转自：http://blog.csdn.net/tspatial_thunder/article/details/5937701
现在的游戏图形部分越来多依靠GPU来渲染绘制。说起GPU先说着色器，着色器是可编程的（programmable），分为顶点着色器和像素着色器。着色器是一段运行在GPU上的程序，可以取代渲染的固定功能流水线中部分功能。
在D3d9c可以用固定管道渲染（fixed function pipeline），也可以用可编程的管道（programmable function pipeline）渲染。下面大概总结下各自对vertex的绘制流程。看看传统的渲染流程和利用可编程管道渲染的异同。
固定管道。
先来看看初始化部分，前5步基本是在构造函数中调用实现。
1: 定义其结构Struct{}
struct Vertex
{
D3DXVECTOR3 pos;
D3DXVECTOR3 normal0;
D3DXVECTOR3 normal1;
D3DXVECTOR3 normal2;
};
2:创建一组D3DVERTEXELEMENT9[]结构的元素描述上面定义的Struct{}，以在后面告诉Directx我们定义什么样的顶点结构。
· D3DVERTEXELEMENT9 decl[] =
· {
· {0, 0, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0},
· {0, 12, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_NORMAL, 0},
· {0, 24, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_NORMAL, 1},
· {0, 36, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_NORMAL, 2},
· D3DDECL_END()
· };
3：创建了D3DVERTEXELEMENT9[]后，我们便可以获取指向类IDirect3DVertexDeclaration9的指针，其是用表示D3DVERTEXELEMENT9[]结构的。我们通过下面的函数来依据D3DVERTEXELEMENT9[]获取指向IDirect3DVertexDeclaration9的指针。
HRESULT IDirect3DDevice9::CreateVertexDeclaration(
CONST D3DVERTEXELEMENT9* pVertexElements,
IDirect3DVertexDeclaration9** ppDecl)
例如：
IDirect3DVertexDeclaration9* d3dVertexDecl = 0;
HR(gd3dDevice->CreateVertexDeclaration(decl, &d3dVertexDecl));
4: 接下来我们为我们的顶点和顶点序号申请一片内存的存储空间
HRESULT IDirect3DDevice9::CreateVertexBuffer(
UINT Length,
DWORD Usage,
DWORD FVF,
D3DPOOL Pool
IDirectSDVertexBuffer9** ppVertexBuffer,
HANDLE* pSharedHandle);
HRESULT IDirect3DDevice9::CreateIndexBuffer(
UINT Length,
DWORD Usage,
D3DFORMAT Format,
D3DPOOL Pool,
IDirectSDIndexBuffer9** ppIndexBuffer,
HANDLE* pSharedHandle);
例如
IDirect3DVertexBuffer9* vb;
gd3dDevice->CreateVertexBuffer( 8 * sizeof( Vertex ), 0, 0,D3DPOOL_MANAGED, &vb, 0);
IDirect3DIndexBuffer9* ib;
gd3dDevice->CreateIndexBuffer(36 * sizeof( WORD ),D3DUSASE_DYNAMIC | D3DUSASE_WRITEONLY, D3DFMTJNDEX16,D3DPOOL_DEFAULT, &ib, 0);
5 再接着获取指向上面申请到的内存（空白）的指针，以对这片区域进行写操作。
HRESULT IDirect3DVertexBuffer9::Lock(
UINT OffsetToLock,
UINT SizeToLock,
BYTE** ppbData,
DWORD Flags);
HRESULT IDirect3DIndexBuffer9::Lock(
UINT OffsetToLock,
UINT SizeToLock,
BYTE** ppbData,
DWORD Flags);
例如
Vertex* vertices;
vb->Lock(0, 0, (void**)&vertices, 0); // lock the entire buffer
Vertices[0] = Vertex(-1.0f, 0.0f, 2.0f); // write vertices to
vertices[1] = Vertex( 0.0f, 1.0f, 2.0f); // the buffer
vertices[2] = Vertex( 1.0f, 0.0f, 2.0f);
vb->Unlock(); // unlock when you're done accessing the buffer
下面几步则是每次系统更新的时候在DrawScene()中调用。
6：想调用Direct3D的draw** 函数了？快了。在之前，我们必须告诉Direct3D去哪里找到顶点和顶点序号的数据。Direct3D的draw**系列函数是从顶点流(vertex stream:a vertex stream is essentially an array of vertex component data)中获得顶点数据的。记得我们将我们的顶点放在vertex buffer中，所以为了能够画vertex buffer中的顶点，我们必须将vertex buffer与某个vertex stream联系上。这可以通过下面的函数实现
HRESULT IDirect3DDevice9::SetStreamSource(
UINT StreamNumber,
IDirect3DVertexBuffer9* pStreamData,
UINT OffsetInBytes,
UINT Stride);
例如，假定vb是指向一个已经用顶点类型为Vertex填好了的vertex buffer。那么我们可以通过下面的句子来将这段buffer与vertex stream联系起来。
