ELVIS: Software Design Research Group

MarkMoir gave a seminar on "Experience with and Potential of Hardware Transactional Memory".

I will briefly summarise our experience with the Hardware Transactional Memory (HTM) feature of Sun's multicore processor code named Rock. This experience has demonstrated significant potential in making it easier to construct concurrent data structures that are scalable, efficient, and correct, but also a number of limitations that prevent us from exploiting the full power of HTM. In the second half of the talk, I will focus on the potential of HTM by considering how it can be used to simplify various concurrent algorithms. The goal of this work has been to explore the assumptions needed about an HTM feature in order to exploit it in various contexts. We hope these observations will help guide designers of future HTM features.

Bio: Mark Moir is a Distinguished Engineer a Sun Microsystems, and is the Principal Investigator of the Scalable Synchronization Research Group at Sun Labs. Moir received the B.Sc. (Hons.) degree in Computer Science from Victoria University of Wellington, New Zealand in 1988, and the Ph.D. degree in Computer Science from the University of North Carolina at Chapel Hill, USA in 1996. From August 1996 until June 2000, he was an assistant professor in the Department of Computer Science at the University of Pittsburgh, after which he joined Sun Labs.

James Skene who is a post-doc with John Hosking at the University of Auckland gave a seminar on "Descript, a declarative, object-oriented modelling language".

Abstract

Software engineering, or the act of producing software, involves capturing decisions and domain knowledge in abstractions. Ultimately, the programs that we write are abstractions, or models, of their every possible execution. But, in the course of writing a program we may also develop other abstractions, such as domain models, class hierarchies, interfaces, domain-specific representations etc.

Ideally, once we have captured some knowledge in an abstraction, we would prefer to reuse that abstraction in other projects, rather than go to the effort of redeveloping it. But, it seems to be difficult to keep track of what artifacts mean over time, and to separate interesting abstractions from non-interesting, project specific ones.

In my talk I will discuss Descript, a language that I am developing with funding from the FRST. Descript is an attempt at a language that will allow software engineers to capture abstractions in such a way that their meaning is clear, retained over time, easy to extract, relatively untainted by quotidian detail, and at the same time still useful. It is object-oriented in the sense that it models objects and relationships in the real world. It is declarative in the sense that it has no primitives for modelling actions. Descript is intended to be useful for programming, modelling and metamodelling. It is also intended to address some perceived deficiencies in existing technologies in the model-driven engineering sphere, such as UML, MOF and OCL.

Elvis arranged for Philippe Kruchten to give a talk at the Wellington Java User Group (JUG) on "Software architecture and agile software development: a clash of two cultures?". February 2010.

Susan Eisenbach from Imperial College, London, England gave a seminar on "Concurrent Programming: Is there a silver bullet?". February 2010.

Elvis had their usual Christmas Party in 2009 at DavidPearce's house. It was a glorious sunny day. Many sausages, steaks, and salads were eaten, not to mention lots of wine and beer were consumed. Fun was had by all. Even Santa made a cameo appearance disguised as a cat, have a look at the photos.