1 ; RUN: opt %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s 2 ; RUN: opt %s -loop-vectorize -force-vector-interleave=2 -force-vector-width=4 -S | FileCheck %s -check-prefix=UNROLL 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 @b = common global [1000 x i32] zeroinitializer, align 16 8 @c = common global [1000 x i32] zeroinitializer, align 16 9 @a = common global [1000 x i32] zeroinitializer, align 16 10 11 ; Generate min.iters.check to skip the vector loop and jump to scalar.ph directly when loop iteration number is less than VF * UF. 12 ; CHECK-LABEL: foo( 13 ; CHECK: %min.iters.check = icmp ult i64 %N, 4 14 ; CHECK: br i1 %min.iters.check, label %scalar.ph, label %min.iters.checked 15 ; UNROLL-LABEL: foo( 16 ; UNROLL: %min.iters.check = icmp ult i64 %N, 8 17 ; UNROLL: br i1 %min.iters.check, label %scalar.ph, label %min.iters.checked 18 19 define void @foo(i64 %N) { 20 entry: 21 %cmp.8 = icmp sgt i64 %N, 0 22 br i1 %cmp.8, label %for.body.preheader, label %for.end 23 24 for.body.preheader: ; preds = %entry 25 br label %for.body 26 27 for.body: ; preds = %for.body, %for.body.preheader 28 %i.09 = phi i64 [ %inc, %for.body ], [ 0, %for.body.preheader ] 29 %arrayidx = getelementptr inbounds [1000 x i32], [1000 x i32]* @b, i64 0, i64 %i.09 30 %tmp = load i32, i32* %arrayidx, align 4 31 %arrayidx1 = getelementptr inbounds [1000 x i32], [1000 x i32]* @c, i64 0, i64 %i.09 32 %tmp1 = load i32, i32* %arrayidx1, align 4 33 %add = add nsw i32 %tmp1, %tmp 34 %arrayidx2 = getelementptr inbounds [1000 x i32], [1000 x i32]* @a, i64 0, i64 %i.09 35 store i32 %add, i32* %arrayidx2, align 4 36 %inc = add nuw nsw i64 %i.09, 1 37 %exitcond = icmp eq i64 %inc, %N 38 br i1 %exitcond, label %for.end.loopexit, label %for.body 39 40 for.end.loopexit: ; preds = %for.body 41 br label %for.end 42 43 for.end: ; preds = %for.end.loopexit, %entry 44 ret void 45 } 46
