Ganesh Swami

I found a really cool library while doing some mesh research: Autopack. In the message passing paradigm of parallelism, keeping tabs on latency is more important than the size of the data being sent. It’s better to send one huge chunk of data instead of many small messages. Unfortunately, without a lot of obfuscation and complexity, it’s inevitable that you send really small messages.

Autopack attempts to solve this automatically. I quote some features from the homepage:

• collective commands that proceed asynchronously
• automatic management of MPI send and receive requests
• management of message buffer memory
• determination of the number of anticipated incoming messages from a set of arbitrary sends
• deterministic message delivery from multiple sources to aid in program testing or debugging

comes handy when you have difficulty deciding on a date.

Let’s let V be potential difference between dating girls & doing something productive, I be current of love, and R be resistance to current flow. The problem with serial dating is the same as with serial circuits. If resistance starts to increase, you’re stuck: because V = IR, and R is increasing, I must decrease to hold the equality. Worse still, because P = IV, you’re just not going to have as much power with the increased resistance.

read it all on bitquabit.

It’s about time I gave an update on my non-linear physics class.

We spent the last few weeks building a framework to analyze partial differential equations with non-linear terms. After the pedagogical material, we went on to our first real example: Faraday waves.

Faraday waves were first described in an appendix to a paper published in the Philosophical Transactions of the Royal Society of London in 1831. These are standing non-linear waves that are generated when an open container with fluid is subject to vertical oscillations. When the oscillations reach a certain threshold, we begin to see an instability on the surface of the fluid. Our professor did a demo for us with two fluids: canola and water. I had posted a video to Faraday waves with corn starch some time back.

A bunch of people had previously observed this phenomena, but Faraday was the one who had described that the oscillation frequency of the waves is half that of the driving frequency. I’ve read this original paper, and he goes into excruciating detail about his experiments. Surprisingly for a physics paper, there wasn’t a single equation.

I’d like to draw attention to the point that we get a frequency that is half that of the driving frequency. This is impossible in a linear system, whose transfer functions have complex exponentials as eigenfunctions. The mixing of the modes of the system is not arbitrary, in fact we determined them after a ton of math.

To finish off, I’m posting a video to Chladni patterns formed with different frequencies. For the sake of saving some brain cells, don’t read the comments on that video.

Pictures courtesy of Jerry Gollub.

Fifteen Mac-vs-PC commercials back to back:

Do you really have to stoop to such low levels?

Chistophe Bisciglia, an engineer from Google, is teaching a course at the University of Washington. The course focuses on problem solving on large-scale clusters. The complete course material is available on the homepage.

I think there’s a huge potential for programs to analyze voluminous amounts of data. Most data analysis is done either using Excel, or using command line tools such as awk and sed. A smart way to distribute the post-processing or analysis would be cool.