John Gustafson (scientist)

For the musician, see John Gustafson (musician).
John Gustafson, circa 2005.

John L. Gustafson (born January 19, 1955) is an American computer scientist and businessman, chiefly known for his work in High Performance Computing (HPC) such as the invention of Gustafson's law, introducing the first commercial computer cluster, measuring with QUIPS, leading the reconstruction of the Atanasoff–Berry computer, inventing the Unum number format and computation system,[1][2] and several awards for computer speedup. Currently he is the Chief Technology Officer at Ceranovo, Inc. pushing the limits of ceramic capacitors.[3] He was the Chief Graphics Product Architect and Senior Fellow at AMD from September 2012 until June 2013, and he previously held the distinguished positions of Director of Intel Labs-SC,[4] CEO of Massively Parallel Technologies, Inc.[5] and CTO at ClearSpeed Technology.[6] Dr. Gustafson holds applied mathematics degrees from the California Institute of Technology and Iowa State University.

Childhood and education

Gustafson was raised in Des Moines, Iowa. After completing a degree in Applied Mathematics at California Institute of Technology in 1977 he moved to Ames, Iowa and completed his M.S. (1981) and Ph.D. (1982) at Iowa State University.

His mother was an electronics technician at Collins Radio and his father was a chemical engineer turned MD, both as a result of World War II. His parents encouraged his scientific explorations at a young age. Assembling radio transmitters, designing and executing chemistry experiments, and making holograms are some of his favorite childhood explorations adding up to over 1000 lab hours before college.[7]

Awards

Unums

Gustafson devised a new format for storing real numbers as a series of binary bits in computers. Normal formats store numbers as a fixed number of bits, for example 64 bits is usual for double-precision floating-point format numbers. Unums on the other hand use a variable number of bits depending on the number of digits required. This can allow them to be smaller than doubles for fast processing and also more precise or larger than the limits for double when desirable.[8]

References

External links

This article is issued from Wikipedia - version of the Thursday, January 07, 2016. The text is available under the Creative Commons Attribution/Share Alike but additional terms may apply for the media files.