IV. Multi-Threading Basics

A. Background

• Multithreaded code allows concurrency of execution w/o “heavy weight” multiple processes. 
• Multiple threads are swapped on and off cpu, or run concurrently on multiple-cpu boxes.
• An app's threads share same process resources – 
  • file handles, 
  • heap memory space, 
  • the resources of the JVM. 
• Each thread has a program counter and stack memory for local data (non-static).
• Synchronization of access to resources and code is necessary. 
  • Synchronization can be thought of as “single-file” access … one at a time.
• Another synchronization concept: thread shutdown. 
  • Controlling thread joins with thread it's stopped - insures proper completion.

B. Classic Concurrency Problem

• The statement k++ is not thread-safe (depending on k). Consider the machine-language implementation –
  • fetch k from memory to register
  • increment k in register
  • store k from register to memory.
• Now, consider two threads, each accessing k - there is a timing issue.

                

• The result of above will be as if only one iteration occurred.
• Solution: make sure T1 is not interrupted between time 1 and time 5.