Intel is working with high-performance computing facilities across the planet to establish Parallel Computing Centers. The goal is modernization. In this blog, Jeff Cogswell looks at what that means in terms of vectorization and multicore development, and considers how we can learn from the work of these centers.
Over the past decade, we have seen multicore processors become commonplace, with most computers today having at least a dual processor. While dual-core and quad-core processors are the norm for everyday computing, scientific and research applications often make use of more advanced processors such as many-core Xeon Phi processors.
Throughout 2014, Intel worked with universities and research institutions throughout the world to establish Intel Parallel Computing Centers. These are high-performance computing centers that already existed and received grants from Intel to take up the issue of modernization of software. But what exactly does modernization mean? In this case, it means improving legacy software to make use of parallel computing technology.
But this doesn’t just mean adapting software to multicore and quadcore processors. Rather, it means adapting software so that it can run on manycore computers, making use of whatever number of cores are available–whether a dual core or a cluster of Xeon Phi coprocessors, with upwards of thousands of cores.
沒有留言:
張貼留言