Go Down

Topic: nanos() [time elapsed since program start in nanoseconds]? (Read 4818 times) previous topic - next topic


Nov 27, 2013, 12:46 am Last Edit: Nov 27, 2013, 05:03 am by titous Reason: 1
hi everyone!

In my current application I'd like to be able to have access to timing in the 10 to 100 ns range.  So I started digging around in the micros() source code and came up with the following:

Code: [Select]
uint32_t nanos( void )
    uint32_t ticks ;
    uint32_t count ;

    do {
        ticks = SysTick->VAL;
    } while (SysTick->CTRL & SysTick_CTRL_COUNTFLAG_Msk);

    // GetTickCount() is millis()
    uint32_t load= (SysTick->LOAD + 1 - ticks);
    uint32_t milliS = GetTickCount() * 1000000;
    uint32_t microS =  load/ (SystemCoreClock/1000000) * 1000;
    uint32_t nanoS = load/ (SystemCoreClock/10000000); // these are actually 100s of nanoseconds; 84MHz  = 11.9ns
    //Serial.print(milliS); Serial.print(" "); Serial.print(microS); Serial.print(" "); Serial.println(nanoS);
    return milliS + microS + nanoS;


It's exactly the micros() source code except that on the return I changed the divisions by 1000.  Could someone please comment on whether this is the right approach?

On a side note: isn't division notoriously slow on the Due, so isn't there a better way to optimize this code for performance?


multiplication and division are in hardware on the Cortex, multiplication is single cycle,
don't know the figure for division though.
[ I will NOT respond to personal messages, I WILL delete them, use the forum please ]

Go Up