Speculative pre-fetching additional line on cache miss if no request pending in out-of-order processor

作者: David Jerome Johnson , Gregg B Lesartre

DOI:

关键词:

摘要: Speculative pre-fetching and pre-flushing of additional cache lines minimize miss latency coherency check an out order instruction execution processor. A pre-fetch/pre-flush slot (DPRESLOT) is provided in a memory queue (MQUEUE) the out-of-order The DPRESLOT monitors transactions between system interface, e.g., bus, address reorder buffer (ARBSLOT) and/or interface (CCCSLOT). When detected, causes one or more addition to data line, which caused current miss, be pre-fetched from hierarchy into (DCACHE) anticipation that would required near future. write back detected as result check, lines, line currently being written back, pre-flushed respective processor owns by requesting logic included prevents request for when another has already been made data.

参考文章(3)
Milford John Peterson, Kevin Arthur Chiarot, Michael John Mayfield, Era Kasturia Nangia, Instruction pre-fetching of a cache line within a processor ,(1995)
Sanjay Raghunath Deshpande, David Mui, Praveen S. Reddy, Method and system for bypass prefetch data path ,(1999)