Welcome!
Operating Systems
Parallelism, concurrency in OS and Hardware

Parallelism, concurrency in OS and Hardware

  flowchart TD

  

subgraph Hardware

    PROC[Processors] --> CORE[Cores per Processor] --> SMT[Hardware Threads]

    SMT --> ILP[Instruction-Level Parallelism /pipeline, superscalar/]

end

  

subgraph OS_Abstractions

    PROG[Program] --> MP

    PROG --> MT

    MP[Multiprogramming] --> PROCOS[Processes] --> THR[Threads]

    MT[Multitasking] --> PROCOS

    THR --> TLP[Thread-Level Parallelism]

end

  

PROCOS -. scheduled on .-> SMT

THR -. scheduled on .-> SMT

TLP -. mapped across .-> SMT

Hardware side:

  • A machine may have multiple processors, each with multiple cores.

  • Each core can support multiple hardware threads (SMT, hyperthreading).

  • Inside a core, instruction-level parallelism (ILP) executes several instructions at once via pipelining or superscalar issue.

OS side:

  • Multiprogramming: many processes are kept in memory. When one blocks on I/O, another is scheduled, ensuring the CPU is not idle.

  • Multitasking (time-sharing): CPU time is sliced among processes, with rapid context switching to give the illusion of simultaneous execution.

  • Each process can contain multiple threads, which are lighter units of execution.

  • Thread-Level Parallelism (TLP): threads from the same process or from different processes can run concurrently, exploiting available cores and hardware threads.

Link between them:

  • The OS scheduler maps both processes and threads onto hardware threads (SMT slots) across cores.

  • Multiprogramming ensures long-term CPU utilization, multitasking ensures short-term responsiveness, and TLP boosts throughput by running many threads in parallel.

  • Hardware provides the substrate: processors, cores, hardware threads, and ILP inside each core.

Processes ERD