S(o)OS Deliverables

S(o)OS D4.2 First Implementation Set: Execution Management - Introduction

Management of execution addresses all aspects related to handling code and data in the distributed environment – this includes all issues from segmenting the code over distributing the segments to their execution. Accordingly, this work also covers aspects related to extending the programming language with segmentation related annotations, such as concurrency and dependency identifications.

First, we describe the programming models that ease and improve parallelization of the code (Section II). To this end, we introduce the programming models and describe how they can be used to improve the usability of parallelization techniques. Next, we proceed to describe code analysis and segmentation which is important for increasing code scalability and heterogeneity, and is also very relevant for the usability of proposed techniques. We follow with the higher level, mathematically-based and thus more abstract view of the parallelization programming models.
The next section discusses the most appropriate programming language extensions that enable easier and, possibly, higher performance use of transactional memory and real-time programming abstractions. We also describe the programming language extensions that specify concurrency to increase the resource usage and overall system utility and performance (Section III).
We also describe the speculative execution in detail, focusing on transactional memory, real-time extensions for transactional memory and speculative memory accesses, which is also known as thread-level speculation (Section IV).
Finally, we describe the operating system kernel primitives for code and resource migration across different CPU cores, possibly residing on different hosts (Section V).

Download

S(o)OS D4.2 First Implementation Set: Execution Management