Browse Prior Art Database

Method for Dynamically Balancing Computer Graphics Workload between Host Workstation Central Processing Unit and Dedicated Graphics Adapter

IP.com Disclosure Number: IPCOM000114904D
Original Publication Date: 1995-Feb-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 4 page(s) / 167K

Publishing Venue

IBM

Related People

Einkauf, MA: AUTHOR [+2]

Abstract

The calculations involved in rendering complex graphics objects or scenes is called the "graphics workload". Other calculations done by a computer system can be thought of as the "non-graphics workload". Typically the graphics workload of a particular application running on a computer system is much greater than the non-graphics workload.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 34% of the total text.

Method for Dynamically Balancing Computer Graphics Workload between
Host Workstation Central Processing Unit and Dedicated Graphics Adapter

      The calculations involved in rendering complex graphics objects
or scenes is called the "graphics workload".  Other calculations done
by a computer system can be thought of as the "non-graphics
workload".
Typically the graphics workload of a particular application running
on a computer system is much greater than the non-graphics workload.

      Various computer graphics system have been designed to deal
with the burdensome graphics workload.  Some systems have no special
hardware dedicated to graphics and rely on the workstation Central
Processing Unit (CPU) to perform all calculations necessary for the
graphics workload.  Though these systems are often relatively
inexpensive, they can be quickly overwhelmed by the graphics
workload.  Other systems employ exotic graphics hardware, with
accompanying software (or microcode) to process the graphics
workload.  These systems excel at rendering graphics quickly, but are
usually quite costly.  Also, when the user does not require complex
graphics, the dedicated graphics hardware is severely under-utilized,
diminishing the value of the system.

      Finally, some systems use both the Host CPU and a graphics
adapter (with graphics hardware and accompanying microcode) to
process the graphics workload.  These systems are moderately priced
and provide moderate graphics performance.  Typically, the graphics
adapter is offered in various configurations resulting in differing
graphics performance.  The adapters will fit workstations of varying
CPU power.  The overall efficiency of the computer system depend on
how graphics system designers choose to balance (i.e., split) the
graphics workload between workstation CPU and adapter.  The "split"
chosen may provide great efficiency for one workstation/adapter
configuration.  However, if a workstation CPU is more powerful than
the CPU in the "optimum" configuration's, the host may be idle at
times, waiting on the graphics adapter to finish its work.  Likewise,
if a less powerful CPU is in use with the same adapter, the adapter
will be idle at times, waiting on the host to give it more work.
Graphics system designers could overcome this efficiency problem by
having unique workstation and adapter software for each
workstation/adapter configuration in an attempt to get the best split
for each configuration.  Although possible, this is clearly becomes
an unmanageable problem if there are more than a few possible
workstation/adapter combinations.

      The system described herein solves the problem of providing
optimum efficiency for the graphics workload, regardless of
workstation CPU processing power and graphics adapter processing
power.  This system 1) determines cpu capability and adapter
capability, 2) determines amount of total work required for a
graphics object, and 3) configures...