HR(gd3dDevice->SetStreamSource(0, vb, 0, sizeof(Vertex)));
7：记得上面我们已经创建了顶点的Declaration来描述我们定义的顶点结构。那么最后我们要做的就是在Direct3D中给这个顶点类型注册，正式告之Direct3D如何处理我们定义的顶点。例如
// d3dVertexDecl is a pointer to an IDirect3DVertexDeclaration9 type
gd3dDevice->SetVertexDeclaration(d3dVertexDecl);
8：最后我们就可以调用Draw**（）来绘制各种图案了。当遇到不同物体需要不同的顶点时，我们就需要重新调用SetStreamSource（）,SetVertexDelaration()等来告之Direct3D。例如下面：
|
// Object 1 uses its own vertex buffer, index buffer, and // vertex declaration gd3dDevice->SetStreamSource( 0, vbl, 0, sizeof( Vertexl ) ); gd3dDevice->SetIndices(indexBufferl); gdSdDevice->SetVertexDeclaration(decl1); // DRAW OBJECT 1 HERE gd3dDevice->DrawPrimitive(...); // Object 2 uses its own vertex buffer, index buffer, and // vertex declaration gd3dDevice->SetStreamSource( 0, vb2, 0, sizeof( Vertex2 ) ); gd3dDevice->SetIndices(indexBuffer2); gdSdDevice->SetVertexDeclaration(decl2); // DRAW OBJECT 2 HERE gd3dDevice->DrawPrimitive(...); // Object 3 uses its own vertex buffer, index buffer, and // vertex declaration gd3dDevice->SetStreamSource( 0, vb3, 0, sizeof( Vertex3 ) ); gd3dDevice->SetIndices(indexBuffer3); gdSdDevice->SetVertexDeclaration(decl3); // DRAW OBJECT 3 HERE gd3dDevice->DrawPrimitive(...); |
可编程管道
可编程管道绘制顶点和固定管道大体上差不多。只是有三个主要的不同点。1:编写HLSL程序，自己制定渲染形式 。2：初始化时创建个与GPU交互的C++接口ID3DXEffect *，并利用ID3DXEffect来对GPU做一些初始化工作，比如建立HLSL中变量与C++中相关变量的联系。3：在最后一步drawScene()中通过ID3DXEffect *来调用HLSL的mhTech来进行绘制。
1：HLSL语言？好大的题目，具体自己找本教科书了。
下面直接贴代码举个简单例子。
uniform extern float4x4 gWVP;
struct OutputVS
{
float4 posH : POSITION0;
};
OutputVS TransformVS(float3 posL : POSITION0)
{
// Zero out our output.
OutputVS outVS = (OutputVS)0;
// Transform to homogeneous clip space.
outVS.posH = mul(float4(posL, 1.0f), gWVP);
// Done--return the output.
return outVS;
}
float4 TransformPS() : COLOR
{
return float4(0.0f, 0.0f, 0.0f, 1.0f);
}
technique TransformTech
{
pass P0
{
// Specify the vertex and pixel shader associated with this pass.
vertexShader = compile vs_3_0 TransformVS();
pixelShader = compile ps_3_0 TransformPS();
// Specify the render/device states associated with this pass.
FillMode = Wireframe;
}
}
2：在d3d9的实例的构造函数或者初始化的时候，我们可以通过调用D3DXCreateEffectFromFile()函数来创建与GPU联系的C++指针D3DXEffect *。
HRESULT D3DXCreateEffectFromFile(
LPDIRECT3DDEVICE9 pDevice,
LPCSTR pSrcFile,
CONST D3DXMACRO* pDefines,
LPD3DXINCLUDE pInclude,
DWORD Flags,
LPD3DXEFFECTPOOL pPool,
LPD3DXEFFECT* ppEffect,
LPD3DXBUFFER *ppCompilationErrors
);
我们可以通过ID3DXEffect *来建立C++程序与HLSL程序的沟通桥梁。而这个桥梁就是变量D3DXHANDLE。
然后我们可以通过以下函数为HLSL每个变量在C++层面创建一个对应变量，即对一个HLSL变量X'建立一个在C++中的对应变量D3DXHANDLE mX。
D3DXHANDLE ID3DXEffect::GetParameterByName( D3DXHANDLE hParent, // scope of variable - parent structure LPCSTR pName // name of variable );
然后通过下面这个系列的函数来设置HLSL中对应hParameter的变量的值。
HRESULT ID3DXEffect::SetValue( D3DXHANDLE hParameter, LPCVOID pData, UINT Bytes);
……
……
HLSL与C++沟通的实例：
// 1 ：C++中的变量
D3DXMATRIX M;
D3DXMatrixRotationY(&M, D3DX_PI);
D3DXVECTOR4 V(x, y, z, 1.0f);
// 2：为HLSL中的变量创建对应的在C++中的D3DXHANDLE系列的变量。 (assume "mtx" and "vec" are parameters // declared in .fx file).
