The maximum size of a state buffer page in bytes.

The alignment in bytes within a state buffer of float or integer data types.

The alignment in bytes of a 4 element vector of floats. While vectors of other element counts are allowed, there are highly unusual packing rules that greatly complicate things and vectors of 4 are preferred - just trim unneeded elements off. (E.g. in Vulkan a float2 needs (address mod 8 == 0), in Direct3D 11 and 12 it needs (address mod 16) to be either (0, 4, or 8), and OpenGL needs (address mod 16 == 0)).

The alignment in bytes of a 4 x 4 matrix of floats. While matrices of other element counts are allowed, there are highly unusual packing rules that greatly complicate things and matrices of 4 x 4 are preferred - just trim unneeded elements off. (E.g. mat3x2 and mat2x3 often have different sizes and alignments on the same engine. On Vulkan Size (24 vs 32) and Alignment (8 vs 16))

The minimum alignment of an entire array in addition to any other restriction on its type. E.g. Vulkan has an entire array alignment rule of 16, so a float[1] array takes up 4 times the space as a single float.

The minimum alignment of each element in the array. E.g. a float[2] and vec2f may not have identical alignment. The 2 floats may not even be next to each other, an alignmentArrayStride of 16 means there's 3 floats padding between the two elements. For simplicitly and portable code it's recommended to avoid arrays of scalars and only have arrays of vec4s, mat4x4s, other arrays, and structures. E.g. float[160] should be vec4f[40].

The minimum alignment of a structure in addition to any other restriction, as "float a, b" and "struct float_t {float my;}; float_t a, b" can have differing alignments for a and b (e.g. 4 vs 16 in Vulkan, but both are 4 in OpenGL 4.3).