A new programming standard, called the Open Spatial Programming Language (or OpenSPL), debuted today “to enable the next generation of high performance parallel spatial computers.”
The open standard was developed by the Open Spatial Programming Language (OpenSPL) consortium, which formed to promote the use of spatial computing among a wide set of users and to standardize the OpenSPL language. The overarching goal of the consortium is for spatial computing to become the industry standard for mission critical computations.
The entire effort hinges on the well-supported idea that the future of computing is parallel. OpenSPL is based on the concept that a program executes in space, rather than in time sequence. It takes the current paradigm and turns it on its head. Operations are assumed to be parallel unless specified as sequential. The consortium explains that this is similar to a factory floor where all operations take place in parallel, with each operation carrying out a piece of the overall process.
“Temporal Programming is a recipe for the execution of actions, whereas Spatial Programming builds a factory to execute the recipe,” is how the backers put it. “Conventional programs execute in 1 dimension, where time progresses forward following the instruction sequence. Spatial programming is programming in 2 dimensions, where data progresses forward in parallel across the fabric of an array or chip.”
It’s a revolutionary view of computing with the potential of yielding significantly higher performance, compute density and energy-efficiency compared to traditional instruction-processor machines. With deterministic throughput and low latency at low power, spatial computing is well-matched to line-rate data processing applications, such as that used in high performance computing, datacenter networking and the Internet of Things.
“OpenSPL enables us to build parallelized applications that fully take advantage of spatial computing technology with the ease of a high-level software project,” notes Ryan Eavy, executive director, Architecture, CME Group. A video on the consortium’s website explains the concept in greater detail.
Spatial programming is also green, according to the OpenSPL consortium, since the amount of computation per cubic foot of datacenter space is maximized. The right applications can experience two orders of magnitude improvement in computational density, reducing power consumption accordingly, say backers.
The OpenSPL consortium includes both industry and academic partners, and offers two levels of membership. Full members participate in all consortium activities, including voting, while observer members participate but may not vote. The consortium was founded by “full members” CME Group, Juniper, Chevron and Maxeler Technologies with Imperial College London, Stanford University, University of Tokyo and Tsinghua University contributing as “observer members.” The consortium is managed by a steering committee, currently chaired by Tamas Nemeth of Chevron.
Activity is building for the effort. The first OpenSPL Summer School will take place July 2014, at Imperial College London. Researchers, students, and members of OpenSPL will gather together to share experiences, work on application development, and continue to advance the space computing paradigm.