D3DXHANDLE hMatrix = mFX->GetParameterByName(0, "mtx");
D3DXHANDLE hVec = mFX->GetParameterByName(0, "vec");
// 3：通过D3DXHANDLE把C++中的值赋给HLSL。
mFX->SetMatrix(hMatrix, &M);
mFX->SetVector(hVec, &V);
全部流程调用实例
void MeshDemo::buildFX()
{
// Create the FX from a .fx file.
ID3DXBuffer* errors = 0;
HR(D3DXCreateEffectFromFile(gd3dDevice, "transform.fx",
0, 0, D3DXSHADER_DEBUG, 0, &mFX, &errors));
if( errors )
MessageBox(0, (char*)errors->GetBufferPointer(), 0, 0);
// Obtain handles.
mhTech = mFX->GetTechniqueByName("TransformTech");
mhWVP = mFX->GetParameterByName(0, "gWVP");
}
最后来比较下在固定管道下与在GPU可编程下的drawScene()函数（自己定义的在周期内调用所有绘图渲染相关函数）的不一样之处。注意黑色加粗的代码。
首先是固定的渲染管道：
void CubeDemo::drawScene()
{
// Clear the backbuffer and depth buffer.
HR(gd3dDevice->Clear(0, 0, D3DCLEAR_TARGET | D3DCLEAR_ZBUFFER, 0xffffffff, 1.0f, 0));
HR(gd3dDevice->BeginScene());
// Let Direct3D know the vertex buffer, index buffer and vertex
// declaration we are using.
HR(gd3dDevice->SetStreamSource(0, mVB, 0, sizeof(VertexPos)));
HR(gd3dDevice->SetIndices(mIB));
HR(gd3dDevice->SetVertexDeclaration(VertexPos::Decl));
// World matrix is identity.
D3DXMATRIX W;
D3DXMatrixIdentity(&W);
HR(gd3dDevice->SetTransform(D3DTS_WORLD, &W));
HR(gd3dDevice->SetTransform(D3DTS_VIEW, &mView));
HR(gd3dDevice->SetTransform(D3DTS_PROJECTION, &mProj));
HR(gd3dDevice->SetRenderState(D3DRS_FILLMODE, D3DFILL_SOLID));
HR(gd3dDevice->DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 8, 0, 12));
mGfxStats->display();
HR(gd3dDevice->EndScene());
// Present the backbuffer.
HR(gd3dDevice->Present(0, 0, 0, 0));
}
再来是可编程管道
void MeshDemo::drawScene()
{
// Clear the backbuffer and depth buffer.
HR(gd3dDevice->Clear(0, 0, D3DCLEAR_TARGET | D3DCLEAR_ZBUFFER, 0xffffffff, 1.0f, 0));
HR(gd3dDevice->BeginScene());
// Let Direct3D know the vertex buffer, index buffer and vertex
// declaration we are using.
HR(gd3dDevice->SetStreamSource(0, mVB, 0, sizeof(VertexPos)));
HR(gd3dDevice->SetIndices(mIB));
HR(gd3dDevice->SetVertexDeclaration(VertexPos::Decl));
// Setup the rendering FX
HR(mFX->SetTechnique(mhTech));
// Begin passes.
UINT numPasses = 0;
HR(mFX->Begin(&numPasses, 0));
for(UINT i = 0; i < numPasses; ++i)
{
HR(mFX->BeginPass(i));
HR(mFX->SetMatrix(mhWVP, &(mView*mProj)));
HR(mFX->CommitChanges());
HR(gd3dDevice->DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, mNumGridVertices, 0, mNumGridTriangles));
HR(mFX->EndPass());
}
HR(mFX->End());
mGfxStats->display();
HR(gd3dDevice->EndScene());
// Present the backbuffer.
HR(gd3dDevice->Present(0, 0, 0, 0));
}