However, correctly threading applications is hard in general, and not all applications can gain significantly from parallelism. But some applications are embarrassingly parallel by their nature, and in this case, breaking down the problems into independent sub-problem is not hard at all, often requiring little more synchronization than waiting for all worker threads to finish.
I have just completed another year in blogging without missing a single scheduled post: This one is the 176th already!
On this last day of ISIEA09, a lot of the participants have already left. This gives me the impression that many participants just left as soon as they had presented their paper. I feel sorry for the last presenters who spoke before almost empty rooms. I think that it is rather rude, nay, even cheap to run away as soon as your paper is presented.
Compounded with the fact that a lot of presenters didn’t even show up—there was a lot of them—the last day at ISIEA09 was rather intimate. On the good side, it meant more time to discuss between talks and some exchanges were rather good. This being said, I think ISIEA should adopt the same ‘no show’ policy other conferences have. Unless you cannot show up because some case of force majeure, your paper is withdrawn from the proceeding and you are banned from the conference for a number of years. That’s harsh, but that prevents situations where 20-25% of the speakers do not even show up.
ISIEA’09 is the most varied conference I attended so far. Usually, conferences are more or less focussed on a specific topic. Here, everything vaguely industrial is a welcome topic: one speaker discussed optimal pump control to avoid cavitation, another an automatic ablution machine using machine vision… I think this kind of conference is very stimulating because you get to discover areas of research you did not necessarily knew about.
I presented my paper on the speeding up of motion estimation in video coding using SIMD instructions and approximate metrics—I will put the paper and the slides online when I return. I think the presentation went pretty well, as it drew some attention.
The guys from the freenode channel #programmeur (irc.freenode.net) are Getting Together this saturday September 5th at 13h00 at the University of Montréal. The program: short (1h) talks about Distributed Version Control software, Numerical Stability, and Secure Networking.
Instructions to get there are here.
The cost of entry is 10$, mostly to pay for renting the room with media equipment. All are welcomed.
Last week, while friends and I were discussing the sensationalistic news coverage of the swine flu pandemic, I was joking that if you were not coughing up bacon, you were probably OK.
I fact, I wasn’t so much joking about the flu itself than about how (dis)information is presented in sensationalist news channels such CNN, Fox, or even Montréal-based LCN. Earlier this week, news were that the flu had already caused tens, maybe hundreds, of deaths, but that data was presented as if, you know, you just catch the swine flu and you die right away from it. However, on Thursday morning, on the radio, I heard that the Mexican authorities recounted “swine flu” death to… less than ten. To really understand what’s going on, you really have to do some research, and sometimes what you discover is radically different from what you’ve been told.
Gerard W. Kelly — Short-Cut Math — Dover, 1984, 112 pp. ISBN 978-0-486-24611-6
This little book—not even 120 pages—will help you sharpen your mental math skills using a series of tricks, strategems, and special cases that will get you to the exact result, or at least to an accurate approximation, with greater speed. The book is not exactly mind-shattering either; still, it may be useful to learn a few new tricks.