Intro to Threads

Notes

• I needed threads to make some demos a little more palatable.
• This is (part of) chapter 4 of the book.
• A thread is a lightweight process
• Originally process threads
  • I knew my process would not use the entire timeslice, but I had multiple things to do
  • So I divide my process into tasks, or threads
  • And manage the scheduling myself.
  • Everything is manage in user space, not in kernel space
  • These are called user threads
• Later this was migrated to the OS.
• Today a thread is the lowest or basic unit of execution.
• Threads are all started, and live within a single process.
• Threads share:
  • code segment
  • data segment (global variables)
  • heap
  • Many things like files, signal masks, ...
• Threads do not share:
  • registers
  • Stack
• Threads can execute concurrently
• The main advantages of threads over processes
  • Automatically shared memory via global variables and memory allocated on the heap.
  • It is much easier, and faster to create a thread than a new process.
  • Since they share memory, threads are less resource intensive
  • Threads may be more responsive than processes, again a resource thing.
• One must be extremely careful however, threads share memory so ...
  • Many traditional unix calls are not reentrant, or can not be run safely by multiple threads.
  • int rand(void) is an example
    • Rand is normally X_n+1 = (aX_n+c)%m
    • Where X is stored in a shared memory location.
    • So we have a problem when multiple threads try to access X at the same time.
  • int rand_r(unsigned int * seedp) is the reentrant version
  • This is true of many functions.
    • Most are function_r
    • And require you to pass the "state" variable that is stored globally.
• This is enought to get us stared, we will look at more details with threads later.