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
|
# luci-profile
`luci-profile` provides profiling related items.
## CircleNodeOrigin
`CircleNodeOrigin` allow us know where some node is originated from.
Let's assume following graph transformations are done.
```
| | |
[node1] --------+ | |
(id = 1) | | |
| +--------> [node5] ----------------> [node6]
| | (origin = [1,2]) (origin = [1,2])
[node2] --------+ | |
(id = 2) | |
| | |
[node3] -----------------> [node3] --------+-------> [node3]
(id = 3) (origin = [3]) | (origin = [3,4])
| | | |
[node4] -----------------> [node4] --------+ |
(id = 4) (origin = [4]) |
| | |
<Circle1> -- optimizer --> <circle2> -- quantizer --> <circle3>
```
The most important purpose of using `CircleNodeOrigin` is preserving origin information.
Following changes show how origin information is preserved even after graph is transformed.
- `node3`
- `node4` is absorbed to **existing** `node3`.
- origin of `node4` is absorbed to origin of `node3`.
- `node5`
- `node1` and `node2` are fused to **newly created** `node5`.
- origin of `node1` and `node2` are inherited to origin of `node4`.
- `node6`
- `node5` is **replaced with newly created** `node6`.
- origin of `node5` is copied to origin of `node6`.
**Therefore, when using `CircleNodeOrigin`, please aware of the most important principle. "Preserve origin information"**
Next items are about implementation details to store the origin information.
### Source Table
Source table includes a set of id and name of origin node.
#### Binary format
```
[ entry_number : uint32_t ]
[ id : uint32_t ][ length : uint32_t ][ data : char * length ] * entry_number
```
- entry_number : The number of entries
- Each entry consists of id, length, and data.
- id : ID of origin node
- length : Length of data
- data : Name of origin node **(null-terminated string)**
#### In-memory format
```cpp
// size = entry_number
std::map<uint32_t /* id */, std::string /* name */>
```
#### Example
Following example means "Name of origin 1 is node1".
```
[Binary Format]
0x01 00 00 00 0x01 00 00 00 0x06 00 00 00 0x6e 0x6f 0x64 0x65 0x31 00
------------- ------------- ------------- ---- ---- ---- ---- ---- ----
entry_number=1 id=1 length=6 'n' 'o' 'd' 'e' '1' '\0'
```
```cpp
[In-memory Format]
std::map<uint32_t, std::string>({1, "node1"});
```
### Op Table
Op table includes a set of id of operation and id(s) of operation's origin nodes.
#### Binary format
Op table is stored in circle file as binary with following format.
```
[ entry_number : uint32_t ]
[ id : uint32_t ][ node_num : uint32_t ][ node_ids : uint32_t * node_num ] * entry_number
```
- entry_number : The number of entries
- Each entry consists of id, node_num, and node_ids.
- id : ID of operation in circle model file
- node_num : The number of operation's origin nodes
- node_ids : Set of IDs of origin nodes
#### In-memory format
```cpp
std::map<uint32_t /* id */, std::set<uint32_t> /* node_ids */>
```
#### Example
Following example means "Operation 5 is originated from origin 1 and origin 2".
```
[Binary Format]
0x01 00 00 00 0x05 00 00 00 0x02 00 00 00 0x01 00 00 00 0x02 00 00 00
------------- ------------- ------------- ---------------------------
entry_number=1 id=5 node_num=2 node_ids : 1, 2
```
```cpp
[In-memory Format]
std::map<uint32_t, std::set<uint32_t>>({5, std::set{1, 2}});
```
|
