EC-9.6 Concurrency, Real Time and the RTOS

Work through these

• The super-loop, and where it breaks

  One loop calling each piece of work in turn is predictable and open to inspection. It breaks when one piece of work is slow and another cannot wait for it, which is a scheduling problem rather than a speed problem.

  NPTEL: Embedded Systems Design (IIT Kharagpur) · Course

• Shared data between a handler and the main flow

  A variable written by a handler and read by the main flow can be read half-updated if it takes more than one instruction. The usual fixes are a single-byte flag, a short section with interrupts held off, or a ring buffer.

  NPTEL: Embedded Systems Design (IIT Kharagpur) · Course

• What real time actually means

  A real-time requirement names a deadline and what happens if it is missed. Hard means a missed deadline is a failure of the system, soft means the result is merely degraded, and most systems contain both kinds.

  NPTEL: Introduction to Embedded System Design (NSUT and IIT Jammu) · Course

• Tasks, the scheduler and the context switch

  An operating system gives each piece of work its own stack and saves and restores registers to swap between them. That swap has a cost in time and memory, and it is what buys you independent pieces of work.

  NPTEL: Embedded System Design with ARM (IIT Kharagpur) · Course

• Semaphores, mutexes and priority inversion

  A mutex protects shared data by making other work wait. If low-priority work holds a mutex that urgent work needs, the urgent work is stuck behind it, which is priority inversion and is why inheritance schemes exist.

  NPTEL: Embedded System Design with ARM (IIT Kharagpur) · Course

• When an operating system is the wrong answer

  It costs memory, adds timing that is harder to predict, and brings a new set of bugs. On a part with two kilobytes of working memory and one job to do, a loop and an interrupt are the better engineering.

  NPTEL: Embedded Systems Design (IIT Kharagpur) · Course