Lightning Engine

Goal: Calculate the sin/cos of an arbitrary angle.
Restrictions:

1. Should be faster than CRT’s sin/cos and faster than FPU’s fsin/fcos
2. The error should be kept to a minimum compared to the above functions
3. Double precision is required
4. Function should be in the form: double func(double), ie. no xmm regs passed to the function, result should be returned on the FPU stack and it should calculate only 1 angle at a time

For the rest of the post we’ll be talking about sin(). Calculating cos(), or even sincos() should be easily derived from the code below. The required changes will be described at the end of the post.
(more…)

Goal: Classify whether a batch of AABBs are completely inside, completely outside or intersecting a frustum (6 planes).
Restrictions:

1. AABBs are defined as (Center, Extent) pairs.
2. All vectors are Vector3f’s

(more…)

Goal: Calculate the Axis Aligned Bounding Box of a point list.
Restrictions:

1. Vertices as Vector3f’s
2. (Optional) Vertex list should be 16-byte aligned

(more…)

Goal: Multiply a batch of Vector3f’s with the same 4×4 matrix.
Restrictions:

1. ‘src’ and ‘dst’ arrays shouldn’t point to the same memory location
2. All pointers should be 16-byte aligned (see below for details on array sizes)
3. Treat Vector3f’s as positions (w = 1.0)
4. Matrix is column major

(more…)