Update docs for Align

ash/src/util.rs

@@ -3,14 +3,13 @@ use std::marker::PhantomData;
 use std::mem::size_of;
 use vk;
 
-/// `Align` handles dynamic alignment. x86 aligns on 4 byte boundries but GPUs
-/// sometimes have different requirements. For example if the user wants to create
-/// a Uniform buffer of a Vec3<f32>, `get_buffer_memory_requirements` might return
-/// an alignment of 16 bytes. A Vec3<f32> has a size of 12 bytes. The memory layout
-/// for a `Vec<Vec3<f32>>` will be contigous, because 4 is a multiple of 12. But Vulkan
-/// expects a 4 byte padding in between each Vec3<f32>, if the alignment is 16 bytes.
-/// `Vec3<f32>, 4bytes, Vec3<f32>, 4bytes, Vec3<f32>...`. Align is able to take a slice
-/// that is allocated on 4 bytes boundries, and insert the correct amount of paddings.
+/// `Align` handles dynamic alignment. The is useful for dynamic uniform buffers where
+/// the alignment might be different. For example a 4x4 f32 matrix has a size of 64 bytes
+/// but the min alignment for a dynamic uniform buffer might be 256 bytes. A slice of `&[Mat4x4<f32>]`
+/// has a memory layout of `[[64 bytes], [64 bytes], [64 bytes]]`, but it might need to have a memory
+/// layout of `[[256 bytes], [256 bytes], [256 bytes]].
+/// `Align::copy_from_slice` will copy a slice of `&[T]` directly into the host memory without
+/// an additional allocation and with the correct alignment.
 #[derive(Debug, Clone)]
 pub struct Align<T> {
     ptr: *mut vk::c_void,