bump mapping, normal mapping, offset mapping, and....
- Thread starter PowerK
- Start date
Well, it's true that there'll be no denormalization since the z-component is derived based on the assumption that x and y are part of a normalized vec3. . . However, the error is always there. Assume, for example, that one is working with normalized vec2. Interpolating between two vec2s will have a denormalized result. Of course, one can simply renormalize a vec2, but when we're trying to derive a third component this is not possible. As such, the result may be a normalized vec3, but it is still derived from an error.
Ostsol said:
objectively speaking you're right - interpolating as vec2 and then deriving the 3rd component such that the norm of the vec3 is 1 is definitely not the mathematically-correct way, yet i doubt it would produce visually-noticible errors (no denormalities), thus would be acceptable for game purposes. of course, it would have been so much nicer if hw did the re-normalisation when sampling from a source to a dot3 operation...
I can agree with that.darkblu said:
How would it determine that? Perhaps for the fixed function pipeline that'd be feasible, but probably not for the programmable pipeline. A special texture lookup op might do the trick -- but rather than crowd the API much more, I think a new texture filtering method would be a better idea. Perhaps. . . cubic interpolation?
ninelven said:
I was a bit harsh, so I deserved that
But on a more serious note, as graphics gets more complex using the correct term will become increasingly important, especially if you talking to 'experts' in the field. When I originally read your message, I had no idea what you were talking about, I couldn't tell if you meant normal maps or APS. The are lots of bad information on the net and being a bit pendantic might help a few people understand whats actually being talked about.