Engineering review at UT, and can we do it too?

Today I attended the UT Computer Architecture and Embedded Processor Industry Review. This was their second industry review, bringing out professors and graduate students in their Electrical and Computer Engineering department to give half-hour talks about their research. Last year's day-long event was tremendous, focusing on circuits and systems. This year's event focused more on research in embedded systems, which happens to be my area of interest. UT is doing this because "it's the right thing to do" (to quote Yale Patt's slide today). Yes, it provides exposure for UT and the potential for additional funding, but moreover it's a way for industry to see what academic researchers are up to, and it helps graduate students mix with their industry counterparts. Win-win.

Here's a couple of my observations:

1. For as beneficial as this event is for keeping professionals up to date on the latest academic research, this is the second year that it was not packed. I'm guessing there were around 100 people in attendance, and that's counting UT students. Why are so many semiconductor professionals not taking advantage of this review? The room should have been packed. It's such a valuable event. Which leads me to my second observation...

2. The review is obviously valuable to provide a glimpse into what our local university is researching. However even more beneficial is understanding how the researcher defined and attacked the problem at hand. As the presenters guide me through their problem analyses, discoveries, and conclusions, I start to see similarities with my own work and analogous problems I have faced. My mind begins buzzing during their presentation as I devise potential solutions on-the-fly (like us engineers do), and then the presenter shows why he/she chose a particular solution. That type of walk-though is useful for any engineer.

In addition, solutions to problems can be generalized. Understanding the approach taken to solve a particular problem often can be applied to similar problems in other domains. For example, the divide and conquer parallelization strategy that attains great speed-up can be found in many places. The notion of branch prediction in microprocessors which exploits the characteristic of locality is now being applied to dynamic power prediction (and as I learned today) more or less in smart DRAM prefetch. It's exciting to realize the common solution in all these problems, and then brainstorm where else it may apply.

OK, I'm obviously happy that UT does this. Now, what I have been pipe-dreaming is to entice door64 members to take part in something similar! Though that sounds monumental, I don't think it is. Thinking the "laid-backness" of BarCamp, but preplanned with presentations written and presented by door64 members who wish discuss an area of their expertise. Why do this? Simple: Enable the community to benefit from someone's expertise, and simultaneously acknowledge the presenters in the community (also beneficial for career advancement). I'm interested in bringing in graduate students to present as well. We can get corporate sponsorship and make it a day-long event. It certainly seems plausible to me.

I've been stewing on this for a long time. What do you think?