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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -march=hexagon < %s | FileCheck %s
define void @f0(ptr %a0, ptr %a1, ptr %a2) #0 {
; CHECK-LABEL: f0:
; CHECK: // %bb.0: // %b0
; CHECK-NEXT: {
; CHECK-NEXT: v0 = vmem(r0+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1 = vmem(r1+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v2.w = vmpye(v0.w,v1.uh)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v2.w += vmpyo(v0.w,v1.h):<<1:sat:shift
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: vmem(r2+#0) = v2
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: jumpr r31
; CHECK-NEXT: }
b0:
%v0 = load <32 x i32>, ptr %a0, align 128
%v1 = load <32 x i32>, ptr %a1, align 128
%v2 = sext <32 x i32> %v0 to <32 x i64>
%v3 = sext <32 x i32> %v1 to <32 x i64>
%v4 = mul nsw <32 x i64> %v2, %v3
%v5 = lshr <32 x i64> %v4, <i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31>
%v6 = trunc <32 x i64> %v5 to <32 x i32>
store <32 x i32> %v6, ptr %a2, align 128
ret void
}
define void @f1(ptr %a0, ptr %a1, ptr %a2) #0 {
; CHECK-LABEL: f1:
; CHECK: // %bb.0: // %b0
; CHECK-NEXT: {
; CHECK-NEXT: v0 = vmem(r0+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1 = vmem(r1+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v2.w = vmpye(v0.w,v1.uh)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v2.w += vmpyo(v0.w,v1.h):<<1:rnd:sat:shift
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: vmem(r2+#0) = v2
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: jumpr r31
; CHECK-NEXT: }
b0:
%v0 = load <32 x i32>, ptr %a0, align 128
%v1 = load <32 x i32>, ptr %a1, align 128
%v2 = sext <32 x i32> %v0 to <32 x i64>
%v3 = sext <32 x i32> %v1 to <32 x i64>
%v4 = mul nsw <32 x i64> %v2, %v3
%v5 = add nsw <32 x i64> %v4, <i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824, i64 1073741824>
%v6 = lshr <32 x i64> %v5, <i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31>
%v7 = trunc <32 x i64> %v6 to <32 x i32>
store <32 x i32> %v7, ptr %a2, align 128
ret void
}
define void @f2(ptr %a0, ptr %a1, ptr %a2) #0 {
; CHECK-LABEL: f2:
; CHECK: // %bb.0: // %b0
; CHECK-NEXT: {
; CHECK-NEXT: v0 = vmem(r1+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: r7 = #124
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: r3 = #15
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1 = vmem(r0+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1:0.w = vmpy(v0.h,v1.h)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1:0 = vshuff(v1,v0,r7)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v0.uw = vlsr(v0.uw,r3)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1.uw = vlsr(v1.uw,r3)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v0.h = vpacke(v1.w,v0.w)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: vmem(r2+#0) = v0
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: jumpr r31
; CHECK-NEXT: }
b0:
%v0 = load <64 x i16>, ptr %a0, align 128
%v1 = load <64 x i16>, ptr %a1, align 128
%v2 = sext <64 x i16> %v0 to <64 x i32>
%v3 = sext <64 x i16> %v1 to <64 x i32>
%v4 = mul nsw <64 x i32> %v2, %v3
%v5 = lshr <64 x i32> %v4, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
%v6 = trunc <64 x i32> %v5 to <64 x i16>
store <64 x i16> %v6, ptr %a2, align 128
ret void
}
define void @f3(ptr %a0, ptr %a1, ptr %a2) #0 {
; CHECK-LABEL: f3:
; CHECK: // %bb.0: // %b0
; CHECK-NEXT: {
; CHECK-NEXT: v0 = vmem(r0+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v1 = vmem(r1+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: v0.h = vmpy(v0.h,v1.h):<<1:rnd:sat
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: vmem(r2+#0) = v0
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: jumpr r31
; CHECK-NEXT: }
b0:
%v0 = load <64 x i16>, ptr %a0, align 128
%v1 = load <64 x i16>, ptr %a1, align 128
%v2 = sext <64 x i16> %v0 to <64 x i32>
%v3 = sext <64 x i16> %v1 to <64 x i32>
%v4 = mul nsw <64 x i32> %v2, %v3
%v5 = add nsw <64 x i32> %v4, <i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384, i32 16384>
%v6 = lshr <64 x i32> %v5, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
%v7 = trunc <64 x i32> %v6 to <64 x i16>
store <64 x i16> %v7, ptr %a2, align 128
ret void
}
attributes #0 = { nounwind "target-features"="+v68,+hvxv68,+hvx-length128b,-packets" }
|
