Closed adrien-ben closed 2 years ago
I think i'm done with this but i'm still unsure about 2 things:
when enabling gpu-allocator
feature you need to create a Renderer using Renderer::with_gpu_allocator
which takes a Arc<Mutex<gpu_allocator::vulkan::Allocator>>
. It works for my usage of the renderer and you don't need to pass a reference to each function call. But I'm not sure it would feat other's usage of the crate or if it is good practice. Any suggestion is welcome.
Should the gpu-allocator
feature disable the default allocator ? It would allow me to get rid of the internal enum allowing multiple allocator implementation at the same time. It does not affect the api but still...
I added vk-mem back but using a forked version for ash compatibility since original repo is not maintained anymore
Reworking the api. The
RendererVkContext
is pretty hard to work with especially when trying to integrate gpu-allocator.