Hi everyone,
I'm working on my own simple abstract rendering API. I would like to implement that API using Direct3D 10 and OpenGL. Currently, I would like to implement input layout abstraction. I have just one question. Why exactly does Direct3D 10 needs compiled shader when creating ID3D10InputLayout instance? Just to compare input layout with shader input? Is it ok to just pass in a compiled "dummy" shader and accept the warning?
Thx in advance
Direct3D 10 InputLayout
Hi everyone,
I'm working on my own simple abstract rendering API. I would like to implement that API using Direct3D 10 and OpenGL. Currently, I would like to implement input layout abstraction. I have just one question. Why exactly does Direct3D 10 needs compiled shader when creating ID3D10InputLayout instance? Just to compare input layout with shader input? Is it ok to just pass in a compiled "dummy" shader and accept the warning?
Thx in advance
Its possible to pass in a dummy one (at least on my nvidia card on my laptop) but i wouldn't recommend it. The reason Microsoft have implemented this layer is that it, in a sense, precompiles the requirements for binding a vertex buffer to a shader, where as in GL and DirectX9 this was done by the driver *every time* you bind a vertex buffer & shader for rendering. I'd recommend mimicing the InputLayout interface in OpenGL, and using that to store VBO offsets per vertex stream etc.
You can also think of this requirement as a form of a unit test. If the input layout changes somehow, or your shader input signature changes somehow, then you will get an error when they don't match anymore - which will save you lots of debugging time down the road.
Is there some architectural problems that are making you not want to use them together?
Is there some architectural problems that are making you not want to use them together?
Author
Hi,
firstly I would like to thank you for your time and responses. Ok, back to topic.
I wanted my render API to allow me usage like this:
1. Create some kind of a vertex layout object (like in Direct3D 10). No compiled shader would by necessary at creation time.
2. Use created layout object with any compatible shader. In debug build I would check shader input against active layout (at shader bind time) by my self.
Is this design/usage a bad idea? This idead seems viable to me. Problem is, that I dont know too much about what is happing under the hood. So may be it is not that viable at all.
Thx
firstly I would like to thank you for your time and responses. Ok, back to topic.
I wanted my render API to allow me usage like this:
1. Create some kind of a vertex layout object (like in Direct3D 10). No compiled shader would by necessary at creation time.
2. Use created layout object with any compatible shader. In debug build I would check shader input against active layout (at shader bind time) by my self.
Is this design/usage a bad idea? This idead seems viable to me. Problem is, that I dont know too much about what is happing under the hood. So may be it is not that viable at all.
Thx
Hi,
firstly I would like to thank you for your time and responses. Ok, back to topic.
I wanted my render API to allow me usage like this:
1. Create some kind of a vertex layout object (like in Direct3D 10). No compiled shader would by necessary at creation time.
2. Use created layout object with any compatible shader. In debug build I would check shader input against active layout (at shader bind time) by my self.
Is this design/usage a bad idea? This idead seems viable to me. Problem is, that I dont know too much about what is happing under the hood. So may be it is not that viable at all.
Thx
Sounds viable, by the look of what you described the DX10 InputLayout would technically reside with the shader, as multiple shaders would refer to the one vertex decl and the IL is what binds the two.
You can do this, but then how will you validate your shaders against the used input layout? Unless you have another way to ensure that they match, then why not just create one at shader compilation time? Unless you create lots and lots of shader objects this shouldn't really pose a problem.
Author
You can do this, but then how will you validate your shaders against the used input layout? Unless you have another way to ensure that they match, then why not just create one at shader compilation time? Unless you create lots and lots of shader objects this shouldn't really pose a problem.
So you are proposing, that I should mimic Direct3D 10 API? Have a method like
CreateVertexLayout(VertexLayoutDescription, CompiledShader)? I was thinking about checking compatibility of VL and Shader in debug build using shader introspection or something like that.Thx
Honestly, I am not up to speed on how OpenGL works with its shaders, so I can't really advise on that part. My suggestion is to use the validation provided in the InputLayout creation mechanism on D3D10. In my D3D11 framework, the input layout is created the first time that a shader is used (this may not be the best practice, but lazy creation of it makes things much simpler). However, if you use other layout objects that are created for multiple shaders, then you are taking the responsibility to ensure that they work with each shader (as opposed to letting the function do it for you).
In your situation, it might make sense to use the shader inspection methods since you are dealing with two APIs that have slightly different functionality.
In your situation, it might make sense to use the shader inspection methods since you are dealing with two APIs that have slightly different functionality.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement