soc virtual prototyping: an approach towards fast system- on-chip solution date – 09 th april 2012...
TRANSCRIPT
SOC Virtual Prototyping: An Approach towards fast System-On-Chip Solution
Date – 09th April 2012
Mamta CHALANA
Tech Leader
ST Microelectronics Pvt. Ltd, Bangalore
2
Agenda
Introduction of System On Chip System On Chip Design Flow What is Virtual prototype ? TLM Methodology What is SystemC ? Image Co-Processor Architecture Image Signal Processor TLM Platform What is TLMdevice? Dataflow of the TLM platform TLMdevice demonstrator TLMdevice Benefits Results & Conclusions Thanks
3
Introduction : System On Chip
How does a system-on-chip like today?
Processor with Embedded SW
InterconnectBus
Memory
Peripherals
Hardware Development
Architecture
Design
Verification
Silicon
Board
System Development
Virtual Prototype
Software Development
Validation
Coding
Architecture
Integration
Time To Market saved
System On Chip Design Flow
Early system integration
Bottlenecks:• Explosive Complexities• Cost• Time to market Pressure
Solution:• Raise the level of abstraction for system level design• Support HW/SW co-development
Benefits of V.P.:• Better Time to Market• Quality Confidence• Communication Point• Reference Simulation
TLM or Virtual Prototype fulfills both conditions as the ideal solution
5
TLM Methodology It is the single
IP/SOC executable reference model that allows: Rationalized modeling effort Consistent development Cross-team communication
TLM Functional Models
Timed Models
Functional Embedded SoftwareFunctional
Verification
Co-simulation
Architecture Analysis
Embedded Software Optimization
6
What is SystemC
SystemC is a modeling platformA set of C++ class libraryC++ extensions to add hardware modeling
constructsSimulation KernelSupports different levels of abstraction like High
Level Functional Models, Loosely Timed Model, Approximate Timed Models, Timed Models, Cycle Accurate models
Models developed in SystemC can act as unique reference model for all teams in product design phase
7
Image Co-Processor Architecture
MicroProcessor (Firmware)
RAWBayer Data
Stats Gathering
Image Processing Pipe (hardware)
Sensor
Control
YUV or RGB
The task of the Image Co-Processor is to process the data from the sensor into a good quality image conforming to a standard format (YUV or RGB)
8
Image Signal Processor TLM Platform
9
What is TLMdevice?
Way to Connect TLM Simulations with real hardware devices like UART, Display and Sensors.
Set Of Libraries. Server-Client Methodology. Connection is based on socket communication. TLMdevice servers can be located anywhere on the network. Drivers are needed to configure sensor, capture the image from
sensor and provide it to the SystemC TLM model. Data can be easily sent and received to/from during TLM simulation
run. The only tools required to develop such TLM IPs are the free-of-
charge open source SystemC 2.0 kernel, C++, GNU compiler & gdb debugger.
10
Dataflow of the platform
TLMdevice_Sensor Server
On Windows
MFC-TestApp
1
‘C’ API I/F
TLM-device DLL
2
‘C’
API
I/FImagingTLMPlatform
TLMdevice_I2C Client on LinuxIPC
I2C Master I/F
Interfaces with the Sensor Board
TLMdevice_RX Client on LinuxIPC
RX I/F
I/F sending final RGB/YUV image file
2
CMOS Sensor Connected with USB Port of PC
I2C TLM IP
RX TLM IP
11
Dataflow of the platform
TLMdevice_GDD_RGB Server
On Solaris
ImagingTLMPlatform TLMdevice_G
DD_RGB/YUV Client On Linux
IPC
Image sent to the display application GDD Frame
TLMdevice_GDD_RGB/YUV Server On Windows
O/P TLM IP
TLMdevice demonstrator
Execution System
Image Post Processing TLM IP
I2C driver
Unmodified Autofocus Software
Sensor I2C
STxP70
IP
Smart DMA
Memory
I2C IP
RX IP
Output IP
Sensor System
Network
Sensor
Sensor API
I2C Client
I2C Server
Sensor Ctrl
Sensor Interface Software
GDD Server
Sensor Out
GDD Client
Sensor RX
RX driver
Network
Output System
Display
Output driver
Image Output
GDD Client
GDD Server
Display
IMG Hard.
TLMdevice
IMG Soft.
13
TLMdevice Benefits
In our Imaging Framework TLM Models were interfaced with the SMIA complaint CMOS sensor using USB interface.
TLM Model was able to process live streamed images from sensor.
Firmware team appreciated this virtual platform because of its integration with real CMOS sensor and speed.
14
Results & Conclusions
Significant Time Gain in Product Tape-out. Successfully validated firmware algorithms like auto white balance, auto
exposure with real sensor streamed images much before availability of chip. Up-to 2 fps frame-rate achieved. Relying on TLMdevice server-client approach, enables servers to be located
anywhere on the network. Easy Integration of ISS debugger made debugging easy and powerful.
SystemC Transaction level models play a key role in different phases of SOC design cycle like architecture models, golden reference models for functional verification, early availability to act as Virtual Prototype for image quality analysis and close loop validations.
Conclusion