8088 instruction stream. The instructions that the NPX interprets as special purpose numer- ics instructions are regarded almost like "no-operations" for the 8088. The NPX con- tains an additional register set of eight 80-bit floating point registers which are mani- pulated with by the additional numerics instructions. Together, the 8088 with the NPX have approximately 100 times the per- formance of a standalone iAPX 88 system for numerics-intensive applications.
I/O Processor
The 8089 lOp, on the other hand, does not receive instructions from the 8088 instruction stream. It is a separate microprocessor with its own instruction set. The lOP is an input/output channel processor and off-loads 110 interfac- ing from the 8088 general purpose CPU. The lOP's instruction set, different from the 8088, is specifically tailored for peripheral control and high speed data transfer. With the lOp, it is
possible to configure a dual-bus system, where the 8089 interfaces with peripheral devices on a separate "local" bus while the 8088 runs its application programs in parallel, interfacing with memories over the system bus.
The lOP has a high-speed direct memory access (DMA) mode that transfers data between memory and peripherals or between memory and memory at 1.25 megabytes per second. The lOP is also capable of on-the-fly processing activities such as masked com- parison operations or data translations. If you have an application that requires very high performance floating point numerics capabilities, numerous peripheral devices, or very high performance peripheral devices, the NPX and lOP should be considered for inclusion in your system. More information on these devices is contained in other manu- als from Intel. This book will focus on single CPU-systems build around the 8088 alone.
LOCAL 1/0 BUS
SYSTEM BUS
ARCHITECTURE EXTENDS FOR EVEN MORE PERFORMANCE
Figure 1-26. iAPX 88 Processor Extensions
