It is currently Sat, 20-10-18, 21:44 GMT

All times are UTC




Post new topic Reply to topic  [ 109 posts ]  Go to page Previous  1, 2, 3, 4, 5, 6, 7, 8  Next
Author Message
 Post subject:
PostPosted: Sat, 29-09-07, 22:11 GMT 
Offline
Site Admin
User avatar

Joined: Fri, 31-08-07, 7:01 GMT
Posts: 4581
Location: Hamburg, Germany
Hey Andy74,

I am VERY pleased to see this and notably that more and more people dare to attempt doing these textures by themselves!

Presumably this weekend I shall release updates of the F-TexTools and the nmtools that are about a factor of TWO faster (largely due to the active profiling help of Christophe Teyssier and dirkpitt!)

Cheers,
Fridger

_________________
Image


Top
 Profile  
 
 Post subject:
PostPosted: Tue, 02-10-07, 7:51 GMT 
t00fri wrote:
Presumably this weekend I shall release updates of the F-TexTools and the nmtools that are about a factor of TWO faster (largely due to the active profiling help of Christophe Teyssier and dirkpitt!)

Too late for me unfortunately, as I have just finished my set of BMNG textures. Earth is looking absolutely amazing now!
Thank you very much for these great tools!

steffens

PS: I didn't take note of the times, sorry


Top
  
 
 Post subject:
PostPosted: Tue, 02-10-07, 8:22 GMT 
Offline
Site Admin
User avatar

Joined: Fri, 31-08-07, 7:01 GMT
Posts: 4581
Location: Hamburg, Germany
steffens,

many thanks for your encouraging report!

The new nmtools-1.5pre1 have been released meanwhile, the new F-TexTools-1.0pre3 will be probably tonight. But there was NO changes in the algorithm and the identity of the images with those from the earlier versions has been checked with nvimgdiff...

Bye Fridger

_________________
Image


Top
 Profile  
 
 Post subject:
PostPosted: Thu, 27-12-07, 10:16 GMT 
Offline
User avatar

Joined: Sun, 23-12-07, 10:13 GMT
Posts: 3
After practicing with the F-Tex/NM Tools for a few days, I finally downloaded the base texture, elevation data, and the watermask. The tools worked without a hitch. I generated and mounted the specular map on the base texture. I used 5 levels for the virtual textures (2048x2048) and converted everything to dds/dxt5nm format. Unfortunately, I didn't benchmark this time around. However, I promise to make another run and document the process. I'll also provide better screen shots. It's late now and I need sleep. :)
Image Image
Thank you t00fri for these excellent tools.

------

Here are some benchmarks:
Code:
Generate level4 tiles (512x2048 to 2048x2048):
  > gzip -dc < ..\world.200409.3x86400x43200.bin.gz | tx2pow2 3 86400 | txtiles 3 65536 4 0
Time: 898.06s ~ 15m

Generate level4 tiles (512x2048 to 2048x2048) 2 separate steps:
  > gzip -dc < ..\world.200409.3x86400x43200.bin.gz | tx2pow2 3 86400 | txtiles 3 65536 4 0 > world.200409.3x65536x32768.bin
Time: 670.67s ~ 11m

  > txtiles 3 65536 4 0 < world.200409.3x65536x32768.bin
Time: 473.98s ~  8m


Generate specmap:
  > gzip -dc < srtm_ramp2.world.86400x43200.bin.gz | specmap 86400 swbd-watermask.1x86400x43200.bin 0.12 1 | tx2pow2 1 86400 > specmap.bin
Time: 423.69s ~  7m

Mount specmap on base texture:
  > gzip -dc < world.200409.3x86400x43200.bin.gz | tx2pow2 3 86400 | tx2rgba 65536 specmap.bin > world.200409.4x65536x32768.bin
Time: 729.75s ~ 12m

