8087

The 8087 Numeric Coprocessor is an optional floating-point unit designed to work with Intel's 8086 and 8088 microprocessors. Introduced in 1980, it accelerates floating-point arithmetic by handling complex numerical calculations that would otherwise burden the host CPU. The processor plugs into the same system bus and communicates with the 8086/8088 to execute floating-point instructions, returning results to memory or registers as required.

Architecture and data formats

The 8087 uses a stack-based register model with eight floating-point registers, ST(0) through ST(7). It operates on data in three precision formats: 32-bit single precision, 64-bit double precision, and 80-bit extended precision, the latter being the most common internal working format. The coprocessor maintains a control word and a status word to manage rounding modes, precision, and exception handling. Typical exceptions include invalid operation, zero divide, overflow, underflow, and precision loss.

Instruction set

The 8087 implements a range of arithmetic and data-movement instructions, plus transcendental functions. In addition to FADD, FSUB, FMUL, and FDIV, it provides operations for comparisons and conversions between formats. Transcendental instructions enable computations such as sine, cosine, and tangent, as well as exponential and logarithmic functions, enabling scientific and engineering software to perform complex calculations more efficiently than with integer-based processing alone.

Interface and influence

When present, the 8087 accelerates software that relies on floating-point math, commonly used in scientific, engineering, and graphics applications of the era. It was later complemented or replaced on many systems by the 80287, 80387, and 80487 coprocessors as Intel expanded the architecture, and floating-point units eventually became integrated into mainstream CPUs.

See also: Floating-point coprocessor, x87 architecture, Intel 80287/80387.