reflection of the incident vector I according to the normal N, N must be normalized

N and I must be normalized

rounds the components of this vector to next multiple of "multiple" (e.g. { 17, 32, 33 }.round_next_multiple(32) == { 32, 32, 64 })

returns the maximum SIMD-width of the device (or 0 if unknown)

returns the corresponding VkBlendOp for the specified BLEND_OP

returns the corresponding VkFormat for the specified IMAGE_TYPE, return empty if there is no matching format

returns the corresponding VkIndexType for the specified INDEX_TYPE

converts a VkObjectType to a human-readable string

return the argument buffer index for the specified buffer index

increments indices dependent on the arg

the max amount of storage we need for compressed data for the given "input_size"

returns the debug name of the specified buffer or "unknown"

creates and initializes the per-thread/thread-local primary command pool/buffers

creates and initializes the per-thread/thread-local secondary command pool/buffers

note that CUDA doesn't have any special argument types and everything is just sized "memory" -> only need to add up sizes