The problem is, i don´t know how to Release/Recreate.

that will be a couple of calls in your graphics library you're using.

they in turn, being based on directx, will call: d3d_device_ptr->CreateTexture (or whatever its called) to create,

and

mytexture-> Release() to destroy.

I have to Dispose and Remake a Texture all the time

no need to allocate and deallocate your buffer each time you want to receive a new frame. once at program (or transmission) start and end is fine .

i personally am unfamiliar with the various create texture from memory functions in directx. you may be able to fill (update) an existing texture from memory (a jpg buffer), or you may have to create and release. the thing is your data is jpg, not directx texture format. so you need to translate somehow. apparently load and create texture support jpg translation. as i said i use bmp myself.

but if create from memory will only make new textures, not update existing ones, then you'll have to release and create with each new frame. or create once, then convert jpg -> texture yourself, then lock and copy. and release once at end.

me personally, i'd do the screen capture to a texture format (a direct draw surface) in ram, then transmit that, read it into an identical data structure on the receiving end, and then just draw it as the texture on a sprite. OTOH, jpg probably gets you really nice compression for transmitting....

But lock and unlock, i just don´t get how to do it:S

once again, those will be api calls in your graphics library.

here's a directx code snippet for lock and unlock on a buffer (not a texture) as an example. textures work the same way:

 

void copy_ground_mesh()
{
LPVOID p;
mesh[FIRSTGROUNDBUF].mesh2.vb->Lock(0,0,&p,D3DLOCK_READONLY);                            
memcpy(gvert2,p,sizeof(vertex0)*GVERT2SIZE);
mesh[FIRSTGROUNDBUF].mesh2.vb->Unlock();                                   
}
 

I am guessing, something like this makes a Texture with Null data.

using (tex = new Texture(device, -2, -2, 1, SharpDX.Direct3D9.Usage.None, SharpDX.Direct3D9.Format.X8R8G8B8, SharpDX.Direct3D9.Pool.Default))

we're both guessing there. time to start reading the docs and find out how sharp dx works. if you discover that you can mix and match sharpdx and real directx code, then i could look it up and probably have an answer in about 5-10 minutes using directx code. or surely someone else here uses sharp dx.

The alteration occurs either on Displaying the Texture, or Converting to a Texture.

well that's a good thing actually. that will be easier to fix than a jpg->texture translation error, once you learn how sharp dx works.

sounds like its time for some quality time with the sharp dx docs.

although it takes a long time (days or weeks), i always read all the relevant parts of the directx docs before i start coding any directx code.

sounds like you haven't done this yet with sharp dx.

The texture object has a LockRectangle method in SharpDX as well as in native D3D. This method takes as a parameter the mip level that you want to lock; the largest mip level (which you typically want to lock) is 0.

Use the overloads that have an "out DataStream" parameter to avoid using IntPtr.

If you would use IntPtr, you could copy data from your own array to the memory address pointed to by the pointer (that is what a pointer does) by using Marshal.Copy (from .net class library) or CopyMemory (import from Win32 API).

If you use the DataStream, you can write your data directly therein from your source array.

In D3D, resource locking has several caveats, especially regarding the pool and usage with which the resource was created. Be sure to read the official D3D docs from MSDN to understand the possible issues. In particular, you cannot lock default pool resources in D3D9 (unless they are marked dynamic), because they are often physically allocated in GPU memory only and D3D runtime simply doesn't have access to the data directly.

You get a datastream when you call LockRectangle; you don't have to initialize one yourself.

DataStream textureData = null;

tex.LockRectangle (0, LockFlags.None, out textureData); // lock the first mip level, no flags

// textureData is now initialized and ready to take in your data

tex.UnlockRectangle(0); // be sure to call this once you're done with the stream

In case you ever do need to allocate a native blob of memory in .net, you can use Marshal.AllocHGlobal method to get a memory chunk (and an IntPtr so you can manipulate it).

Many interoperation scenarios in .net involve using IntPtr, so it is useful to become familiar with it.

Texture data would be a byte array that contains the raw texel data. It doesn't understand JPEG at all.

Before this, you need to decode the jpeg file to a pixel array. This is how you'd do it in .net. It is admittely unfortunate that D3D does not supply this as a convenience (with, for example, something like UpdateSurfaceFromMemory), since it supports similar operations with texture creation.

Once you get the BitmapSource from the decoder, you could use its CopyPixels method with the destination IntPtr (that you get from locking the texture surface) to neatly move your stuff to D3D texture.

The byte layout of the data is expected to match the actual format of the resource (which you specify during its creation). For example, R8G8B8A8 format expects one pixel to be 4 bytes, with red, green, blue and alpha adjacent to each other. Pixels are adjacent to each other in memory. Scanlines (horizontal lines pixels) may or may not be adjacent to each other in memory; the "pitch" you get from locking the rectangle is the distance between adjacent scanlines, and is always AT LEAST width * numBytesPerPixel (but don't rely on any specific pitch).