New Release D3D9Client Development

[fred18]

Hello guys,

so, following to this conversation https://www.orbiter-forum.com/showthread.php?p=591865&postcount=40

I am here asking if possible to implement the following function in a test version of D3D9.

given a set of values which is:
- center coordinates
- size
- altitude
- shape factor

a rectangular terrain area will be rendered simply flat with possibly its edges smoothly linked to the surrounding and this rectangle will have the following characteristics:
- center at center coordinates
- long side = 2*size
- short side = shape factor * long side
- altitude = constant altitude as per input value

This flat area should be rendered overriding the terrain elevation tiles, simply when it's time to be rendered it will be flat instead of following terrain tiles.

If the set of values is removed then the regular terrain shall be rendered

I gave a look at the code of the tile manager of D3D9 and I am available to participate actively also to the coding part, but I need at least help to get a grasp of what is going on in those functions before I could touch anything there.

Will someone be so kind to make this test or to help me through it?

thank you very much in advance for any attention

Fred

 

[jarmonik]

EDIT: So after some more tinkering, it seems that this issue only occurs when using a Generic MFD size of 3 or less (which I do on a regular basis), otherwise the MFDs appear just as they should. However, this still isn't the case in Orbiter 2010 where MFDs appear smooth regardless of the selected size.

I can reproduce the problem and I think I know what is causing it. Haven't had time to fix it.

---------- Post added at 07:48 ---------- Previous post was at 07:42 ----------

so, following to this conversation https://www.orbiter-forum.com/showthread.php?p=591865&postcount=40

I am here asking if possible to implement the following function in a test version of D3D9.

This flat area should be rendered overriding the terrain elevation tiles, simply when it's time to be rendered it will be flat instead of following terrain tiles.

If the set of values is removed then the regular terrain shall be rendered

I gave a look at the code of the tile manager of D3D9 and I am available to participate actively also to the coding part, but I need at least help to get a grasp of what is going on in those functions before I could touch anything there.

If you want to flatten a terrain in D3D9 then the easiest way is to do it in a vertex shader especcially if you can settle down to a circle instead of a square as an shape. You could feed a reference point (i.e. a vector [x, y, z, radius]) to a shader and flatten the terrain within the reference radius and smoothed the edges with vertex blending. Even if relatively easy to implement it has one major flaw that the orbiter's surface collision logic wouldn't take the modified terrain in count.

So, it would be more practical to create somekind of patch utility to apply a terrain flattening described in a configuration file to elevation data on a harddrive. Which would work without any modifications to the Orbiter or to the Client.

Back when I was working with the surface micro-textures, I got an idea of micro-elevation which would require some kind of tile-server in order to be implementable. Right now the Client requests a tile elevation file with it's tile-number from a harddrive. If a tile-server would be in place then the request/number would be passed to the server which would then load the file and preprocess the elevation data using a pixel-shader before forwarding it to a terrain manager for rendering. This would allow us to do more extensive real-time processing with the elevetion data. This idea came to a halt due to fact that the Orbiter's surface collision logic wouldn't know about the modified terrain.

So, technically it would sound more practical approach if the Orbiter would request the surface elevation data from the Client directly instead of loading it from a data file. Client could easily return elevation and normal for a requested lng, lat.

 

[fred18]

If you want to flatten a terrain in D3D9 then the easiest way is to do it in a vertex shader especcially if you can settle down to a circle instead of a square as an shape. You could feed a reference point (i.e. a vector [x, y, z, radius]) to a shader and flatten the terrain within the reference radius and smoothed the edges with vertex blending. Even if relatively easy to implement it has one major flaw that the orbiter's surface collision logic wouldn't take the modified terrain in count.

So, it would be more practical to create somekind of patch utility to apply a terrain flattening described in a configuration file to elevation data on a harddrive. Which would work without any modifications to the Orbiter or to the Client.

Back when I was working with the surface micro-textures, I got an idea of micro-elevation which would require some kind of tile-server in order to be implementable. Right now the Client requests a tile elevation file with it's tile-number from a harddrive. If a tile-server would be in place then the request/number would be passed to the server which would then load the file and preprocess the elevation data using a pixel-shader before forwarding it to a terrain manager for rendering. This would allow us to do more extensive real-time processing with the elevetion data. This idea came to a halt due to fact that the Orbiter's surface collision logic wouldn't know about the modified terrain.

So, technically it would sound more practical approach if the Orbiter would request the surface elevation data from the Client directly instead of loading it from a data file. Client could easily return elevation and normal for a requested lng, lat.

