TechOnLine Publication Date: Mar. 6, 2000
Automated IP Integration: Gaining Cycle
Time Advantage
Dr. Stefan
Scharfenberg
Motorola, SoCDT
With increasing pressure to shorten design
cycle times, system integrators have realized the advantage of
reusing blocks of intellectual property (IP). Today, two forces are
pushing designers to optimize reuse in their designs to meet
shortened cycle times. The first, miniaturization, has resulted in
ever increasing chip complexity. As chip geometries have gotten
smaller and smaller, designers can pack more features and IP
modules into the same geography, an additional pressure on design
cycle times. This has liberated system-on-chip (SoC) designs to
grow larger and larger.
The second force, consumerization, rewards first-to-market
manufacturers that can optimize its design performance and feature
sets and thus, capture market share. With market share, the
manufacturer can realize volume production efficiencies. Thus,
manufacturers continually introduce product development processes
that shorten time-to-market. Consumerization has also pushed
manufacturers to shorten design cycle times to as little as three
to six months to offer the advanced feature sets that will retain
or increase market share.
Since the rate of increasing chip complexity is outpacing design
productivity, product manufacturers have recognized that rapidly
integrating and reusing IP blocks are key strategies to overcome
this imbalance. Reusing IP also offers system designers an avenue
to overcome the apparent hurdles and expense of a rapid
time-to-market design sequence.
This article will present Motorola's IP integration flow,
particularly in the context of our recently released Semiconductor
Reuse Standards (SRS). The flow itself is conceptually easy to
understand (Figure 1). However, successful reuse implementation of
an IP block involves some interlocking dependencies between process
steps. Effective socketization, for example, requires the IP
owner's attention to the tools used to integrate the IP from a raw
block through socketization, certification, the repository, and
finally, integration into the final design.
Semiconductor Reuse Standards (SRS)
The SRS provide the playing field, the rules, and the interface
methodology between stages of the IP flow. These standards create
the backbone for our multi-site, multi-technology global
development community. Automated tools provide the infrastructure
for socketization and repository, while the certification and
integration steps have partially automated tools currently
available.
At the beginning of the design process, any SoC design team will
map out the IP blocks to be included. These IP blocks come from
either internal or external sources. Regardless of the source, each
IP block in the design must be socketized. This means that it must
be prepared for optimized integration in conformity with the
SRS.
IP Socketization
SoC designs can incorporate modular IP blocks that are either
fully customized (hard IP), fully synthesizable (soft IP), or a
hybrid of both approaches. The IP creator's objective is to create
blocks that system designers (the customers) can successfully
integrate into their designs. To properly integrate the IP block,
the system designer will require a set of "views." Only the IP
creator has sufficient knowledge of the IP to create these views,
eliminate false paths during timing characterization, and determine
the correctness of the results.
These views provide the designer with the appropriate
information to reuse and integrate the block into the design. A
capable View Generation System (VGS) will create and provide the
designer these views ready for use by the integration tool. IP
creators and designers have successfully used VGS to characterize a
number of Motorola cores and large blocks. However, creating a
package of views requires knowledge in three areas: the IP
integration system, the standard cell synthesis library, and the IP
block itself. The following outline summarizes the points required
in each of these areas:
- IP integration system
- list of tools and views
- naming convention of views and directories
- directory structure
- special data formats, e.g., format of compiled and protected
simulation models
- platform dependent data.
- Standard cell synthesis library
- list of PVT (process, voltage, temperature points)
- measure thresholds for delays and slopes
- layer list for physical layout (DRC, LVS)
- naming convention of subcircuits.
- Knowledge about the IP block
- specifications
- operating temperature range
- operating voltage range
- output loads that can be driven
- input slopes that can be tolerated
- clock frequency
- clock tree recognition
- false path elimination
- verification of characterization results.
Ideally, the CAD system will already be properly configured to
automatically provide the required views in a plug-and-play manner.
The VGS system performs all of the tasks associated with the IP
integration system (#1 above) and the standard
cell synthesis library (#2 above). Rules to cover
these two task areas are built into the system or configured using
configuration files. These rules are based on Motorola's SRS. The
tasks in the knowledge about IP (#3 above) area
are partially automated by using template forms and template
configuration files.
