This document discusses the process of writing code for embedded systems. It covers topics such as choosing what and how to write code, reviewing code for correctness, optimizing code for aspects like CPU clocks and memory, choosing development platforms and tools, debugging and simulating code, and finally testing code on actual silicon hardware. The overall process takes the writer from initial code writing through multiple stages of review, optimization, and testing before final deployment on embedded hardware.

1. WRITE ONCE ON SILICON AND PLAY IT FOR EVER
: AN EMBEDDED JOURNEY
Sandip Ray
April 29, 2011

2. WHAT TO WRITE ?
2/7/2013
 Any thing that comes to your mind.
 Any language you can choose.
 Write a small C code of your favorite game.
 Or your favorite algorithm that you have learned in
your computer science class.
2

3. HOW TO WRITE ?
2/7/2013
 Create a Flow Chart of what is inside your
algorithm.
Start
Is
Temp End
> 90 C
Generate
3
Heat

4. IS YOUR CODE CORRECT ?
2/7/2013
 Review 2 times before a bug catches your eye.
 Share the code with your peer to review it once
again.
 Compile the code.
 0 Warnings, 0 Errors !!!!
 Too good, you have written an excellent code.
 If errors are there, you have to go through your
code and try to fix it.
 If warnings are there, you have to understand the
source of it and try to avoid it.
4

5. THINK OF OPTIMIZATION
2/7/2013
 Optimization in terms of :
 CPU Clocks
 Code Memory
 Data Memory
 Power etc.
5

6. OPTIMIZATION IN TERMS OF CPU CLOCKS
2/7/2013
 for (i=0;i<100;i++)
{
 for(j=0;j<50;j++)
 {
 for(k=0;k<25;k++)
 {
 A[i][j][k]= B[i][j]*8 + C[i]*4;
 }
 }
 }
6

7. OPTIMIZATION IN TERMS OF CPU CLOCKS
2/7/2013
 for (i=0;i<100;i++)
{
M= C[i]<<2;
 for(j=0;j<50;j++)
 {
 N= B[i][j]<<3;
 for(k=0;k<25;k++)
 {
 A[i][j][k]= M + N;
 }
 }
 } 7

8. OPTIMIZATION IN TERMS OF CODE
2/7/2013
 *ptr++ = 10;
 *ptr++ = 20;
 *ptr++ = 30;
 *ptr++ = 40;
for(i=0;i<4;i++)
{
 *ptr++ = (10*i);
 }
8

9. OPTIMIZATION IN TERMS OF DATA
2/7/2013
for(i=0;i<4;i++)
{
 B[i]=A[i]*3;
 for(j=0;j<10;j++)
 {
C[i] [j]= B[i]*5+2;
}
}
9

10. OPTIMIZATION IN TERMS OF DATA
2/7/2013
for(i=0;i<4;i++)
{
 for(j=0;j<10;j++)
 {
C[i][j] = A[i]*15+2;
}
}
10

11. WHAT IS YOUR PLATFORM ?
2/7/2013
 X86
 ARM
 TI C55x
 TI C64x
 ADI Blackfin
11

12. WHAT ARE CODE GENERATION TOOLS ?
2/7/2013
 Compiler
 Assembler
 Archiver
 Linker
 Debugger
 Simulator
 Profiler
 Hex Dump Utility
12

13. COMPILER
2/7/2013
 Compiles the code from high level language to
assembly language.
 Example : C, C++, Fortran, Java
13

14. ASSEMBLER
2/7/2013
 Assembles the assembly code to binary opcodes
and operands.
 MOV #40, A
 SUB A, B, A
 LD *A++, C
14

15. LINKER
2/7/2013
 Links the code as per the provided code and data
memory map.
 Example : TI C64x Linker, gcc
15

16. DEBUGGER
2/7/2013
 To find the bug or fault in the code.
 Run
 Single Step
 Run Till This Point
 Step Over
 Step Through
 Breakpoint
 Log
 Example : TI Code Composer Studio IDE, gdb. 16

17. SIMULATOR
2/7/2013
 Functional Simulator
 Instruction Set Simulator
 Cycle Accurate Simulator
 Example : TI Code Composer Studio IDE
17

18. IS YOUR CODE RUNNING FINE ON SIMULATOR
?
2/7/2013
 Debug till the point your expected output doesn’t
appear on the screen.
 Ahhhhh…… Finally it is coming !!!
18

19. WANT TO TRY IT OUT ON SILICON ?
2/7/2013
 Yes !!! You can run the code on silicon(provided it is
not a mission critical one)
 Connect the hardware board with serial port,
parallel port or PCI through JTAG.
 Load the object file on processor.
 Press Run and see if expected output are coming
on the screen.
 If yes, you are done !!!
 And don’t forget to check the performance of your
code by using a profiler. 19

20. 2/7/2013
THANK YOU
20