Karthik Bhat | 88db86d | 2015-03-06 10:11:25 +0000 | [diff] [blame] | 1 | ; RUN: opt < %s -basicaa -loop-interchange -S | FileCheck %s |
| 2 | ;; These are test that fail to interchange due to current limitation. This will go off once we extend the loop interchange pass. |
| 3 | |
| 4 | target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" |
| 5 | target triple = "x86_64-unknown-linux-gnu" |
| 6 | |
| 7 | @A = common global [100 x [100 x i32]] zeroinitializer |
| 8 | @B = common global [100 x [100 x [100 x i32]]] zeroinitializer |
| 9 | |
| 10 | ;;--------------------------------------Test case 01------------------------------------ |
| 11 | ;; [FIXME] This loop though valid is currently not interchanged due to the limitation that we cannot split the inner loop latch due to multiple use of inner induction |
| 12 | ;; variable.(used to increment the loop counter and to access A[j+1][i+1] |
| 13 | ;; for(int i=0;i<N-1;i++) |
| 14 | ;; for(int j=1;j<N-1;j++) |
| 15 | ;; A[j+1][i+1] = A[j+1][i+1] + k; |
| 16 | |
| 17 | define void @interchange_01(i32 %k, i32 %N) { |
| 18 | entry: |
| 19 | %sub = add nsw i32 %N, -1 |
| 20 | %cmp26 = icmp sgt i32 %N, 1 |
| 21 | br i1 %cmp26, label %for.cond1.preheader.lr.ph, label %for.end17 |
| 22 | |
| 23 | for.cond1.preheader.lr.ph: |
| 24 | %cmp324 = icmp sgt i32 %sub, 1 |
| 25 | %0 = add i32 %N, -2 |
| 26 | %1 = sext i32 %sub to i64 |
| 27 | br label %for.cond1.preheader |
| 28 | |
| 29 | for.cond.loopexit: |
| 30 | %cmp = icmp slt i64 %indvars.iv.next29, %1 |
| 31 | br i1 %cmp, label %for.cond1.preheader, label %for.end17 |
| 32 | |
| 33 | for.cond1.preheader: |
| 34 | %indvars.iv28 = phi i64 [ 0, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next29, %for.cond.loopexit ] |
| 35 | %indvars.iv.next29 = add nuw nsw i64 %indvars.iv28, 1 |
| 36 | br i1 %cmp324, label %for.body4, label %for.cond.loopexit |
| 37 | |
| 38 | for.body4: |
| 39 | %indvars.iv = phi i64 [ %indvars.iv.next, %for.body4 ], [ 1, %for.cond1.preheader ] |
| 40 | %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 |
| 41 | %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %indvars.iv.next, i64 %indvars.iv.next29 |
| 42 | %2 = load i32, i32* %arrayidx7 |
| 43 | %add8 = add nsw i32 %2, %k |
| 44 | store i32 %add8, i32* %arrayidx7 |
| 45 | %lftr.wideiv = trunc i64 %indvars.iv to i32 |
| 46 | %exitcond = icmp eq i32 %lftr.wideiv, %0 |
| 47 | br i1 %exitcond, label %for.cond.loopexit, label %for.body4 |
| 48 | |
| 49 | for.end17: |
| 50 | ret void |
| 51 | } |
| 52 | ;; Inner loop not split so it is not interchanged. |
| 53 | ; CHECK-LABEL: @interchange_01 |
| 54 | ; CHECK: for.body4: |
| 55 | ; CHECK-NEXT: %indvars.iv = phi i64 [ %indvars.iv.next, %for.body4 ], [ 1, %for.body4.preheader ] |
| 56 | ; CHECK-NEXT: %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 |
| 57 | ; CHECK-NEXT: %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %indvars.iv.next, i64 %indvars.iv.next29 |
| 58 | |