Generate level4 tiles:
  > gzip -dc < ..\world.200409.4x65536x32768.bin.gz | txtiles 4 65536 4 0
Time: 515.31s ~ 8.5m

Compress 512 tiles to dxt3 format:
  $ time for x in *.png; do nvcompress -bc2 -nomips $x; done
Time: 62m

Generate level4 normalmap tiles:
  > gzip -dc < ..\srtm_ramp2.world.86400x43200.bin.gz | nm2pow2 86400 1 | nmtiles 4 6378.14 65536 2.5
Time: 705.47s ~ 12m

Compress 512 tiles to dxt5nm format:
  $ time for x in *.png; do nvcompress -bc3n -nomips $x; done
Time: 12m


Image Image

_________________
AMD Athlon 64 3200+ (Orleans) 1GB Ram
GeForce 8400 GS 256 MB (163.71)
Windows XP Pro SP2
Celestia 1.5.0pre5


Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 1:33 GMT 
Offline

Joined: Thu, 21-02-08, 23:26 GMT
Posts: 16
I ran my new scripts now for f-tex and nmtools but I get the following picture:
Around the Equator there seems to be missing a lot. Furthermore there´s no specular reflection at all. The normalmap and the night texture match with the RGBA on the northern hemisphere but not on the southern.
I used the recommended SWBD watermap from cartrite.
Maybe there´s an error in my code ?
Code:
gzip -dc swbd-watermask.1x86400x43200.bin.gz

gzip -dc < srtm_ramp2.world.86400x43200.bin.gz | specmap 86400 swbd-watermask.1x86400x43200.bin 0.12 1 > specmap.bin

gzip -dc < world.200406.3x86400x43200.bin.gz | tx2rgba 86400 specmap.bin | tx2pow2 4 86400 > level5\world.200406.4x65536x32768.RGBA.bin

Image

I can´t seem to get the point, maybe the watermap is not made for this way of using it like the original bmng watermap ?
Is dxt3 the correct compression for the RGBA tiles ?

Jörg


Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 10:42 GMT 
Offline
User avatar

Joined: Thu, 25-10-07, 15:20 GMT
Posts: 992
Location: NE PA, USA
guest_jo,
What does the command look like that you used to run txtiles. It should look "something" and maybe not exactly like this.
Code:
cd level5
txtiles 4 65536  5  2.7 < world.200406.4x65536x32768.RGBA.bin
. The important part is
Code:
txtiles 4 65536  5
where "4" is the number of layers (channels), 65536 is the width, and 5 is the level your running txtiles on.
cartrite


Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 15:11 GMT 
Offline

Joined: Thu, 21-02-08, 23:26 GMT
Posts: 16
I ´ll give you my code.
Here´s another picture of one of the tiles in dxt3 with gimp :

I suppose the specmap is inverted somehow ?
Image
Code:
tx2half 4 65536 < level5\world.200406.4x65536x32768.RGBA.bin > level4\world.200406.4x32768x16384.RGBA.bin

cd level5
txtiles 4 65536 5 < world.200406.4x65536x32768.RGBA.bin
for %%x in (tx*.png) do nvcompress -nomips -bc2 %%x


I´ll add my ctx- file I´m new with celestia 1.5. Had 1.41 all the time...
Code:
VirtualTexture
{
        ImageDirectory "world200406_64kdds"
        BaseSplit 0
        TileSize 1024
        TileType "dds"
}

the 2.7 was the compression I think but it was crecommended to let it on default for subsequent dxt compression.

The problem shows up regardless of the displayed tile-level so I suppose the mistake happens before the complete texture is cut in half.

I am a bit helpless with this problem... is it possibly a problem with byteswap 1 instead of 0 ? I have Intel PC.

Jörg

Edit: I just looked through the code of Orphu of io some posts before.
He used tx2pow2 already for the specmap with channels =1 and then for the base texture with channels=3. But does this make sense? Why not producing specmap in 86400 and then use tx2pow2 after mounting it with channels=4 ?


Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 16:42 GMT 
Offline
User avatar

Joined: Thu, 25-10-07, 15:20 GMT
Posts: 992
Location: NE PA, USA
The specular part of the image you posted looks correct. The problem is with the boundaries. That tile should extend farther north. What does your png file look like. Here is what my tx_1_1.png looks like from level1.

Image

cartrite


Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 17:06 GMT 
Offline
User avatar

Joined: Thu, 25-10-07, 15:20 GMT
Posts: 992
Location: NE PA, USA
This code should produce all levels on a windows machine in png format. Then you can compress to the dxt3 format.
Code:
gzip -dc < world.200406.3x86400x43200.bin.gz | tx2rgba 86400 specmap.bin | tx2pow2 4 86400 > level5\world.200406.4x65536x32768.RGBA.bin
tx2half 4 65536 < level5\world.200406.4x65536x32768.RGBA.bin > level4\world.200406.4x32768x16384.RGBA.bin
cd level5
txtiles 4 65536  5  2.7 < world.200406.4x65536x32768.RGBA.bin
tx2half 4 32768 < level4\world.200406.4x32768x16384.RGBA.bin > level3\world.200406.4x16384x8192.RGBA.bin
cd ..\level4
txtiles 4 32768  4 2.7 < world.200406.4x32768x16384.RGBA.bin
tx2half 4 16384 < level3\world.200406.4x16384x8192.RGBA.bin > level2\world.200406.4x8192x4096.RGBA.bin
cd ..\level3
txtiles 4 16384  3 2.7 < world.200406.4x16384x8192.RGBA.bin 
tx2half 4   8192 < level2\world.200406.4x8192x4096.RGBA.bin > level1\world.200406.4x4096x2048.RGBA.bin
cd ..\level2
txtiles 4   8192  2 2.7 < world.200406.4x8192x4096.RGBA.bin 
tx2half 4   4096 < level1\world.200406.4x4096x2048.RGBA.bin > level0\world.200406.4x2048x1024.RGBA.bin
cd ..\level1
txtiles 4   4096  1 2.7 < world.200406.4x4096x2048.RGBA.bin
cd ..\level0
txtiles 4   2048  0 2.7 < world.200406.4x2048x1024.RGBA.bin

EDIT Make sure that you are not using Basic or Multitexture render path. Spec maps are not rendered in those paths. You can change by "ctrl v" or menu.
cartrite


Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 17:33 GMT 
Offline
User avatar

Joined: Thu, 25-10-07, 15:20 GMT
Posts: 992
Location: NE PA, USA
You may also need to add specular color and specular power to you ssc file.
Code:
AltSurface "BlueMarble-June" "Sol/Earth"
{
      Color [ 0.85 0.85 1.0 ]
      Texture "bluemarble.ctx"
      NormalMap "normals.*"
      SpecularPower 25.0
      SpecularColor [ 0.8 0.8 0.85 ]
}



Top
 Profile  
 
 Post subject:
PostPosted: Sat, 01-03-08, 23:44 GMT 
Offline

Joined: Thu, 21-02-08, 23:26 GMT
Posts: 16
Hello cartrite,
first I want to thank you for spending your time and for your help.
I am on track now

Image

The problem with specularity is solved. It were the additional lines in your last post that were missing in my altsurface ssc. I thought the lines in solarsys.ssc were enough.

Concerning the scrambled tiles:
I didn´t use your code but exactly mine without the nvcompress ( and delete png ) line.
The resulting png tiles are absolutely correct but thats only the first step.
Why does this
Code:
for %%x in (tx*.png) do nvcompress -nomips -bc2 %%x
destroy the tiles ?
actually they had the correct 1024x1024 size.
And why was this cut in africa always at the same location from level 0 to level5 ?
I don´t understand that but I´ll do another test:
I´ll compress the correct pngs now to find out if its the CUDA compession thats making me trouble here.
So I´ll take level0 without CUDA and level1 with CUDA. We´ll see...

