Just a qestion, have you tested that your log and exp implementations are really faste

Here's a quick test on the T3.2, 96Mhz, Faster: <div class="snippet-clipboard-cont

log/exp about audioeffectdynamics HOT 2 OPEN

FrankBoesing commented on June 2, 2024

log/exp

from audioeffectdynamics.

Comments (2)

MarkzP commented on June 2, 2024

Here's a quick test on the T3.2, 96Mhz, Faster:

void setup()
{
  unsigned long start;
  float f;
  float q;

  randomSeed(12345);
  Serial.println("Go!");
  start = micros();

  for (int i = 0; i < 1000000; ++i)
  {
    f = random(10000) / 10000;

    q = log2f(f);
       
    (void)q;
  }

  unsigned long t = micros() - start;
  Serial.print("Done: "); Serial.print(t); Serial.println();
}

Here are the results:

    q = log(f);                 // 7557535
    q = log2f(f);               // 7808295
    q = log2f_approx(f);        // 5178506

    q = exp(f);                 // 9019425
    q = expf(f);                // 6514553
    q = expf_approx(f);         // 5555512

I would say it's a modest improvement at best on the T3.2, but still worth it IMHO

from audioeffectdynamics.

MarkzP commented on June 2, 2024

Now I need to repeat this on a T3.5 (just need to find one!)

from audioeffectdynamics.

Recommend Projects