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
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
|
; RUN: llc -march=hexagon -enable-pipeliner=false -hexagon-vector-combine=false < %s | FileCheck %s
; Test that the vsplat and vmemu are not all serialized due to chain edges
; caused by the hasSideEffects flag. The exact code generation may change
; due to the scheduling changes, but we shouldn't see a series of
; vsplat and vmemu instructions that each occur in a single packet.
; CHECK: loop0(.LBB0_[[LOOP:.]],
; CHECK: .LBB0_[[LOOP]]:
; CHECK: vsplat
; CHECK-NEXT: vsplat
; CHECK: vsplat
; CHECK-NEXT: vsplat
; CHECK: endloop0
@g0 = global [256 x i8] c"^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^00226644,,..**8888::66,,,,&&^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^22000022..4444>>::8888**..^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^<<66220000226644<<>>::^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^>><<446622000022>>", align 64
; Function Attrs: nounwind
define void @f0(i16** noalias nocapture readonly %a0, i16* noalias nocapture readonly %a1, i32* noalias nocapture %a2, i32 %a3, i32 %a4, i32 %a5, i32 %a6) #0 {
b0:
%v0 = load <16 x i32>, <16 x i32>* bitcast ([256 x i8]* @g0 to <16 x i32>*), align 64, !tbaa !0
%v1 = load <16 x i32>, <16 x i32>* bitcast (i8* getelementptr inbounds ([256 x i8], [256 x i8]* @g0, i32 0, i32 64) to <16 x i32>*), align 64, !tbaa !0
%v2 = load <16 x i32>, <16 x i32>* bitcast (i8* getelementptr inbounds ([256 x i8], [256 x i8]* @g0, i32 0, i32 128) to <16 x i32>*), align 64, !tbaa !0
%v3 = load <16 x i32>, <16 x i32>* bitcast (i8* getelementptr inbounds ([256 x i8], [256 x i8]* @g0, i32 0, i32 192) to <16 x i32>*), align 64, !tbaa !0
%v4 = icmp sgt i32 %a5, 0
br i1 %v4, label %b1, label %b5
b1: ; preds = %b0
%v5 = bitcast i32* %a2 to <16 x i32>*
%v6 = tail call <16 x i32> @llvm.hexagon.V6.vd0()
%v7 = bitcast i16* %a1 to i64*
%v8 = mul nsw i32 %a3, 4
%v9 = add i32 %v8, %a6
%v10 = add i32 %v9, 32
%v11 = add i32 %a5, -1
br label %b2
b2: ; preds = %b4, %b1
%v12 = phi i32 [ 0, %b1 ], [ %v59, %b4 ]
%v13 = phi <16 x i32>* [ %v5, %b1 ], [ %v58, %b4 ]
%v14 = getelementptr i16*, i16** %a0, i32 %v12
br label %b3
b3: ; preds = %b3, %b2
%v15 = phi i16** [ %v14, %b2 ], [ %v57, %b3 ]
%v16 = phi i32 [ 0, %b2 ], [ %v55, %b3 ]
%v17 = phi i64* [ %v7, %b2 ], [ %v23, %b3 ]
%v18 = phi <16 x i32> [ %v6, %b2 ], [ %v54, %b3 ]
%v19 = load i16*, i16** %v15, align 4, !tbaa !3
%v20 = getelementptr inbounds i16, i16* %v19, i32 %v9
%v21 = getelementptr inbounds i64, i64* %v17, i32 1
%v22 = load i64, i64* %v17, align 8, !tbaa !0
%v23 = getelementptr inbounds i64, i64* %v17, i32 2
%v24 = load i64, i64* %v21, align 8, !tbaa !0
%v25 = trunc i64 %v22 to i32
%v26 = lshr i64 %v22, 32
%v27 = tail call <16 x i32> @llvm.hexagon.V6.lvsplatw(i32 %v25)
%v28 = trunc i64 %v26 to i32
%v29 = tail call <16 x i32> @llvm.hexagon.V6.lvsplatw(i32 %v28)
%v30 = trunc i64 %v24 to i32
%v31 = lshr i64 %v24, 32
%v32 = tail call <16 x i32> @llvm.hexagon.V6.lvsplatw(i32 %v30)
%v33 = trunc i64 %v31 to i32
%v34 = tail call <16 x i32> @llvm.hexagon.V6.lvsplatw(i32 %v33)
%v35 = bitcast i16* %v20 to <16 x i32>*
%v36 = load <16 x i32>, <16 x i32>* %v35, align 4, !tbaa !0
%v37 = getelementptr inbounds i16, i16* %v19, i32 %v10
%v38 = bitcast i16* %v37 to <16 x i32>*
%v39 = load <16 x i32>, <16 x i32>* %v38, align 4, !tbaa !0
%v40 = tail call <16 x i32> @llvm.hexagon.V6.vpackeh(<16 x i32> %v39, <16 x i32> %v36)
%v41 = tail call <16 x i32> @llvm.hexagon.V6.vpackeh(<16 x i32> %v40, <16 x i32> %v40)
%v42 = tail call <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32> %v41, <16 x i32> %v0)
%v43 = tail call <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32> %v41, <16 x i32> %v1)
%v44 = tail call <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32> %v41, <16 x i32> %v2)
%v45 = tail call <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32> %v41, <16 x i32> %v3)
%v46 = tail call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %v27, <16 x i32> %v42)
%v47 = tail call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %v29, <16 x i32> %v43)
%v48 = tail call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %v32, <16 x i32> %v44)
%v49 = tail call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %v34, <16 x i32> %v45)
%v50 = tail call <16 x i32> @llvm.hexagon.V6.vdmpyhvsat(<16 x i32> %v46, <16 x i32> %v46)
%v51 = tail call <16 x i32> @llvm.hexagon.V6.vdmpyhvsat.acc(<16 x i32> %v50, <16 x i32> %v47, <16 x i32> %v47)
%v52 = tail call <16 x i32> @llvm.hexagon.V6.vdmpyhvsat.acc(<16 x i32> %v51, <16 x i32> %v48, <16 x i32> %v48)
%v53 = tail call <16 x i32> @llvm.hexagon.V6.vdmpyhvsat.acc(<16 x i32> %v52, <16 x i32> %v49, <16 x i32> %v49)
%v54 = tail call <16 x i32> @llvm.hexagon.V6.vasrw.acc(<16 x i32> %v18, <16 x i32> %v53, i32 6)
%v55 = add nsw i32 %v16, 1
%v56 = icmp eq i32 %v16, 7
%v57 = getelementptr i16*, i16** %v15, i32 1
br i1 %v56, label %b4, label %b3
b4: ; preds = %b3
%v58 = getelementptr inbounds <16 x i32>, <16 x i32>* %v13, i32 1
store <16 x i32> %v54, <16 x i32>* %v13, align 64, !tbaa !0
%v59 = add nsw i32 %v12, 1
%v60 = icmp eq i32 %v12, %v11
br i1 %v60, label %b5, label %b2
b5: ; preds = %b4, %b0
ret void
}
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vd0() #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.lvsplatw(i32) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vpackeh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdelta(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhvsat(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vdmpyhvsat.acc(<16 x i32>, <16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vasrw.acc(<16 x i32>, <16 x i32>, i32) #1
attributes #0 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length64b" }
attributes #1 = { nounwind readnone }
!0 = !{!1, !1, i64 0}
!1 = !{!"omnipotent char", !2, i64 0}
!2 = !{!"Simple C/C++ TBAA"}
!3 = !{!4, !4, i64 0}
!4 = !{!"any pointer", !1, i64 0}
|
