To develop or buy
a Verification IP
Atul Bhatia, Director, nSys Design Systems
Independent
Interpretation
The most important benefit of buying the VIP rather than making
it is that a commercial VIP provides a totally independent, clear
and unambiguous interpretation of the specifications of a protocol
for which it is designed. Until the time when the specifications
themselves are written such that they have no ambiguity, this
independent interpretation is invaluable. It gets even more important
when the specifications are those of a complex protocol. In case
the commercial VIP is already proven by use with RTL designs of
other users, the value increases to several times of its cost
to the user. This is because the user is assured inter-operability
and compliance due to the use of this VIP with other designs.
Availability of Test Suites
If the VIP is available not only with the Bus Function Models,
Monitor and Protocol Checker but also with Test Suites, then the
user is able to save considerable time that would have been spent
in writing the test cases that are available as part of the Test
Suites. Not only are the test cases difficult to identify and
time-consuming, most engineers write them grudgingly. Availability
of test cases in source code makes it easier to modify and create
additional cases unique to the user’s designs.
Packaging
Just as a home-made dish cannot be packaged as attractively as
the one that is available from a restaurant, an in-house VIP will
also not be packaged as well as a commercial VIP. Some of the
distinct features of the packaging that are very difficult to
incorporate in an in-house VIP are: availability of a well-documented
User Manual, Flash-based tutorials, comprehensive FAQs based upon
queries faced by existing users, a self-service bug tracking portal,
Application notes and a well designed GUI. In case the VIP provider
is offering a family of VIPs, the additional features of packaging
would be: consistency of interface, installation, operation, and
documentation across the VIP family. Since the APIs for a family
would be well thought out and consistent across the family of
VIP, it would reduce the requirement of learning the API while
using additional interfaces.
FPGA Designs
At the other end of the spectrum of the develop-or-buy decision
are the FPGA developers who do not even consider using simulation
for verification. They feel that they would find the bugs at the
system level in any case and the bug would just require a change
in the RTL rather than throw away the chip as would be the case
for an ASIC developer. This belittles the complexity of FPGA designs
brought about by increase in their size. Fortunately, FPGA developers
are becoming aware of this and have begun verifying their designs
extensively in simulation too. Especially since it is very easy
to compute that the cost of a license of a commercial VIP may
be less than the cost of one engineering man-month. We have not
factored in the impact of the cost of delay in product which will
be many times this cost.
RTL IP
One of the ways of overcoming the well known productivity gap
is the use of Design or RTL IP. A 2002 study by Collett International
Research revealed that 14 percent of all chips that failed had
bugs in reused components or imported IP. The other key problem
while using an IP is that its integration with rest of the design
has to be thoroughly tested. Thus a VIP should be able to verify
the IP at the block level and also provide features for System
level verification to ensure the correctness of integration of
the IP with rest of the design.
Bundled VIP
Some situations when the develop-or-buy decisions get difficult
to make are when an RTL IP has been used for a while and can be
expected to be bug-free or when an RTL IP is available with a
VIP. It has been observed that even if an RTL IP is field-proven,
it may still have several bugs that are brought out by a new design
(as luck would have it). A free VIP that may be available with
RTL IP unless it is from a 3rd party vendor is not good enough
for use in a commercial project. In the case when a 3rd party
VIP is available bundled with the IP, it is likely to have a reduced
feature set sufficient to demonstrate the working of the IP and
the VIP while the features required to perform verification of
the IP as well as its integration with the rest of the design
may not be available. In fact, the extra effort spent in using
the free or scaled-down VIP is several times the cost of a good
3rd party full feature VIP.
Conclusion
Verification productivity that is required to meet the challenges
of tomorrow can only be met by incorporating widely used and proven
VIPs in the verification flow. A family of VIPs having consistency
of interface as well as look and feel increases this productivity
across projects. The case for developing a VIP can only be made
for proprietary interfaces. For all other interfaces, it is better
to buy rather than develop a VIP.
