Patent · US Expired

Data processing with first level cache bypassing after a data transfer becomes excessively long

US5729713A · kind A · utility

14Cited by
17References
13Claims
0Family size

Assignee

Inventor

Key dates

Filing dateMar 27, 1995
Grant dateMar 17, 1998
Priority date
Expiry dateMar 27, 2015

Classification

  • Technology area (CPC Y)Emerging Cross-Sectional Technologies
  • CPC primaryY02D10/00
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

The hit rate of a cache memory (43) is improved by monitoring data transfer commands on a command bus (51) by Non-Cache circuitry (45A). Cache data replacements are inhibited after a consecutive sequence of data transfers which exceeds a threshold number of data transfers are detected by Non-Cache circuitry (45A). The threshold number is selected to be an amount of data transfers which is large enough to imply that a large block of data is being transferred. Such large data blocks tend to flush the cache and reduce subsequent cache hit rate. Other sources of cache inhibit signals may be included, such as System Cache Enable (SKEN), to inhibit caching for other reasons, such as when non-cacheable areas such as video memory are being accessed. Inhibiting useless cache data replacements in this manner improves hit rate and reduces power consumption.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.