Basic Smart Placement

Table of Contents

• Introduction
• Supported Combinations
  • Building benchmark application
• Steps to Run the Example
• Sample Output
• Description
  • Process of smart placement
    • Critical Function Identification
    • Priority Assignment
    • Annotating functions
    • Linker change
  • How Smart Placement improves code performance?
  • How Cache Miss ratio is also improved?
• Conclusion

Introduction

This example provides a basic overview of applying smart placement and compares run time of functions with smart placement and without smart placement.

The Aim of this example is to:

1. Showcase the process of smart placement in simple terms
2. How Smart Placement improves code performance?
3. How Cache Miss ratio is also improved?

More on smart placement can be read at Smart Placement

Supported Combinations

Parameter	Value
CPU + OS	r5fss0-0 nortos
Toolchain	>= ti-arm-clang
Boards	am243x-evm, am243x-lp
Example folder	examples/kernel/nortos/basic_smart_placement

Building benchmark application

To build this application, compiler ti-cgt-armllvm >= 3.2.0 LTS or later is required. Application can be compiled using make command.

Steps to Run the Example

• When using CCS projects to build, import the CCS project for the required combination and build it using the CCS project menu (see Using SDK with CCS Projects).
• When using makefiles to build, note the required combination and build using make command (see Using SDK with Makefiles)
• Launch a CCS debug session and run the executable, see CCS Launch, Load and Run

Sample Output

Shown below is a sample output when the application is run,

Profile Point: Without Smart Placement
Cycle Count: 9964
No. Of CPU instructions executed Count: 3660
ICache Miss Count: 227
ICache Access Count: 9973
 
Profile Point: With Smart Placement
Cycle Count: 3906
No. Of CPU instructions executed Count: 3663
ICache Miss Count: 1
ICache Access Count: 3905

Description

When the program runs, it first calls basic_smart_placement_main function which internally calls function_f1 and annotated_function_f1.

PMU has been used for profiling execution of function_f1 and annotated_function_f1.

Process of smart placement

Following steps can be followed to apply smart placement:

Critical Function Identification

Here, critical functions are manually found using inspection. Because this is manual process, criticality of a function is determined purely based on the knowledge of software/firmware/usecase.

Priority Assignment

Following is critical function's priority table.

Function Name	Priority
annotated_function_f1	1
annotated_function_f2	2
annotated_function_f3	2
annotated_function_f4	2

Here note that from the manual inspection it has been identified that annotated_function_f1 is more critical than rest of the functions therefore, it has given more a priority number which is smaller than other function's priority number. Because rest of the critical functions are all relatively not as of same priority so they are clubbed together by given them same priority number.

What priority number to assign to which critical function is also a manual process which involves a good understanding of firmware.

Annotating functions

Once the above table has been formed, each function is annotated as shown:

void __attribute__((local(1))) annotated_function_f1(void);
void __attribute__((local(2))) annotated_function_f2(void);
void __attribute__((local(2))) annotated_function_f3(void);
void __attribute__((local(2))) annotated_function_f4(void);

Note:

1. Attributes are added in front, and won't work if following is tried:

   void annotated_function_f1(void) __ attribute __((local(1)));

2. by specifying local, all these functions are indicated to be placed in local memory.

Linker change

Here it has to be made sure that following 3 lines are added/present in the SECTIONS of linker.cmd:

.TI.local   : {} >> R5F_TCMA | R5F_TCMB | MSRAM
.TI.onchip  : {} >> MSRAM | FLASH
.TI.offchip : {} > FLASH

The above lines basically channeling all the functions that are annotated to be in local memory into TCM memory and if total size of the functions that are marked local is more than the size of R5F_TCMA then all the functions that could be placed in R5F_TCMA will be placed in R5F_TCMA and rest of functions will be moved in R5F_TCMB and even if it still fills R5F_TCMB then remaining function will be moved to MSRAM.

Similar treatment is for all the functions that are marked onchip, however, they should never be placed in any TCM otherwise it will be logically wrong.

Also all functions which are marked offchip, should be placed in external FLASH.

How Smart Placement improves code performance?

From the linker, onchip marked functions are routed to TCMA/B and TCM are fastest memories. Therefore, by annotating the critical functions directly from the source, code is placed in faster memory and hence runtime performance has been improved.

This can be seen from the above sample output. For same number of instructions executed, CPU cycles taken to execute functions without smart placement is 3 times more than with smart placement.

How Cache Miss ratio is also improved?

Again, TCMs are never cached. Therefore, for the critical functions which are not cached well, placing them in TCM will help in improving cache rate and ultimately CPI of function.

Conclusion

This example shows how to manually apply smart placement and shows how it improves code performance with ease-of-use because of it allows placement of function at source level.