A note on C99_and_julia_introduction, a short course repo I put together in Feb–Mar 2021.

The premise

If you teach C99 and Julia in parallel, at the numerics layer, the languages agree more than students expect. Both give you column-major arrays (Julia by default, C if you decide to), both let you reason in terms of strided memory, both expose the actual machine representation. The things that differ — GC, dispatch, the notebook-friendly tooling — are not the things that matter when you’re teaching what a for loop over a matrix actually does.

What the course covered

• Arrays, indexing, strides. Drawing the same memory diagram for both languages.
• Loops vs broadcast. Showing that Julia’s . broadcast lowers to a loop you could have written in C.
• A small numerical problem (something along a Cholesky or a Jacobi iteration) written both ways, benchmarked both ways, profiled both ways.
• Where Julia’s zero-cost-abstraction promise turns out to cost runtime — specifically when you hit a type-unstable path and the JIT falls back to boxed values.

What I’d revise

The course predates a lot of the modern C++ / Rust / Julia interop story. If I were teaching this today, I’d probably add a Rust column and put them side by side in three. Also the Julia compile-time-vs-first-run-time story has changed enough with native-code caching that the “just wait for compilation” slide needs rewriting.

Code / slides scaffolding @ github.com/zeta1999/C99_and_julia_introduction.