Can you explain the simd layer?

[btgvesta]

Can you please explain the SIMD layer in the C project?

[bagaffey]

Yes, like you said, this is an abstraction layer for using x86 SSE intrinsics. I am only using the 4 floats at a time operations.
This is used for various math and vector operations.
This makes using these operations easier to use than writing out raw _mm_mul_ps(a, b) calls everywhere throughout the code.

It creates 4 reusable types and functions (operations) to use with those types. They are:

1. f32_4x 4 element vector with 32-bit floating point values or 32-bit integers, either are supported.
2. v3_4x - this is traditional vector with 3 scalars, it stores 4 of them batched together for SIMD registers and operations.
3. v4_4x - Just like the v3_4x but for vectors with 4 scalar properties.

There are then a bunch of operations defined to use with these structs. There are also wrapper functions around intrinsics, masking and selection functions. There are also utility macros for shuffling.