Hardware architecture
The hardware architecture provides several capabilities
that are required for multiprocessing.
Recent PC hardware architectures
include a number of performance enhancements
(such as per-processor data/instruction caches,
fast memory/CPU bus,
and high-speed I/O systems)
that were once found only in the large-scale environments
of minicomputers and mainframes.
SCO multiprocessing systems
use the following hardware features:
-
The same byte is visible at the same address
for all processors,
and all memory is accessible by all processors.
-
A transparent data/instruction cache
that can compensate for the fact that
bus and memory speeds are generally slower
than processor speeds.
A cache reduces bus contention
by allowing memory fetches to be completed
without requiring the CPU to wait for the bus
or for the transfer of data to main memory.
Reducing bus contention also allows
all the CPUs faster access to the bus
when main memory access is required,
which improves general system performance.
Hardware caches with high hit rates
improve both local CPU performance
and overall system performance.
-
All CPUs can interrupt each other.
This is required for priority-based scheduling,
optimal clock interrupt servicing,
and distributed device driver execution.
© 2005 The SCO Group, Inc. All rights reserved.