The annual Advent of Code is coming up soon! It’s a series of daily programming puzzles beginning December 1st and ending on Christmas. The puzzles start off easy and get progressively harder and more interesting as it goes on. They are typically more open-ended, practical and fun compared to pure algorithmic contest type problems like LeetCode. It’s a great way to experiment with new programming languages & concepts alongside others solving the same problems.

Recap of 2021

So in 2021 I decided to try and finish all the days in Rust - starting with no knowledge of Rust. It wasn’t the cleanest or most idiomatic code, and it was definitely a bumpy ride, as you can see from my commit messages:

>  day8: "part 1 works"                           # denial
> day13: "dumb broken"                            # anger
> day18: "stupid fish"                            # anger
> day18: "f*** parser combinators"                # anger
> day18: "borrowchecker why"                      # bargaining
> day20: "fin. right answer, wrong test result.." # acceptance

but i finished it! Given I was learning Rust and fighting the borrow checker as I went, i’m pretty happy with the result.

AoC 2023

My main goals for advent of code this year are:

  1. Speed

  2. Speed

  3. Learning New Stuff

2 speeds? DOUBLE the speed?? Let me explain..

Speed (1)

So my #1 goal is programming speed - getting a working solution as fast as possible. I think speed is an underrated skill and being able to program simple things quickly and easily translates to the ability to do more complex things.

Speed (2)

My other speed goal is runtime. Hanging out in one of the Rust Discord channels in 2021 and seeing the fastest solutions was inspiring. One repo with well optimized solutions can complete all problems in 2022 and 2021 in 10ms each! With that in mind I’ve got some code ready to benchmark Rust code using criterion and Python code using pyperf. These can arguably be benchmarked much more simply by just running time on the command line or something but frankly this is cooler and probably more accurate & extensible. For example I can test individual functions and these tools can likely ignore the overhead of starting a program (which would be significant especially in Python) and measure the core functionality alone. (perchance?)

Frankly there’s a lot I don’t know here (if you couldn’t tell from the above) so its just something I hope to pick up on as I try to optimize my solutions and compare with other solutions.

New stuff

And for the final goal, there are a few languages and concepts I want to maybe try and learn:

  • SIMD/Vectorization - I’m interested in learning more about parallelism at all levels - up from instruction-level parallelism via SIMD up to high performance computing tools, maybe like OpenMPI. Using OpenMPI is probably overkill for AoC though, so I’ll stick to trying to SIMD :) Hopefully there will be 1-2 opportunities to apply this during AoC, although I’m not sure given that I know very little about SIMD!!

  • JAX - an autodiff and numerical computing library that is roughly in the same space as numpy and pytorch. This is also in the parallelism bucket - it seems to have some nice facilities for vectorization and parallelization. It might be a more user friendly way to do some SIMD!

  • miniKanren - a logic programming language. I’m interested in learning more about biology and bioinformatics, and I heard about this becuase of mediKanren, a tool Matt Might wrote to diagnose his own son suffering from a rare condition - two mutations in the same gene, NGLY1, something that had never been seen before. There was no name, no literature and no treatment for this condition - his tools helped him treat his son! miniKanren supports a few languages including Python but it seems best suited to lisps so I might try Clojure.

  • Javascript - I may try to build a few cool visualizations in Javscript.

Overall it’s a strange jumble of technologies that caught my eye, some of which have niche applications so we will see how many I can actually use significantly during AoC.

Hey that wraps up my first blog post! Anyways, here’s Wonderwall: