VulkanMemoryAllocator VS wgpu

Hello - working on a Vulkan renderer and I ran into a topic I don't understand. I use the Vulkan Memory Allocator and upon updating my local targeted Vulkan SDK to 1.3.250 from 1.3.204, I receive exactly the same error as this user: https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator/issues/280 Error: The procedure entry point vkGetDeviceBufferMemoryRequirements could not be located in the DLL.

For problem 1 I would recommend looking into Vulkan Memory Allocator: https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator The general idea is to allocate large chunks of memory, then sub-allocate out of those when buffers are required. This is nice for several reasons: - Memory allocation can be slow and involve system calls. Sub-allocating does not require system calls - Your memory will be more cache-friendly as all your buffers will be relatively adjacent - There is a finite number of unique allocations you can have. Performing larger allocations keeps your usage lower

Integrated Vulkan Memory Allocator, making all GPU memory allocations much easier to work with.

Seems like it was initially added more than 2 years ago, they really should release more often :P

Use VMA, originally created by AMD, never worry about device memory and device memory alignment again.

I am using VMA for memory allocation with the VMA_MEMORY_USAGE_GPU_TO_CPU usage flag. That means that the flags VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT and VK_MEMORY_PROPERTY_HOST_CACHED_BIT are already set.

Repo for Vulkan Memory mangement (AMD): https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator

Since it's a header-only lib, you could always try cloning the latest from [GitHub](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator) and see if the problem continues with that version.

Alternatively, you could just use an existing allocator library, a popular choice is VMA: https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator

Personally, I'd recommend (VMA)[https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator] if you don't want to deal with all of this yourself.

Oh look it's subgroup support landing last week: https://github.com/gfx-rs/wgpu/pull/5301

And wgpu has been doing this for years. Things like descriptor indexing are not exposed to the web but used by Rust (mostly) engines on native.

https://wgpu.rs/

If they used https://wgpu.rs/ they would get directx and metal for free (:

IIRC it was delayed multiple times. I think the first intent to ship from chrome was before 100 but they kept pushing it off. Firefox still does not support it. There are projects like wgpu[0] that wrap provide a higher level API and I have used some projects using it with no issues. WFIW I didn't see any issue with OP's demo either.

[0] https://github.com/gfx-rs/wgpu

wgpu, the Rust WebGPU implementation is the bee's knees. https://wgpu.rs/ You can use it beyond the web.

There is anything like wgpu.rs for Zig? wgpu.rs is an abstraction on top of Vulkan, Metal, DirectX, etc...