SDL Based Auto Code Generation: A Multi Beneficial Approach

1

• SDL Based Auto Code Generation A Multi
  Beneficial Approach
• Bhaskar Rao.G
• Software Engineering Group,
• Motorola India Electronics Ltd.
• Banglore, India - 560 042.
• email gbr_at_miel.mot.com.

2
Agenda

• Introduction to SDL and Benefits ( 3 min.)
• Experiences and Results (17 min.)
• Pilot Project
• GPRS Layer Study
• Problems and Solutions
• Challenges Remaining (3 min.)
• Conclusions (2 min.)

3
Introduction to SDL and Benefits
4
SDL, Is it only Auto Code Generation?

• Methodology
• Integrated Methods and Tools for Software
  Development
• Requirements - MSC
• SAD - State Charts/UML/SDL
• Coding - SDL, C, C, Java ltAUTO CODE gt
• Unit and Integration Testing - Simulation
  Validation
• System Testing - TTCN/SDL

5
Auto Code Generation from SDL
SDL Code
Manual C Code env
SDT Auto Code Tool
C Compiler
SDL Code
Gen. code
H2SDL
Executable
Run Time Library
C Headers
Simulation, Validation, UNIX, NT, pSOS, VxWorks
etc..
6
SDL Usage Perspectives

• Design Language
• Implementation Methodology
• Infrastructure
• Handsets?
• Test Development Language
• Strong data features
• Better tool support compared to TTCN tools

7
Range of Benefits Observed

• Cycle Time Improvement 1.2 to 2X
• Productivity 1.2 to 2X
• Effort Reduction 1.2 to 2X
• Quality 2 to 5X
• Performance PROBLEMS NOT NOTICED

8
Experiences and Results
9
Pilot Project Architecture
10
Pilot Project

• EIA 709.1 Control Network Protocol
• Network Management Layer
• Provides services to the application
• Uses either Transport layer or Session layer
• Session Layer
• Provides session services
• Transport layer
• Acknowledged/Un-acknowledged repeat data transfer
• Network layer
• Interacts with the driver
• Driver
• send and receive messages

11
SDL Architecture

Application Interface
Network Management Process
Controller
Controller
Transmit side (SL, TL)
Receive Side (SL, TL)
Controller
Network Layer
Driver Interface
12
Features of the Architecture

• Layer independent architecture
• Adaptable to any changes
• Cadvanced Code Generator (Telelogic Tau)
• lt 1 year project

13
Pilot Results
14
Observations .

• Simulation based testing has been proved very
  effective
• Less than average of the organization
• Script language
• Real Time Simulation
• Traces
• Automatic support of targeting for Windows, pSOS
  etc.
• Significantly impacted customer satisfaction

15
Problems ...

• Code Size??????
• 3X
• System Testing
• Interpreted the AUTO CODE
• Less productive

16
Reasons for Code Bloat

• Architecture
• Too Modular
• Layer independent
• Tool translation of SDL support functions
• Assignment, Equality etc.
• StructVari StructVar2
• if(StructVari StructVar2)
• Overhead is 400 bytes per structure
• Translation of Header files to C code
• Hdr.h gtgtgt GenHdr.c

17
Optimization Results on one layer
18
Optimization Features

• C code mix is increased to 65
• Pointers are used
• Negative impact on Productivity and Quality
• Effort increased to locate the error

19
Plans After Pilot

• Used in Infrastructure projects
• Results are encouraging
• Code size is not CRITICAL
• Handsets?????

20
Further study

• Try new tools
• Cmicro Code generator
• Proprietary generator
• Optimize with the help of the vendors
• Excellent help

21
Optimization Results (UNIX)
22
Optimization Results (UNIX) contd..
23
Auto Code Generation from SDL (Optimization)
SDT Auto Code Tool
SDL Code
Manual C Code env
C Compiler
SDL Code
Gen. code
SCRIPTS (Optimization)
H2SDL
Executable
Run Time Library
C Headers
Simulation, Validation, UNIX, NT, pSOS, VxWorks
etc..
24
Optimization of Code

• Data Types defined only in SDL
• NON OO
• Minimize C Code Integration
• Through common types between SDL and C (Octet
  Unsigned Char)
• Scripts
• Script to convert C constants to SDL macros
• Script to convert STRUCTURES to ADTs
• Compilation Switches for Optimization
• XNOUSEOFREAL, XNOUSEOFOBJECTIDENTIFIER etc.

25
Size of Utility Functions

• Default Functions
• Size 16 Type FUNC yDef_zF_MyStruct1
• Size 16 Type FUNC yDef_zG_MyStruct5
• Equal Functions
• Size 192 Type FUNC yEq_zF_MyStruct1
• Size 192 Type FUNC yEq_zG_MyStruct5
• Make Functions
• Size 152 Type FUNC yMake_zF_MyStruct1
• Size 152 Type FUNC yMake_zG_MyStruct5

26
After Optimization

• newtype MyStruct1 struct
• a1, a2, a3, a4, a5 Integer
• /ADT(E(H) D(H) K(H) H)
• TYPE
• define yDef_(MyStruct1)(yVar)
• define yMake_(MyStruct1)(yVar1, yVar2)
• define yEq_(MyStruct1) (E1, E2)/
• endnewtype

27
Size analysis command

• Usage
• gt nm -f ltexec_file_namegt
• Displays Type, Size etc. of all symbols
• NOTYPE no type was
  specified
• OBJECT a data object such
  as an array or variable in bytes
• FUNC a function or
  other executable code in bytes
• SECTION a section symbol
• FILE name of the
  source file.

28
Challenges Remaining
29
No Wider Acceptance??

• Not Part of Curriculum
• Industry has to invest
• Technical
• Performance and Code Size issues
• Debugging the generated code (off site and
  onsite)
• Performance, Code Size VS READABILITY of SDL
• Customer Related
• Cost of the tools
• Lack of familiarity and Expertise
• Risk of failure

30
100 Real time Software Development with SDL
Methodology
VISION
VISION
31
For Realizing the Vision

• Effective Algorithms for mapping SDL code to
  implementation language
• Effective Approaches to handle on-site testing

32
Conclusions

• SDL is a Methodology but only tool for auto code
• Positive impact on Cycle Time, Quality and
  Productivity
• Code Size Overheads could be brought down with a
  few restrictions in SDL
• Needs improvement for 100 usage

33
Q A
Q A
Q A
Q A
Q A

• Questions????

Q A
Q A
Q A
Q A
Q A
Q A
Q A
Q A
34
Thank you

• Bhaskar Rao.G
• Motorola India Electronics Ltd.,
• SE Group,
• Banglore, India- 560 042
• gbr_at_miel.mot.com