Browse Prior Art Database

Alternating Scheme for Generating Register Rename Addresses

IP.com Disclosure Number: IPCOM000107451D
Original Publication Date: 1992-Feb-01
Included in the Prior Art Database: 2005-Mar-21
Document File: 1 page(s) / 63K

Publishing Venue

IBM

Related People

Levitan, DS: AUTHOR [+2]

Abstract

Register rename addresses or tags are allocated to tag requests by assigning each available tag to each request in order until no more are available. The size of the logic required to assign tags to requests grows exponentially with the number of tags and the number of requests. When a system has a large number of requests, say 16, and of available tags, say 16, then the logic array to allocate the tags cannot be economically implemented. This invention describes a method of efficiently splitting up the tag assignment logic into a manageable size.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 68% of the total text.

Alternating Scheme for Generating Register Rename Addresses

      Register rename addresses or tags are allocated to tag requests
by assigning each available tag to each request in order until no
more are available.  The size of the logic required to assign tags to
requests grows exponentially with the number of tags and the number
of requests. When a system has a large number of requests, say 16,
and of available tags, say 16, then the logic array to allocate the
tags cannot be economically implemented.  This invention describes a
method of efficiently splitting up the tag assignment logic into a
manageable size.

      In this invention the number of request and available tags are
subgrouped into two or more sets.  The set of available tags, which
contains the largest number of tags, is then gated to the assignment
array for the first (highest importance) set of tag requests.  The
second or other sets of available tags are gated to the lower
importance tag request assignment arrays at the same time.  The
gating is controlled by a priority logic which determines which group
of available tag groups is largest.

      Since the number of requests and available tags in each
subgroup are smaller, the assignment array size for each subgroup
thus decreases dramatically.  Since, in normal operation, sufficient
available tags are present in one of the subgroups to satisfy the
high priority request group, system performance is minimally reduced.

      In the ...