I. J. Communications, Network and System Sciences, 2008, 2, 105-206
Published Online May 2008 in SciRes (http://www.SRPublishing.org/journal/ijcns/).
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
A Co-verification Method Based on TWCNP-OS for Two-
way Cable Network SOC
Chong LI1, Xiaotong ZHANG, Yadong WAN, Qin WANG
Department of Computer Science and Technology
University of Science and Technology Beijing, Beijing, P.R.China
E-mail: 1lichong0564@163.com
Abstract
Co-verification is the key step of software and hardware codesign on SOC. This paper presents a hw/sw co-
verification methodology based on TWCNP-OS, a Linux-based operating system designed for FPGA-based
platform of two-way cable network (TWCNP) SOC. By implementing HAL (hardware Abstraction level)
specially, which is the communications interface between hardware and software, we offer a homogeneous
Linux interface for both software and hardware processes. Hardware processes inherit the same level of
service from kernel, as typical Linux software processes by HAL. The familiar and language independent
Linux kernel interface facilitates easy design reuse and rapid application development. The hw/sw
Architecture of TWCNP and design flow of TWCNP-OS are presented on detail. A software and hardware
co-verification method using TWCNP-OS is proposed, through the integrated using of Godson-I test board
and TWCNP, which realizes the combination of design and verification. It is not a replacement of the co-
verification with generic RTOS modeling, but is complementary to them. Performance analysis of our
current implementation and our experience with developing this system based on TWCNP-OS will be
presented. Most importantly, since the introduction of TWCNP-OS to our FPGA-based platform, we have
observed increased productivity among high-level application developers who have little experience in
FPGA application design.
Keywords: CM, SOPC, MIPS, Co-verification, FPGA, HAL
1. Introduction
Now CATV mainly transmits via analog signals in some
regions and countries. In order to realize A/D
transmission, two-way network chip based on HFC
(Hybrid Fiber-Coax) is becoming the core technology
and key equipment in interactive digital television and
other integrated digital services. SOC enables the
realization of all these functions in one chip and has
reusable technology of IP module to develop the SOC
which has the independent intellectual property has
important meaning.
The hardware and software codesign is commonly
used in SOC design. It is different from traditional
design methods that simulation and verification is used
to find and rectify mistakes in time and optimized the
system in the end. With the complexity increasing in
SOC design, scientific design method and high efficient
verification are growing more important [1]. Hw/sw co-
verification does not just occur at the system integration
point but rather throughout the design process. A high
efficient verification platform should provide simple and
scientific verification environment to ensure the
performance of whole system. Co-verification includes
hardware verification, software verification and
software-hardware interface verification. Using
concurrent design and co-verification to shorten project
period and combine hardware behavior model with
software running environment, construct a high efficient
co-verification platform has been the research hotspot
recently.
While traditional hw/sw codesign researches have
produced encouraging results in the area of hw/sw
partitioning, cosimulate, cosynthesis, and co-verification,
most of them rely on self-contained design environments
that are based on their specific input languages or library
API’s [2]. As a result, migrating existing software
200 C. LI ET AL.
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
designs to two-way cable network platform using these
traditional hw/sw codesign methodologies would have
incurred major re-engineering efforts, including learning
a new language and API, getting familiar with a new
design environment and reimplementing existing designs
in the new language environment.
So, an easy to use hw/sw interface that allows rapid
application development and migration should be (1)
familiar and intuitive to both software and hardware
engineers; and (2) language independent. We achieve
this goal by setting hw/sw boundary at the embedded
operating system kernel level.
Embedded operating systems (EOS) have become
one of the most important components of embedded
systems due to the growing complexity of the system
functionalities as well as the increasing time-to-market
pressures. With this trend, a demand for fast
cosimulation of hardware and embedded software
including an EOS is also becoming stronger in order to
validate the functionality of the overall systems. With
this trend, a demand for fast cosimulation of hardware
and embedded software including an EOS is also
becoming stronger in order to validate the functionality
of the overall systems.
