Motorola 68881
The building blocks MC68881 and MC68882 from Motorola (later Freescale ) were floating - coprocessors , which together with the processors MC68020 and MC68030 could be used at the coprocessor interface, or as a companion device for any other processors.
With the MC68881 / MC68882 , the instruction set of the MC68020 or MC68030 has been expanded to include various floating point operations, as well as eight floating point registers, each with a length of 80 bits and the associated status register. You implement the IEEE 754 standard largely and correctly. The number formats supported are simple (single, 32 bit), double (double, 64 bit) and extended (extended, 80 bit) precision as well as packed BCD representation.
In addition to the basic arithmetic operations, trigonometric and exponential functions and format conversions are also supported.
There is relatively little difference between the MC68881 and MC68882 . It is practically the same chip, but the MC68882 has a revised microcode that allows some functions to be carried out in parallel internally, e.g. For example, format conversions can be processed in parallel with arithmetic operations. This allows the MC68882 to perform slightly better than the MC68881 for certain operations .
The floating point unit was integrated in the MC68040 processor following the MC68030 . The instruction set is largely compatible with the MC68881 / MC68882 . For example, the trigonometric functions on the MC68040 are no longer carried out as commands , but rather using a function library.
The FPUs were used with Atari MegaSTE , Atari TT , Atari Falcon 030 , Apple Macintosh II , the Amiga 3000 , and some Amiga accelerator cards (especially for the Amiga 1200 and Amiga 2000 ).
Web links
- MC68000 Family Manual, also describes MC68881 / MC68882 (PDF file; 2.28 MB)