Verification - PowerPoint PPT Presentation

1 / 20
About This Presentation
Title:

Verification

Description:

Title: PowerPoint Presentation Author: Joanne E. DeGroat Last modified by: Joanne Degroat Created Date: 3/21/2001 4:02:40 PM Document presentation format – PowerPoint PPT presentation

Number of Views:48
Avg rating:3.0/5.0
Slides: 21
Provided by: Joann258
Category:

less

Transcript and Presenter's Notes

Title: Verification


1
Verification The importance
2
Project 1 is underway
  • Open discussion of the Verification Plan for
    project 1.
  • Discussion.

3
Verification The Importance
  • Verification model
  • Verification approaches
  • Verification issues
  • Versus testing
  • Design for verification
  • Reuse
  • Type I II mistakes
  • Reference Writing Testbenches Functional
    Verification of HDL Models, Janick Bergeron,
    Kluwer Academic Press, 2000. for all figures in
    this lecture.

4
Importance of Verification
  • Verification consumes 70 of design effort and is
    often 80 of total code.
  • Typical textbooks on VHDL or Verilog treat
    verification and testbench creation only very
    lightly
  • Navabi textbook used in EE762
  • 3 pages on verification
  • 4 pages on testbenches
  • In todays environment verification resides
    squarely on the critical path.

5
Importance (cont)
  • Verification is currently the target of several
    new tools
  • That reduce to overall verification time by
    enabling parallelism of effort
  • Write and debug testbenches in parallel with each
    other and with implementation of design
  • That enable higher levels of abstraction
  • Work more efficiently as low-levels of detail are
    ignored.
  • This is acceptable as long as reduction in
    control and observability are acceptable and
    chosen wisely
  • That employ automation
  • Must fully understand what tool is and is not
    capable of.

6
Reconvergence Model
  • A conceptual representation of the verification
    process.
  • One of the most important questions is What are
    you verifying?

7
Reconvergence Model (cont)
  • Purpose of verification - ensure result of some
    transformation is as intended or expected, i.e.,
    reconcile result with starting point.
  • Examples of possible transformations
  • RTL coding of a design from a specification
  • Insertion of a scan chain
  • Synthesizing code into a gate-level netlist
  • Layout of a gate-level netlist in the target
    technology

8
Human Intervention
  • Necessary for a group or individual to interpret
    a specification
  • If the same individual does both the
    interpretation of the specification and the
    verification of the coding the common origin is
    the interpretation not the specification. If the
    interpretation is wrong

9
Reducing human-itroduced errors
  • Human-introduced errors reduced through
  • Automation Take human intervention completely
    out of the process (or kind of still there in
    the specification and design of the automation
    process)
  • Poka-Yoka Reduce human intervention to simple
    and foolproof steps. Usually last step toward
    complete automation. Verification process
    remains an art.
  • Redundancy Every transformation accomplished is
    either verified independently by another
    individual or separate transformations are done
    and the results compared.

10
Verification Approaches
  • Formal Verification - Equivalence Checking
  • Mathematically prove origin and output are
    logically equivalent and transformation preserved
    functionality
  • Keeps synthesis tool honest, i.e., identify when
    the synthesis tool produces an incorrect result.

11
Verification Approaches
  • Formal Verification - Model Checking
  • Assertion or characteristics of a design are
    formally proven or disproved.
  • Example Are all states in a state machine
    reachable?

12
Verification Approaches
  • Functional Verification
  • Main purpose is to insure design implements
    intended functionality
  • Only shows that design meets intent of
    specification but cannot prove it.
  • Can prove that a design does not implement
    intended function by finding a single case.

13
Verification Approaches
  • Testbench Generation
  • Use code coverage metrics and the source code
    under analysis to generate a testbench that
    either meets some code coverage or exercises the
    design to violate a property.
  • Actually do little to contribute to the
    verification process

14
Functional Verification Approaches
  • Black Box verification
  • Performed without knowledge of actual
    implementation
  • Access to I/O interface only
  • No access to internal signals and no knowledge of
    structure and implementation
  • Only approach possible if functional verification
    was implemented in parallel while the design was
    being done

15
Functional Verification Approaches (continued)
  • White Box verification
  • Full visibility and controllability of internal
    structure and signals
  • Tightly integrated with a particular
    implementation
  • Gray Box verification
  • Compromise between the other two
  • Uses the I/O interface but also has knowledge of
    significant features

16
Testing versus Verification
  • Verification Ensure design meets its functional
    intent
  • Testing Ensure part was manufactured correctly

17
Design for Verification
  • Just as design-for-test requires addition of
    circuitry, add non-functional features to
    facilitate verification
  • As you do a design need to ask not only
  • what is this supposed to do?
  • but also
  • how is this going to be verified?

18
Design Reuse
  • Major obstacle to reuse is cultural!!
  • Engineers have little willingness to incorporate
    an unknown design into their own. Do not trust
    that the design is as good to the one they can
    produce.
  • Requires trust in the design to be used
  • Trustworthiness can be demonstrated through a
    proper verification process

19
Type I and Type II mistakes
  • Type I False negatives verification finding
    errors where none exist Type I errors are easy
    to identify
  • Type II False positives MOST SERIOUS failed
    to identify an error where one exists

20
Project 1 and Type I/II
  • When you find discrepancies what is approach
  • Establish if reference model or RTL model is the
    source of the error.
  • High probability that it is the RTL model, but
    that is not a certainty! ?
  • Elevates analysis of errors to a critical level
    in the verification effort.
Write a Comment
User Comments (0)
About PowerShow.com