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
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
| ; RUN: llc -march=hexagon -O2 < %s
; REQUIRES: asserts
; Test that we generate the correct Phi names in the epilog when we need
; to reuse an existing Phi. This bug caused an assert in live variable
; analysis because the wrong virtual register was used.
; The bug occurs when a Phi references another Phi, and referent Phi
; value is used in multiple stages. When this occurs, the referring Phi
; can reuse one of the new values. We have code that deals with this in the
; kernel, but this case can occur in the epilog too.
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32>, <16 x i32>) #0
; Function Attrs: nounwind readnone
declare <32 x i32> @llvm.hexagon.V6.vmpabus.acc(<32 x i32>, <32 x i32>, i32) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vlalignbi(<16 x i32>, <16 x i32>, i32) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.lo(<32 x i32>) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.hi(<32 x i32>) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32>, <16 x i32>, i32) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32>, <16 x i32>) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32>, <16 x i32>, i32) #0
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vshuffob(<16 x i32>, <16 x i32>) #0
; Function Attrs: nounwind
define void @f0(i8* noalias nocapture readonly %a0, i32 %a1, i32 %a2) #1 {
b0:
%v0 = mul nsw i32 %a1, 2
br i1 undef, label %b1, label %b5
b1: ; preds = %b0
%v1 = getelementptr inbounds i8, i8* %a0, i32 %v0
%v2 = icmp sgt i32 %a2, 64
%v3 = add i32 %v0, 64
%v4 = add i32 %a1, 64
%v5 = sub i32 64, %a1
%v6 = sub i32 64, %v0
br i1 %v2, label %b2, label %b4
b2: ; preds = %b1
%v7 = getelementptr inbounds i8, i8* %v1, i32 %v3
%v8 = getelementptr inbounds i8, i8* %v1, i32 %v4
%v9 = getelementptr inbounds i8, i8* %v1, i32 64
%v10 = getelementptr inbounds i8, i8* %v1, i32 %v5
%v11 = getelementptr inbounds i8, i8* %v1, i32 %v6
%v12 = bitcast i8* %v7 to <16 x i32>*
%v13 = bitcast i8* %v8 to <16 x i32>*
%v14 = bitcast i8* %v9 to <16 x i32>*
%v15 = bitcast i8* %v10 to <16 x i32>*
%v16 = bitcast i8* %v11 to <16 x i32>*
br label %b3
b3: ; preds = %b3, %b2
%v17 = phi <16 x i32>* [ null, %b2 ], [ %v52, %b3 ]
%v18 = phi <16 x i32>* [ %v12, %b2 ], [ %v34, %b3 ]
%v19 = phi <16 x i32>* [ %v13, %b2 ], [ %v32, %b3 ]
%v20 = phi <16 x i32>* [ %v14, %b2 ], [ %v30, %b3 ]
%v21 = phi <16 x i32>* [ %v15, %b2 ], [ %v28, %b3 ]
%v22 = phi <16 x i32>* [ %v16, %b2 ], [ %v26, %b3 ]
%v23 = phi <32 x i32> [ undef, %b2 ], [ %v37, %b3 ]
%v24 = phi <32 x i32> [ zeroinitializer, %b2 ], [ %v23, %b3 ]
%v25 = phi i32 [ %a2, %b2 ], [ %v53, %b3 ]
%v26 = getelementptr inbounds <16 x i32>, <16 x i32>* %v22, i32 1
%v27 = load <16 x i32>, <16 x i32>* %v22, align 64
%v28 = getelementptr inbounds <16 x i32>, <16 x i32>* %v21, i32 1
%v29 = load <16 x i32>, <16 x i32>* %v21, align 64
%v30 = getelementptr inbounds <16 x i32>, <16 x i32>* %v20, i32 1
%v31 = load <16 x i32>, <16 x i32>* %v20, align 64
%v32 = getelementptr inbounds <16 x i32>, <16 x i32>* %v19, i32 1
%v33 = load <16 x i32>, <16 x i32>* %v19, align 64
%v34 = getelementptr inbounds <16 x i32>, <16 x i32>* %v18, i32 1
%v35 = load <16 x i32>, <16 x i32>* %v18, align 64
%v36 = tail call <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32> %v33, <16 x i32> %v29) #3
%v37 = tail call <32 x i32> @llvm.hexagon.V6.vmpabus.acc(<32 x i32> undef, <32 x i32> %v36, i32 67372036) #3
