- Lock and fill again the vertex buffer anytime you want to change the transparency. It is better to keep a copy of the buffer content in memory (that means to remember the vertices array from your code also outside of the InitGraphics function), modify the transparency values in this copy, then lock the buffer, copy the memory content to the buffer (memcpy), unlock.
- Use your own vertex shader.
You would create the vertex buffer only once at the begining and just change its content during rendering.
Dynamic vertex buffers created with D3DUSAGE_DYNAMIC | D3DUSAGE_WRITEONLY usage flags are meant to be updated frequently, even every frame. Notice the writeonly flag, that's related to what I said - you should have a local copy of the content and only write to the buffer, never read from it. In this case you can use this flag and everything will be faster.
Be aware that D3DUSAGE_DYNAMIC | D3DUSAGE_WRITEONLY will mean creating in D3DPOOL_DEFAULT, which in turn means getting into the realms of lost device handling.
For the number of vertexes in your buffer you could just stay with D3DPOOL_MANAGED and lock/update/unlock as required. With a managed buffer D3D will create both a system memory and a video memory copy of the buffer, but perform lock/unlock operations on the system memory copy and automatically handle updating the video memory copy for you.
Yes, it will slow down rendering. By much? Maybe, maybe not. For drawing a single quad you're very likely to not even notice it - your main bottleneck is almost definitely elsewhere.
Another option is to use DrawPrimitiveUP instead of DrawPrimitive. This will let you source your vertex data from your own memory instead of from a vertex buffer, so you don't even need to worry about lock/unlock behaviour. Again, it's going to be slower, but again, I highly doubt it would be measurably so in your use case.
Both of these approaches are ones that you may occasionally read dire warnings about - don't use them, they're slow, etc. You may get the impression that means that your performance is going to collapse to something like single-digit framerates. Don't - it won't. For your kind of basic use case (drawing a single quad) they will be more than fast enough - if you can measure maybe 10% performance drop from them then you're close to worst case - it's only when you get to non-trivial stuff that you need to avoid them (it's still a good habit to learn how to use vertex buffers properly though - DrawPrimitiveUP in particular is most definitely not recommended as a general-case solution to the problem of handling vertex buffer updates, but if it's suitable for a particular use-case then why not use it? See further http://tomsdxfaq.blogspot.com/2006_04_01_archive.html#114482755569486635.)
One final note - none of the above is really necessary. What you really need to do is set your TextureStageState to blend with a constant color (D3DTA_CONSTANT), set your constant color (D3DTSS_CONSTANT) and away you go - no need to modify the vertex colors at all. It's a long long long time since I've done this kind of fixed pipeline setup (I'd just write a shader these days) so I'm afraid I can't be more specific, provide code examples, etc.
Direct3D has need of instancing, but we do not. We have plenty of glVertexAttrib calls.
I am now using D3DTA_TFACTOR instead and its working for changing the texture alpha blending during rendering!
The only problem I am having now is that I need to get rid of the black in the texture, I tried using color key in D3DXCreateTextureFromFileEx() but it's not work.
An example of the texture that I want to get rid of the black from it:
You should create the texture with alpha channel (RGBA instead of RGB or XRGB) in your image editor. PNG supports it.
This aplha channel will mean which pixels in the texture are transparent and how much (as opposed to color key which can control only which pixels are fully transparent and which fully opaque). And the transparency in your vertex structure will mean how transparent is the whole quad with this texture.
To be honest this is the first time I've heard about D3DTA_TFACTOR Maybe you could also try to move from the fixed function pipeline to shaders. As I see even the fixed function pipeline can produce quite complicated effects, but you need to know how to set this and that etc. Getting into shaders may be hard, but then you'll realise that you have a huge flexibility in your hands and you'll never want to come back ;) (Just a suggestion.)
Maybe you could also try to move from the fixed function pipeline to shaders. As I see even the fixed function pipeline can produce quite complicated effects, but you need to know how to set this and that etc. Getting into shaders may be hard, but then you'll realise that you have a huge flexibility in your hands and you'll never want to come back ;) (Just a suggestion.)
