Quote:Original post by manxankur
I don't want to have to unlock and lock the vertex buffer to edit the texture coordinates every time as this requires a lot of the computers resources (so i've heard). Is there another way of doing this?

The standard way to do this is to use a dynamic vertex buffer. If your vertex information is changing from frame to frame, you don't really have much choice.

Your other option might be to use the texture transform matrix. You could apply the correct scale/shift to the texture matrix and then render your geometry and then set it back to identity; this way I think you could get it done without a VB lock. But, you have to break the batch to do this.

Also FYI I'm not sure that your statement 'This way i can tell the SetTexture function which triangles to draw' makes any sense. D3D is a state machine, and SetTexture just sets the current texture. All drawing is done through DrawPrimitive-type calls.

Quote:Original post by manxankur
There is one way i think could work but it requires a dynamic way of making vertex buffers so that i can make the right amount of polygons e.g. 2 triangles per sprite. This way i can tell the SetTexture function which triangles to draw.

Your vertex buffer can be extra large. The arguments to Draw[Indexed]Primitive tell D3D which parts of the buffer to use. Generally you have a dynamic buffer large enough to hold a few hundred to a few thousand quads/sprites at once. You can draw before it's full, then append more data on to the end, and draw some more. It's best not to draw each sprite by itself, as Draw calls are expensive.

When using a dynamic VB, it must be in pool default (so you'll have to handle a lost device. Kill the VB and remake it when you are able to Reset).

When locking a dynamic VB you'll use one of two flags.
DISCARD = I'd like to start writing to the start of the buffer.

NOOVERWRITE = I'd like to add stuff to the end of the buffer. I promise I won't touch anything written previously (ie: I promise I won't screw up a buffered draw call).


DISCARD *usually* also lets you have control of the VB immediately. If a VB is being used by a buffered draw call, D3D will actually switch things around internally to give you a new VB. If you're drawing many batches per frame, you want to have a larger VB and use NOOVERWRITE sometimes. There are a limited number of "rename buffers" D3D can give you, so you don't always want to lock with DISCARD unless you've only got 1 or 2 draw calls per frame. If you run out of buffers, D3D will wait until one is free.

NOOVERWRITE lets you have control of the VB immediately, as you've promised to not interfere with anything.

With static vertex buffers you always wait until the GPU has finished processing before you get the buffer.