In this paper, we present TWCNP-OS, an operating
system designed specifically for FPGA-based platform
of two-way cable network (TWCNP) SOC, which is key
device based on DOCSIS (Data-Over-Cable Service
Interface Specifications) on HFC network whose CPU is
Godson-I. Under TWCNP-OS, hardware and software
share the same familiar Linux interface and the same
level of support from the OS kernel. We use the concept
of hardware process [3], which is the same as a normal
Linux process except its “program” is an FPGA
hardware design instead of software program. HAL is
implemented, and communications between hardware
and software are accomplished through it, which uses
conventional Linux inter-process communication (IPC)
mechanisms, such as shared file, pipe, shared memory,
signal, and message passing. Hardware processes have
access to system resources as their software counterparts,
such as the general file system, standard input, standard
output.
By building a cross GNU/GCC compiler for Godson-
I, we can develop software in host machine
(PC).Software designs can be developed in C/C++
language development environment a designer is
familiar with. For hardware designs to communicate
with the kernel, TWCNP-OS defines a standard message
passing network that resembles the software system call
interface by maintaining the hw/sw interface-HAL at the
kernel level. This standardized network allows hardware
designs be developed in any hardware language
environment of choice.
The remainder of this paper is organized as follows.
Section 2 surveys related work on hw/sw cosimulate/co-
verification with RTOS supports, and analyses the
difference all of them including our TWCNP. Section 3
introduces the system architecture, particularly the
software and hardware architecture of CM platform in
tow-way CATV. Section 4 elaborates on the co-
verification approach based on TWCNP-OS, and its
detailed design process is presented. Section 5, we give
the system testing and performance comparison of the
platform designed by this method.
2. Related Work
Hw/sw co-verification has been studied around the
world for more than a decade. Simulator is an usual
method of hw/sw co-verification, which must join with
other parts of the system verification problem to create a
complete verification solution. A number of commercial
cosimulators have come onto the market. The
cosimulators are currently one of indispensable CAD
tools in the design of embedded systems and systems-
on-chip.
There is no single hardware-software co-simulation
problem and as a result there is no single tool or
technology that can successfully solve all the problems
associated with co-verification of hardware and software.
Each developer must trade off the desired performance
against the level of accuracy. Each developer must trade
off the desired performance against the level of accuracy.
For hardware designers the trade-off is principally
between software simulation and the relatively faster
technologies of hardware acceleration, emulation or
rapid prototyping. For software developers of embedded
system application software the primary need is
simulation speed and the trade-off is among a variety of
simulation approaches which achieve greater degrees of
speed at the cost of diminishing the accuracy of the
modeling of the hardware.
Figure 1. Trade-off between simulation accuracy and speed
HW emulation is a very important alternative to
overcome the SW speed problems [4]. A co-verification
platform using C++ simulator and FGPA (Field-
Programmable Gate Arrays) emulator is presented in [5],
gets a good simulation speed, and provides an
accuracy/efficiency tradeoff through various abstraction
levels (Figure 1). When the C/C++ abstraction is at the
algorithm level, the simulation speed can reach about 1
MHz, but the simulation results have no cycle accuracy.
A CO-VERIFICATION METHOD BASED ON TWCNP-OS FOR TWO-WAY CABLE NETWORK SOC 201
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
The speed of hardware emulator is typically up to 1
MHz, and they preserve cycle accuracy.
However, since the cosimulators do not feature
explicit supports for RTOSs, a designer needs to use an
ISS in many cases to run embedded software including
an RTOS. Due to the slow execution speed, extensive
simulation of large software is impossible.
In the recent years, several research efforts have been
made to model RTOSs for system-level design and
cosimulation. SoCOS presented in [6] is a system-level
design environment where the OSAPI library provides
generic RTOS system calls to application software.