Timing Characterization
The integration flow process that incorporates a block of
reusable IP into a new design depends heavily on timing
characterization. These timing views meet a variety of
purposes:
- Digital timing simulation
- Synthesis of the surrounding logic
- Static timing analysis of the SoC
- Timing driven place and route.
Our design flow automation process recognized that timing
characterization is a natural extension of the IP creator's
analysis approach. To meet these needs, we developed a timing
characterization flow using Synopsys' transistor-level static
timing tool, Pathmill, and gate-level timing tool, PrimeTime.
Simulation Models and Other Views
VGS also generates compiled and protected digital simulation
models. Pre-compiled simulation models integrate into the target
design environment more easily. Compiled simulation models can be
generated from either the behavioral or gate level description and
are platform and operating system dependent.
The VGS also generates many small files that are nonetheless
important for seamless integration into the target SoC design
environment. These views include the following: Verilog stub model,
Eel (Edif external library) file, black-box ATPG model, and TLF and
CLF files for place and route files. In some cases, the fully
socketized IP block can bypass certification and the repository
steps and be immediately integrated.
IP Certification
Certification determines whether an IP block is reusable. It
also provides metrics that indicate the quality of the block for
reuse. Most of the certification is done before the IP enters the
IP repository. Additional certification may be done while the IP is
in the repository. We use the Interra product called Spyglass,
which operates much like a spell checker to check the block for
compliance with the SRS for each view. The certification team uses
the VGS to test for consistency between IP views and also tests for
compatibility with tools, libraries, and hardware platforms. The
team also prepares a prediction of the quality and reliability of
the IP block (deliverables, usage, and defect history) called
"grading."
IP Repository
The IP repository provides an infrastructure where IP creators
can offer and system designers (IP users) can acquire IP blocks for
reuse. The repository system allows participants to share
information in the form of soft/firm/hard IPs of many different
types. The repository provides global access for our multi-site
development community of system integrators. It allows these
integrators to quickly and easily search, select, and integrate
available IP blocks into their designs. The repository
infrastructure consists of:
- Multiple IP vault servers that store the IP and
deliverables
- A meta data relational database for easy search by a variety of
user-selected parameters
- The meta database also initiates downloads from the appropriate
vault server
- A Web application server to query the database and generate
results on a Web page.
The repository also allows IP creators to upload blocks
including all deliverables into the IP vault and also upload the
necessary meta data to the meta data server. The creator can use a
special GUI (graphical user interface) to configure new data or
edit existing meta data files.
IP Integration
We have created a common global design environment based on the
capabilities of the best-in-class design and analysis tools. The
design environment supports an ASIC style synthesis and integration
flow. All supported tools in the design flow are encapsulated and
have a GUI that simplifies usage for an inexperienced user. This
environment also flexibly supports multiple technologies and
synthesis libraries.
This environment supports all of our major core architectures
and main IC design areas and includes a full set of user
documentation and tutorials. Our mission has been to integrate
inherent support for industry design and reuse standards and our
internal SRS. SRS version 2.0 covers 11 areas, of which the
following areas were released to the public in December 1999:
- IP Block Deliverable Standards
- Documentation Standards
- HDL Coding Standards
- Peripheral Interface Standards/IP Bus Specification.
Summary
IP creators and IP consumers now face accelerated design cycle
times as short as three to six months. To accomplish this task in
the face of continually smaller chip geometries requires an
integration methodology. As a logical evolution from our internally
developed Semiconductor Reuse Standards, we developed an IP
integration flow from selection of the IP block through
socketization, certification, repository, and integration.
Best-in-class design and analysis tools automate many of these
steps in our current design environment and further automated steps
are in development. With this approach, we have developed a
multi-user, multi-technology worldwide design environment equipped
with GUI interfaces for ease-of-use by even novice users.
About the Author
Stefan Scharfenberg studied Electrical
Engineering at the Technical University of Aachen. He received his
Diploma in 1991 and finished 1995 with a Ph.D. in Mixed-Level
Circuit-Device Simulation. Since September 1995 he has been working
for Motorola SPS Munich covering timing simulation, timing
characterization, and core integration. Stefan is project leader of
the Motorola View-Generation-System and responsible for timing
analysis within the Motorola Design Environment.