Does MoltenVK allow vertexAttributeAccessBeyondStride if the stride is zero?

[FilippoLeon]

I'm getting this validation and I'm not sure if MoltenVk would accept such a case (is not crashing and working), or if the validation in indeed correct.

Validation Error: [ VUID-VkVertexInputAttributeDescription-vertexAttributeAccessBeyondStride-04457 ] | MessageID = 0x8690bfd0 | vkCreateGraphicsPipelines(): pCreateInfos[0] (portability error): attribute.offset (0) + sizeof(vertex_description.format) (4) is larger than the vertex stride (0).
+ The Vulkan spec states: If the VK_KHR_portability_subset extension is enabled, and VkPhysicalDevicePortabilitySubsetFeaturesKHR::vertexAttributeAccessBeyondStride is VK_FALSE, the sum of offset plus the size of the vertex attribute data described by format must not be greater than stride in the VkVertexInputBindingDescription referenced in binding (https://vulkan.lunarg.com/doc/view/1.3.296.0/mac/1.3-extensions/vkspec.html#VUID-VkVertexInputAttributeDescription-vertexAttributeAccessBeyondStride-04457)

Or is the validation being overzealous?

[cdavis5e]

tl;dr I think the spec is too strict here.

MoltenVK does support this case; this is because, in Metal, a Vulkan stride of zero corresponds to MTLStepFunctionConstant, where the restriction on attribute offsets and buffer strides doesn't apply. However, a strict reading of the spec would indicate that an attribute's offset must never exceed the buffer stride when vertexAttributeAccessBeyondStride is false, even when the stride is zero. Therefore, the validation is correct according to the spec. But this is too strict for what Metal supports. Unfortunately, I don't have any pull with the Vulkan committee anymore, so I can't get the spec fixed here.

[FilippoLeon]

that's what I suspected, thanks

I think the "spirit" of the spec would be to accept such a case.

I will see to open a ticket for the vulkan specification itself.