AM2634: Instructions and clock calculation

Rohal

Intellectual 495 points

Part Number: AM2634

Tool/software:

Dear Team,

I want to calculate program execution time to define the state machine cycles for our project.

I want to know how to calculate program execution time in CCS

SDK - mcu_plus_sdk_am263x_09_02_00_56

lets take a example:

void initializeStringArray() {

char stringArray[10][100];

char buff[100] = { 0 };

// Initialize the string array

for (int i = 0; i < 10; ++i) {

for (int j = 0; j < 100; j++) {

stringArray[i][j] = j;

}

stringArray[i][99] = '\0';

}

we have to calculatre the time taken to execute the above code.

over 1 year ago

+1 Kier over 1 year ago

Mastermind 9520 points

Here's how to get the number of execution cycles: AM263x MCU+ SDK: Cycle Counter (ti.com)

0 Rohal over 1 year ago in reply to Kier

Intellectual 495 points

thanks for your support & answer

0 Rohal over 1 year ago in reply to Kier

Intellectual 495 points

Dear Kier,

I have 2 observations

1. WIthout Optimization:-

CycleCounterP_reset();

cycleCountBefore = CycleCounterP_getCount32();
initializeStringArray(); // User functon
cycleCountAfter = CycleCounterP_getCount32();

void initializeStringArray() {

char stringArray[10][100];
char buff[100] = { 0 };

// Initialize the string array
for (int i = 0; i < 10; ++i) {
for (int j = 0; j < 100; j++) {
stringArray[i][j] = j;
}
stringArray[i][99] = '\0';
}
}

without optimzation its showing 23154 clocks that is irrelevant, how its possible to take that much of clock

2. With Optimizatoin:-

and with optimization its showing 9 clock only that's also irrelevant

So kindly let me know how to test task execution time

/------------------------- Unoptimized code assemble --------------------------/

47 void initializeStringArray() {
initializeStringArray():
70051ed0: B580 push {r7, r14}
70051ed2: F5AD6D8B sub.w r13, r13, #0x458
70051ed6: A803 add r0, r13, #0xc
70051ed8: 2164 movs r1, #0x64
50 char buff[100] = { 0 };
70051eda: F001EB36 blx __aeabi_memclr
70051ede: 2000 movs r0, #0
54 for (int i = 0; i < 10; ++i) {
70051ee0: 9002 str r0, [r13, #8]
70051ee2: E7FF b #0x70051ee4
70051ee4: 9802 ldr r0, [r13, #8]
70051ee6: 2809 cmp r0, #9
70051ee8: DC20 bgt #0x70051f2c
70051eea: E7FF b #0x70051eec
70051eec: 2000 movs r0, #0
55 for (int j = 0; j < 100; j++) {
70051eee: 9001 str r0, [r13, #4]
70051ef0: E7FF b #0x70051ef2
70051ef2: 9801 ldr r0, [r13, #4]
70051ef4: 2863 cmp r0, #0x63
70051ef6: DC0C bgt #0x70051f12
70051ef8: E7FF b #0x70051efa
56 stringArray[i][j] = j;
70051efa: 9801 ldr r0, [r13, #4]
70051efc: 9902 ldr r1, [r13, #8]
70051efe: AB1C add r3, r13, #0x70
70051f00: 2264 movs r2, #0x64
70051f02: FB013102 mla r1, r1, r2, r3
70051f06: 5408 strb r0, [r1, r0]
57 }
70051f08: E7FF b #0x70051f0a
55 for (int j = 0; j < 100; j++) {
70051f0a: 9801 ldr r0, [r13, #4]
70051f0c: 3001 adds r0, #1
70051f0e: 9001 str r0, [r13, #4]
70051f10: E7EF b #0x70051ef2
58 stringArray[i][99] = '\0';
70051f12: 9802 ldr r0, [r13, #8]
70051f14: AA1C add r2, r13, #0x70
70051f16: 2164 movs r1, #0x64
70051f18: FB002101 mla r1, r0, r1, r2
70051f1c: 2000 movs r0, #0
70051f1e: F8810063 strb.w r0, [r1, #0x63]
59 }
70051f22: E7FF b #0x70051f24
54 for (int i = 0; i < 10; ++i) {
70051f24: 9802 ldr r0, [r13, #8]
70051f26: 3001 adds r0, #1
70051f28: 9002 str r0, [r13, #8]
70051f2a: E7DB b #0x70051ee4
60 }
70051f2c: F50D6D8B add.w r13, r13, #0x458
70051f30: BD80 pop {r7, pc}
145 }

/-------------------------------------------------------------------------------------------/

+1 Akshit Pareek over 1 year ago in reply to Rohal

TI__Expert 4230 points

Hi Vikas,

As you have tried in your code, the usage of Cycle counter is correct.

uint32_t start, end, overhead;

    /* Calculate overhead */
    CycleCounterP_reset();

    start = CycleCounterP_getCount32();
    end = CycleCounterP_getCount32();
    overhead = end - start;

    DebugP_log("Total Overhead: %d Cycles\r\n", overhead);


As it's a 32 bit counter, you can handle overflow like: 

uint32_t cycleCountBefore, cycleCountAfter, cpuCycles;

    /* enable and reset CPU cycle coutner */
    CycleCounterP_reset();

    cycleCountBefore = CycleCounterP_getCount32();

    /* call functions to profile */

    cycleCountAfter = CycleCounterP_getCount32();

    /* Check for overflow and wrap around.
     *
     * This logic will only work for one overflow.
     * If multiple overflows happen during the profile period,
     * then CPU cycles count will be wrong.
     */
    if (cycleCountAfter > cycleCountBefore)
    {
        cpuCycles = cycleCountAfter - cycleCountBefore;
    }
    else
    {
        cpuCycles = (0xFFFFFFFFU - cycleCountBefore) + cycleCountAfter;
    }

If measurement of longer durations is needed, it is recommended to use the ClockP_getTimeUsec() API from the Clock module.

You can also use the CCS in-built profiling options; Please refer to the following links for more information:

Regards,
Akshit

Arm-based microcontrollers

Arm-based microcontrollers forum

AM2634: Instructions and clock calculation