OSAPI is a virtual RTOS to enable native execution of
embedded software. After simulation-based validation,
the OSAPI calls are replaced with the system calls of the
actual RTOS used in the final implementation to obtain
the final software code. In [7], a similar approach is
presented. A main difference is that the RTOS model in
[7] is build upon an existing system-level design
language, i.e., SpecC [8], so that existing CAD tools
such as simulators can be used. Techniques presented in
[9] also use the SpecC language to model the preemptive
behavior. In [10], an OS model is proposed for fast and
time-accurate cosimulation. The model focuses on
accurately modeling the RTOS overhead during task
execution as well as the preemptive behavior. In [11], a
method which automatically generates RTOS-dependent
software from SystemC description is presented. The
method replaces SystemC’s constructs for concurrency
and communication with corresponding RTOS service
calls. In this sense, it can be considered that SystemC
involves a simple RTOS model in itself. With the
development of Embedded OS, [12] gives us a RTOS-
centric hw/sw co-simulator, which features co-
simulation with functional simulation models of
hardware written in C/C++ and co-simulation with HDL
simulators, supports a complete simulation model of an
RTOS based on uITRON [13].
A common weakness of these OS models is that they
support only a limited set of RTOS services in order to
make the models generic and independent of specific
RTOSs. for example, [12] have more than 80 service
calls but only for µITRON-based RTOSs platform.
However, the RTOS model in [7] supports only 16
service calls. These services may need to be fully
utilized in order to write high quality software.
Therefore, it is easily imagined that the quality of
software automatically generated by these previous
methods is lower than that of hand-crafted RTOS-
dependent software. Instead, we have developed fully
supports RTOS services by porting the standard
MIPS/Linux kernel to our platform, such high-quality
software can be designed and simulated. But there are a
weakness for hardware-dependent, and have to put a
Virtual hardware level in HAL when some hardware
component haven’t finished.
TWCNP shares a similar design philosophy as
BORPH [3] in providing a unifying coarse-grain
hardware (FPGA) and software component interface.
They are not a complete system to perform typical
hw/sw codesign tasks such as partitioning, cosynthesis,
cosimulate, or verification. Instead, by providing basic
Linux OS services, it acts as a platform on which these
tasks can be carried out. In fact, our TWCNP is not a
replacement of the cosimulators with generic RTOS
modeling, but is complementary to them.
The main contribution of TWCNP-OS is that by
leveraging conventional Linux semantics to FPGA-
based platform of two-way cable network (TWCNP)
SOC, it provides a unique, unified environment for both
FPGA and software application designers. The Linux
semantics is familiar to developers across many research
domains, thus lowering the barrier-to-entry into FPGA-
based SOC. Furthermore, since TWCNP-OS is
implemented as an extended Linux kernel, a TWCNP-
OS managed system may leverage all commodity Linux
software applications for developing, testing,
benchmarking, and deploying FPGA applications.
3. Two-way CM Platform Architecture
The CM (cable modem) platform is based on HFC
network, which can increase the transmission quality
efficiently, and integrate the CATV and internet digital
network. Using the CM can expediently develop many
services based on Internet protocol, such as VoIP, VOD,
HDTV and high-speed web browsing. The design of CM
includes three parts: evaluation board design, embedded
software design and SOC design. Verification is used
through the whole process of SOC development. The
software and hardware co-verification is stressed in this
paper.
Figure 2. Overall frame of FPGA emulation board
The hardware architecture of CM is described in
Figure 2. It includes three modules:
1) EuroDOCSIS protocol processor: It includes
independent developed physical layer modem
(Jupiter), EuroDOCSIS MAC protocol processor and
radio frequency interface circuit.
2) BUS and peripheral equipment module:
Interconnection through AMBA bus and peripheral
equipment includes IIC, GPIO, UART, PWM and
Ethernet MAC.
202 C. LI ET AL.
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
3) Godson-I CPU: Godson-I IP core, Based on MIPS
instruction, can apply to both the universal and
embedded system, has a good compatibility.
