| Eric Christopher | cee313d | 2019-04-17 04:52:47 +0000 | [diff] [blame] | 1 | ; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -dce -instcombine -S | FileCheck %s |
| 2 | |
| 3 | target 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-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" |
| 4 | |
| 5 | @a = common global [2048 x i32] zeroinitializer, align 16 |
| 6 | @b = common global [2048 x i32] zeroinitializer, align 16 |
| 7 | @c = common global [2048 x i32] zeroinitializer, align 16 |
| 8 | |
| 9 | ;CHECK-LABEL: @example1( |
| 10 | ;CHECK: load <4 x i32> |
| 11 | ;CHECK: ret void |
| 12 | define void @example1() nounwind uwtable ssp { |
| 13 | br label %1 |
| 14 | |
| 15 | ; <label>:1 ; preds = %1, %0 |
| 16 | %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ] |
| 17 | %2 = getelementptr inbounds [2048 x i32], [2048 x i32]* @b, i64 0, i64 %indvars.iv |
| 18 | %3 = load i32, i32* %2, align 4 |
| 19 | %4 = getelementptr inbounds [2048 x i32], [2048 x i32]* @c, i64 0, i64 %indvars.iv |
| 20 | %5 = load i32, i32* %4, align 4 |
| 21 | %6 = add nsw i32 %5, %3 |
| 22 | %7 = getelementptr inbounds [2048 x i32], [2048 x i32]* @a, i64 0, i64 %indvars.iv |
| 23 | store i32 %6, i32* %7, align 4 |
| 24 | %indvars.iv.next = add i64 %indvars.iv, 1 |
| 25 | %lftr.wideiv = trunc i64 %indvars.iv.next to i32 |
| 26 | %exitcond = icmp eq i32 %lftr.wideiv, 8 ; <----- A really small trip count |
| 27 | br i1 %exitcond, label %8, label %1 ; w/o scalar iteration overhead. |
| 28 | |
| 29 | ; <label>:8 ; preds = %1 |
| 30 | ret void |
| 31 | } |
| 32 | |
| 33 | ;CHECK-LABEL: @bound1( |
| 34 | ;CHECK-NOT: load <4 x i32> |
| 35 | ;CHECK: ret void |
| 36 | define void @bound1(i32 %k) nounwind uwtable ssp { |
| 37 | br label %1 |
| 38 | |
| 39 | ; <label>:1 ; preds = %1, %0 |
| 40 | %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ] |
| 41 | %2 = getelementptr inbounds [2048 x i32], [2048 x i32]* @b, i64 0, i64 %indvars.iv |
| 42 | %3 = load i32, i32* %2, align 4 |
| 43 | %4 = getelementptr inbounds [2048 x i32], [2048 x i32]* @c, i64 0, i64 %indvars.iv |
| 44 | %5 = load i32, i32* %4, align 4 |
| 45 | %6 = add nsw i32 %5, %3 |
| 46 | %7 = getelementptr inbounds [2048 x i32], [2048 x i32]* @a, i64 0, i64 %indvars.iv |
| 47 | store i32 %6, i32* %7, align 4 |
| 48 | %indvars.iv.next = add i64 %indvars.iv, 1 |
| 49 | %lftr.wideiv = trunc i64 %indvars.iv.next to i32 |
| 50 | %large = icmp sge i32 %lftr.wideiv, 8 |
| 51 | %exitcond = icmp eq i32 %lftr.wideiv, %k |
| 52 | %realexit = or i1 %large, %exitcond |
| 53 | br i1 %realexit, label %8, label %1 |
| 54 | |
| 55 | ; <label>:8 ; preds = %1 |
| 56 | ret void |
| 57 | } |