Intel 8087 Chip's Floating-Point Addition and Subtraction Microcode Examined

Introduction of the Intel 8087

In 1980, Intel released the 8087, a math coprocessor designed to enhance floating-point calculations in systems like the IBM PC. This chip provided a substantial performance increase over the 8086 processor, particularly for operations such as addition, which saw speed improvements of almost 100 times.

Internal Architecture and Microcode

The 8087's internal structure was examined by decapping a chip and imaging its die. The central microcode ROM contains 1648 micro-instructions that govern the chip's operations. The datapath, located in the bottom half of the chip, is responsible for performing 80-bit floating-point calculations.

Key Components for Arithmetic

Within the datapath, several functional blocks are crucial for addition and subtraction. The integer adder is a central component, processing arguments from registers and storing results in the sum register. A large shifter component allows rapid bit shifting for value alignment, while leading zero circuitry counts zero bits for efficient processing. The exponent converter manages exponents and detects overflow or underflow conditions. These specialized circuits contributed to the 8087's high performance in floating-point arithmetic.

Importance of Floating-Point Numbers

While computers primarily use integers, many applications require fractional numbers and a wider numerical range than integers provide. Floating-point numbers address these needs by offering fractional representation over an extensive range, making them essential for complex computations.