porting to d3d11

1) : VPOS not supported for pixel-position in fragment program
NOT FUNCTIONAL workaround: (see https://gist.github.com/aras-p/8e0c061d4e2490c1f8e2 )
#if defined(SHADER_API_D3D11)
#define VPOS_TYPE float4
#define VPOS_SEM WPOS
#else
#define VPOS_TYPE float2
#define VPOS_SEM VPOS
#endif
half4 frag( v2f IN, VPOS_TYPE in_vpos : VPOS_SEM ) : COLOR
{ ... }

// ENDED UP NOT USING VPOS ON ANY PLATFORMS, CALCULATING SCREENPOS USING UnityCg.cginc->ComputeScreenPos()


2) surface shaders with vertex-programs relying on implicit generation of uv_MainTex fails,
   due to vertex-shader not manually filling out struct.
   - it can't be 'inout' as 'out' is forced by surface shaders.

workaround: as per Aras' comment: add UNITY_INITIALIZE_OUTPUT( Input, IN ) to all verts.
#define D3D11_HACKAROUND IN.uv_MainTex = float2(0,0);

void myVertexProgram( inout appdata_full v, out Input IN )
{
	#if defined( D3D11_HACKAROUND )
	D3D11_HACKAROUND
	#endif //D3D11_HACKAROUND
	...
}

// =========
// from auto-generated surface-shader:

struct Input {
	float2 uv_MainTex	: TEXCOORD0; //note: set automagically
	float fogfact		: TEXCOORD1;
};

	void vert_fog( inout appdata_full v, out Input IN )
	{
		IN.fogfact = calcFogFactor( v.vertex.xyz );
	}


struct v2f_surf {
  float4 pos : SV_POSITION;
  float2 pack0 : TEXCOORD0;
  float cust_fogfact : TEXCOORD1;
  fixed3 normal : TEXCOORD2;
};
float4 _MainTex_ST;
v2f_surf vert_surf (appdata_full v) {
  v2f_surf o;
  Input customInputData;
  vert_fog (v, customInputData); //NOTE: uv_MainTex not filled out manually, as will be transferred from pack0 later.
  o.cust_fogfact = customInputData.fogfact;
  o.pos = mul (UNITY_MATRIX_MVP, v.vertex);
  o.pack0.xy = TRANSFORM_TEX(v.texcoord, _MainTex);
  o.normal = mul((float3x3)_Object2World, SCALED_NORMAL);
  return o;
}


3)
- this causes an internal error in unity, as it allocates memory for a vec3 but tries to set a vec4
uniform float3 _SphereScale;
mpb.AddVector( "_SphereScale", new Vector4() );

workaround: just declare as float4 _SphereScale;

3.1) same thing happens for 
Properties {
	_MyVector( "itsavector", Vector) = (0.25, 0.25, 0.25, 0.25)
}
Pass {
float3 _MyVector; // <-- this should be float4 or d3d11 will keel over
...
}


4) UNITY_MATRIX_MVP somewhat different...
workaround: TBD

// DENORMALS

5) denormals are handled differently (or at all) in d3d11
- divide by zero blows up..!

6) ...something about large numbers (seen with exp2(-large_negative_number)) - min(anyvalue, brokennumber) appears to fix the denormals(!?!)

7) normalize(0) on d3d9 returns 1, on d3d11 it returns 0

I'd suggest just adding a UNITY_INITIALIZE_OUTPUT(Input, IN); as first thing in myVertexProgram