Transcript Functional Verification of Hardware Designs

Functional Verification of
Hardware Designs


EE764 – Functional Verification of
Hardware Designs
Course Objectives




Learn to use verification tools and
experiment on actual circuits designed in
industry (or equal to)
Learn to plan & carry out effective
functional verification of a design
Learn to work in teams to debug designs
Both code and debug HDL design code
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-1-
Functional Verification of
Hardware Designs (cont)

By the end of the course the student
will have



Verified several examples provided by the
instructor (some may come from industry)
Have an understanding of why verification
is important
Have an understanding of the level of
complexity present in the verification of
modern microcircuits
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-2-
Syllabus/Grading/Project


Syllabus
Projects
-
Project 1 – Floating Point Adder
Project 2 – Develop a package for injected error
testing.
-
-
-
Students will work on projects 1, 2 in a group of two. Each
is due 2 weeks from when assigned
Project 3 – Fault tolerant circuit example which will
include combinational and possibly sequential
components.
Project 4 – Floating Point Execution Sub Unit.
-
Students will work on projects 3 and 4 in a group of three.
Each will be due three weeks from when assigned.
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-3-
History 101

VHDL 1st standardized in 1987 by IEEE



VHDL about to be standardized again and will
include many new aspects, several aimed
specifically aimed toward verification (PSL)
Verilog 1st standardized in 1995 by IEEE
Hardware Description Languages were the
beginning of several new advances in chip
and system design



Simulation of chips and systems
Synthesis of digital hardware
Verification
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-4-
History continued

1990 focus – Does chip work at all?


2000 focus – Does chip work in the
system environment as specified?
Today – more efficient verification
than in 2000.
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-5-
Level of complexity problem
Note that
this was
a 2003
chart!!!
There has
been no
slowdown
in fielded
systems.
10 core Xeon
2,600,000,000
Transistors
On a 512 mm2
Die in a 32nm
Process
~400,000,000
gates
From www.chipexpress.com
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-6-
Today’s design issue


Gate counts and system complexity growing
exponentially
Bulk of time in design of an new IC is spent
on verification


In most companies there is at least a 1-to-1 ratio
of design engineers to verification engineers.
Real problem is not how to create the 12
million (or more) gate IC but how to verify it.
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-7-
What is verification?



Verification is not a testbench or a series of
testbenches.
Verification is a process used to demonstrate
the functional correctness of a design. It is
the act of ensuring that the logic design
conforms to the specifications.
Verification does not insure that the
specification is correct. (Next real advance in
design will likely be specification languages! –
Note new topics in text – Open Vera and e!)
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-8-
Testbenches


VHDL (and Verilog) testbenches refer to
the code used to create an input
sequence to the design under test
(DUT) and optionally observe the
response.
In EE762 the testbenches do both and
the error signal indicates that the
results do not equal what is expected.
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-9-
Verification testbenches


In verification the testbench provides the
inputs and monitors the outputs.
The challenge of verification is to determine
what input patterns to supply and what is the
expected output of a properly working design
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-10-
EE762 testbenches


The testbenches used in EE762 check the
essentials of basic correct operation and are
less than a typical verification testbench. The
goal of the EE762 testbenches is to catch
typical mistakes, check the operation in
central and corner cases, and provide
feedback to the student (the error signal).
They are somewhat less than a testbench to
be used for production of an IC.
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-11-
Verification Cycle
DESIGN
Develop Environment
Create Testplan
Debug Design
Escape Analysis
Regression
Hardware Debug
RIT
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-12-
Verification Testplan






Schedule
Required tools
Input and completion criteria
Specific tests
What is expected to be found with each
test
What’s not covered by the tests
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-13-
Other parts of cycle

RIT – stands for Release Interface Tape



RIT is sending the design to the chip fab.
Chip is fabricated and then IC is tested.
Escape Analysis – Fully understand any bugs
in fabricated part.


A critical part of the IC verification process
Reproduce in simulation if possible


Lack of ability to reproduce means fix cannot be verified
Could misunderstand the bug
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-14-
Todays Work

Form into your group of two. (or three)



No groups of 4 permitted!
Read Chapter 1 of text
Assignment will be covered Wednesday.


Verification of a Floating Point Adder.
Behavioral and synthesis model provided.
Lecture 1 – Overview (rSp06)
©2008 Joanne DeGroat, ECE, OSU
-15-