%v38 = tail call <16 x i32> @llvm.hexagon.V6.lo(<32 x i32> %v23) #3
%v39 = tail call <16 x i32> @llvm.hexagon.V6.lo(<32 x i32> %v24) #3
%v40 = tail call <16 x i32> @llvm.hexagon.V6.vlalignbi(<16 x i32> %v38, <16 x i32> %v39, i32 2) #3
%v41 = tail call <16 x i32> @llvm.hexagon.V6.lo(<32 x i32> %v37) #3
%v42 = tail call <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32> %v41, <16 x i32> %v38, i32 2) #3
%v43 = tail call <16 x i32> @llvm.hexagon.V6.hi(<32 x i32> %v37) #3
%v44 = tail call <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32> %v43, <16 x i32> undef, i32 2) #3
%v45 = tail call <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32> %v40, <16 x i32> %v42) #3
%v46 = tail call <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32> %v45, <16 x i32> %v38, i32 101058054) #3
%v47 = tail call <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32> %v46, <16 x i32> zeroinitializer, i32 67372036) #3
%v48 = tail call <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32> undef, <16 x i32> %v44) #3
%v49 = tail call <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32> %v48, <16 x i32> undef, i32 101058054) #3
%v50 = tail call <16 x i32> @llvm.hexagon.V6.vmpyiwb.acc(<16 x i32> %v49, <16 x i32> zeroinitializer, i32 67372036) #3
%v51 = tail call <16 x i32> @llvm.hexagon.V6.vshuffob(<16 x i32> %v50, <16 x i32> %v47) #3
%v52 = getelementptr inbounds <16 x i32>, <16 x i32>* %v17, i32 1
store <16 x i32> %v51, <16 x i32>* %v17, align 64
%v53 = add nsw i32 %v25, -64
%v54 = icmp sgt i32 %v53, 64
br i1 %v54, label %b3, label %b4
b4: ; preds = %b3, %b1
unreachable
b5: ; preds = %b0
ret void
}
; Function Attrs: nounwind
define void @f1(i32 %a0, i32* %a1) #1 {
b0:
%v0 = ptrtoint i32* %a1 to i32
%v1 = ashr i32 %a0, 1
%v2 = tail call i32 @llvm.hexagon.A2.combine.ll(i32 undef, i32 undef)
br i1 undef, label %b1, label %b2
b1: ; preds = %b1, %b0
br i1 undef, label %b1, label %b2
b2: ; preds = %b2, %b1, %b0
%v3 = phi i64 [ %v11, %b2 ], [ undef, %b0 ], [ undef, %b1 ]
%v4 = phi i32 [ %v12, %b2 ], [ 0, %b0 ], [ undef, %b1 ]
%v5 = phi i32 [ %v6, %b2 ], [ %v2, %b0 ], [ undef, %b1 ]
%v6 = phi i32 [ %v10, %b2 ], [ undef, %b0 ], [ undef, %b1 ]
%v7 = tail call i64 @llvm.hexagon.M2.vdmacs.s0(i64 undef, i64 %v3, i64 undef)
%v8 = tail call i64 @llvm.hexagon.A2.combinew(i32 %v5, i32 %v5)
%v9 = tail call i64 @llvm.hexagon.M2.vdmacs.s0(i64 undef, i64 %v8, i64 undef)
%v10 = tail call i32 @llvm.hexagon.A2.combine.ll(i32 0, i32 undef)
%v11 = tail call i64 @llvm.hexagon.A2.combinew(i32 %v10, i32 %v6)
%v12 = add nsw i32 %v4, 1
%v13 = icmp eq i32 %v12, %v1
br i1 %v13, label %b3, label %b2
b3: ; preds = %b2
%v14 = phi i64 [ %v9, %b2 ]
%v15 = phi i64 [ %v7, %b2 ]
%v16 = trunc i64 %v14 to i32
%v17 = trunc i64 %v15 to i32
%v18 = inttoptr i32 %v0 to i32*
store i32 %v17, i32* %v18, align 4
%v19 = bitcast i8* undef to i32*
store i32 %v16, i32* %v19, align 4
call void @llvm.trap()
unreachable
}
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.A2.combine.ll(i32, i32) #0
; Function Attrs: nounwind readnone
declare i64 @llvm.hexagon.A2.combinew(i32, i32) #0
; Function Attrs: nounwind readnone
declare i64 @llvm.hexagon.M2.vdmacs.s0(i64, i64, i64) #0
; Function Attrs: noreturn nounwind
declare void @llvm.trap() #2
attributes #0 = { nounwind readnone }
attributes #1 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length64b" }
attributes #2 = { noreturn nounwind }
attributes #3 = { nounwind }
|