54 %add8 = add nsw i32 %2, %k 55 store i32 %add8, ptr %arrayidx7 56 %lftr.wideiv = trunc i64 %indvars.iv to i32 57 %exitcond = icmp eq i32 %lftr.wideiv, %0 58 br i1 %exitcond, label %for.cond.loopexit, label %for.body4 59 60 for.end17: 61 ret void 62} 63 64; When currently cannot interchange this loop, because transform currently 65; expects the latches to be the exiting blocks too. 66 67; IR-LABEL: @interchange_02 68; IR-NOT: split 69; 70; CHECK: Name: ExitingNotLatch 71; CHECK-NEXT: Function: interchange_02 72define void @interchange_02(i64 %k, i64 %N) { 73entry: 74 br label %for1.header 75 76for1.header: 77 %j23 = phi i64 [ 0, %entry ], [ %j.next24, %for1.inc10 ] 78 br label %for2 79 80for2: 81 %j = phi i64 [ %j.next, %latch ], [ 0, %for1.header ] 82 %arrayidx5 = getelementptr inbounds [100 x [100 x i64]], ptr @C, i64 0, i64 %j, i64 %j23 83 %lv = load i64, ptr %arrayidx5 84 %add = add nsw i64 %lv, %k 85 store i64 %add, ptr %arrayidx5 86 %exitcond = icmp eq i64 %j, 99 87 br i1 %exitcond, label %for1.inc10, label %latch 88latch: 89 %j.next = add nuw nsw i64 %j, 1 90 br label %for2 91 92for1.inc10: 93 %j.next24 = add nuw nsw i64 %j23, 1 94 %exitcond26 = icmp eq i64 %j23, 99 95 br i1 %exitcond26, label %for.end12, label %for1.header 96 97for.end12: 98 ret void 99} 100