We probably should, but I think we're also allowed to just clamp to [0, 1]. It's not immediately clear to me whether a corresponding caps bit exists for this in d3d12.
If, as it seems, there isn't one, then I think `DepthBoundsTestSupported` should be intended as advertising support for an unrestricted range, so shouldn't be set unless we're able to uphold to that promise (i.e., `VK_EXT_depth_range_unrestricted` is available). Why is clamping to [0, 1] a good solution? The application is likely to misrender if we change the requested parameters under its feet.
I'm mostly going by the line "The runtime will not clamp or validate the input, but implementations may clamp to the range [0,1] if necessary." from https://microsoft.github.io/DirectX-Specs/d3d/DepthBoundsTest.html#detailed-.... That seems slightly ambiguous though; it's not immediately clear to me whether "necessary" there refers to e.g. limitations of the depth format in use, or limitations of the GPU more generally. (I.e., the question seems somewhat moot for depth formats like DXGI_FORMAT_D24_UNORM_S8_UINT that can't represent values outside [0,1].) The "Testing" section doesn't appear to address that question either.
-
Henri Verbeet (＠hverbeet)