Anticipated Performance Development

Presentation Transcript

Anticipated Performance Development

TOP 500 Performance Projection

The Exponential Growth of Computing, 1900-1998 Pentium II PC Cray 1 IBM 360 Model 75 IBM 704 ENIAC Bell Calculator Model 1 Hollerith Tabulator Adapted from Kurzweil, The Age of Spiritual Machines

The Exponential Growth of Computing, 1900-2100 Adapted from Kurzweil, The Age of Spiritual Machines

Growth of Computing Power and "Mental Power" Hans Moravec, CACM 10, 2003, pp 90-97

Why this shortsighted view isn't right Unsuitability of Current Architectures Teraflop frameworks are centered around exceeding expectations in figuring; just a single of the six (or eight) measurements of human knowledge Fundamental absence of scientific models for intellectual procedures That's the reason we are not utilizing the most intense PCs today for subjective errands Complexity limits We don't know yet how to model turbulence, how then do we demonstrate thought?

"The PC show turns out not to be useful in clarifying what individuals really do when they think and see" Hubert Dreyfus, pg.189 Example: one of the greatest examples of overcoming adversity of machine knowledge, the chess PC "Dark Blue", did not show us anything about how a chess grandmaster considers.

Six Dimensions of Intelligence Verbal-Linguistic ability to think in words and to utilize dialect to express and acknowledge complex ideas Logical-Mathematical makes it conceivable to figure, measure, consider suggestions and speculations, and do complex numerical operations Spatial capacity to think and orientate in physical three-dimensional condition Bodily-Kinesthetic capacity to control protests and adjust physical abilities Musical affectability to pitch, song, beat, and tone Interpersonal c apacity to comprehend and cooperate adequately with others Howard Gardner.  Frames of Mind:  The Theory of Multiple Intelligences .  New York:  Basic Books, 1983, 1993.

Current State of Supercomputers

Retina to Visual Cortex Mapping http://cgl.elte.hu/~racz/santafe.html

Building New Models About 1/3 of human mind is presumably devoted towards handling of visual data We have just exceptionally simple learning of the standards for human vision processing Research extend by Don Glaser at UC Berkeley explores mapping from retina to visual cortex Attempt to show "optical fantasies" and basic development of articles in the visual cortex Current models constrained to around 10**5 neurons Supercomputer extend at NERSC in 2005

Fourth Observation about CSE 4. There are limitless ranges of science and building where CSE has not started to have an effect current rundown of CSE applications is nearly the same as 15 years prior current arrangement of designs is catching just a little subset of human psychological capacities in numerous logical zones there is as yet a practically total nonappearance of computational models See additionally: Y. Deng, J. Glimm, and D. H. Sharp, Perspectives on Parallel Computing, Daedalus Vol 12 (1992) 31-52.

Major Application Areas of CSE Science Global atmosphere demonstrating Astrophysical displaying Biology: genomics, protein collapsing, medicate outline Computational science Computational material sciences and nanosciences Engineering Crash reenactment Semiconductor plan Earthquake and basic demonstrating Computational liquid progression Combustion Business Financial and monetary displaying Transaction preparing, web administrations, and web crawlers Defense Nuclear weapons—test by reproductions Cryptography This rundown from 2004 is indistinguishable to a rundown from 1992!

Your vote checks A Zettaflops PC will have new smart conduct. The initially supported Petaflops application will utilize MPI. The initially maintained Exaflops application will utilize MPI.