并发编程：理论，框架与实践

应在宽邀请为其团队分享该topic

为什么需要并发，又为什么需要并发控制？
- 充分利用某些资源
- 避免状态破坏
- more？！
资源并发访问的策略
- 悲观策略
  - lock based concurrency(theory)
  - java.util.concurrent (framework)
  - 锁还是不锁，这是个问题，锁多还是锁少，也是个问题(practice)
- 乐观策略
  - lock free concurrency(theory) : CAS
  - disruptor (framework)
  - 并发度高，还是并发度低的时候使用，这是个问题(practice)
我拆我拆我拆拆拆
- task-based concurrency (theory)
  - Runnable | Callable(model)
  - Executor|ExecutorService(framework)
- data-based concurrency(theory)
  - Actor (model)
  - Akka(framework)
从个人到群体，从单机到分布式
- divide and conquer
  - map reduce pattern
  - master-worker pattern
- swarm framework(move computation instead of data)
从软件到硬件
- GPU
  - CUDA, jcuda, scuda
  - floating point computation, e.g. image reader and processing
- PPU
  - 物理计算
want to know more ?
- agent
- STM
- data flow