Is 10 minute compile time normal for Arduino nano 33 BLE?

Hello!

I am currently following the edge impulse coursera course (It is excellent by the way) and trying to implement the magic wand / motion detector on the Arduino nano ble 33.

One big issue I have is that compiling the arduino code with the motion inference library takes about 10 minutes on a Ryzen 5 3600, is this normal expected behaviour?

I am using Win 10, Arduino IDE 2.0 (to solve the to long path error) and I have disabled real time protection in windows defender because I saw anti-malware executable fight a lot with the compiler.

Despite of this I still have ~10 min compile time. When I look at performance graphs the first 9 minutes only utilizes 2 CPU cores, while the last minute utilizes all cores to almost max.

So, I wonder if this is something normal and expected or if it is worthwhile to try to troubleshoot and solve this?

1 Like

I have no idea what they do but indeed the Arduino IDE has shown ridiculous compile times on some Windows systems… I’m not entirely sure why. They should be using all available cores (see e.g. https://github.com/arduino/Arduino/issues/8589). Maybe process affinity is not set to all CPUs?

1 Like

Hello, in my case with a more modest PC with i3 in W10 it is taking about 24 minutes each compilation with Arduino 1.8.5. I really love Edge Impulse but it drives me crazy having to use Arduino on W10. I don’t know how to solve it any other way.

@tototek Yeah I completely agree! I have no idea what’s happening here but I remember having similar issues with compiling Mbed OS on Windows. Users have been reporting that the Arduino online IDE works much better.

1 Like

True that. The offline IDE is simply too slow for a compile. I suggest moving to the Arduino Cloud IDE which gets the work done in under a minute.

@Morne try PlatformIO.

Check also Use PlatformIO to Build Your tinyML Projects (Edge Impulse Blog)

1 Like