8 pass relief mapping?

[charliejay]

I seem to recall that somebody mentioned an 8-pass or multipass relief mapping shader algorithm (probably in one of the several big relief mapping threads), but have been utterly unable to find it, despite searching this board and google and going through the big threads).

Had a look at the multipass link that SKoder posted in the later of fpo's large threads (Curved relief mapping with correct silhouettes), but got an error saying that the file didn't exist.

Does this ring a bell with anyone, and if so, could you post a link or a reference, please?

Thanks, CJ

[SKoder]

4 pass relief mapping:
http://www.webfile.ru/640627
On the center of a page click Download

[holdeWaldfee]

A bit OT again:

I don't understand why someting like relief mapping should be used.

True vertex meshes with rasterized surfaces are more efficient to compute, excepth for cases where the vertex count comes very close to the fragment count.
Or do I misunderstand something?

[charliejay]

SKoder:

Nice one, many thanks!

Linear search seems to be easiest (and only?) way of doing R.M.on HW that supports it. Mine doesn't, more's the pity, so I'm searching (probably in vain) for an alternate method...

holdeWaldfee:

Consider how much bandwith/memory you use up with a mesh - a relief mapped texture gives you the same or better detail, for four bytes per sample point, a mesh is 12 bytes per sample point before you even get colour...

CJ

[Chuck0]

and you get level of detail nearly for free (only as many fragments are rasterized as needed... with a mesh you have to use explicit LOD techniques that can be quite costly)