The software architecture which is described in
Figure 3, includes three main parts:
1) Operating System: based on MIPS/Linux
configuration, and making Linux run on CM based
on Godson-I, including Bootloader and OS kernel, It
supports memory management, file management, task
schedule, and modularization.
2) Hardware driver: it ensures hardware working high-
efficiently and properly. The driver such as GPIO,
IIC, network adapter can either be compiled into OS
kernel or be loaded dynamically.
3) EuroDocsis protocol stack and CM application:
realizing central functions defined in EuroDocsis, it
provides software guarantee to bridge CM and CMTS
(cable modem terminal system). Application software
defines a set of user interface and ensures function
maintenance and management to CM.
Figure 3. Software frame of two-way cable network CM
OS is the cornerstone of the software design platform,
which manages all software and hardware directly,
provides interface to application and system call. It is
also used to validate hardware and software function,
guide the hardware design to be more reasonable.
4. Co-verification Based on TWCNP-OS
There are two common verification methods on soc
design.
1) The first is software verification which simulates the
system behavior through establishing emulation
model. Through this method, system’s detailed status
should be observed but the run time is much longer.
ISS (Instruction set simulation) is an example of it.
2) The second is the hardware verification. Now the
prevalent one is the SOPC based on FPGA, which
can online program and has accurate clock cycle.
This way has a higher emulation speed, but the
system’s status cannot be observed. It includes three
aspects: software verification, hardware verification,
and software-hardware interface verification.
Combining the merits of these two kinds of method, a
co-verification method based on TWCNP-OS is
proposed in this section, which is supposed that the
hw/sw partitioning and hardware cosynthesis have
finished. Firstly, we introduce the co-verification
platform, give a general idea of porting the MIPS/Linux
to our SOC. Secondly, Software and Hardware co-
verification based on TWCNP-OS will be introduced.
The co-verification platform (Figure 4) is composed
of host, two self-existent target machines and
equipments which connects them each other, such as
UART, USB and network. Host is a common PC, which
provides cross-compilation environment, console and
HDL simulator. Godson-I test board and FPGA
development board (TWCNP) constitute target machine.
Godson-I test board adopts Godson-I CPU and it8172G
chip sets, MIPS instruction set and Linux OS can be run
in it. This one is mainly used in the development of
DOCSIS protocol stack. Being composed of FPGA
vp200, peripherals and interfaces, FPGA board is the
hardware verification platform of the whole SOC system,
which is mainly used in the verification of EuroDOCSIS
protocol processor and related hardware logic.
Figure 4. Co-verification platform of the TWCNP
4.1. Realization of TWCNP-OS
TWCNP-OS is an operating system designed for
TWCNP based on MIPS processor. It extends a standard
Linux kernel to include support for FPGA’s in our
platform. Treating FPGA’s as both coprocessors and
processor, TWCNP-OS treats FPGA’s as hardware
process in the system as normal computational resources.
The interface between hardware and software is HAL.
All processes can therefore be either software programs,
or hardware designs running on FPGA’s. Therefore, to
the rest of the system, communicating with a hardware
process is no different from communicating with a
normal Linux process. This homogeneous handling of
hardware and software in the kernel forms the
foundation of coarse grain hw/sw codesign boundary.
Figure 5 depicts this conceptual block diagram.
Figure 5. TWCNP-OS extends a traditional Linux system
with hardware process support. HW/SW processes share
the same I/O interface and communicate by HAL
A CO-VERIFICATION METHOD BASED ON TWCNP-OS FOR TWO-WAY CABLE NETWORK SOC 203
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
Like embedded OS, TWCNP-OS includes cross-
compiler, Bootloader and OS Kernel too. Cross-
compiler is necessary to design an embedded OS, which
is a compiler capable of creating executable code for a
platform other than the one on which the compiler is run.
