|
    
|
|
HPCC seminar videotapes4D Data assimilationSpeaker: Peter Lyster Atmospheric 4D Data Assimilation involves combining observations with forecast models to form a best estimate of the atmosphere's state over time. Lyster's Earth and Space Sciences project Grand Challenge team is developing parallel versions of methods in use at NASA's Data Assimilation Office. The Physical-space Statistical Analysis system will be used for mission support and reanalysis of climate data over 10 to 15 years. Lyster's project also encompasses development and scientific assessment of the more rigorous Kalman Filter used to assimilate stratospheric chemical constituents such as ozone. These parallelized techniques will be critical for NASA's Earth Observing System, which will collect nearly one trillion bytes of data per day early in the next century.
|
Analysis of subwavelength optical elementsSpeaker: Dr. Dennis Prather
A burgeoning development of computational systems strives to improve computer performance by getting more out of each processor's clock cycle, instead of simply increasing its clock speed. The idea is to increase the throughput efficiency of a CPU by increasing the number of functional units within it. A functional unit is an adder or shift register.However, because these units share memory and simultaneously perform tasks within a single clock cycle, they require the exchange of data at fractions of a clock cycle. This exchange is complicated and ultimately limited by the delay of conventional metallic-based interconnects. UD researchers are developing an alternate approach that uses wavelength-scale diffractive optical elements, which are capable of being integrated with active optical devices on a commensurate scale-the micron scale. These elements control and redirect light among functional units, thus creating an optical interconnect. This allows for the transfer of data at the speed of light.
|
|||
|
|
