Techniques for Implementing Concurrent Exceptions in C++
Rintala, Matti (2012)
Rintala, Matti
Tampere University of Technology
2012
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:ISBN:978-952-15-2922-1
https://urn.fi/URN:ISBN:978-952-15-2922-1
Tiivistelmä
In recent years, concurrent programming has become more and more important. Multi-core processors and distributed programming allow the use of real-world parallelism for increased computing power. Graphical user interfaces in modern applications benefit from concurrency which allows them to stay responsive in all situations. Concurrency support has been added to many programming languages, libraries and frameworks.
While exceptions are widely used in sequential programming, many concurrent programming languages and libraries provide little or no support for concurrent exception handling. This is also true for the C++ programming language, which is widely used in the industry for system programming, mobile and embedded applications, as well as high-performance computing, server and traditional desktop applications. The 2003 version of the C++ standard provides no support for concurrency, and the new C++11 standard only supports thread-based concurrency in a shared address space.
Procedure and method calls across address space boundaries require support for serialisation. Such C++ libraries exist for serialisation of parameters and return values, but serialisation of exceptions is more complicated. Types of passed exceptions are not known at compile-time, and the exceptions may be thrown by third-party code.
Concurrency also complicates exception handling itself. It makes it possible for several exceptions to be thrown concurrently and end up in the same process. This scenario is not supported in most current programming languages, especially C++.
This thesis analyses problems in concurrent exception handling and presents mechanisms for solving them. The solution includes automatic serialisation of C++ exceptions for RPC, and exception reduction, future groups and compound exceptions for concurrent exception handling. The usability and performance of the mechanisms are measured and discussed using a use case application.
Mechanisms for concurrent exception handling are provided using a library approach (i.e., without extending the language itself). Template metaprogramming is used in the solutions to automate mechanisms as much as possible. Solutions to the problems given in this thesis can be used in other programming languages as well.
While exceptions are widely used in sequential programming, many concurrent programming languages and libraries provide little or no support for concurrent exception handling. This is also true for the C++ programming language, which is widely used in the industry for system programming, mobile and embedded applications, as well as high-performance computing, server and traditional desktop applications. The 2003 version of the C++ standard provides no support for concurrency, and the new C++11 standard only supports thread-based concurrency in a shared address space.
Procedure and method calls across address space boundaries require support for serialisation. Such C++ libraries exist for serialisation of parameters and return values, but serialisation of exceptions is more complicated. Types of passed exceptions are not known at compile-time, and the exceptions may be thrown by third-party code.
Concurrency also complicates exception handling itself. It makes it possible for several exceptions to be thrown concurrently and end up in the same process. This scenario is not supported in most current programming languages, especially C++.
This thesis analyses problems in concurrent exception handling and presents mechanisms for solving them. The solution includes automatic serialisation of C++ exceptions for RPC, and exception reduction, future groups and compound exceptions for concurrent exception handling. The usability and performance of the mechanisms are measured and discussed using a use case application.
Mechanisms for concurrent exception handling are provided using a library approach (i.e., without extending the language itself). Template metaprogramming is used in the solutions to automate mechanisms as much as possible. Solutions to the problems given in this thesis can be used in other programming languages as well.
Kokoelmat
- Väitöskirjat [4894]