It is a tool that one must use for a platform where it is
inconvenient or impossible to compile on that platform,
like microcontrollers that run with a minimal amount of
memory for their own purpose. Whole process of
building a cross-compiler can find in [15].
Bootloader is used to boot the machine, such as
initializing basic CPU registers, UART, SDRAM
controller, and copying OS kernel from Flash to memory.
Its design is derived from PMON [14], which is a
freeware ROM-monitor developed for early LSI Logic
MIPS R3000 evaluation boards. Since its creation,
PMON has become a very common firmware for MIPS
evaluation boards and development systems. It contains
support for several boards, R4000-style exceptions, and
uses the GNU make system. So the major revision is the
address of hardware components. There are three
important parts will be down:
1) CPU register in Godson-I have to be set up the
normal value, such as interrupt, timer, cache, memory
management;
2) The drivers of hardware components should be taken
effect to use UART, serials and console etc. We can
debug and interact with target machine.
3) Only after moving the kernel from Flash kernel space
to memory, the kernel can boot and manage whole
system.
Linux is one of the common embedded OS. For
MIPS architecture of Godson-I, OS mainly adopts
MIPS/Linux schema, using ANSI C, MIPS assembly
language and cross-compiler-MIPS-GCC 3.23 as SDK.
TWCNP-OS is divided into two parts. The hardware
independent part is realized by MIPS assembly language.
The primary design flow is described in figure 6. We
will introduce the design of TWCNP-OS on the basis of
MIPS/Linux kernel source code in the following part.
Overall design steps as following:
1) Hello World! — Get board setup, serial porting
working, and print out “Hello, world!” through the
serial port.
2) Get early printk working — Make the first TWCNP-
OS image and see the printk output from kernel.
3) Serial driver and serial console — Get the real printk()
function working with the serial console.
4) KGDB — KGDB can be enormously helpful in our
development.
5) CPU support — Because Godson-I CPU is not
currently supported, we need to add new code that
supports it in arch directory of Linux source codes.
6) Board specific support — Create your board-specific
directory. Setup interrupts routing/handling and
kernel timer services.
7) HAL — It makes the upper layer software be
independent of the bottom hardware and manages or
simulates a large number of the bottom hardware.
8) Ethernet drivers — We should already have the serial
port working before attempting this. With ethernet
driver working, we can set up a NFS (network file
system) root file system which gives you a fully
working Linux user space.
9) ROMFS root file system — Alternatively you can
create a user space file system as a ROMFS image
stored in a ramdisk, or Flash root file system.
Here we will emphatically present HAL [19] (Figure
6(b)), which is the interface between Hardware and
software, which includes device driver, configuration
manager, control manager and data manager and VHL
(virtual hardware level). It includes all the functions that
manage and schedule the hardware (hardware process)
uniformly for the SOC, and can support a many of
different hardware devices or virtual hardware devices
(VHL). So we can send messages to hardware using
standard system call just as to a software process by
HALIF (HAL interface). Overall control module is the
key of HAL implementation. Configuring manager takes
charge of configuration channel. Control manager
supplies control channel. Data manager takes charge of
two-way data channel. Driver module regards all
hardware devices which are managed by HAL as one
device, and its run driver is regard as a hardware process.
Figure 6. Design flow of the TWCNP-OS, comparing to
generic Linux kernel, we give a HAL on kernel
4.2. Software Verification Based on TWCNP-OS
Software is an important part of the system, verification
of software is necessary. Software verification mainly
uses ISS, though it has accurate clock cycle, the
simulation speed is comparatively slow. The software
verification proposed in this paper simulates hardware
on target machine. Host and Godson-I test board to
validate the validity of software directly on function
level. Software verification is mainly used to DOCSIS
protocol stack module and the relative CM application
program. According to DOCSIS specification, DOCSIS
protocol stack module is divided into hardware-
dependent part and hardware-independent part. The
hardware-independent software realizes user level
204 C. LI ET AL.
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
function such as VOD etc.
