The Godot game engine has spent the past number of months collaborating with Google and The Forge to bring performance optimizations to their Vulkan back-end.
Some of the accomplished work includes utilizing unified membry architecture (UMA) buffers where possible, replacing large push constants with dynamic uniform buffers, optimizing descriptor sets and descriptor set batching, and optimizing swapchain operations.
There’s also been some Android-specific work such as supporting the Android Thermal API and Swappy frame-pacing with the Google AGDK.
In testing on mobile devices, the improved code is showing a consistent 10~20% lower GPU frame-time.
The work is currently maintained in a separate code branch and there is this big pull request with the optimizations made over the past several months.
More details on the work of this Vulkan collaboration with Google and The Forge can be found via the Godot Engine blog.
The original article contains 210 words, the summary contains 142 words. Saved 32%. I’m a bot and I’m open source!
This is the best summary I could come up with:
The Godot game engine has spent the past number of months collaborating with Google and The Forge to bring performance optimizations to their Vulkan back-end.
Some of the accomplished work includes utilizing unified membry architecture (UMA) buffers where possible, replacing large push constants with dynamic uniform buffers, optimizing descriptor sets and descriptor set batching, and optimizing swapchain operations.
There’s also been some Android-specific work such as supporting the Android Thermal API and Swappy frame-pacing with the Google AGDK.
In testing on mobile devices, the improved code is showing a consistent 10~20% lower GPU frame-time.
The work is currently maintained in a separate code branch and there is this big pull request with the optimizations made over the past several months.
More details on the work of this Vulkan collaboration with Google and The Forge can be found via the Godot Engine blog.
The original article contains 210 words, the summary contains 142 words. Saved 32%. I’m a bot and I’m open source!