A new subimage function which would wrap the region being uploaded at the edges of the destination texture - so if the region overlapped the right edge of the texture, the overlapping portion would be copied to the left edge, instead of being clamped.
There are render-to-texture workarounds for this, but they involve a temporary scrap texture to put the region data into first.
This would help with geometry clipmaps.
I have seen bad suggestions before, but this could actually be useful, though i don’t think it will make it into gl2.x.
But i do think there is room for a much more powerful set of glTexSubImage2D like tools in Longs Peak.
I won’t hold my breath.
As usual, because there’s a work around people will be against it.
The work arounds are render-to-texture, or several glPixelStore+glTexSubImage2D calls.
It couldn’t hurt to make it a bit more flexible, especially when you still have to rewrite certain functions to accommodate for the new object model.
I’d only want this added if there’s possible server-side support for it…if it’s going to be the driver hiding the details then maybe it shouldn’t be exposed.
I’m becoming a bit of a purest about this new GL API.
Sounds like it’s easy to put in hw.
What are you doing that requires this?
V-Man : anything with very large textures. It is a way to “virtualize” by hand the texture memory.
Clipmaps , " megatextures " and of course geometry clipmaps as expressed by the OP.
Regarding using this for software clipmaps, I think there may be some patent issues that would get in the way, not that I’m a big fan of software patents.
I can find the patent number if anyone wants, but look for “Universal Texture” if you want to find it yourself.
Who do you reckon has the patent?
Microsoft? SGI? Google? Keyhole? ID? NASA?
You can implement something similar to geometry clipmaps in many different ways. It would be impossible to patent it.
Funny. It might be Vicarious Visions, actually. They bought a lot of assets from Intrinsic.
As for the rest, Keyhole doesn’t exist anymore. Google seems to patent just hardware. ID doesn’t patent anything (good for them). I sure hope NASA doesn’t. And SGI traded most of their good ones to Microsoft, God knows why…
I have no idea if anyone would notice or complain. But I can imagine the mere existence of protected IP causing problems for a new GL extension. I hope it wouldn’t. But I’m just mentioning it out of concern.
(personally, I wouldn’t mind UT becoming public domain at this point)
So who should I make out the cheque to?
If you’re making out checks, I’ll take one. I can’t help on the IP issue though.
If this extension moves along, the author(s) can email me privately and I can ask someone who might know. [just to be clear, Chris Tanner is the inventor. I was just kidding about the check ;]
I was under the impression that Microsoft and other large companies were in the habit of applying for patents as more of a defensive maneuver than anything else. In the event of being sued, they could play the counter-suit card, with a patent (pending).
With varied interpretations and implementations of this basic algorithm, I imagine the legal waters get very murky indeed. Take the relatively new twist on the basic theme going by the name of “Seamless Patches”, which uses a quadtree and a different tile layout…