Parallel application are becoming even more fine-grain and differentiated in the activities they perform so that task-parallelism is gaining an increased consensus in the choice of the programming model that best suits their needs. In this regard, several programming languages, libraries and parallel programming environments have been presented in literature in order to make the most of the underlying hardware resources by relying on both the user-level thread technologies and the OS-managed threads. Nevertheless, the applications execution's state evolves very quickly at run-time while adequate software supports are currently lacking at all. We propose therefore a paradigm shift in the execution of the application's tasks based on the innovative user-level micro-thread (ULMT) technology which differs from the classical user-level thread technology which is not application-transparant and still bound to the program control-flow graph (CFG). Our approach does not require the programmer to insert function calls or compiler directives while asynchronously introduces changes of the execution trajectory of threads in arbitrary points in time to allow the applications to control their own evolution. This can be accomplished by offering library APIs to the programmers which will be responsible only of registering minimal informations in the startup phase while all the operations aimed at optimizing the application execution's state will be automatically and timely performed at run-time.
The proposed research has as its main goal that of providing a new paradigm for the execution of parallel applications. According to this new paradigm, one thread activity is admitted to shift out from its current execution path in favour of an arbitrary different one, while the suspended activity will be able to resume later along the execution path of the same thread or another with no loss of the achieved progresses. Such kind of schema is completely innovative and never presented before in the stae-of-the-art of parallel computing, even in a form that is similar to the one described in this project.