B3CC: Concurrency01: Introduction

Trevor L. McDonell

Utrecht University, B2 2019-2020

Today

1. Introduction

2. What is concurrency?

2

Course structure

3

henlo

• Trevor L. McDonell

- BBG 5.68

- t.l.mcdonell@uu.nl mention [INFOB3CC] in the subject

• Office hours (may change)

- Mon 13:00 - 15:00

4ig: jasper.samoyed

Topics

• Program a multithreaded application

- Managing threads

- Synchronise with locks, etc.

- Software transactional memory

- Parallelism

• Work & span

• Concurrent algorithms / data structures

6

Goals

• By the end of the course you should be able to:

- Design and implement a multithreaded application

- Understand the difference between concurrency and parallelism

- Reason about the properties/complexity of parallel algorithms

7

Haskell

8https://xkcd.com/1312/

Homepage

• http://www.cs.uu.nl/docs/vakken/b3cc/

- Feel free to let me know if there are broken links, missing slides, etc.

9

Resources

• Parallel and Concurrent Programming in Haskellhttps://simonmar.github.io/pages/pcph.html

• #infob3cc slack channel

• Many more on the website http://www.cs.uu.nl/docs/vakken/b3cc/resources.html

10

Sessions

• Lectures:

- Tue: 13:15 - 15:00 (KBG Cosmos)

- Thu: 9:00 - 10:45 (KBG Cosmos)

• Work collage:

- Tue: 15:15 - 17:00

- Thu: 11:00 - 12:45

• Participation in lectures is expected (please ask questions!)

11

Course components

• Exam (50%)

- Mid session exam: 19-12-2019 (40%) (EDUC Gamma 8:30 - 10:30)

- Final exam: 30-01-2020 (60%) (EDUC Gamma 8:30 - 10:30)

• Practicals (50%)

- Assignment 1: 30-11-2019 (individual) (20%)

- Assignment 2: 21-12-2019 (may change!) (in pairs) (40%)

- Assignment 3: 01-02-2020 (may change!) (in pairs) (40%)

12

If you wait until the last minute,

it only takes a minute to do.

— Key Skills for Professionals, 2013, pp65

Software installation

• A recent version of GHC

• I recommend using stack

- https://docs.haskellstack.org/en/stable/README/

13

Concurrency control

14

What is concurrency?

• Consider multiple tasks being executed by the computer…

- Tasks are concurrent with respect to each other if:

• They may be executed out-of-order

• Implies they can be executed at the same time, but this is not required

- Concurrency: deal with lots of things at once

15

What is parallelism?

• Consider multiple tasks being executed by the computer…

- Tasks are parallel if they are executed simultaneously:

• Requires multiple processing elements

• The primary motivation for parallel programming is to reduce the overall running time (wall clock) of the program: parallel execution

- Parallelism: do lots of things at once

16

Question

• What does it mean for an application to be concurrent but not parallel?

- Give an example

• What does it mean for an application to be parallel but not concurrent?

- Give an example

17

Concurrency vs. Parallelism

• Concurrency: composition of independently executing processes

• Parallelism: simultaneous execution of (possibly related) computations

18

Concurrency

• Programming with multiple threads of control

- A tool for structuring programs with multiple interactions

- Examples: GUI, web server, different tasks in a game engine loop, …

• There is no single right answer. In this course we will discuss several approaches: it is up to you to pick which is right for your application

19

Concurrency control

• Concurrency appears in many contexts:

- Multithreading: concurrent threads share an address space

- Multiprogramming: concurrent processes execute on a uniprocessor

- Multiprocessing: concurrent processes on a multiprocessor

- Distributed processing: concurrent processes executing on multiple nodes connected by a network

• Concurrency is also used in different forms:

- Multiple applications (multiprogramming)

- Structured applications (applications is a set of threads/processes)

- Operating system structure (OS is a set of threads/processes)20

Concurrency control

• Concurrent processes (threads) need special support

- Communication among processes

- Allocation of processor time

- Sharing of resources

- Synchronisation of multiple processes

• Concurrency can be dangerous to the unwary programmer:

- Sharing global resources (order of read & write operations)

- Management of allocation of resources (danger of deadlock)

- Programming errors are difficult to locate (Heisenbugs)21

Example: access to a global queue

• Inserting:

22

head last

Example: access to a global queue

• Inserting:

