Interim Report

Team Number: 006
School Name: Albuquerque Academy
Area of Science: Bioinformatics
Project Title: Parallelizing a CRI-Map Program
Project Abstract: http://mode.lanl.k12.nm.us/97.98/abstracts/006.html
Interim Report: http://mode.lanl.k12.nm.us/97.98/interims/006.html
Final Report: http://mode.lanl.k12.nm.us/97.98/finalreports/006/finalreport.html

Our problem is to take the code CRI-Map, and rewrite it to run in parallel on symmetric multi-processors (SMP's). We have downloaded the code and have begun to instrument it to find the timecode. We have also compiled the single processor version and ran the test problem that came with the code. We expect the multi processor version to yield the same processor version can not handle.

The first proble we encountered was finding that a team at Rockfeller University had done a similar problem with the Parallel Virtual Machine (PVM) network. PVM does significantly decrease the time needed to run the problem, but we are hoping that a multi-processor supercomputer will have the architecture to make the time needed even less. There might also be sample problems that will not be able to run on a single processor computer or on PVM that we can run on the supercomputer, problems we are trying to find from our present contacts.

The second problem we encountered was the question of whether we should continue with the CRI-Map problem after finding out about PVM. The speed of PVM was a significant acheivement over the single processor speed we had originally planned to conquer. Now the only problem that we seem to be solving is that of time. We are hoping that the difference in time will be significant enough so that we are actually solving a problem. Taking the advice of our advisor we considered some other problems that still did not seem as promising at this time as the CRI-Map. Another barrier we hope to surmount is the complexity computers or on PVM that will run on the supercomputer as an added goal.

The process we are in at the present is finding which parts of the CRI-Map code we will parallelize. We are mapping out the code tree, finding out which functions or subroutines the main block calls and the functions those functions call and so on. Once we have finished the tree we will test each function and subroutine and find out how much time it takes we will parallelize those functions to make them faster. We do not have time to parallelize the whole code so we will change the slowest parts.

We believe that we have enough time to complete the most necessary parts of this project. The more time we have the more parts of the code and find one good sample problem. We expect to get a faster completion speed and the results of the computation of some complex problems.

Team Members:

Sponsoring Teacher(s):

Project Advisor(s):

New Mexico High School Supercomputing Challenge