Audio DSP
A soundtrack computed rather than recorded is a signal processing problem with an audible pass mark. These articles cover the parts of software synthesis where the theory is not optional: why a naive sawtooth aliases and what a two-sample polynomial correction does about it, how a filter keeps its tuning near Nyquist, and how the arithmetic of an envelope shapes what a listener actually hears.
1 article
PolyBLEP: two samples that fix oscillator aliasing
A naive sawtooth at 1199.71 Hz puts alias energy 15.15 dB under its harmonics. A two-sample polynomial correction pushes it to 31.71 dB, for six flops
Related topics
Demoscene
Demos written in C with no assets at all: every pixel and every audio sample computed at runtime, and the techniques that make that possible offline.
Systems Programming
C and C++ close to the machine: memory layout, atomics and thread pools, what the compiler is allowed to assume, and measurements that settle it.