Information Technology Reference
In-Depth Information
Testbench
Reference Model
Stimuli
Source
Checker
Driver
DUV
Monitor
FIGURE 19.11
Basic test bench model.
The test bench architecture used in this verification method is characterized for
modularity and reusability of its components. The test bench model comprises all
elements required to stimulate and check the proper operation of the design under
verification (DUV); the DUV is an RTL description.
Figure 19.11 shows the basic test bench model in which the stimuli source is
based on aid tools and applies pseudorandom-generated test cases to both the DUV
and the reference model, a module with a behavioral description at a higher level of
abstraction. The driver and monitor are blocks aimed to convert the transaction-level
data to RTL signals and vice versa. Outputs from the simulation performed on both
the reference model and the RTL modules are compared, and outcomes on coverage
are computed and presented in the checker.
The designer must carefully plan aspects of the coverage model and the stimuli
source. The stimuli can be classified in the following categories:
Directed cases, whose responses previously are known (e.g., compliance test)
Real cases dealing with expected stimuli for the system under normal conditions
of operation
Corner cases, aimed to put the system on additional stress (e.g., boundary
conditions, design discontinuities, etc.)
Random stimuli, determined by using probability functions (Bergeron, 2003)
Moving to the coverage related to the strategy, the coverage is an aspect that
represents the completeness of the simulation, being particularly important when
random stimuli are applied. Functional coverage usually is considered the most
relevant type because it directly represents the objectives of the verification process,
and it is limited by project deadlines.
Each engineer has his own verification coverage measurement metrics. Thus, to
deal with the complexity of a problem, the engineer follows some generic steps for
functional coverage. The steps are as follows:
A judicious selection must be made on a set of parameters associated with input
and output data, for instance, the size of packets (words with specific meaning) as
keys, passwords, and so on.
For every selected parameter, the designer must form groups defined by ranges of
values it may assume, following a distribution considered relevant.
Search WWH ::
Custom Search