Working with large floating point numbers.

[lordreas]

Hello there,

I'm struggling with calculations involving large floats. Running the following code on my Wemos Lolin32 Lite on Arduino featuring an ESP-32 Rev1

Code: [Select all] [Expand/Collapse]

1. double test = 2.;
2.  
3. void setup() {
4.   // put your setup code here, to run once:
5.   Serial.begin(31250);
6. }
7.  
8. void loop() {
9.   // put your main code here, to run repeatedly:
10.   test*=1.1;
11.   Serial.println(test);
12. }

GeSHi © Codebox Plus Extension

leading to the following output after some time:

Code: Select all

3400938557.93
3741032413.72
4115135655.09
ovf
ovf
ovf

4115135655.09 is way below theoretical limits of float (3e38) let alone double (1e308). What's happening here? I know Esp32 is not the best number for fast or precise FLOPs, but something else but precision seems to be wrong here.

I would be glad if someone could reproduce this / help me / hint me at other resources, thanks in advance!

[ESP_Sprite]

That overflow seems to happen at the limit of a 32-bit integer. I'd guess the double variable itself doesn't overflow; it's the println routine that messes up. Perhaps you can see what happens if you replace that with a printf("%f\n", test); ?