Browse Prior Art Database

Technique for the Enhancement of Event-Driven Simulation Effectiveness

IP.com Disclosure Number: IPCOM000105340D
Original Publication Date: 1993-Jul-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 6 page(s) / 135K

Publishing Venue

IBM

Related People

Cash, RC: AUTHOR

Abstract

The accuracy of computer simulation of digital devices is limited by the assumptions made for the fixed timing of the stimuli provided to the circuit-under-test (CUT) by the testcases and behavioral models. Described below are two models that can be placed between the CUT and the testcases or behaviorals to randomly vary the timing of the stimuli.

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

Technique for the Enhancement of Event-Driven Simulation Effectiveness

      The accuracy of computer simulation of digital devices is
limited by the assumptions made for the fixed timing of the stimuli
provided to the circuit-under-test (CUT) by the testcases and
behavioral models.  Described below are two models that can be placed
between the CUT and the testcases or behaviorals to randomly vary the
timing of the stimuli.

      Event-driven computer simulation of digital devices relies upon
the accuracy of the circuit delays assumed for the circuit-under-test
(CUT) and the accuracy of the circuit delays of the other devices to
which it is connected.  Ideally, a CUT should be simulated with CUTs
of the other devices with which it will eventually be used.  However,
for a variety of reasons, the stimuli which would normally be
provided by the other CUTs must often be synthetically produced by
user-provided testcases (programs) and/or behavioral models of the
other CUTs.  Although the delay models to be introduced below will
accommodate input from either testcases or behavioral models, for
simplicity assume that all the CUT's stimuli are provided by
behavioral models.

      When coding a behavioral the user makes assumptions about the
relative timing of the various signals the model will be supplying to
the CUT.  This timing is hard coded into the behavioral so that it
generates exactly the same timing on its outputs every time some
sequence 'A' is invoked.  Unfortunately, the relative timing of the
sequence is unlikely to closely match the timing of the actual
hardware that the behavioral is intended to represent, especially if
the model is a generic representation of a class of devices.

      To introduce a degree of variability into the sample sequence
'A', this disclosure proposes the placement of random delay
generation blocks on selected lines between the behavioral model and
the CUT as shown in Fig. 1.  These random delay generators are
themselves a type of behavioral model.  Block X is a model for
unidirectional signals and block Y is a model for bidirectional
signals.  The notation X-1, X...