In the traditional software and hardware concurrent
design, the design of hardware-dependent software can
not progress until the accomplishment of the hardware
design. VHL can efficiently resolve this problem. The
VHL established on OS can be one or a group of virtual
hardware, which simulate hardware to carry out function
verification. VHL is extension of driver. it can not only
apply uniform interface to application, drive hardware,
but also response according to hardware’s actual
characteristic. If the hardware design has been
accomplished, we only need to simply change the VHL
to drivers and the software need not to be changed. Thus,
hardware-dependent software design and verification
can be processed without hardware environment which
has great meaning.
Figure 7. The sketch map of software co-verification, the
left part of dotted line denotes the hw-dependent software,
the right one is hw-independent software. Each uses a
different verification method
TWCNP-OS make it possible to use VHL in software
verification. The DOCSIS protocol stack and CM
application program validated on development board can
drive hardware on TWCNP-OS. Godson-I test board is
mainly used in software verification. Target host is the
auxiliary equipment which provides convenience for the
verification. The software verification flow is described
in Figure 7, the broken line divided software into
hardware-dependent part and hardware-independent part.
The hardware-independent part is validated directly on
development board and run on target FPGA board. The
hardware-dependent part is mainly in DOCSIS protocol
stack module which needs the support of MAC co-
processor. Using modularize mechanism to establish
virtual MAC co-processor on VHL can simulate MAC
co-processor to realize software-hardware interaction.
Thus the software is independent on or weak dependent
on hardware. Virtual MAC co-processor construct
standard data, notify the hardware-independent part in
DOCSIS module through interrupt to process data
analysis. The verification of CM application program
use C/S mode, the server responses to the client by
simulating CMTS.
4.3. Hardware Verification Based on
TWCNP-OS
Hardware is the support platform of embedded system.
Hardware verification validates through simulating each
unit’s behavior and mainly has two modes. One mode is
pre-simulation which simulate all hardware modules to
validate if the hardware accord with the design
requirement. The common simulation tool is ModelSim
[17], NC-verilog and NC-VHDL [18]. The other mode is
integrated post-simulation which can put composite
latency files into integrated simulation model to estimate
the effect brought out by gateway delay. The hardware
verification introduced in the following base on the
accomplishment of pre-simulation and is in the domain
of FPGA-based post-simulation.
Any verification should carry out according to some
standards. We adopt TWCNP-OS and DOCSIS protocol
stack module as test program to validate hardware.
TWCNP-OS which bases on Linux structure has high
reliability and stability through long time verification.
DOCSIS protocol stack module has passed the software
verification. We can validate the MAC co-processor as
long as modify the virtual MAC co-processor.
The TWCNP-OS-based hardware verification has
high stability and simulation speed. At the time of the
accomplishment of verification of hardware and
software, the hw/sw interface has been validated. This
verification method is on the basis of passing pre-
simulation for each hardware module and validates the
function of hardware system. For the complexity of the
system, the hardware system is divided into two parts:
MAC co-processor and other hardware logics. The
hardware verification has three steps:
1) The first step is the verification of MAC co-processor.
We first connect FPGA with the bus of Godson-I test
board, download the hardware logic of MAC
coprocessor to FPGA, regard MAC co-processor as a
peripheral. Then modify the virtual MAC co-
processor, delete its function of simulation hardware,
make it be a driver to control hardware. At last,
validate it through DOCSIS protocol stack. Thus we
can easily find out errors in it. Furthermore, the
concurrent process of hardware verification can
improve the efficiency.
2) The second step is validating other hardware logic
and software-hardware interface using TWCNP-OS
and FPGA board. The hardware is validated
respectively by calling the drivers compiled to drive
hardware.
3) The third step is the system verification on FPGA
board by using TWCNP-OS and software module
after integrating the hardware logic. The function and
performance test is processed in this step. The
detailed process is introduced in next section.
