Browse Prior Art Database

Transparent Cache Reload Buffer Design

IP.com Disclosure Number: IPCOM000100672D
Original Publication Date: 1990-May-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 3 page(s) / 86K

Publishing Venue

IBM

Related People

Thatcher, LE: AUTHOR [+2]

Abstract

Disclosed is a transparent Cache Reload Buffer (CRB) design. In processor systems involving cache designs the reloading of a cache line impacts processor performance in a number of ways. This article describes a CRB design rich in function as a novel solution to increase performance.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 79% of the total text.

Transparent Cache Reload Buffer Design

       Disclosed is a transparent Cache Reload Buffer (CRB)
design. In processor systems involving cache designs the reloading of
a cache line impacts processor performance in a number of ways.  This
article describes a CRB design rich in function as a novel solution
to increase performance.

      This processor design implements a highly functional CRB design
to reduce performance bottlenecks.  In this design the processor has
the capability to perform read/write functions directly to the CRB.
Not, as in typical designs, only to the cache array.  This gives the
processor the ability to execute LOAD/STORE instructions directly to
the CRB.

      The unloading of the CRB to the cache array is simplified in
this design since the CRB presents no performance problems to the
processor. On the first cache miss, the CRB is loaded with a cache
line from memory.  Since the processor can execute LOAD/STORE
instructions to the CRB, the CRB need not be unloaded to the cache
array until second cache miss sequence.  Here, the bandwidth of the
cache has not been impacted by the unloading of the CRB, since during
a cache miss sequence the cache is available for unloading of the
CRB.

      Fig. 1 shows the CRB data flow path.  This illustrates the
processor data paths to perform read/write functions to the CRB and
unloading the CRB to the cache array.  Fig. 2 diagrams the processor
target compare logic.  This function indicat...