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

Means for Reducing Paging Overhead

IP.com Disclosure Number: IPCOM000101800D
Original Publication Date: 1990-Sep-01
Included in the Prior Art Database: 2005-Mar-16
Document File: 2 page(s) / 71K

Publishing Venue

IBM

Related People

Milenkovic, M: AUTHOR [+2]

Abstract

This article describes a means to force context swaps to occur when a process is likely to have a small number of active pages in active memory.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Means for Reducing Paging Overhead

       This article describes a means to force context swaps to
occur when a process is likely to have a small number of active pages
in active memory.

      The basic idea is to recognize when a process has abandoned
pages in main memory and to use such an event to trigger a context
swap if one is pending.  The objective is to reduce the overhead when
restarting a process by reducing the number of pages that have to be
reloaded.  This will improve performance of a paged virtual-memory
system.

      Description of the Swapping Mechanism The basic idea is to arm
a context swap to occur at the end of a time quantum, but to defer
the swap at the end of time quantum until the program reaches a point
at which a swap is not very costly, or until it reaches the end of
a second time quantum that suspends the program unconditionally.  A
typical instruction that reduces swapping overhead is a subroutine
entry or exit, a block entry or exit, or a loop exit.  For each of
these instructions, the current set of active pages in memory and in
cache will tend to change.  Thus, if program execution is suspended
at this point, when the program is restarted it will load its new
context anyway, and no or little work will be expended restoring the
pages and cache lines that will not be referenced again in the
immediate future.

      There are several events that can be used to trigger context
swaps in paged virtual-memory systems, and the relative utility of
the trigger events differs for paging and for cache.  For example, a
subroutine entry may indicate a change of footprint size in cache,
but it does not necessarily change the number of active pages in real
memory, because the caller may not be reentered in the short term but
...