I am currently working with the DryWetMidi package to control the KORG nanoKONTROL2 MIDI board with C#/.Net and I noticed that the memory of my application will increase when I fire a lot of MIDI messages (in other words: when I move the faders a lot).

For inspection purposes, I wrote this simple code, which basically does nothing but initializing the DryWetMidi package to start listening to MIDI messages send by my device. Then I have 100 seconds to try out various faders/buttons/knobs (and send a lot of MIDI messages)

static async Task Main(string[] args)
{
    InputDevice nano = InputDevice.GetByName("nanoKONTROL2");
    nano.StartEventsListening();

    Console.WriteLine("start");
    for (int i = 1; i < 101; i  )
    {
        await Task.Delay(1000);
        Console.WriteLine(i   " seconds passed");
    }
    Console.WriteLine("end");

    nano.StopEventsListening();
    nano.Dispose();

}

The memory starts increasing by approx. 15MB when I move the faders a lot. The memory only increases when I move anything (when the board sends MIDI messages)! Since the memory does not increase further after these 15MB, I assume that there are some logging processes going on in the background? Or is it some kind of memory leak?

Now my question:

Is there any way to prevent this memory increase?

CodePudding user response：

I'm the author of the DryWetMIDI library. I've passed MIDI events receiving through .NET memory profiler and here my conclusions:

1. Yes, memory increased due to new instances of MidiEvent class created by DryWetMIDI when new data arrived from a device.
2. GC works as expected. If memory not released, then GC thinks it's not needed right now. In my tests I sent a lot of note events, memory grows and then released.

I don't see any problems. There are no memory leaks and of course there is no any logging within the library. What you see is how .NET memory management works.