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
| ; RUN: llc -march=hexagon -enable-pipeliner -hexagon-expand-condsets=0 < %s
; REQUIRES: asserts
; Disable expand-condsets because it will assert on undefined registers.
; Another test that the pipeliner doesn't ICE when reusing a
; PHI in the epilog code.
@g0 = external global [18 x i16], align 8
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.S2.asl.r.r.sat(i32, i32) #0
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.A2.sxth(i32) #0
; Function Attrs: nounwind
define void @f0() #1 {
b0:
%v0 = alloca [166 x i32], align 8
br label %b1
b1: ; preds = %b1, %b0
%v1 = icmp eq i16 undef, 0
br i1 %v1, label %b2, label %b1
b2: ; preds = %b1
br i1 undef, label %b3, label %b4
b3: ; preds = %b3, %b2
%v2 = add i32 0, 2
br i1 undef, label %b3, label %b4
b4: ; preds = %b3, %b2
%v3 = phi i32* [ undef, %b2 ], [ undef, %b3 ]
%v4 = phi i32 [ 0, %b2 ], [ %v2, %b3 ]
%v5 = getelementptr [18 x i16], [18 x i16]* @g0, i32 0, i32 undef
br label %b5
b5: ; preds = %b5, %b4
%v6 = phi i16 [ 0, %b4 ], [ %v17, %b5 ]
%v7 = phi i16 [ undef, %b4 ], [ %v6, %b5 ]
%v8 = phi i32 [ %v4, %b4 ], [ %v35, %b5 ]
%v9 = phi i32* [ %v3, %b4 ], [ undef, %b5 ]
%v10 = phi i16* [ undef, %b4 ], [ %v12, %b5 ]
%v11 = add i32 %v8, 0
%v12 = getelementptr inbounds i16, i16* %v10, i32 1
%v13 = sext i16 %v7 to i32
%v14 = call i32 @llvm.hexagon.M2.mpy.acc.sat.ll.s1(i32 0, i32 %v13, i32 undef)
%v15 = getelementptr inbounds i16, i16* %v10, i32 2
%v16 = call i32 @llvm.hexagon.M2.mpy.acc.sat.ll.s1(i32 %v14, i32 undef, i32 undef)
%v17 = load i16, i16* %v15, align 2, !tbaa !0
%v18 = sext i16 %v17 to i32
%v19 = call i32 @llvm.hexagon.M2.mpy.acc.sat.ll.s1(i32 %v16, i32 %v18, i32 undef)
%v20 = call i32 @llvm.hexagon.S2.asr.r.r.sat(i32 %v19, i32 undef)
%v21 = getelementptr [166 x i32], [166 x i32]* %v0, i32 0, i32 %v11
%v22 = load i32, i32* %v21, align 4
%v23 = call i64 @llvm.hexagon.M2.dpmpyss.s0(i32 %v22, i32 undef)
%v24 = call i64 @llvm.hexagon.S2.asr.i.p(i64 %v23, i32 15)
%v25 = call i32 @llvm.hexagon.A2.sat(i64 %v24)
%v26 = call i32 @llvm.hexagon.A2.subsat(i32 %v20, i32 %v25)
%v27 = load i16, i16* %v5, align 4
%v28 = sext i16 %v27 to i32
%v29 = call i32 @llvm.hexagon.A2.sxth(i32 %v28)
%v30 = call i64 @llvm.hexagon.M2.dpmpyss.s0(i32 undef, i32 %v29)
%v31 = call i64 @llvm.hexagon.S2.asr.i.p(i64 %v30, i32 15)
%v32 = call i32 @llvm.hexagon.A2.sat(i64 %v31)
%v33 = call i32 @llvm.hexagon.A2.subsat(i32 %v26, i32 %v32)
%v34 = call i32 @llvm.hexagon.S2.asl.r.r.sat(i32 %v33, i32 undef)
store i32 %v34, i32* %v9, align 4, !tbaa !4
%v35 = add i32 %v8, 1
%v36 = icmp eq i32 %v35, 164
br i1 %v36, label %b6, label %b5
b6: ; preds = %b5
call void @llvm.trap()
unreachable
}
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.M2.mpy.acc.sat.ll.s1(i32, i32, i32) #0
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.S2.asr.r.r.sat(i32, i32) #0
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.A2.sat(i64) #0
; Function Attrs: nounwind readnone
declare i64 @llvm.hexagon.S2.asr.i.p(i64, i32) #0
; Function Attrs: nounwind readnone
declare i64 @llvm.hexagon.M2.dpmpyss.s0(i32, i32) #0
; Function Attrs: nounwind readnone
declare i32 @llvm.hexagon.A2.subsat(i32, i32) #0
; Function Attrs: noreturn nounwind
declare void @llvm.trap() #2
attributes #0 = { nounwind readnone }
attributes #1 = { nounwind "target-cpu"="hexagonv55" }
attributes #2 = { noreturn nounwind }
!0 = !{!1, !1, i64 0}
!1 = !{!"short", !2, i64 0}
!2 = !{!"omnipotent char", !3, i64 0}
!3 = !{!"Simple C/C++ TBAA"}
!4 = !{!5, !5, i64 0}
!5 = !{!"long", !2, i64 0}
|