I never done a shader in my life so I don't know where to start from :embarrassed:

anyway you are surely right that having a client which stands in between Orbiter core and the planetary textures would be a very good organization. Actually the point would all be there: put something in the middle between the hard drive files and orbiter to override the terrain definition when (where) needed. I was thinking about creating temporary files that orbiter will then parse and acquire, but I don't know if that is feasible

 

[Donamy]

Or in vain, if future changes in Orbiter break it.

 

[4throck]

put something in the middle between the hard drive files and orbiter to override the terrain definition

Elev_mod files.... they already exist.


Don't know it the graphic client could be used to generate them.
If it could, it would be great. Do it all from within Orbiter...

 

[Face]

anyway you are surely right that having a client which stands in between Orbiter core and the planetary textures would be a very good organization. Actually the point would all be there: put something in the middle between the hard drive files and orbiter to override the terrain definition when (where) needed. I was thinking about creating temporary files that orbiter will then parse and acquire, but I don't know if that is feasible

If you put some area definitions into a text file, you can parse it and use the information at the end of "bool SurfTile::LoadElevationData ()" inside of Surfmgr2.cpp to manipulate the "elev" field in the tile, which is nothing but an 259x259 array of INT16. It is what the renderer will eventually put on screen.

So there you can iterate over all the shape definitions, check if they are relevant for the tile at hand (tile boundaries), then project the shape onto the tile and set the covered pixel to the altitude value you want.

It will not make the collision thing work, but it will demonstrate what the idea could accomplish in terms of developer experience. It will also demonstrate the impact on the frame-rate.

 

[fred18]

Thank you, I will surely try this, I am very curious to see where it does end up. I have an issue anyway: I can't compile the D3D9 client, it seems that I don't have the right redistributables, but if I try to install them I get the error s1023 or something like that which means that I have a newer set. I use VS community 2019. Anyone knows a workaround?

 

[kuddel]

To compile D3D9Client, the redistributables are not enough! You need the DirectX SDK.
See Orbitersdk\D3D9Client\README.txt for further details.


[*] ...some points in that document are not very up to date I must say :blush:, it was last updated May 2014!
Let's see if I can find the SDK Link for you...
This should be it:
https://www.microsoft.com/en-us/download/details.aspx?id=6812

 

[fred18]

To compile D3D9Client, the redistributables are not enough! You need the DirectX SDK.
See Orbitersdk\D3D9Client\README.txt for further details.


[*] ...some points in that document are not very up to date I must say :blush:, it was last updated May 2014!
Let's see if I can find the SDK Link for you...
This should be it:
https://www.microsoft.com/en-us/download/details.aspx?id=6812

thank you very much but that is what I was talking about:

 

[kuddel]

It's a known issue (nowadays):

https://support.microsoft.com/de-de...or-when-you-install-the-directx-sdk-june-2010

https://walbourn.github.io/known-issue-directx-sdk-june-2010-setup-and-the-s1023-error/

---------- Post added at 22:31 ---------- Previous post was at 22:27 ----------

It looks like I have to do a whole new chapter on "how to setup & build D3D9Client"

 

[fred18]

It looks like I have to do a whole new chapter on "how to setup & build D3D9Client"

yes, please... I managed to get the dxsdk but then I had to do this:

https://stackoverflow.com/questions/17015088/missing-files-directx-sdk-d3dx9-lib-d3dx9-h

And now I am fighting against the NVAPI because I get a bunch of lnk2019 errors relevant to that but he is winning...

---------- Post added 13th Apr 2019 at 00:00 ---------- Previous post was 12th Apr 2019 at 23:56 ----------

I did it, adding the nvapi.lib file to the project...

 

[DaveS]

yes, please... I managed to get the dxsdk but then I had to do this:

https://stackoverflow.com/questions/17015088/missing-files-directx-sdk-d3dx9-lib-d3dx9-h

And now I am fighting against the NVAPI because I get a bunch of lnk2019 errors relevant to that but he is winning...

For the NVAPI thing, just create a blank header file and name it nvapi.h. This will bypass that and link everything just fine.

 

[fred18]

anyway, even if I manage to compile the code, when I try to run it I get this error

