Supercomputing 2012

The FRCRC had a booth at Supercomputing 2012 where we hosted the winners of the HPC 2012 student poster competition.

The FRCRC will have a booth at Supercomputing 2012! Come visit us at booth 3847 at SC12 in Salt Lake City Utah November 12-15, 2012.  For more information please visit the Supercomputing 12 Official Site (opens in a new window.)

The winners of the HPC 2012 student poster competition will be in the booth with their posters.  The winning posters are below, click the image to download a full size PDF.


HPC 2012 Student Poster Competition winners

 


Torey Semi, Colorado School of Mines:

CaAl2Si2 is the prototype of the CaAl2Si2 class of Zintl structures established to be useful as thermoelectrics.  We propose that CaAl2Si2 be interpreted as an ordinary covalently bonded, tetrahedrally coordinated quasi-semiconductor consisting of a large distortion of the wurtzite structure with the almost fully ionized Ca inserted at an interstitial site.  We support this interpretation via a structural mapping and calculations for a charged binary primitive  cell.

Our intent is to explain its unusual structure, the origin of its semimetallic behavior, the basis for its stability and the effect of substituting  other column II atoms for Ca.

 

 

Andy Stone, Colorado State University:

This poster describes a library and code rewriting tool for easily implementing semi-regular grid computations.  Semi-regular grids are seen in many Earth science computations and consist of sets of regular grids that may connect to one another in an irregular fashion.  With the abstractions in our GridLib library scientists are able to separately specify grid connectivity from algorithm.  A separated specification helps to improve code clarity and maintainability.

 

Ben Greer, University of Colorado Boulder: 

A common method of numerically solving partial differential equations is the Alternating Direction Implicit (ADI) method. At the heart of this algorithm is a series of recurrence relations that make ADI non-trivial to run in parallel. We have implemented and tested a method for parallelizing recurrence relations known as the pre-propagation and message vectorization method and applied it to a two-dimensional ADI problem. We compare our results to a more commonly used method and find better scaling with processor count.

 

Jared Baker, University of Wyoming:

This study involves the aerodynamic enhancement of wind turbines by computationally investigating the usage of an attached spoiler to the root sections of wind turbine blades which is biologically inspired from birds.