5. System Testing and Performance
Comparison
The whole system on the actual running environment for
A CO-VERIFICATION METHOD BASED ON TWCNP-OS FOR TWO-WAY CABLE NETWORK SOC 205
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
testing is the last step of verification. Therefore, we will
give a topology of cable network (Figure 8) and
construct a test environment (Figure 9) accordingly.
Figure 8. The topology of cable network
Figure 9. Performance test environment of CM
Stability, time delay and network congestion are the
performance target of CM. Stability is a basic target for
any device. Here it means data can be transferred to
CMTS safely without losing data. Time delay is another
target to estimate performance, if it can’t meet the some
special requirement of the system, it will be no-good.
We will test the response time between CPE (customer
premises equipment) and PC. Network congestion is an
important problem, it is good rule to evaluate whether
the device is excellent or not. Because some equipment
can work well without network congestion, but it will
induce packages losing when it happens network
congestion. So it is a complemental test to stability.
In the following parts in this section, we will give
some tests to get the above performance targets
respectively in TWCNP and SM5100. SM5100 is a
mainstream CM produced by Motorola, and its
performance can delegate a standard of CM in a certain
sense. Special declaration, test results will be influenced
by real environment.
Test 1
To get the response time of CM, We send 100000 IP
data packages by ping command from CPE to PC via
TWCNP and SB5100 CM respectively, each package is
74bytes, the results as table 1.
Table1. Functional test and comparison of CM emulation
platform
Test
platform
ave
T ans
T ans
T (max)p
TWCNP 12ms 10ms 17ms 0
SM5100 9ms 6ms 15ms 0
Parameters
s
um
N=100000packet size=74 bytes
The following is the formula of calculate different
response time. Here, Tans is response time, p is rate of
loss package, Tave is average response time, Tsent is time
to reach destination, Tresv is response data package cost
time, Nlost is number of loss package, Nsum is total
number of package.
ans sent resv
T=T+T (1)
%100×=
N
N
sum
lost
P (2)
[] []
()
()
P
N
ii
N
N
TT
T
lost
sum
i
resvsent
ave
sum
−×
+
=
=1
1 (3)
From testing results of Table 1, we can get that the
average network response performance is similar
between TWCNP and SM5100.
Test 2
To get network congestion and stability results, we
fabricate a mass of normal at the speed of 1.4 MB/s and
congestion data packets at a speed beyond upper line by
CommView on CPE, and send them to PC via TWCNP
an SM5100 about 24 hours. At the same time, we will
capture the throughput on PC.
Figure 10. The stability test of TWCNP
Figure 11. Data throughput analyses of TWCNP and
Motor 5100 CM
Sampling 120 minutes normal packets (Figure 10)
206 C. LI ET AL.
Copyright © 2008 SciRes. I. J. Communications, Network and System Sciences, 2008, 2, 105-206
and Sampling 90s’ data on congestion time (Figure 11),
we get that throughput of TWCNP is still up to 1.52M
B/s, MOTOROLA 5100s’ is 1.67 M B/s under network
congestion. TWCNP can transfer all packets to PC at
same speed. The result of this case study demonstrates
that it gets a high stable throughput.
From the above tests, we can find that the
performance of our TWCNP is equivalent or a little low
to SM5100. The main reason is that the hardware of
them is not on the same level, Godson-I CPU (only 26M
Hz) frequency is far lower comparing to
SM5100’processor, so throughput should be lower. But
it still exerts a good performance.
6. Conclusions
This paper presents a hw/sw co-verification
methodology based on TWCNP-OS, a Linux-based
operating system designed for FPGA-based platform of
two-way cable network (TWCNP) SOC, and then
particularly expatiates on the software and hardware co-
verification method. Through the implementation of
TWCNP-OS on TWCNP and development on host/two
targets, we realize the combination of design and
verification. It is not a replacement of the co-verification
with generic RTOS modeling, but is complementary to
them.