---------- Post added at 00:21 ---------- Previous post was at 00:19 ----------

from orbiter.log

000000.000: **** Creating simulation session
000000.000: D3D9: [DirectX 9 Initialized]
000000.000: D3D9: 3D-Adapter = NVIDIA GeForce GTX 1080
000000.000: D3D9: MaxTextureWidth........: 16384
000000.000: D3D9: MaxTextureHeight.......: 16384
000000.000: D3D9: MaxTextureRepeat.......: 8192
000000.000: D3D9: VolTexAddressCaps......: 0x3F
000000.000: D3D9: NumSimultaneousRTs.....: 4
000000.000: D3D9: VertexDeclCaps.........: 0x30F
000000.000: D3D9: XNA Math Support.......: Yes
000000.000: D3D9: Vertex Texture.........: Yes
000000.000: D3D9: Shadow Mapping.........: Yes
000000.000: D3D9: D3DFMT_A16B16G16R16F...: Yes
000000.000: D3D9: D3DFMT_A32B32G32R32F...: Yes
000000.000: D3D9: D3DFMT_D32F_LOCKABLE...: Yes
000000.000: D3D9: D3DFMT_A2R10G10B10.....: Yes
000000.000: D3D9: D3DDTCAPS_DEC3N........: No
000000.000: D3D9: D3DDTCAPS_FLOAT16_2....: Yes
000000.000: D3D9: D3DDTCAPS_FLOAT16_4....: Yes
000000.000: D3D9: Available Texture Memory = 4059 MB
000000.000: D3D9: [3DDevice Initialized]
000000.000: D3D9: ERROR: D3D9Effect.cpp Line:252 Error:-2147467259 D3DXCreateEffectFromFileA(pDev, name, macro, 0, D3DXSHADER_NO_PRESHADER, 0, &FX, &errors)
000000.000: D3D9: ERROR: Effect Error: C:\Orbiter2016\Modules\D3D9Client\D3D9Client.fx(96,33): error X3004: undeclared identifier 'KERNEL_SIZE'
C:\Orbiter2016\Modules\D3D9Client\D3D9Client.fx(96,26): error X3074: 'kernel': implicit array missing initial value

 

[jarmonik]

anyway, even if I manage to compile the code, when I try to run it I get this error

KERNEL_SIZE is a macro defined in D3D9Effect.cpp line 324 (Orbiter 2016 branch). Why the effect compiler doesn't recognize it is unknown. Could it be some kind of codepage/charset thing ? You can replace the KERNEL_SIZE with value 27 and see if the compiler hangs into an other macro.

 

[Face]

Thank you, I will surely try this, I am very curious to see where it does end up. I have an issue anyway: I can't compile the D3D9 client, it seems that I don't have the right redistributables, but if I try to install them I get the error s1023 or something like that which means that I have a newer set. I use VS community 2019. Anyone knows a workaround?

No clue. On that note, I only managed to get the 2015 solution to compile in recent versions. All others have either some weird platform problems or right out crash the appropriate VS version.

 

[fred18]

KERNEL_SIZE is a macro defined in D3D9Effect.cpp line 324 (Orbiter 2016 branch). Why the effect compiler doesn't recognize it is unknown. Could it be some kind of codepage/charset thing ? You can replace the KERNEL_SIZE with value 27 and see if the compiler hangs into an other macro.

actually the D3D9Effect.cpp that I have in the package have the following lines as 323,324 and 325:

eTuneEnabled  = FX->GetParameterByName(0,"gTuneEnabled");
// General parameters -------------------------------------------------- 
eSpecularMode = FX->GetParameterByName(0,"gSpecMode");

 

[jarmonik]

actually the D3D9Effect.cpp that I have in the package have the following lines as 323,324 and 325:

eTuneEnabled  = FX->GetParameterByName(0,"gTuneEnabled");
// General parameters -------------------------------------------------- 
eSpecularMode = FX->GetParameterByName(0,"gSpecMode");

This isn't right. You must have an old version. The latest are 1138 and 1137. From where did you got it ?
This is where you should get it: svn://mirror.orbiter-radio.co.uk/D3D9client/trunk

 

[fred18]

This isn't right. You must have an old version. The latest are 1138 and 1137. From where did you got it ?
This is where you should get it: svn://mirror.orbiter-radio.co.uk/D3D9client/trunk

I got it from the website : http://users.kymp.net/~p501474a/D3D9Client/

scrolling down, additional downloads - downloads for orbiter2016 r1137

the following file:

http://users.kymp.net/~p501474a/D3D9Client/D3D9ClientR3.8-forOrbiter2016(r1137).zip

 

[Face]

I got it from the website : http://users.kymp.net/~p501474a/D3D9Client/

scrolling down, additional downloads - downloads for orbiter2016 r1137

the following file:

http://users.kymp.net/~p501474a/D3D9Client/D3D9ClientR3.8-forOrbiter2016(r1137).zip

Any reason why you don't use the SVN repo?

 

[fred18]

Any reason why you don't use the SVN repo?

I'm not very accustomed to it, since there is a website I usually get the files from there. Anyway I never thought there was a needing to use svn in this case. I just opened the svn repo and downloaded the updated files. Now I have to setup the project to make it compile again, but at least i checked and there are the macros needed.