HP Vectra XU 6/XXX Optimization Manual page 38

Hp vectra xu 6/xxx - guide to optimization performance
Hide thumbs Also See for Vectra XU 6/XXX:
Table of Contents

Advertisement

Dedicated Level-Two Cache Bus
The Pentium Pro increases the performance of its level-two cache by accessing it over a
dedicated, 64-bit internal bus. This means that accesses to the cache will cause no contention
with any operations on the local bus. It also ensures that the level-two cache is able to operate
at the full internal clock speed of the processor.
Cache Coherency
One important implication of using cache memories is that duplicate copies of data exist in
different parts of your PC. With duplicate copies of the same data, a mechanism must be used
to ensure coherency between the copies. This is particularly true for dual processor
configurations, where duplications can exist not only between a processor cache and memory,
but also between the caches in the two processors.
The Pentium Pro's level-two data cache is a copy-back cache, meaning that if the processor
writes data that produces a cache hit, then that data will be written directly in the level-two
cache. The cache will then copy the data back to memory at a later time. Until the cache
copies the data back to memory, the copy of the data held in memory is no longer valid.
If another device in your PC requests this data from memory, the Pentium Pro processor must
be able to intercept the request so that it can supply the correct version of the data. To meet
this requirement, the Pentium Pro processor uses two cache coherency mechanisms: the MESI
(Modified, Exclusive, Shared, Invalid) protocol; and bus snooping.
The MESI Protocol
The MESI protocol defines the status of data in the cache using four states:
Modified — the data is valid, but different to that in memory (the data in memory is
invalid)
Exclusive — the data is valid and contained in one processor cache only

Advertisement

Table of Contents
loading

This manual is also suitable for:

Vectra xu

Table of Contents