Floating-point unit

A processor component that performs arithmetic on floating-point numbers, including operations such as addition, multiplication, and division. Its hardware implements numerical formats and rounding rules.

IEEE 754: A technical standard defining binary and decimal floating-point formats, arithmetic behavior, rounding, and exceptional values. Its rules determine the results and edge cases the unit must produce.
Significand: The digit sequence in a floating-point representation that carries a number’s precision. Arithmetic circuits operate on significands after accounting for their exponents.
Graphics processing unit: A processor designed to execute many operations in parallel, especially for graphics and data-parallel workloads. GPUs contain extensive floating-point hardware for graphics and scientific computation.
Integer arithmetic: Arithmetic on whole-number representations, usually with fixed-width binary values. It uses different representations and hardware behavior from floating-point arithmetic.
Intel 8087: An early x87 floating-point coprocessor introduced by Intel in 1980 for the 8086 and 8088 processors. It exemplifies a separate chip that supplied floating-point capability to a general-purpose CPU.
Floating-point arithmetic: Arithmetic on numbers represented with a finite significand and a scaled exponent. The unit is hardware for carrying out these operations.
Exponent: A value that scales a floating-point significand by a power of the format’s base. Exponent comparison and adjustment are central to aligning and normalizing operands.
Scientific computing: The use of computation to model, analyze, or solve problems in science and engineering. Its simulations and numerical methods rely heavily on floating-point operations.
Fixed-point arithmetic: Arithmetic on scaled integers whose binary or decimal point position is fixed by convention. It trades floating-point range for simpler, predictable scaling and hardware.
x87: The instruction-set architecture for Intel’s legacy floating-point operations, originating with the 8087 coprocessor. Its stack-based operations shaped desktop floating-point programming for decades.

Linked from 3 pages