XNA 4.0 beta: EffectPass not a simple wrapper!

　　So, the beta is out!

      XNA 4.0进行了大幅更新，大部分改动都不错，不过一直有2点让我非常不爽：1. render target必须和某个depth buffer绑定到一起；2. 没有了SetxxShader和SetShaderConstant等底层函数。前者对deffered shading非常不友好，后者让我写的整个材质系统都不能用了。相比之下，后者更让我担心，因为CTP中的Effect和EffectPass仅仅是DX Effect系统之上的一个简单wrapper众所周知，DirectX中Effect的性能一直是被人诟病的地方。好在Shawn告诉我：

“The Windows EffectPass.Apply implementation in the CTP is not complete, and many optimizations not yet implemented. The CTP implementation is functionally correct, but not from a performance standpoint, so you should not worry too much about Windows performance in the CTP.”

     似乎Beta版已经解决了这个问题，4.0中的EffectPass不再是DX Effect上的一个简单wrapper，xna team做了很大的努力减少状态改变，以至于PIX分析xna 4.0的程序时有可能出错，Shawn在blog中说：

"The problem occurs because PIX does not support some crazy tricks that GS4 uses to optimize state management" 。

      下面是同一段代码在CTP和beta版下的对比：

//test code//
.....set parameter for first object
effectPass.Apply();
graphicsDevice.DrawIndexedPrimitives(PrimitiveType.TriangleList, 0, 0,vertices.Count, 0, primitiveCount);
//only change world matrix and draw the same object again
effect.World = Matrix.CreateTranslation(1, 0, 0);
effectPass.Apply();
graphicsDevice.DrawIndexedPrimitives(PrimitiveType.TriangleList, 0, 0,vertices.Count, 0, primitiveCount);



//Pix data for ctp:
ID3DXEffect::BeginPass(0)                          
ID3DXEffect::EndPass()                         
ID3DXEffect::End()                         
IDirect3DDevice9::BeginScene()                         
IDirect3DDevice9::DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 24, 0, 12)//draw the 1 primitive                  
ID3DXEffect::SetMatrix(0xFF6B4D6B, 0x0039E6F0)                     
ID3DXEffect::SetMatrix(0xFF6B4E13, 0x0039E684)                     
ID3DXEffect::SetMatrix(0xFF6B4DBF, 0x0039E684)                     
ID3DXEffect::Begin(0x0039E7EC, 0x00000001)                 
ID3DXEffect::BeginPass(0)                          
 IDirect3DDevice9::SetVertexShader(0x06487820)                     //redundance!!!!!!
 IDirect3DDevice9::SetVertexShaderConstantF(0, 0x015851C8, 1)      //redundance!!!!!!!!      
 IDirect3DDevice9::SetVertexShaderConstantF(14, 0x015852A8, 1)     //redundance !!!!!!    
 IDirect3DDevice9::SetVertexShaderConstantF(15, 0x015852B8, 4)             
 IDirect3DDevice9::SetVertexShaderConstantF(19, 0x015852F8, 3)             
 IDirect3DDevice9::SetVertexShaderConstantF(23, 0x01585338, 3)             
 IDirect3DDevice9::SetPixelShader(0x06486AE8)   3914655838         //redundance!!!!!
 IDirect3DDevice9::SetPixelShaderConstantF(0, 0x01586D40, 15)      //redundance!!!!!!    
ID3DXEffect::EndPass()                         
ID3DXEffect::End()                     
IDirect3DDevice9::DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 24, 0, 12)// draw the 2 primitive      


//Pix data for beta with d3dx function call
ID3DXEffect::BeginPass(0)                
IDirect3DDevice9::BeginScene()
IDirect3DDevice9::DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 24, 0, 12)        
ID3DXEffect::SetMatrix(0xFF84502B, 0x0023E444)                    
ID3DXEffect::SetMatrix(0xFF8450D3, 0x0023E3D8)                    
ID3DXEffect::SetMatrix(0xFF84507F, 0x0023E3D8)                    
ID3DXEffect::CommitChanges()                    
  IDirect3DDevice9::SetVertexShaderConstantF(15, 0x00FB5668, 4)
  IDirect3DDevice9::SetVertexShaderConstantF(19, 0x00FB56A8, 3)
  IDirect3DDevice9::SetVertexShaderConstantF(23, 0x00FB56E8, 3)
IDirect3DDevice9::DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 24, 0, 12)

//Pix data for beta without d3dx function:
IDirect3DDevice9::BeginScene()                
IDirect3DDevice9::DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 24, 0, 12)
IDirect3DDevice9::SetVertexShaderConstantF(15, 0x007D5660, 4)
IDirect3DDevice9::SetVertexShaderConstantF(19, 0x007D56A0, 3)
IDirect3DDevice9::SetVertexShaderConstantF(23, 0x007D56E0, 3)
IDirect3DDevice9::DrawIndexedPrimitive(D3DPT_TRIANGLELIST, 0, 0, 24, 0, 12)                    

     可以看到beta版中，完全过滤了冗余的状态设置，测试代码中的effect.World会更新world marix，worldViewProjMatrix以及WorldInverseTranspose，因此渲染第二个对象前出现3个参数更新是正确的。测试程序里只渲染了2个图形，因此，BeginScene在第一次DP之前出现也是正确的。