1; RUN: opt < %s -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s 2 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 16 6 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 void 16define void @inc(i32 %n) nounwind uwtable noinline ssp { 17 %1 = icmp sgt i32 %n, 0 18 br i1 %1, label %.lr.ph, label %._crit_edge 19 20.lr.ph: ; preds = %0, %.lr.ph 21 %indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ] 22 %2 = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %indvars.iv 23 %3 = load i32, ptr %2, align 4 24 %4 = trunc i64 %indvars.iv to i32 25 %5 = add nsw i32 %3, %4 26 store i32 %5, ptr %2, align 4 27 %indvars.iv.next = add i64 %indvars.iv, 1 28 %lftr.wideiv = trunc i64 %indvars.iv.next to i32 29 %exitcond = icmp eq i32 %lftr.wideiv, %n 30 br i1 %exitcond, label %._crit_edge, label %.lr.ph 31 32._crit_edge: ; preds = %.lr.ph, %0 33 ret void 34} 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 i32 44define i32 @histogram(ptr nocapture noalias %A, ptr nocapture noalias %B, i32 %n) nounwind uwtable ssp { 45entry: 46 %cmp6 = icmp sgt i32 %n, 0 47 br i1 %cmp6, label %for.body, label %for.end 48 49for.body: ; preds = %entry, %for.body 50 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] 51 %arrayidx = getelementptr inbounds i32, ptr %B, i64 %indvars.iv 52 %0 = load i32, ptr %arrayidx, align 4 53 %idxprom1 = sext i32 %0 to i64 54 %arrayidx2 = getelementptr inbounds i32, ptr %A, i64 %idxprom1 55 %1 = load i32, ptr %arrayidx2, align 4 56 %inc = add nsw i32 %1, 1 57 store i32 %inc, ptr %arrayidx2, align 4 58 %indvars.iv.next = add i64 %indvars.iv, 1 59 %lftr.wideiv = trunc i64 %indvars.iv.next to i32 60 %exitcond = icmp eq i32 %lftr.wideiv, %n 61 br i1 %exitcond, label %for.end, label %for.body 62 63for.end: ; preds = %for.body, %entry 64 ret i32 0 65} 66