it is likely that “choke points”, or highly congested
regions may be created, leading to a core that is
unroutable for many applications.
Pre-Fabrication Verification
Perhaps the most challenging issue will be that of pre-
fabrication verification. At the time an SoC with a
programmable logic core is designed, the circuitry that
is to be implemented in the programmable portion is
unknown. To perform a system-level simulation of the
entire SoC, at least a behavioural description of at least
one possible configuration of the programmable logic
core is required. For example, when simulating the
interface chip of Figure 2, it is necessary to have a
behavioural description of at least one implementation
of the network interface in order to simulate the entire
chip. A more thorough verification job would verify
the circuit with a number of different network interface
circuits. Thus, there is a tradeoff between the amount
of verification and the time needed to perform the
verification. A study of this tradeoff is essential.
For some verification tasks, a behavioural description
will not be sufficient. Consider the case of power grid
sizing. This task involves the analysis of the power
requirements of each core, an IR-drop analysis, and the
creation of a power grid that can meet these power
requirements. In a programmable logic core, however,
the amount of power required depends on the digital
logic implemented in the core. To ensure that sufficient
power is supplied to the core, it is likely necessary to
take into account the worst-case power requirements of
the core. The development of worst-case power
models for these cores needs to be created.
Timing analysis is another important issue. The load
of wires entering and exiting a programmable logic core
depends on the digital logic implemented in the core.
One option is to create a fully buffered (and likely
registered) interface between the programmable logic
and the fixed logic. The development of such an
interface, and its interaction with a timing analysis tool
is an important research area.
Post-Fabrication Verification
Once the chip has been fabricated, the designer must
create logic to be implemented in the programmable
logic core. Although the core can be reprogrammed
many times, development of this logic will be
simplified if simulation models of the rest of the SoC
are available. These simulation models may be already
be provided by the core vendors, but their integration
into a system which can simulate both fixed and
programmable logic is an open research area. In the
end, it may be necessary to have programmable IP with
programmable on-chip test structures to fully exploit
this approach.
Applications
A fourth issue that must be studied is the identification
of which sorts of applications can make use of a
programmable logic core, and how they can use it.
Summary
It is clear that the ability to make post-fabrication
changes will become essential as time-to-market
pressures become increasingly tight, and as the
complexities of integrated circuits increase.
Programmable logic cores is a natural way to support
these post-fabrication changes. One of the primary
ways in which these cores will be used is to allow
designers to leave certain aspects of the design
unspecified until after fabrication. The use of these
cores in platform-based design and as a programmable
test controller may also be very valuable.
Although these programmable cores have been
available for approximately a year, their use is not yet
mainstream. The verification issue is an important
concern that must be addressed. However, possibly the
most important issue will be how the cores can be used
within specific applications. Programmable logic adds
another dimension that designers must come to grips
with before the full potential of these cores can be
realized.
References
[1] C. Matsumoto, “LSI Logic ASICs to add Programmable
Logic Cores”, E.E. Times, August 29, 1999.
[2] S. Ohr, “ADI Taps Systolix Processor Array,” E.E. Times,
April 21, 2000.
[3] “Lucent Introduces ORCA Series 4 FPGA,” Programmable
Logic News and Views, pages 7-11.
[4] R. Merritt, “QuickLogic Steps up Merger of FPGA with IP
Cores – DSP First Target,” E.E. Times, August 9, 2000.
[5] “Embedded FPGA Cores Enable Programmable ASICs,
ASSPs,” E.E. Times, September 20, 1999.
[6] C. Matsumoto, “Actel Plans to Produce FPGAs as ASIC
Cores,” E.E. Times, June 5, 2000.
[7] C.Matsumoto,“StartupPutsaFreshSpinonProgrammable
Cores,” E.E. Times, September 15, 2000.
[8] V.Betz,J.Rose,A.Marquardt,“Architecture and CAD for
Deep-Submicron FPGAs,” Kluwer Academic Pub., 1999.
[9] V. Betz and J. Rose, ``Directional Bias and Non-Uniformity
in FPGA Global Routing Architectures,'' ICCAD, 1996,
pages 652 - 659.