Microcode inside the Intel 8087 floating-point chip: register exchange | Hacker News

Early floating-point processors in the 1970s leaned on microprograms because fast cache-based instruction execution was too expensive.

The FPS AP-120B is cited as an early VLIW-like design, showing how accelerator-style ideas emerged before modern RISC chips.

Later RISC and VLIW CPUs moved to single floating-point instructions backed by more complex hardware, improving throughput.

The shift from microcode to dedicated FP hardware reflects falling cache costs and rising processor capability over time.