Documentation around CanUseSmallAlignment being overly conservative

[shubham2398]

D3D12MemoryAllocator/src/D3D12MemAlloc.cpp

Line 771 in 0fa62ed

// This algorithm is overly conservative.

As per https://learn.microsoft.com/en-us/windows/win32/api/d3d12/ns-d3d12-d3d12_resource_desc#alignment , the tiling logic specifies constraints similar to the ones implemented in CanUseSmallAlignment. It also says that -

The runtime will use an architecture-independent mechanism of size-estimation, that mimics the way standard swizzle and D3D12 tiled resources are sized.

So why does the documentation mention that this function is overly conservative? Is it because not all pixel formats are specified in GetBitsPerPixel or some other reason is there too?

[sawickiap]

Yes, this disclaimer is there because for some pixel formats or other parameters this function may return false even for formats that could formally use small alignment, according to the Microsoft spec. I try to make it as close as possible, though. I'm not sure if anything is still missing.

By the way, I recommend predefining #define D3D12MA_USE_SMALL_RESOURCE_PLACEMENT_ALIGNMENT 2. Then, the function is not used, and the library always queries D3D12MA code whether small alignment is supported.

[shubham2398]

Thank you @sawickiap for your response.
Yes, we have tried using #define D3D12MA_USE_SMALL_RESOURCE_PLACEMENT_ALIGNMENT 2 but it causes the debug layer to throw errors. And we want to avoid unnecessary calls to the device if the same can be computed mathematically, without generating errors on the debug layer.
But I get your point. Thank you!