I went to two seminars in the IIT CS department on Thursday. It was just me and about fifteen grad students, a couple professors, and maybe a few confused undergrads. The talks were over my head, but not far enough over that I didn't understand them; just far enough over that I didn't fully grasp the context in which the points made should rest.
Talk 1: controller/hardware level memory allocation, and the correct (best worstcase scenario) way to redirect addressing functions between banks (because a bank based on an 8bit system with a 16bit address, which is the current norm, ends up only having 64KB of address space, and therefore there must be more than one bank... ) There was some discussion on NP hardness. Now, I had a bit of an introduction to NP[n] with Nazli and databases, but it was cursory. Here, have a semiuseful wiki page. http://en.wikipedia.org/wiki/NP-hard
There was a debate about how this scheme works with vortexes. I don't know what a vortex is in this context; I'm under the impression that it's a pattern of behavior of a hardware system, or possibly an equation describing the behavior of a bank-selector-direction planning schema. Some googling brought up these links.
"The "Vortex" processor is a general purpose CPU with a novel architecture and instruction set." http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4137031&userType=inst
"Vortex is a language-independent optimizing compiler infrastructure for object-oriented and other high-level languages." http://www.cs.washington.edu/research/projects/cecil/www/vortex.html
Talk 2: WLAN, network access points, accessibility, and algorithms (and business plans!) which will enable some users to be hop points, i..e transfer traffic through themselves (like using your smartphone to provide a mini wifi network for the guy with the netbook next to you on the train) in an automated way which compensates the guy with the phone for providing access to the guy with the netbook.
Calculating how much someone should be compensated for how much bandwith (and he didn't even mention battery life issues) and how to auto-sense things like which device is closest to the network connection point and how to minimize the number of hops... how to make a grid of access towers such that they're on optimal frequencies for not interfering with each other (because you know that the main wireless frequencies overlap with each other, so you shouldn't have a tower broadcasting on frequency 1 next to a tower broadcasting on frequency 1... much better to have a 1 next to a 4, etc.)
There were graphs showing optimal arrangements and equations showing how the interference amount was calculated. Throughput tradeoff with interference, charts of data. MUCA, MICA, RC, ADJ. (what are those? In context?) Company-set limits on maximum transfer point usage, etc to keep the cost curve in the right place. Large graph of user benefit, transfer point cost, transfer point benefit, and company benefit/cost...
I have seen the future, and it's based on equations.
Maybe you can get more out of the lecture descriptions than I can.
There are two more next week which I won't be able to go to because I'll be at #ghc10
Persistent links here: