Threads & Concurrency

1. Multicore Programming

1.1. Concurrency supports more than one task making progress

1.2. Parallelism implies a system can perform more than one task simultaneously

1.2.1. device drivers

1.2.2. file systems

1.2.3. virtual memory manager

1.2.4. windowing system

1.2.5. security services

1.3. Multicore or multiprocessor systems putting pressure on programmers, challenges include:

1.3.1. Dividing activities

1.3.2. Balance

1.3.3. Data splitting

1.3.4. Data dependency

2. Implicit Threading

2.1. Growing in popularity as numbers of threads increase, program correctness more difficult with explicit threads

2.2. Creation and management of threads done by compilers and run-time libraries rather than programmers

2.2.1. OpenMP

2.3. Five methods explored

2.3.1. Fork-Join

2.3.2. Thread Pools

3. Scheduler Activations

4. Both M:M and Two-level models require communication to maintain the appropriate number of kernel threads allocated to the application

5. Scheduler activations provide upcalls - a communication mechanism from the kernel to the upcall handler in the thread library

6. This communication allows an application to maintain the correct number kernel threads

7. Multithreading Models

8. Many-to-One

9. One-to-One

10. Many-to-Many

11. User Threads and Kernel Threads

12. User threads - management done by user- level threads library

13. Three primary thread libraries

14. Kernel threads - Supported by the Kernel

15. Motivation

15.1. Multiple tasks with the application can be implemented by separate threads

15.1.1. Update display

15.1.2. Fetch data

15.1.3. Spell checking

15.1.4. Answer a network request

15.2. Can simplify code, increase efficiency

15.3. Kernels are generally multithreaded

15.4. Most modern applications are multithreaded

16. Benefits

16.1. Scalability – process can take advantage of multicore architectures

16.2. Responsiveness – may allow continued execution if part of process is blocked, especially important for user interfaces

16.3. Resource Sharing – threads share resources of process, easier than shared memory or message passing

16.4. Economy – cheaper than process creation, thread switching lower overhead than context switching

17. System Calls

17.1. provide interface between a running program and the OS.

17.2. types of system calls

17.2.1. file management

17.2.2. deivce management

17.2.3. information maintenance

17.2.4. communications