Multithreading Programming in C++ Training Course

Concurrency challenges and issues

• Hardware platform limitations - hardware threads, pipelines, cache coherence
• Elementary concepts related to multithreaded programming (correctness of multithreaded programs, liveness, starvation, deadlock, race conditions, atomicity, memory models, Flynn's taxonomy)
• Available mechanisms in C++ for implementing multithreading and multiprocessing
• Implementation of the standard library - elements from C++14 to the present day (C++2a)
• Implementation available in Boost - differences compared to the standard library

Processes and threads

• Multiprocess vs. multithreaded applications
• Operating system and threads (brief overview of available implementations)
• Synchronization mechanisms for processes and threads
• Structural implementation of synchronization
• Specifics of multithreaded implementation in the object-oriented approach (the role of the RAII pattern)

Memory model in C++

• Objects and memory locations
• Operation sequencing and enforcement
• Relationships: synchronization, happens-before, ordering of operations for atomic actions, release sequences
• Memory barriers
• Atomic operations

Thread management in C++14 and newer standards

• The std::thread class and thread execution
• Thread identification
• Joining and detaching threads
• Threads and standard containers
• Exception handling in C++14 threads

Mutual exclusion and critical sections

• Basic principles of mutual exclusion
• Concepts (types) of available blocking objects
• Quick introduction to time handling in C++14 (chrono library)
• Classes std::mutex, std::timed_mutex, std::shared_mutex and derivatives
• RAII pattern in lock handling (std::lock_guard, std::unique_lock)
• Dependent and deferred locks (std::lock)
• Ownership and shared ownership of locks and threads
• Readers-writers problem
• One-time initialization
• Thread-local (private) memory (std::thread_local)

Locks and thread synchronization

• Operations on semaphores
• Recursive locks
• Locks with defined lock levels
• Decision variables and cooperation with locks
• Implementation of active and passive barriers (Idle Waits, Busy Waits)
• Resuming thread execution
• Communication between threads (Actor model and CSP)

Asynchronous tasks

• Asynchronous events and their handling
• Futures
• Promises
• Packaged tasks

Atomic types

• Standard atomic types
• Universal atomic flag (std::atomic_flag)
• The "std::atomic<T>" class template and its specializations
• Free utility functions for std::atomic types

Specifics of designing multithreaded APIs

• Designing structures using locks - coarse-grained and fine-grained locks
• Performance of concurrent code - estimation and measurement techniques
• Data partitioning
• Scalability of solutions
• Error handling and exceptional situations in application architecture

Testing and debugging multithreaded applications

• Code reviews focused on multithreaded operations
• Specifics of testing multithreaded applications
• Unit tests
• Increasing the likelihood of failure scenarios for testing purposes
• Formal methods for testing multithreaded applications
• Multithreaded application architecture patterns versus testing and debugging (overview)

Advanced thread management

• Dependent threads (Saga pattern)
• Avoiding contention in data access
• Controlled thread shutdown and interruption
• Interrupting blocking calls
• Interruption handling
• Task stealing
• Interrupting tasks during application shutdown

Multithreading design patterns

• Types and purposes of multithreading patterns
• Scope of pattern application
• Thread Pool
• Active Object pattern
• Thread-Safe Queue
• Producer-Consumer pattern
• Monitor pattern
• Multithread safety of GoF patterns