What I found out already is that CUDA will not jump in when I compress to dxt5nm. Maybe that is not supported. It would be sad if dxt3 wouldn´t give usable results with CUDA. The speed is so nice.

Starting from the world and srtm and watermap bin.gz files to the complete dxt3 / dxt5nm sets of textures I needed
world 64k dxt3: 39 min.
srtm dxt5nm: 28 min.

I´ll post the result of my CUDA tests.

Jörg


Top
 Profile  
 
 Post subject:
PostPosted: Sun, 02-03-08, 0:13 GMT 
Offline
User avatar

Joined: Thu, 25-10-07, 15:20 GMT
Posts: 992
Location: NE PA, USA
I never used nvcompress so I'm not sure what could be going on. It does sound like some kind of bug though with nvcompress or Cuda if the png files are alright.
EDIT I just tried to build the latest nvidia texture tools 2.0.1 that was released a couple of days ago and I still can't build it on Linux. It almost finished but got errors while building the ui. Anyhow I was going to give the Windows version a try. guest_jo, if the nvcompress is giving you problems, maybe you should try this newer version. I think there is a newer version of Cuda too.
cartrite


Top
 Profile  
 
 Post subject:
PostPosted: Sun, 02-03-08, 1:38 GMT 
Offline

Joined: Thu, 21-02-08, 23:26 GMT
Posts: 16
It is done.
level0 nocuda
level1 cuda
level2 nocuda
level3 cuda
level4 cuda
level5 cuda 2048 tiles from png to dxt3 in 12 minutes 36 seconds :shock:

On first sight everything looks good so I thought that converting each level seperately seems to work better than in one rush of a batch file... ...but then in level 5 I found problems : the tiles with the ending _31 have some stripes.
This is how it looks:
Image

I´m a bit tired now, maybe tomorrow I´ll start again from png and do the complete set without CUDA ...
Ah and I have to try new version of CUDA too (if available...) mine is from around Feb 20th. I definitely vote for a day to have 26 hours ... :wink:

Have a good night ,

Jörg from Germany at 2.30 am


Top
 Profile  
 
 Post subject:
PostPosted: Sun, 02-03-08, 7:49 GMT 
Offline
User avatar

Joined: Thu, 25-10-07, 15:20 GMT
Posts: 992
Location: NE PA, USA
guest_jo,
I got the nvidia texture tools working on Windows and Linux. I think I saw something that may be causing your problems. I noticed strange effects when I did the normalmap in dxt5nm and the ext was *.dds. The file names had to be *.dxt5nm and the line "TileType "dxt5nm" had to be in the ctx file for the normalmap. If you didn't do this already, try it.
Quote:
Ah and I have to try new version of CUDA too (if available...) mine is from around Feb 20th.

EDIT About Cuda, The version I used was 1.1 and I'm not sure when it was updated but Feb 20 is pretty recent. The texture tools were just updated to 2.0.1 a few days ago on Feb 28.
cartrite


Top
 Profile  
 
 Post subject:
PostPosted: Sun, 02-03-08, 18:06 GMT 
Offline

Joined: Thu, 21-02-08, 23:26 GMT
Posts: 16
Hi cartrite,
my ssc was already correct with dxt5nm.
Anyways I made two ssc´s: One with normalmap and one without. So I can switch instantly to see the difference.
The lines in file tx.._31 remain in any case. I´m just running the script again without CUDA but I suppose it will finish quite late today ( tomorrow ? ).

Meanwhile I checked the version history of Nvidia Texture Tools and CUDA but for windows users without any experience in compiling there are no new versions available.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 109 posts ]  Go to page Previous  1, 2, 3, 4, 5, 6, 7, 8  Next

All times are UTC


Who is online

Users browsing this forum: No registered users and 2 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group