The experiment makes it clear that the verification
platform is simple-built and cost-effective. The scientific
verification method shortens the development cycle and
the validated system is stable and can be applied in the
verification of other embedded system. Most importantly,
since the introduction of TWCNP-OS to our FPGA-
based platform, we have observed increased productivity
among high-level application developers who have little
experience in FPGA application design.
7. References
[1] N. Ohba and K. Takano, “An SoC design methodology
using FPGAs and embedded microprocessors,”
Proceedings of the 41st annual conference on Design
automation, San Diego, CA, USA, June 07–11, 2004.
[2] Habibi and S. Tahar, “Design and verification of
SystemC transaction-level models,” IEEE Trans. VLSI
Syst., 14(1): pp. 57–68, January 2006.
[3] H. So, A. Tkachenko, and R. Brodersen, “A Unified
Hardware/Software Runtime Environment for FPGA-
Based Reconfigurable Computers using BORPH,”
CODES+ISSS, 2006.
[4] D. Atienza, P.G. Del Valle, G. Paci, et al., “A fast
HW/SW FPGA-based thermal emulation framework for
multi-processor system-on-chip,” Proceedings of the
43rd annual conference on Design automation, San
Francisco, CA, USA, July 24–28, 2006.
[5] Y. Nakamura, K. Hosokawa, I. Kuroda, K. et al., “A fast
hardware/software co-verification method for system-on-
a-chip by using a C/C++ simulator and FPGA emulator
with shared register communication,” Proceedings of the
41st Design Automation Conference (DAC’04), San
Diego, Calif., USA, pp. 299–304, June 2004.
[6] D. Desmet, D. Verkest, and H. De Man, “Operating
system based software generation for systems-on-chip,”
Proceedings of Design Automation Conference (DAC),
2000.
[7] A. Gerstlauer, H. Yu, and D. Gajski, “RTOS modeling
for system level design,” Proceedings of Design
Automation and Test in Europe (DATE), Embedded
Software Forum, 2003.
[8] SpecC Technology Open Consortium,
http://www.specc.org/.
[9] H. Tomiyama, Y. Cao, and K. Murakami, “Modeling
fixedpriority preemptive multi-task systems in SpecC,”
Proceedings of Workshop on Synthesis and System
Integration of Mixed Technologies (SASIMI), 2001.
[10] Y. Yi, D. Kim, and S. Ha, “Virtual synchronization
technique with OS modeling for fast and time-accurate
cosimulation,” Proceedings of International Conference
on Hardware/Software Codesign and System Synthesis
(CODES+ISSS), 2003.
[11] F. Herrera, H. Posadas, P. Sanchez, and E. Villar,
“Systematic embedded software generation from
SystemC,” Proceedings of Design Automation and Test
in Europe (DATE), Embedded Software Forum, 2003.
[12] S. Honda, T. Wakabayashi, H. Tomiyama, et al., “RTOS-
centric hardware/software cosimulator for embedded
system design,” Proceedings of the 2nd IEEE/ACM/IFIP
international conference on Hardware/software codesign
and system synthesis, Stockholm, Sweden, September
08–10, 2004.
[13] ITRON, http://www.assoc.tron.org/itron/.
[14] http://www.linux-mips.org/wiki/PMON.
[15] http://cross-lfs.org/view/svn/mips/index.html.
[16] http://linux.junsun.net/porting-howto/.
[17] Mentor Graphics Corporation, http://www.mentor.com.
[18] C. Wang, X.G. Xue, et al., “FPGA/CPLD Design TOOL-
Specification for Xilinx ISE 5.X,” Posts and Telecom
Press, 2003.
[19] S. Yoo, I. Bacivarov, A. Bouchima, et al., “Building fast
and accurate SW simulation models based on hardware
abstraction layer and simulation environment abstraction
layer,” Proceedings of the Design, Automation and Test
in Europe (DATE’03), Munich, Germany, pp. 500–506,
March 3–7, 2003.