Skip to content

Commit bdc2011

Browse files
0xrinegade0xrinegade
authored
Merge pull request #5 from openSVM/openSVM_lessvm_issue_4_d999e602
fix bugs (Run ID: openSVM_lessvm_issue_4_d999e602)
2 parents 1fc3353 + b6b1b5e commit bdc2011

File tree

9 files changed

+1288
-7
lines changed

9 files changed

+1288
-7
lines changed

‎docs/implementation_details.md‎

Lines changed: 186 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,186 @@
1+
# LessVM Implementation Details
2+
3+
This document provides details about the implementation of LessVM, focusing on recent changes and optimizations.
4+
5+
## Table of Contents
6+
7+
- [SIMD Vector Addition](#simd-vector-addition)
8+
- [Data Structure Store](#data-structure-store)
9+
- [Memory Management](#memory-management)
10+
- [Opcode Implementations](#opcode-implementations)
11+
- [Solana Account Operations](#solana-account-operations)
12+
- [Graph Operations](#graph-operations)
13+
- [OHLCV Operations](#ohlcv-operations)
14+
- [Hypergraph Operations](#hypergraph-operations)
15+
16+
## SIMD Vector Addition
17+
18+
The `vector_add` function uses SIMD (Single Instruction, Multiple Data) instructions to efficiently add multiple values at once. The implementation loads two vectors from the stack and adds them together.
19+
20+
```mermaid
21+
sequenceDiagram
22+
participant VM
23+
VM->>VM: _mm256_loadu_si256(stack[top] to values2)
24+
VM->>VM: stack.pop() x 4
25+
VM->>VM: _mm256_loadu_si256(stack[top] to values1)
26+
VM->>VM: result = _mm256_add_epi64(values1, values2)
27+
VM->>VM: stack.push(result) x 4
28+
```
29+
30+
The function works as follows:
31+
1. Load the second vector (values2) from the top of the stack
32+
2. Pop the first 4 values from the stack
33+
3. Load the first vector (values1) from the new top of the stack
34+
4. Add the two vectors together using SIMD instructions
35+
5. Push the result back onto the stack
36+
37+
This approach ensures that we're adding two different vectors together, rather than adding a vector to itself.
38+
39+
## Data Structure Store
40+
41+
The `DataStructureStore` manages various data structures used by the VM, including BTreeMaps, Tries, Graphs, OHLCV, and Hypergraphs.
42+
43+
```mermaid
44+
classDiagram
45+
class DataStructureStore {
46+
btrees: Vec<Option<BTreeMapDS>>
47+
tries: Vec<Option<TrieDS>>
48+
graphs: Vec<Option<GraphDS>>
49+
ohlcvs: Vec<Option<OHLCVDS>>
50+
hypergraphs: Vec<Option<HypergraphDS>>
51+
+new() DataStructureStore
52+
+ensure_capacity(ds_type: DataStructureType, id: usize) void
53+
}
54+
note for DataStructureStore "Stores different types of data structures."
55+
```
56+
57+
The `ensure_capacity` method ensures that the vectors have enough capacity to store a data structure at a specific index. If the index is beyond the current length of the vector, the vector is resized to accommodate the new index.
58+
59+
```mermaid
60+
sequenceDiagram
61+
participant VM
62+
participant DataStructureStore
63+
VM->>DataStructureStore: ensure_capacity(DataStructureType::BTreeMap, 5)
64+
alt Vector length < 6
65+
DataStructureStore->>DataStructureStore: resize_with(6, || None)
66+
else Vector length >= 6
67+
DataStructureStore->>DataStructureStore: Do nothing
68+
end
69+
```
70+
71+
## Memory Management
72+
73+
The memory management system has been optimized to use a more efficient growth strategy. Instead of doubling the capacity when more space is needed, the new implementation grows the memory by 50% or to the required size, whichever is larger.
74+
75+
```mermaid
76+
sequenceDiagram
77+
participant Memory
78+
participant Vec
79+
Memory->>Memory: ensure_capacity(required_size)
