About
机器学习
1.
PyTorch
❱
1.1.
pytorch pocket reference 笔记
1.2.
mini torch
2.
Tengine
3.
NLP
❱
3.1.
Bert
4.
nncase
❱
4.1.
comile流程
4.2.
k210 standalone sdk
4.3.
face detect example
5.
Ray
❱
5.1.
Ray Paper
5.2.
Ray Remote
6.
Tensorflow
❱
6.1.
tensorflow inside
❱
6.1.1.
Executor: 执行Computation Sub Graph
❱
6.1.1.1.
SubGraph预处理：Node/NodeItem/TaggedNode
6.1.1.2.
Flow control op: switch/merge/enter/exit/nextIteration
6.1.1.3.
Frame: ControlFlowInfo/FrameInfo/FrameState/IterationState
6.1.2.
DirectSession: 单机执行computation graph
6.1.3.
RendezVous：跨设备，跨主机通信
6.1.4.
Device：计算单元抽象(CPU/GPU)
6.2.
tensorflow模型对接工程优化
❱
6.2.1.
将keras模型导出为tf frozen graph
6.2.2.
使用dataset iterator 优化keras model预测的吞吐量
6.2.3.
统计cpu/gpu 负载率脚本
7.
pthread
❱
7.1.
Pthread Primer笔记
7.2.
Pthread线程生命周期
7.3.
Pthread线程同步
Rust
8.
Yew
9.
Axum
10.
Hyper
11.
tokio
❱
11.1.
Executor
❱
11.1.1.
Park
11.1.2.
thread pool
11.2.
driver
11.3.
io
11.4.
codec
11.5.
channel
11.6.
waker
12.
crossbeam
❱
12.1.
SkipList(draft)
13.
draft Rust
GO
14.
Go
❱
14.1.
Runtime PGM调度模型
14.2.
Goroutine Stack
14.3.
Memory分配
14.4.
GC
14.5.
Context
14.6.
defer-panic-recover
数据库
15.
LevelDB
❱
15.1.
draft
15.2.
代码模块间关系
15.3.
Write 流程
15.4.
Read 流程
15.5.
SSTable 文件格式和读写
15.6.
versionset和Manifest
15.7.
Do Compact
15.8.
Iterator迭代器
15.9.
Bloom filter
16.
RocksDB
❱
16.1.
draft
16.2.
主要struct引用关系
16.3.
Write Ahead Log
16.4.
write 并发控制
16.5.
后台flush和compact线程
16.6.
Leveled Compaction Picker
16.7.
read 流程
16.8.
Blob
16.9.
事务
❱
16.9.1.
Optimistic Transaction
16.9.2.
Transaction lock mgr
16.9.3.
two phase commit
17.
ClickHouse(daft)
❱
17.1.
Server main
17.2.
block
17.3.
blockio
17.4.
Storage
❱
17.4.1.
MergeTreeData
17.4.2.
StorageMergeTree
PingCAP
18.
TiDB
❱
18.1.
tidb学习资料整理
18.2.
Server Main Loop
18.3.
Insert 语句
18.4.
Select 语句
18.5.
数据类型
18.6.
expression
18.7.
Online DDL
❱
18.7.1.
Schema 存储
18.7.2.
Schema Cache和加载
18.7.3.
Schema Modification
18.7.4.
Online Schema Change
18.8.
统计信息
❱
18.8.1.
基本概念
18.8.2.
stats tables
18.8.3.
Analyze
18.8.4.
Query Feedback
18.8.5.
统计信息使用场景
18.9.
Logical Optimize
18.10.
Physical Optimize
18.11.
DataSource
❱
18.11.1.
buildDataSource
18.11.2.
索引范围计算
18.11.3.
table/index存储编码
18.11.4.
paritionProcessor
18.11.5.
PredicatePushDown
18.11.6.
Physical Optimize
18.11.7.
Executors
18.12.
DistSQL
❱
18.12.1.
ReginCache
18.12.2.
TiKV GRPC Client
18.12.3.
CopTask
18.12.4.
CopIteratorWorker
18.12.5.
Coprocessor
18.13.
Join
❱
18.13.1.
Join算法
18.13.2.
Logical Optimize
18.13.3.
Physical Optimize
18.13.4.
Executor: Hash Join
18.13.5.
Executor: Merge Join
18.13.6.
Executor: Index Lookup Join
18.14.
Agg
❱
18.14.1.
AggFunc
18.14.2.
Executor: HashAgg
18.14.3.
Executor: StreamAgg
19.
PD
20.
TiKV
❱
20.1.
raft-rs
❱
20.1.1.
RawNode
20.1.2.
Storage
20.1.3.
ProgressTracker
20.1.4.
Hearbeat
20.1.5.
Election
20.1.6.
LogEntry
20.1.7.
Snapshot
20.1.8.
ConfChange
20.1.9.
ReadIndex
20.2.
RaftKV
❱
20.2.1.
BatchSystem
20.2.2.
RaftMessage
20.2.3.
Raft Client
20.2.4.
Region
20.2.5.
PeerStorage
20.2.6.
Thread Local Engine
20.2.7.
Leader Lease
20.2.8.
Read Index
20.2.9.
Async Snapshot
20.2.10.
Async Write
20.2.11.
Region Epoch
20.2.12.
Conf Change
20.2.13.
Split Region
20.2.14.
Merge Region
20.3.
Storage
❱
20.3.1.
Percolator
20.3.2.
Percolator In TiKV
20.3.3.
2PC
20.3.4.
AsyncCommit
20.3.5.
OnePC
20.3.6.
悲观事务
20.3.7.
Resolve Lock
20.3.8.
Scheduler
20.3.9.
Wait Lock
20.3.10.
死锁检测
20.3.11.
TiDB分组提交
20.3.12.
Scanner
20.4.
Coprocessor
❱
20.4.1.
RpnExpression
20.4.2.
AggrFunction
20.4.3.
BatchExecutor
❱
20.4.3.1.
RangesScanner
20.4.3.2.
ScanExecutor
20.4.3.3.
Selection
20.4.3.4.
AggExecutor
20.5.
Performance(draft)
20.6.
yatp
21.
Bevy
❱
21.1.
bevy draft
22.
Blender
❱
22.1.
blender manual notes
23.
Trash
❱
23.1.
python
❱
23.1.1.
records
23.2.
react
❱
23.2.1.
从jsx到html dom的流程分析
23.3.
Godot
❱
23.3.1.
godot 学习笔记
23.3.2.
gdquest tutorial
23.4.
Kafka
❱
23.4.1.
client: producer
23.4.2.
group coordinator
23.4.3.
produce and fetch
23.4.4.
log
23.4.5.
Partition
23.4.6.
Controller
❱
23.4.6.1.
通信管理 channelManager
23.4.6.2.
选举
23.4.6.3.
zk监听处理
23.4.7.
副本迁移
23.4.8.
Partition/Replica状态机
23.4.9.
事务
23.4.10.
Stream
23.5.
hotspot
❱
23.5.1.
osx下编译调试hotspot
23.5.2.
jvm的初始化时创建的线程
23.5.3.
class文件的加载和执行
23.6.
Paper notes
❱
23.6.1.
raft paper
23.6.2.
draft: gfs
23.6.3.
draft: Bw-tree(draft)
Light (default)
Rust
Coal
Navy
Ayu
blog
Ray
相关资料整理
官方doc
github
ray paper
主要组件
Gcs: Global Control State, 存储了代码，输入参数，返回值
Raylet:Local Scheduler, Worker通过Raylet和Gcs通信。
Redis