Page 1 of 1

Square shaped areas in texture

PostPosted: Wed Jul 18, 2007 2:07 pm
by Tormented
First, the project details:

HF: 1024x1024 pixels
AM: 32x, tilesize 256, temperate climate
TN: 32x, tilesize 256, no bump
LM: 32x, tilesize 256, no water effects, no bump, and no cast shadows
TX: 32x, tilesize 1024, 8x anti-aliasing

In the sandy underwater parts of the texture there are very noticeable square shaped areas with sharp edges.

Image

PostPosted: Sun Jul 22, 2007 10:21 pm
by Aaron
Hi Tormented,

Thanks for the bug report. I'll look into it.

Best regards,
Aaron.

PostPosted: Tue Aug 14, 2007 1:12 pm
by Aaron
Hi Tormented,

I apologise that I haven't got back to you sooner on this bug.

I've had a look through my 32x res megatexture maps, and I've not (yet) found this problem. Can you tell me which version/build of L3DT you were using? (in the help->about box) Also, were these blocks present in the same region of the attributes map?

My line of questioning here is to determine whether the bug is in the attributes map algorithm, or in the texture anti-aliasing algorithm, and if it's the latter, then to determine which version of the anti-aliasing algorithm is at fault (it has been revised twice). I suspect (hope?) that the fault was in an earlier iteration of the texture anti-aliasing algorithm.

Then again, I must admit I haven't tested the megatexturing using a tile size of 256 for the attributes map, and nor have I recently tested the attributes map generation at 32x resolution (I normally use 16x for the AM, and 32x for the rest). On the face of it, these settings shouldn't be a problem, but I'll have to test it to be sure. I'll It looks like my poor computer has some serious work ahead of it...

Best regards,
Aaron.

PostPosted: Tue Aug 14, 2007 2:01 pm
by Aaron
Hi Tormented,

Please disregard the last post. I just managed to reproduce this bug; it's a problem in the high-res attributes map calculation. I'm on the case!

Best regards,
Aaron.

PostPosted: Wed Aug 15, 2007 11:12 am
by Aaron
Hi Tormented,

I've found and fixed the problem. In the interests of speed, I was using a 3x3 sampling routine where I should have been using 4x4. Fortunately, this was only a one-line fix, which I've added to the very latest dev build. Please let me know if problems persist.

Best regards,
Aaron.