Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Secondary Targets for I-Shadow Prefetching

IP.com Disclosure Number: IPCOM000043914D
Original Publication Date: 1984-Oct-01
Included in the Prior Art Database: 2005-Feb-05
Document File: 1 page(s) / 11K

Publishing Venue

IBM

Related People

Pomerene, JH: AUTHOR [+4]

Abstract

Analysis of the performance of the I-Shadow shows that roughly 40% of all targets point to the current MRU (Most Recently Used) within congruence class) cache lines. The ability to prefetch with high coverage is limited by this phenomenon. An enhancement to coverage does occur with an additional second level (Double Target Set) for all I-Shadow positions, but this can be improved upon by using the MRU-Target Hit as a trigger. Although a secondary I-Shadow defined along the lines similar to the primary I-Shadow but accessed or these triggering events is sufficient, a simple approach which handles these events more symmetrically involves duplicate entries in the primary shadow. Only one of the entries will be accessed on a I-Shadow Hit (the one which is MRU). In the event of an MRU-Target Hit, the second entry is accessed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 100% of the total text.

Page 1 of 1

Secondary Targets for I-Shadow Prefetching

Analysis of the performance of the I-Shadow shows that roughly 40% of all targets point to the current MRU (Most Recently Used) within congruence class) cache lines. The ability to prefetch with high coverage is limited by this phenomenon. An enhancement to coverage does occur with an additional second level (Double Target Set) for all I-Shadow positions, but this can be improved upon by using the MRU-Target Hit as a trigger. Although a secondary I- Shadow defined along the lines similar to the primary I-Shadow but accessed or these triggering events is sufficient, a simple approach which handles these events more symmetrically involves duplicate entries in the primary shadow. Only one of the entries will be accessed on a I-Shadow Hit (the one which is MRU). In the event of an MRU-Target Hit, the second entry is accessed. If none is present, the event causes the creation of duplicate first level and different second level entry. Since confirmation involves both (Key, Target), the duplicate entry at the first level creates no problem, since the second level is guaranteed to be different.

1