shader - Invariant and precise keywords in GLSL -

I am trying to understand these two concepts. I am studying manual very brief on them and things like multi-chess algorithms are new to me

when they say that

I want some examples (no code) That would require me to use the immutable or precise variable, just to get a general idea. P>

GLSL does not guarantee that there will be two equal computing values ​​in different shaders

I do not know how and when I need 2 equals Computation in different shaders will be

These keywords are necessary to avoid "cracking" and other visual artifacts which may occur in various Computing is being done in a little different way in the shaders.

Rendered scenes are made up of a large number of "primitives" - small pieces of image (usually triangle, sometimes more complex "patch") that are sung each and every image is stored To do with Where two preferences are omitted, it is important that both shadars are included Absolutely Calculate the boundary in the same way if not, things sometimes due to floating-point golfing May be different, which can cause "cracks", so that the background leaks and appears.

For the simple (top + piece) sha ader setup, the irreversible keyword is sufficient for this - you usually show it at the peak position computations, and possibly other For variations that have a spherical sensitivity that can cause artifacts.

When you add to the geometry and soldation shaders, things get worse because the boundaries between patch / premiums can be complex size rather than a straight line. This is often the case that a tesselian shader will calculate a boundary in one direction, while the adjacent will calculate it in each other, so the calculation is completely symmetrical and the bit-similar result is important to produce both methods. This is where the precise keyword comes in the game, generally when you have computing like * b + c * d with the exact which is called MUL + FMA Can not be compiled in form, but instead should be MUL + MUL + ADD.