66 lines · plain
1; RUN: opt < %s -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s2 3target 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 166 7; This is the loop.8; for (i=0; i<n; i++){9; a[i] += i;10; }11;CHECK-LABEL: @inc(12;CHECK: load <4 x i32>13;CHECK: add nsw <4 x i32>14;CHECK: store <4 x i32>15;CHECK: ret void16define void @inc(i32 %n) nounwind uwtable noinline ssp {17 %1 = icmp sgt i32 %n, 018 br i1 %1, label %.lr.ph, label %._crit_edge19 20.lr.ph: ; preds = %0, %.lr.ph21 %indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ]22 %2 = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %indvars.iv23 %3 = load i32, ptr %2, align 424 %4 = trunc i64 %indvars.iv to i3225 %5 = add nsw i32 %3, %426 store i32 %5, ptr %2, align 427 %indvars.iv.next = add i64 %indvars.iv, 128 %lftr.wideiv = trunc i64 %indvars.iv.next to i3229 %exitcond = icmp eq i32 %lftr.wideiv, %n30 br i1 %exitcond, label %._crit_edge, label %.lr.ph31 32._crit_edge: ; preds = %.lr.ph, %033 ret void34}35 36; Can't vectorize this loop because the access to A[X] is non-linear.37;38; for (i = 0; i < n; ++i) {39; A[B[i]]++;40;41;CHECK-LABEL: @histogram(42;CHECK-NOT: <4 x i32>43;CHECK: ret i3244define i32 @histogram(ptr nocapture noalias %A, ptr nocapture noalias %B, i32 %n) nounwind uwtable ssp {45entry:46 %cmp6 = icmp sgt i32 %n, 047 br i1 %cmp6, label %for.body, label %for.end48 49for.body: ; preds = %entry, %for.body50 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]51 %arrayidx = getelementptr inbounds i32, ptr %B, i64 %indvars.iv52 %0 = load i32, ptr %arrayidx, align 453 %idxprom1 = sext i32 %0 to i6454 %arrayidx2 = getelementptr inbounds i32, ptr %A, i64 %idxprom155 %1 = load i32, ptr %arrayidx2, align 456 %inc = add nsw i32 %1, 157 store i32 %inc, ptr %arrayidx2, align 458 %indvars.iv.next = add i64 %indvars.iv, 159 %lftr.wideiv = trunc i64 %indvars.iv.next to i3260 %exitcond = icmp eq i32 %lftr.wideiv, %n61 br i1 %exitcond, label %for.end, label %for.body62 63for.end: ; preds = %for.body, %entry64 ret i32 065}66