Advanced Computing Technologies
Investigating novel computer architectures, computer languages, systems software, performance analysis, and algorithms for next-generation HPC and edge computing.
Over the past 30 years, advances in computer technology allowed processor performance and functionality to double every two years a – trend known as Moore’s Law – and computational and experimental science have leveraged this trend in the broad computing industry with little change to their supporting software environment. But this era of predictable processor improvements is ending. Computer chip manufacturing techniques are reaching the limits of the atomic scale, and this shift will significantly impact the designs of future high performance computers (HPCs) and the software infrastructure required to use the nation’s scientific supercomputing facilities effectively.
As we enter the era of “extreme heterogeneity,” computer vendors are pursuing systems built from combinations of different types of processors to improve capabilities, boost performance, and meet energy efficiency goals. External economic drivers, like the marketplace growth of data-centric machine learning and artificial intelligence, will result in scientific supercomputers built from a custom aggregation of components, thus increasing the difficulty and complexity of developing scientific software.
At Berkeley Lab, our researchers are developing open source software environments and working with vendors to adapt to these complex changes in the design of future supercomputers, which is key to improving the nation’s scientific advancement and innovation rate.