There are a couple of things you want to have a look at, the first is the ID3DXSprite Interface, this object is designed for just the purpose you're talking about, however the documentation for it is scant to say the least. The interface renders the texture straight out to screen space and is compatible with alpha blending and such like. It also has methods for rotating and scaling the textures.

The other way (the harder way) is a litle laborious. Firstly, why bother lighting your textures? Just deactivate the lighting renderstate (D3DRS_LIGHTING) before you render them. You may also want to deactivate fog (D3DRS_FOG) and the Z Buffer (D3DRS_ZENABLE, D3DRS_ZWRITEENABLE). Next you need to set up an orthographic projection matrix (a projection matrix that does not take into account perspective). For example:

D3DXMATRIX ortho;
D3DXMatrixOrthoLH(&ortho,100.0f,100.0f,0.0f,1.0f);

Generates an orthographic matrix with the origin on the bottom left (I think, this whole left hand/right hand thing confuses me ) You can now take a single vertex buffer holding a square with the following corner coordinates (rendered as a triangle strip):

0,0,0
1,0,0
0,1,0
1,1,0

And use that to render all of your textures, just translate and scale the VB to the screen. The following optimised function will generate the kind of world matrix you want to do this:




Note the 'in' matrix must either be an identity matrix or a matrix generated by the same function from an identity matrix.

Of course you may be asking 'Why Bother?' and well you might, the advantage of this approach is that the orthographic matrix's coordinate system will always be the same regardless of the screen resolution, therefore all your positioning and scaling can be kept the same if you resize the screen.

[edited by - bligblug on October 3, 2002 10:52:56 AM]

Andrew,

The function D3DXMatrixOrthoLH() should be all you need to build that ortho matrix. You could do it by hand, but why bother? I assume you're using DirectX 8+, although your mention of IM and RM (immediate/retained mode?) makes me uncertain...

As for your second question, bligbug's approach would give you a view volume of 100 by 100 (these are specified in the D3DXMatrixOrthoLH() call). This means you can make your textured rectangles have sizes anywhere from 0x0 (invisible, of course) to 100x100 (filling the entire screen).

So, he's created a single 1x1 rectangle, which of course isn't too useful by itself since it's so tiny. However, he then uses a matrix to scale the rectangle to whatever dimensions it needs to be at render time. His scale factors would be from 1 to 100.

--Hoozit.

[edited by - HoozitWhatzit on October 3, 2002 1:38:30 PM]

I went through the whole four months of building a whole GUI. Allow me to toss a few free tips: The Ortho approach is a definite plus. I''ve done it with dynamic vertexes (one quad of vertexes per rectangle on screen) that I''d only lock and alter when the rectangles change coords, and got very good performance, but alas I can''t rotate, scale, twist or mangle my vertexes as freely as matrixes allow.

Go for Ortho!

Also, it''s great to have all the coordinates ''virtual'', winding up on screen at the same place regardless of resolution. (I used 10000x10000 virtual coords m''self) You might want to make it go from -1.0f to 1.0f and you''ll get free compatibility with the DirectX SDK samples D3DFont code, they use -1 to 1 for their scaled font.
One problem I ran into with virtual coords was when I tried to get pixel-precise. I realized that my using rectangles with ''Left, Top, Right, Bottom'' data caused trouble when scaling things. Whereas using rectangles with ''Left, Top, Width, Height'' would have saved me so many hassles! That''s my tip: Use Height-Width when you get around to saving rectangle sizes.

Why not just use DX8?