- Create new object

23

head last

Example: access to a global queue

• Inserting:

- Create new object

- Set last!->next to &new

24

head last

Example: access to a global queue

• Inserting:

- Create new object

- Set last!->next to &new

- Set last to &new

25

head last

Example: concurrent access to a global queue

• Thread A: • Thread B:

26

head last

Example: concurrent access to a global queue

• Thread A:

- Create new object

• Thread B:

27

head last

Example: concurrent access to a global queue

• Thread A:

- Create new object- Set last!->next to &new

• Thread B:

28

head last

Example: concurrent access to a global queue

• Thread A:

- Create new object- Set last!->next to &new

• Thread B:

- Create new object

29

head last

Example: concurrent access to a global queue

• Thread A:

- Create new object- Set last!->next to &new

• Thread B:

- Create new object- Set last!->next to &new

30

head last

Example: concurrent access to a global queue

• Thread A:

- Create new object- Set last!->next to &new

• Thread B:

- Create new object- Set last!->next to &new- Set last to &new

31

head last

Example: concurrent access to a global queue

• Thread A:

- Create new object- Set last!->next to &new

- Set last to &new

• Thread B:

- Create new object- Set last!->next to &new- Set last to &new

32

head last

Example: concurrent access to a global queue

• Lessons learned

- We have to control access to shared resources (such as shared variables)

- We can do this by controlling access to the code utilising those shared resources: critical sections

33

head last

Example: concurrent access to a global queue

• Only one thread at a time should have access to the queue:

- Thread A creates a new object, sets last!->next pointer

- Thread A is suspended

- Thread B is scheduled: since Thread A is currently in insert, has to wait

- Thread A is resumed, the data structure is in the same state as it was when it was suspended

- Thread A completes operation

- Thread B is allowed to execute insert

34

Concurrency control

• Processes can

- Compete for resources

• Processes may not be aware of each other

• Execute must not be affected by each other

• OS is responsible for controlling access

- Cooperate by sharing a common resource

• Programmer responsible for controlling access

• Hardware / OS / programming language may provide support

• Threads of a process usually do not compete, but cooperate35

Concurrency control

• We face three control problems:

- Mutual exclusion: critical resources !=> critical sections

• Only one process at a time is allowed in a critical section

• e.g. only one process at a time is allowed to send commands to the GPU

- Deadlock: e.g. two processes and two resources

- Starvation: e.g. three processes compete for a single resource

36

Requirements for mutual exclusion

• Implementation:

- Only one thread at a time is allowed in the critical section for a resource

- No deadlock or starvation

- A thread must not be delayed access to a critical section when there is no other thread using it

- A thread that halts in its non-critical section must do so without interfering with other threads

- No assumptions made about relative thread speed or number of processes

• Usage

- A thread remains inside its critical section for a finite time only

- No potentially blocking operations should be executed inside the critical section

- No deadlock or starvation 37

Mutual exclusion

• Three ways to satisfy the implementation requirements:

- Software approach: put responsibility on the processes themselves

- Systems approach: provide support within the OS or programming language

- Hardware approach: special-purpose machine instructions

38

Therac-25

• Computer controlled medical radiation device (1982)

- Two operating modes: a low-current electron beam; or high-energy x-rays

• Involved in at least six incidents, resulting in serious injury or death

- A race condition could cause the high-power electron beam to be administered directly to the patient

- Resulted in radiation doses 100x higher than normal

- Additional problems related to poor software development practices

39https://en.wikipedia.org/wiki/Therac-25

Northeast blackout (2003)

• Widespread power outage throughout USA and Canada

- Second most widespread blackout in history (at the time)

- Affected an estimated 10M people in Ontario and 45M in 8 US states

- A race condition prevented an alarm from going off

- Operators were unaware of the need to redistribute power—a minor problem—which cascaded into complete collapse of the electrical grid

40https://en.wikipedia.org/wiki/Northeast_blackout_of_2003

Mars Pathfinder

• Launched by NASA in 1996

• Sojourner became the first rover to operate outside the Earth-Moon system

• Control computer contained a priority inversion bug

- Triggered under certain high loads causing a system reset

- Successfully patched remotely

41https://en.wikipedia.org/wiki/Mars_Pathfinder

Next time…

• Haskell refresh

• Make sure you have access to a modern Haskell installation

42

Photo by Jf Brou

tot ziens