What do the ddx and ddy values do in this AABB ray intersect algorithm?

Question

Does anyone know what the ddx and ddy values do in the AABB ray intersect algorithm? Taken from the following site http://www.blitzbasic.com/codearcs/codearcs.php?code=1029 (show below).

Local txmin#,txmax#,tymin#,tymax#

// rox, rdx are the ray origin on the x axis, and ray delta on the x axis ... y-axis is roy and rdy

Local ddx# =1.0/(rox-rdx)
Local ddy# =1.0/(roy-rdy)

If ddx >= 0
    txmin = (bminx - rox) * ddx
    txmax = (bmaxx - rox) * ddx
Else
    txmin = (bmaxx - rox) * ddx
    txmax = (bminx - rox) * ddx
EndIf

If ddy >= 0
    tymin = (bminy - roy) * ddy
    tymax = (bmaxy - roy) * ddy
Else
    tymin = (bmaxy - roy) * ddy
    tymax = (bminy - roy) * ddy
EndIf

If ( (txmin > tymax) Or (tymin > txmax) ) Return 0

If (tymin > txmin) txmin = tymin
If (tymax < txmax) txmax = tymax

Local tzmin#,tzmax#
Local ddz# =1.0/(roz-rdz)

If ddz >= 0
    tzmin = (bminz - roz) * ddz
    tzmax = (bmaxz - roz) * ddz
Else
    tzmin = (bmaxz - roz) * ddz
    tzmax = (bminz - roz) * ddz
EndIf

If (txmin > tzmax) Or (tzmin > txmax) Return 0

Return 1

Answer 1

(rox-rdx, roy-rdy) is the vector from the destination point to the origin point of the ray. ddx and ddy are the inverse of those 2 coordinates.

The inversion was used as a precomputation in order to only have to use a multiplication (by those inverses) instead of a division in the rest of the function. Computers compute multiplications faster than divisions.