Very excited as I've just figured out the reason why previous attempts to optimize my Pocket sprite drawing code actually made it slower, is because modulo is extremely slow on this vexriscv softCPU. This is SO unsurprising in retrospect but I hadn't realized it until now. Already getting significant speedups with minor tweaks and I think I can get even more.
I got too used to the "nothing is slow except branches and cache misses" era of CPUs. I was seduced by the decadence of the modern world!
oblomov