mason

correct me if i'm wrong, but NVIDIA's bandwidthTest program included in the CUDA SDK takes single timing measurements and reports them. if there's any noise in the measurements, these single reports may be misleading. i wrote my own bandwidth test program that takes 100 measurements and spews out the resulting data for analysis. with my program at least, there is plenty of noise if you measure unpinned rather than pinned transfers.

For tips on designing a software system, I read The Mythical Man-Month. I recommend it, but here are my notes in case you just want the juicy bits:

Google Docs and Evernote will both store your text data on their servers. I've used both for keeping my diary, but I'm moving over to Evernote completely now. Here's a python program I wrote that downloads all of your Google Docs from a particular folder and puts them into an Evernote notebook: gdocs2evernote.zip (Windows only)

A calculator that could do arithmetic on time, like 5:13*2 + 10:50 = 21:16, would be handy. So I built it. Download Timelab to try it! You'll need python to run Timelab (macs already have this installed).

If you discard color, a 100x100 pixel image has 10,000 pieces of information--one for each pixel. Consider each image a data point. To plot them, you'd need graph paper with 10,000 axes, because each pixel's value tells you the position of the image (data point) along the corresponding dimension. And you thought images were 2-D.

As a final project for my computer graphics class last semester, I implemented a fast version of a technique called "seam carving". Seam carving resizes an image by removing the least noticeable bits. I designed my seam carver to do this resizing on video streaming in from a webcam in such a way that you could remove an object from the video. Read on for my writeup, or just download the code and try it for yourself.

Windows Update kept trying to upgrade my graphics card (Intel 945GM in Lenovo X60 tablet) and failing. I tried manually upgrading the driver with ones from both Intel and Lenovo, and neither worked. Here are the steps that solved my problem:

Context-free grammars (CFGs) can be used as a model for natural language. That means they capture the structure of language in a relatively simple form that can be reasoned about and used to generate new instances of natural language. My natural language processing book says CFGs were proven insufficient for modeling German spoken in Zurich (Speech and Language Processing (2nd Edition), p.538), but they're good enough to bring you the following silly, random, grammatical sentences:

One day Soshi was walking on the bank of a river with a friend. "How delightfully the fishes are enjoying themselves in the water!" exclaimed Soshi. His friend spake to him thus: "You are not a fish; how do you know that the fishes are enjoying themselves?" "You are not myself," returned Soshi; "how do you know that I do not know that the fishes are enjoying themselves?"

I ran the half-marathon in San Francisco this past summer. It was amazing. Afterwards I wanted pictures to remember it by--specifically, pictures that had me in them to remember my experience. Before the race, MarathonFoto informed me that "you and your family will not have to worry about trying to capture your special race moments" because they would take pictures. Great! They neglected to mention those pictures would cost $40 each to download.

I liked the pictures, but that heads-up they gave before the race was slimy. I wondered, "how can I make a better, cheaper service for finding photos of myself (and put these guys out of business :P)?" The result is bibfind.com: a crowd-sourced photo tagging system that uses the bib numbers worn by runners to identify them in photos.