1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
| ; RUN: llc -march=hexagon -O2 -disable-block-placement < %s | FileCheck %s
; Disable block placement because it intereferes with the generated code.
; CHECK: if (p{{[0-9]*}}) jump:nt .LBB0_2
; CHECK-NEXT: v{{[0-9]*}} = vmem(r{{[0-9]*}}+#0)
; CHECK-NEXT: }
target triple = "hexagon"
@g0 = global <16 x i32> zeroinitializer, align 64
; Function Attrs: nounwind
define void @f0(i16* nocapture readonly %a0, i32 %a1, i32 %a2, i16* nocapture %a3) #0 {
b0:
%v0 = mul i32 %a2, -2
%v1 = add i32 %v0, 64
%v2 = bitcast i16* %a3 to <16 x i32>*
%v3 = load <16 x i32>, <16 x i32>* @g0, align 64
%v4 = sdiv i32 %a1, 32
%v5 = icmp sgt i32 %a1, 31
br i1 %v5, label %b1, label %b4
b1: ; preds = %b0
%v6 = bitcast i16* %a0 to <16 x i32>*
%v7 = icmp sgt i32 %a1, 63
%v8 = mul i32 %v4, 32
%v9 = select i1 %v7, i32 %v8, i32 32
%v10 = getelementptr i16, i16* %a3, i32 %v9
br label %b2
b2: ; preds = %b2, %b1
%v11 = phi i32 [ 0, %b1 ], [ %v19, %b2 ]
%v12 = phi <16 x i32> [ %v3, %b1 ], [ %v16, %b2 ]
%v13 = phi <16 x i32>* [ %v2, %b1 ], [ %v18, %b2 ]
%v14 = phi <16 x i32>* [ %v6, %b1 ], [ %v15, %b2 ]
%v15 = getelementptr inbounds <16 x i32>, <16 x i32>* %v14, i32 1
%v16 = load <16 x i32>, <16 x i32>* %v14, align 64
%v17 = tail call <16 x i32> @llvm.hexagon.V6.valignb(<16 x i32> %v16, <16 x i32> %v12, i32 %v1)
%v18 = getelementptr inbounds <16 x i32>, <16 x i32>* %v13, i32 1
store <16 x i32> %v17, <16 x i32>* %v13, align 64
%v19 = add nsw i32 %v11, 1
%v20 = icmp slt i32 %v19, %v4
br i1 %v20, label %b2, label %b3
b3: ; preds = %b2
%v21 = bitcast i16* %v10 to <16 x i32>*
%v22 = load <16 x i32>, <16 x i32>* @g0, align 64
br label %b4
b4: ; preds = %b3, %b0
%v23 = phi <16 x i32> [ %v22, %b3 ], [ %v3, %b0 ]
%v24 = phi <16 x i32> [ %v16, %b3 ], [ %v3, %b0 ]
%v25 = phi <16 x i32>* [ %v21, %b3 ], [ %v2, %b0 ]
%v26 = tail call <16 x i32> @llvm.hexagon.V6.valignb(<16 x i32> %v23, <16 x i32> %v24, i32 %v1)
store <16 x i32> %v26, <16 x i32>* %v25, align 64
ret void
}
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.valignb(<16 x i32>, <16 x i32>, i32) #1
attributes #0 = { nounwind "target-cpu"="hexagonv62" "target-features"="+hvxv62,+hvx-length64b" }
attributes #1 = { nounwind readnone }
|