I recently stumbled across a post about some crazy optimization that clang does to divisions by a constant. If you aren’t interested in reading it yourself, the summary is as follows: Arbitrary integer division is slow. Division by powers of 2 is fast. Given a divisor , the compiler finds some such that approximates . This approximation gives exact results for any 32-bit integer. [Read More]

This might sound like a strong opinion, but I’m just going to put it out there: Python should make tzinfo mandatory on all datetime objects. To be fair, that’s just an overzealous suggestion prompted by my frustration after spending two full days debugging timestamp misbehaviors. There are plenty of practical reasons to keep timezone-agnostic datetimes around. Some projects will never need timestamp localization, and requiring them to use tzinfo everywhere will only needlessly complicate things. However, if you think you might ever need to deal with timezones in your application, then you must plan to deal with them from the start. My real proposition is that a team should assess its needs and set internal standards regarding the use of timestamps before beginning a project. That’s more reasonable, I think. [Read More]

Recently, another student asked me to set up a PostgreSQL instance that they could use for some data mining. I initially put the instance on a HDD, but the dataset was quite large and the import was incredibly slow. I installed the only SSD I had available (120 GB), and it sped up the import for the first few tables. However, this turned out to not be enough space. I did not want to move the database permanently back to the HDD, as this would mean slow I/O. I also was not about to go buy another SSD. I had heard of bcache, a Linux kernel module that lets a SSD act as a cache for a larger HDD. This seemed like the most appropriate solution – most of the data would fit in the SSD, but the backing HDD would be necessary for the rest of it. This article explains how to set up a bcache instance in this scenario. This tutorial is written for Ubuntu Desktop 16.04.1 (Xenial), but it likely applies to more recent versions as well as Ubuntu Server. [Read More]

Suppose you have to run some program with 100 different sets of parameters. You might automate this job using a bash script like this: ARGS=("-foo 123" "-bar 456" "-baz 789") for a in "${ARGS[@]}"; do my-program $a done The problem with this type of construction in bash is that only one process will run at a time. If your program isn’t already parallel, you can speed up execution by running multiple jobs at a time. This isn’t easy in bash, but fortunately Python’s multiprocessing library makes it quite simple. [Read More]

I recently consulted on a project involving embedded devices. Like most early-stage embedded endeavors, it currently consists of an Arduino and a bunch of off-the-shelf peripherals. During the project, I developed two small libraries (unrelated to the main focus of the project) which I’m open-sourcing today. [Read More]