Floating Point Coprocessors

Floating point coprocessor The designer of microprocessors wants to expand the instruction set nearly indefinitely, but it is limited by the amount of silicon available (not to mention testability and complexity issues) . Therefore, a true microprocessor represents a compromise between what is desirable and acceptable for most chip users. For example, the 68020 microprocessor is not optimized for calculations that require much science.

The 16-bit Intel x86 processor (including 80386) does not include a floating-point unit (FPU). Intel introduced 8087, 80187, 80287, and 80387 math coprocessors to add hardware floating point and override capabilities to CPUs from 8086 to 80386. 8087 is used in 8086/8088 and 80186/80188, it is used in 80186 instead of 80188, 80877 is used in 80286, 80387 in 80386 is used. The combination of x86 CPU and x87 coprocessor forms a single part. Multichip: Two chips are programmed into one unit using a single integrated instruction set. The 8087 and 80187 coprocessors are parallel to the data bus and address buses of their parent processors and execute direct instructions on them.

Early microprocessors did not actually have floating point functions and were exact integer computers. Floating-point calculations are usually done in the form of mathematical coprocessors, with separate dedicated hardware. Shortly thereafter, the transistor size became smaller, the floating point unit could be placed directly on the main CPU chip, and the latest integer / floating point microprocessor was born. Of course, adding floating point hardware means adding floating point instructions. In the x86 world, this means the introduction of the x87 floating point architecture and its (currently desperate) stack based register model.