Page 7 of 18

Re: mapconv

Posted: 06 Feb 2008, 00:09
by user
testing new -z function it doesnt works as i expected, but it works,

just write a line on file that has something like this:

nvdxt.exe -file temp\*.png [options]

i can try to make the -file not required but it will take some time,

and dxt3 compression creates 1,3mb files.

Re: mapconv

Posted: 06 Feb 2008, 00:18
by Argh
The main reason to use DXT3 is for quality, not because it's small. And I wouldn't worry about getting rid of the -file thing, I just want to test this, see if I can get your version to render a map without the tile problems, frankly.

If this -z stuff works, though, I guarantee people will want to try various options. It'll be really, really nice just to have the ability to tweak how the compressor handles stuff- that, by itself, is a major big deal, because with different compressor settings, you can really boost performance, and beauty. I'm going to check the map code, and see if it just uses GL_MIPMAP_NEAREST when rendering... I'll bet it does, and if so, then we can actually test using a single mip-level vs. multiple, etc. in terms of performance vs. beauty.

Re: mapconv

Posted: 06 Feb 2008, 00:21
by user
so the -file requirement is not a problem?

just tested one dxt3, slow but got some good results, going to test a few more times then relase,

yes -z works,

Re: mapconv

Posted: 06 Feb 2008, 00:26
by Argh
Sweet. Finally, a line-free map format. Having texture-tweaking stuff on top of that is going to be awesome.

Re: mapconv

Posted: 06 Feb 2008, 00:42
by user
it is already line free, texcompress was the solution, that is what i thought, i did not tested nvdxt correctly to see if there were options that would not give lines.

and other tex compressing programs will be suported too so more options for texturing.

these are the tex compressing programs that i know that can be suported:

texcompress,
nvdxt,
nvcompress,
texconv,

Re: mapconv

Posted: 06 Feb 2008, 00:47
by Argh
So, um, does texcompress no longer cause tile issues? Because if I have to pick between tile issues and lines, I'm gonna have to pick lines. I strongly suspect that the lines are due to the fade being applied, which was causing artifacts when GL_MIPMAP_NEAREST was being applied... we'll see...

Re: mapconv

Posted: 06 Feb 2008, 00:50
by user
ever tried zooming in to max on a dds tile from the old mapconv?

you will see that the colours are different than they should,

get the old mapconv and try it.

Re: mapconv

Posted: 06 Feb 2008, 00:55
by Argh
That's what I figured, since when I read the code, it was just using DXT1, no other arguments, and the default is to use fade. The config I posted last should cure that...

Re: mapconv

Posted: 06 Feb 2008, 01:00
by user
testing it, and remember to put -file in the filename parameter,

maybe the parameter -mitchell should be used too, because its the best scaling filter for mip mapping.

Re: mapconv

Posted: 06 Feb 2008, 01:11
by Argh
Great. Can't wait to try this out!

Re: mapconv

Posted: 06 Feb 2008, 12:19
by user
it is done now, testing a few more times.

Re: mapconv

Posted: 06 Feb 2008, 13:33
by user
i will name it 0.9 because 0.8b sounds like a bugfix relase,

it will now come with nvcompress.exe which can compress dds 12.5
times faster, and -z is working.

Re: mapconv

Posted: 06 Feb 2008, 14:35
by user
ok here is 0.9:

http://www.unknown-files.net/spring/4041/mapconv09/#

made some really cool changes on this one, such as -z

-z gives you full control on texture creation part, this means choose from a variety of texture programs and their options, for information on how to use it look at the readme.

one of the programs is nvcompress, it compresses faster than nvdxt,
and if your GPU suports cuda it will compress about 12x faster, to get cuda go to:

http://www.nvidia.com/object/cuda_home.html

to see if your GPU suports cuda go to:

http://www.nvidia.com/object/cuda_learn_products.html

Re: mapconv

Posted: 06 Feb 2008, 16:24
by user
dxt3 looks better than dxt1, i am testing all the compression types.

and tree placement is much better in this mapconv, i made a small map
with about 4708 trees, the old one would remove most of the trees if
they were placed very close.

you must see how dxt3 compressed maps looks in game, its really good,
it hurts my eyes.

Re: mapconv

Posted: 06 Feb 2008, 17:16
by Argh
Testing now...

Re: mapconv

Posted: 06 Feb 2008, 17:40
by user
im testing every option, sharpenStrong with point filtering looks really bad, i will try dxt1 with high quality

Re: mapconv

Posted: 06 Feb 2008, 17:50
by Argh
Um, I can't seem to get nvdxt working... any ideas what I'm supposed to use here? It always errors out at the end:

MapConv script:

Code: Select all

mapconv.exe  -x 101 -n 100 -m metal.png -a height.png -t terrain.png -f feature.png -c 0.5 -o WorldBuilder_GPL.smf -q 100 -z test.txt
nvdxt script (test.txt):

Code: Select all

nvdxt.exe temp\*.png -quality_production -nmips 4 -dither -dxt3 -fadeamount 0 -outfile temp

Re: mapconv

Posted: 06 Feb 2008, 17:56
by Argh
Oh, duh, I forgot the -file parameter :oops: let's see if this runs...

Re: mapconv

Posted: 06 Feb 2008, 17:56
by user
dxt3 and dxt5 are not working for me, they give some very weird results ,but the good thing is that they dont make the lines.

i will post some screenshots.

and the line must be only one line,

Re: mapconv

Posted: 06 Feb 2008, 17:59
by Argh
Can you show me a working config script, for any of these? That last try didn't work, either...