It still boggles my mind that Chris Sawyer wrote Roller Coaster Tycoon was written in pure assembly, only failing back to C to handle windows sound apis. The game is a masterpiece, it defined an entire game genre for decades. And RCT ran smooth on my pentium 2 400mhz processor, even with hundreds of guests and dozens of rides. I believe there are many magnitudes of performance hiding in poor software abstractions in modern software.

What I'd really love to know is how the game was ported to Android/iOS recently, because as far as I can tell the game is literally a 1:1 copy, so I wonder if they wrote some kind of x86 assembly interpreter for it? I cannot imagine they just remade the game in some other engine and achieved this level of accuracy.