1; RUN: llc -march=hexagon < %s | FileCheck %s
2; CHECK: vaslh
3
4target datalayout = "e-p:32:32:32-i64:64:64-i32:32:32-i16:16:16-i1:32:32-f64:64:64-f32:32:32-v64:64:64-v32:32:32-a0:0-n16:32"
5target triple = "hexagon-unknown-linux-gnu"
6
7define void @foo(i16* nocapture %v) nounwind {
8entry:
9  %p_arrayidx = getelementptr i16, i16* %v, i32 4
10  %vector_ptr = bitcast i16* %p_arrayidx to <4 x i16>*
11  %_p_vec_full = load <4 x i16>, <4 x i16>* %vector_ptr, align 2
12  %_high_half = shufflevector <4 x i16> %_p_vec_full, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
13  %_low_half = shufflevector <4 x i16> %_p_vec_full, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
14  %0 = sext <2 x i16> %_low_half to <2 x i32>
15  %1 = sext <2 x i16> %_high_half to <2 x i32>
16  %shr6p_vec = shl <2 x i32> %0, <i32 2, i32 2>
17  %shr6p_vec19 = shl <2 x i32> %1, <i32 2, i32 2>
18  %addp_vec = add <2 x i32> %shr6p_vec, <i32 34, i32 34>
19  %addp_vec20 = add <2 x i32> %shr6p_vec19, <i32 34, i32 34>
20  %vector_ptr21 = bitcast i16* %v to <4 x i16>*
21  %_p_vec_full22 = load <4 x i16>, <4 x i16>* %vector_ptr21, align 2
22  %_high_half23 = shufflevector <4 x i16> %_p_vec_full22, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
23  %_low_half24 = shufflevector <4 x i16> %_p_vec_full22, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
24  %2 = zext <2 x i16> %_low_half24 to <2 x i32>
25  %3 = zext <2 x i16> %_high_half23 to <2 x i32>
26  %add3p_vec = add <2 x i32> %addp_vec, %2
27  %add3p_vec25 = add <2 x i32> %addp_vec20, %3
28  %4 = trunc <2 x i32> %add3p_vec to <2 x i16>
29  %5 = trunc <2 x i32> %add3p_vec25 to <2 x i16>
30  %_combined_vec = shufflevector <2 x i16> %4, <2 x i16> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
31  store <4 x i16> %_combined_vec, <4 x i16>* %vector_ptr21, align 2
32  ret void
33}
34