1; RUN: llc -march=hexagon -mcpu=hexagonv5 -enable-pipeliner < %s | FileCheck %s
2; RUN: llc -march=hexagon -mcpu=hexagonv5 -O2 < %s | FileCheck %s
3; RUN: llc -march=hexagon -mcpu=hexagonv5 -O3 < %s | FileCheck %s
4;
5; Check that we pipeline a vectorized dot product in a single packet.
6;
7; CHECK: {
8; CHECK: += mpyi
9; CHECK: += mpyi
10; CHECK: memd
11; CHECK: memd
12; CHECK: }      :endloop0
13
14@a = common global [5000 x i32] zeroinitializer, align 8
15@b = common global [5000 x i32] zeroinitializer, align 8
16
17define i32 @vecMultGlobal() {
18entry:
19  br label %polly.loop_body
20
21polly.loop_after:
22  %0 = extractelement <2 x i32> %addp_vec, i32 0
23  %1 = extractelement <2 x i32> %addp_vec, i32 1
24  %add_sum = add i32 %0, %1
25  ret i32 %add_sum
26
27polly.loop_body:
28  %polly.loopiv13 = phi i32 [ 0, %entry ], [ %polly.next_loopiv, %polly.loop_body ]
29  %reduction.012 = phi <2 x i32> [ zeroinitializer, %entry ], [ %addp_vec, %polly.loop_body ]
30  %polly.next_loopiv = add nsw i32 %polly.loopiv13, 2
31  %p_arrayidx1 = getelementptr [5000 x i32], [5000 x i32]* @b, i32 0, i32 %polly.loopiv13
32  %p_arrayidx = getelementptr [5000 x i32], [5000 x i32]* @a, i32 0, i32 %polly.loopiv13
33  %vector_ptr = bitcast i32* %p_arrayidx1 to <2 x i32>*
34  %_p_vec_full = load <2 x i32>, <2 x i32>* %vector_ptr, align 8
35  %vector_ptr7 = bitcast i32* %p_arrayidx to <2 x i32>*
36  %_p_vec_full8 = load <2 x i32>, <2 x i32>* %vector_ptr7, align 8
37  %mulp_vec = mul <2 x i32> %_p_vec_full8, %_p_vec_full
38  %addp_vec = add <2 x i32> %mulp_vec, %reduction.012
39  %2 = icmp slt i32 %polly.next_loopiv, 5000
40  br i1 %2, label %polly.loop_body, label %polly.loop_after
41}
42