I'm planning on creating a real-time audio editing and sequencing application in C# using mainly the cscore library for handling the audio on the backend, and then Avalonia for the frontend (using Avalonia.Markup.Declarative, although not sure that matters a huge amount). I'm expecting that the front-end is going to require occasional GC cleanup during normal operation/use, and would like to make sure that it won't affect the backend during GC - if the audio backend has to pause during GC there may be audible freezing or lag when this occurs. It's feasible for me to write code using pre-allocated memory or manual memory management for the backend, but it's not feasible for Avalonia, without modifying almost the entirety of its internal logic and massively changing how the I interact with the framework relative to the official docs (may not be possible, period - I've not really looked into it much).
Is there a way to create separate heaps/GC per thread, so that the backend threads can have their own separate GC not influenced by the UI, or would I have to create two separate applications in a server/client style and do some form of IPC between them. Alternatively, is there a way to mark backend code that requires no GC interaction to continue running during a GC clean-up, if I know for a fact it won't be affected. If so, how would be best to go about it, given the potentially large amounts of arbitrary data I'd like to transfer between them.
I don't think, that you can have two GCs for one app. But, you can make two apps:
For user you can make an shortcut on Desktop for your appui.exe. When the UI app is starting, it should check, do we have started appbackend.exe, or no. If no - start it and wait for loading.
As a result, you will get two absolutely separated apps, with different GC. You can setup communications between apps using any network protocol, or you can take JetBrains.Rd