Many of the big scientific computing problems can be solved using matrix mathematics. One of my favorite problems to tackle is implementation of vector times matrix = vector. This has utility in many places, one of which is inference in neural networks.

Since an ARM processor has some support for vector and matrix mathematics, I started wondering what were the size limitations we would hit when trying to solve big problems. Here are some notes:

“Get out”

That was my sister, telling me via chat to get out of the cab I had just gotten in to. In her defense, I had screwed up her instructions, which was to go to the Imperial Cab counter at the airport, prepay for a ride to her apartment, and in no case was I to get directly into a cab.

What a way to spend an evening.

As I attempted to rewrite some Gitlab automation from bash to python, I stumbled over what should have been a trivial thing: collapsing section headers. They were not trivial.

One of my new years resolutions this year was to try and replace bash scripting with python in some of my projects. The reasons for this include the ability to use objects to group parameters together.

At today’s (Jan 19, 2025) Python over Coffee meetup, Ricardo suggested that I look at the plumbum library as a tool to help make that transition. In my first hack at using it, I came across an issue: I need to read in environment variables.

If you can’t read the text on the default index page of your Apache HTTPD server, you can make it larger with a .htaccess file.

cat .htaccess 
IndexOptions NameWidth=80

If you put this in the root of the troublesome tree, it will server for that directory and all directories below it.

Not quite as succinct as I would have liked, but this does work. This is a one liner in Bash, and I had hoped to do it with a single python call once I had the repo object established. In general, diffs are not simple to do on blob objects. I had trouble using the difflib and set approaches to diffing the two files, possibly due to the fact that I was originally not decoding the streams. Thus all the splitlines work probably can be simplified.