Day 3

@10:36 Prof. Demmel is giving a talk on the computational patterns for parallel programming. All Greek to me – architecture people practically live with matrices and their multiplications :( Well, probably its my fault; missed the first part of the talk again! Why do people need to start working so early in the day anyway? :-S

Anyway, his conclusions are

• Fast code must minimize communication or message passing in the MPI model,
• The rest seemed (at least to me) on the same line,
• And of course, reuse other people’s code instead writing libraries by yourself.

@11:14 ParLab faculties are getting ready to talk on what they are working on. The first of the four speakers is Nelson Morgan. He is going to give an overview of his work on speech recognition – in “real” meetings. The aim is to create real-time searchable meeting notes/excerpts in real-life situation without relying on heavy equipment and in noisy-environment. Lots of signal-processing, probability estimation, and decoding.

@11:31 Prof. Keaveny is going to talk on application of parallel computing in medicine and medical imaging for personalizing health solutions. He works on biomechanics – failures of organs specially bones and blood vessels.

The objective of his project is to perform blood flow simulation in 10 minute timescale to find out the location of stroke in the brain so that the time window to provide help for a stroke patient increases.

@11:43 Prof. Bodik is presenting a ParLab project on parallel mobile web browser.There is a nice connection between compilers and browsers.

@12:07 Finally, David Wessel on parallel music and audio applications. As far as I can understand, he wants to create new musical instruments using parallel computation. The biggest challenge in real-time musical instrument is very low latency. There are some more stuff he is talking about, that seems too hard to comprehend in such a short notice :|

@03:49 Back for the last session after skipping some in the middle. Matei is going to talk about MapReduce and Hadoop.

@04:45 Matei is almost finishing up. His talk was on Hadoop, its derivates, and tools required to use it. I would say he made his points across the room, at least to me. Well done!

@04:55 So we are done!