80+
alt required_size > data.len()
81+
Memory->>Memory: new_capacity = (data.len() * 3 / 2).max(required_size)
82+
Memory->>Vec: resize(new_capacity, 0)
83+
else required_size <= data.len()
84+
Memory->>Memory: Do nothing
85+
end
86+
```
87+
88+
This approach reduces memory waste while still providing amortized constant-time operations.
89+
90+
## Opcode Implementations
91+
92+
### Solana Account Operations
93+
94+
The following Solana account operations have been implemented:
95+
96+
- `GetBalance`: Gets the balance (lamports) of an account
97+
- `GetOwner`: Gets the owner of an account
98+
- `IsWritable`: Checks if an account is writable
99+
- `IsSigner`: Checks if an account is a signer
100+
101+
```mermaid
102+
sequenceDiagram
103+
participant VM
104+
participant Account
105+
VM->>VM: Pop account_idx from stack
106+
VM->>VM: Check if account_idx is valid
107+
alt Valid account_idx
108+
VM->>Account: Get account information
109+
Account->>VM: Return account information
110+
VM->>VM: Push result to stack
111+
else Invalid account_idx
112+
VM->>VM: Return InvalidAccount error
113+
end
114+
```
115+
116+
### Graph Operations
117+
118+
The following graph operations have been implemented:
119+
120+
- `GraphAddEdge`: Adds an edge to a graph
121+
- `GraphGetNode`: Gets the value of a node
122+
- `GraphSetNode`: Sets the value of a node
123+
- `GraphGetNeighbors`: Gets the neighbors of a node
124+
- `GraphBfs`: Performs a breadth-first search starting from a node
125+
- `GraphClear`: Clears a graph
126+
127+
```mermaid
128+
sequenceDiagram
129+
participant VM
130+
participant DataStructureStore
131+
participant GraphDS
132+
VM->>VM: Pop weight, to, from, and id from stack
133+
VM->>DataStructureStore: Access graph with id
134+
alt Graph exists
135+
DataStructureStore->>GraphDS: add_edge(from, to, weight)
136+
GraphDS-->>DataStructureStore: Result
137+
else Graph does not exist
138+
DataStructureStore-->>VM: Error: InvalidDataStructureOperation
139+
end
140+
```
141+
142+
### OHLCV Operations
143+
144+
The following OHLCV (Open-High-Low-Close-Volume) operations have been implemented:
145+
146+
- `OhlcvAddBar`: Adds a bar to an OHLCV
147+
- `OhlcvGetBar`: Gets a bar from an OHLCV
148+
- `OhlcvSma`: Calculates the Simple Moving Average (SMA) of an OHLCV
149+
150+
```mermaid
151+
sequenceDiagram
152+
participant VM
153+
participant DataStructureStore
154+
participant OHLCVDS
155+
VM->>VM: Pop timestamp, open, high, low, close, volume, and id from stack
156+
VM->>DataStructureStore: Access ohlcv with id
157+
alt OHLCV exists
158+
DataStructureStore->>OHLCVDS: add_entry(timestamp, open, high, low, close, volume)
159+
OHLCVDS-->>DataStructureStore: Result
160+
else OHLCV does not exist
161+
DataStructureStore-->>VM: Error: InvalidDataStructureOperation
162+
end
163+
```
164+
165+
### Hypergraph Operations
166+
167+
The following hypergraph operations have been implemented:
168+
169+
- `HyperAddNode`: Adds a node to a hypergraph
170+
- `HyperAddEdge`: Adds an edge to a hypergraph
171+
- `HyperAddNodeToEdge`: Adds a node to an edge in a hypergraph
172+
173+
```mermaid
174+
sequenceDiagram
175+
participant VM
176+
participant DataStructureStore
177+
participant HypergraphDS
178+
VM->>VM: Pop node_id, edge_id, and id from stack
179+
VM->>DataStructureStore: Access hypergraph with id
180+
alt Hypergraph exists
181+
DataStructureStore->>HypergraphDS: add_node_to_edge(edge_id, node_id)
182+
HypergraphDS-->>DataStructureStore: Result
183+
else Hypergraph does not exist
184+
DataStructureStore-->>VM: Error: InvalidDataStructureOperation
185+
end
186+
```

0 commit comments

Comments
 (0)