Parallel processing calls for a Fortress mentality

When you're the inventor of one of the most successful and influential programming languages of the last decade, what do you do for an encore? Judging by demos at the recent JavaOne conference, if you're Sun Microsystems, you invent another programming language. But Fortress is more than just another syntax to remember. Instead, Sun Labs is taking lessons learned from Java and applying them to the thornier probl

When you're the inventor of one of the most successful and influential programming languages of the last decade, what do you do for an encore? Judging by demos at the recent JavaOne conference, if you're Sun Microsystems, you invent another programming language.

But Fortress is more than just another syntax to remember. Instead, Sun Labs is taking lessons learned from Java and applying them to the thornier problems of application development for HPC (high-performance computing). If it succeeds, Fortress could become the language of choice for the new computing era.

The problem with most existing languages is that they were designed for an earlier generation of machines. Processing resources were scarce. Desktop computers typically gave you only one CPU to worry about, and parallel supercomputers enjoyed the rarified air of academia, where coders were accustomed to jumping through mathematical hoops to get their applications running efficiently.

Flash forward to today, however, and the computing landscape looks considerably different. Low-cost, commodity processors, clustered together over high-speed networks, have made HPC accessible to nearly anyone. The lowliest desktop PC might be powered by a multicore CPU. There's a boundless world of processing power out there, waiting to be tapped. Unfortunately, the popular programming languages of today simply weren't designed for the parallel-processing model.

Take, for example, the humble for-next loop. It seems unambiguous enough. If I tell the computer to perform the same task a hundred times, the program loops through exactly 100 cycles and then proceeds to the next instruction.

In real life, however, a task that can be described as "doing the same thing 100 times" is often more nuanced than it seems. If my task is to fetch water from a well, for example, one person armed with one bucket could make the trip 100 times to complete the task. But if I had a hundred buckets, and a hundred strong backs to carry them, I could complete the mission in one one-hundredth of the time. Barring any other bottlenecks, there's nothing about a trip to fetch water from a well that depends on anyone else's trip to the well; you can complete all the trips at roughly the same time.

This, of course, is exactly the same kind of efficiency that you get from parallel computing. But writing algorithms that can take advantage of parallel processors isn't easy in today's languages. There's nothing about the grammar of a for-next statement that lets the system know that it's OK to parallelize that loop. Instead, you have to manage the calculus of making it parallelizable by hand. Lose track of what's happening in the system and you can quickly end up in a "Sorcerer's Apprentice" scenario.

In Fortress, on the other hand, language constructs such as for-next loops are parallelizable by default. The Fortress specification supports the concept of transactions within the language itself, which means that complex calculations can be computed as atomic units, independent of any other program threads that might be running.

To aid developers in conceptualizing complex parallel-processing applications, Fortress' syntax is based on mathematical notation. In Fortress, programmers don't need to translate abstract concepts such as physical units into representations designed for the machine. Wherever possible, the goal of the language is to do away with obsolete programming concepts and concentrate on making the coder's job easier. In addition, Fortress was designed to be a "growable" language, to which new features can be added easily.

Don't plan on writing your next major business application in Fortress just yet, however. So far the language exists mostly on paper. A reference interpreter that implements most of the core language features is available from the Fortress project community site, but this is still bleeding-edge stuff.

Still, as multiple cores replace megahertz as the new benchmark of CPU power, a new kind of software development needs to emerge to ensure that software can continue to take advantage of the latest hardware. If work on the project continues apace, Fortress could very easily become one of the most important developments in computer science since parallel processing itself. I'll be keeping my eye on this one.

Join the discussion
Be the first to comment on this article. Our Commenting Policies