Concerns regarding performance throughout framework, future goals

[bjadamson]

Concerns regarding performance throughout framework, future goals

[ebkalderon]

@bjadamson You have some valid points here, and I agree that resolving them is definitely on the radar, especially the point about reducing our dependence on locks where possible. Indeed, the shared Context might need to slim down a bit to reduce contention and improve performace. For example, removing the renderer from the context and placing it directly into the Rendering processor where it belongs comes to mind.

However, given that most games are big blobs of shared mutable state, there is only so much we can do on this front. The idea that state changes are collected at the end of the frame and processed all at once with correct ordering was actually considered at one point during issue #10, but in the end the majority seemed to favor the mutex-based approach for various reasons, mostly due to inevitable latency.

Thankfully, Amethyst is still a young project and as such can afford to experiment on different approaches without really worrying (or perhaps caring?) about breakage until we find what works best.

[Xaeroxe]

We use next to 0 Arcs and Mutexes today so I'm closing this.