sqwarmed/sdk_src/materialsystem/stdshaders/common_fxc.h

376 lines
12 KiB
C

//========== Copyright (c) Valve Corporation, All rights reserved. ==========//
//
// Purpose:
//
// $NoKeywords: $
//
//===========================================================================//
#ifndef COMMON_FXC_H_
#define COMMON_FXC_H_
#include "common_pragmas.h"
#include "common_hlsl_cpp_consts.h"
#ifdef NV3X
# define HALF half
# define HALF2 half2
# define HALF3 half3
# define HALF4 half4
# define HALF3x3 half3x3
# define HALF3x4 half3x4
# define HALF4x3 half4x3
# define HALF_CONSTANT( _constant ) ((HALF)_constant)
#else
# define HALF float
# define HALF2 float2
# define HALF3 float3
# define HALF4 float4
# define HALF3x3 float3x3
# define HALF3x4 float3x4
# define HALF4x3 float4x3
# define HALF_CONSTANT( _constant ) _constant
#endif
#define FP16_MAX 65504.0f
// This is where all common code for both vertex and pixel shaders.
#define OO_SQRT_3 0.57735025882720947f
static const HALF3 bumpBasis[3] = {
HALF3( 0.81649661064147949f, 0.0f, OO_SQRT_3 ),
HALF3( -0.40824833512306213f, 0.70710676908493042f, OO_SQRT_3 ),
HALF3( -0.40824821591377258f, -0.7071068286895752f, OO_SQRT_3 )
};
static const HALF3 bumpBasisTranspose[3] = {
HALF3( 0.81649661064147949f, -0.40824833512306213f, -0.40824833512306213f ),
HALF3( 0.0f, 0.70710676908493042f, -0.7071068286895752f ),
HALF3( OO_SQRT_3, OO_SQRT_3, OO_SQRT_3 )
};
#if defined( _X360 )
#define REVERSE_DEPTH_ON_X360 //uncomment to use D3DFMT_D24FS8 with an inverted depth viewport for better performance. Keep this in sync with the same named #define in public/shaderapi/shareddefs.h
//Note that the reversal happens in the viewport. So ONLY reading back from a depth texture should be affected. Projected math is unaffected.
#endif
bool IsX360( void )
{
#if defined( _X360 )
return true;
#else
return false;
#endif
}
HALF3 CalcReflectionVectorNormalized( HALF3 normal, HALF3 eyeVector )
{
// FIXME: might be better of normalizing with a normalizing cube map and
// get rid of the dot( normal, normal )
// compute reflection vector r = 2 * ((n dot v)/(n dot n)) n - v
return 2.0 * ( dot( normal, eyeVector ) / dot( normal, normal ) ) * normal - eyeVector;
}
HALF3 CalcReflectionVectorUnnormalized( HALF3 normal, HALF3 eyeVector )
{
// FIXME: might be better of normalizing with a normalizing cube map and
// get rid of the dot( normal, normal )
// compute reflection vector r = 2 * ((n dot v)/(n dot n)) n - v
// multiply all values through by N.N. uniformly scaling reflection vector won't affect result
// since it is used in a cubemap lookup
return (2.0*(dot( normal, eyeVector ))*normal) - (dot( normal, normal )*eyeVector);
}
float3 HuePreservingColorClamp( float3 c )
{
// Get the max of all of the color components and a specified maximum amount
float maximum = max( max( c.x, c.y ), max( c.z, 1.0f ) );
return (c / maximum);
}
HALF3 HuePreservingColorClamp( HALF3 c, HALF maxVal )
{
// Get the max of all of the color components and a specified maximum amount
float maximum = max( max( c.x, c.y ), max( c.z, maxVal ) );
return (c * ( maxVal / maximum ) );
}
#if (AA_CLAMP==1)
HALF2 ComputeLightmapCoordinates( HALF4 Lightmap1and2Coord, HALF2 Lightmap3Coord )
{
HALF2 result = saturate(Lightmap1and2Coord.xy) * Lightmap1and2Coord.wz * 0.99;
result += Lightmap3Coord;
return result;
}
void ComputeBumpedLightmapCoordinates( HALF4 Lightmap1and2Coord, HALF2 Lightmap3Coord,
out HALF2 bumpCoord1,
out HALF2 bumpCoord2,
out HALF2 bumpCoord3 )
{
HALF2 result = saturate(Lightmap1and2Coord.xy) * Lightmap1and2Coord.wz * 0.99;
result += Lightmap3Coord;
bumpCoord1 = result + HALF2(Lightmap1and2Coord.z, 0);
bumpCoord2 = result + 2*HALF2(Lightmap1and2Coord.z, 0);
bumpCoord3 = result + 3*HALF2(Lightmap1and2Coord.z, 0);
}
#else
HALF2 ComputeLightmapCoordinates( HALF4 Lightmap1and2Coord, HALF2 Lightmap3Coord )
{
return Lightmap1and2Coord.xy;
}
void ComputeBumpedLightmapCoordinates( HALF4 Lightmap1and2Coord, HALF2 Lightmap3Coord,
out HALF2 bumpCoord1,
out HALF2 bumpCoord2,
out HALF2 bumpCoord3 )
{
bumpCoord1 = Lightmap1and2Coord.xy;
bumpCoord2 = Lightmap1and2Coord.wz; // reversed order!!!
bumpCoord3 = Lightmap3Coord.xy;
}
#endif
// Versions of matrix multiply functions which force HLSL compiler to explictly use DOTs,
// not giving it the option of using MAD expansion. In a perfect world, the compiler would
// always pick the best strategy, and these shouldn't be needed.. but.. well.. umm..
//
// lorenmcq
float3 mul3x3(float3 v, float3x3 m)
{
#if !defined( _X360 )
return float3(dot(v, transpose(m)[0]), dot(v, transpose(m)[1]), dot(v, transpose(m)[2]));
#else
// xbox360 fxc.exe (new back end) borks with transposes, generates bad code
return mul( v, m );
#endif
}
float3 mul4x3(float4 v, float4x3 m)
{
#if !defined( _X360 )
return float3(dot(v, transpose(m)[0]), dot(v, transpose(m)[1]), dot(v, transpose(m)[2]));
#else
// xbox360 fxc.exe (new back end) borks with transposes, generates bad code
return mul( v, m );
#endif
}
float3 DecompressHDR( float4 input )
{
return input.rgb * input.a * MAX_HDR_OVERBRIGHT;
}
float4 CompressHDR( float3 input )
{
// FIXME: want to use min so that we clamp to white, but what happens if we
// have an albedo component that's less than 1/MAX_HDR_OVERBRIGHT?
// float fMax = max( max( color.r, color.g ), color.b );
float4 output;
float fMax = min( min( input.r, input.g ), input.b );
if( fMax > 1.0f )
{
float oofMax = 1.0f / fMax;
output.rgb = oofMax * input.rgb;
output.a = min( fMax / MAX_HDR_OVERBRIGHT, 1.0f );
}
else
{
output.rgb = input.rgb;
output.a = 0.0f;
}
return output;
}
// 2.2 gamma conversion routines
float LinearToGamma( const float f1linear )
{
return pow( f1linear, 1.0f / 2.2f );
}
float3 LinearToGamma( const float3 f3linear )
{
return pow( f3linear, 1.0f / 2.2f );
}
float4 LinearToGamma( const float4 f4linear )
{
return float4( pow( f4linear.xyz, 1.0f / 2.2f ), f4linear.w );
}
float GammaToLinear( const float gamma )
{
return pow( gamma, 2.2f );
}
float3 GammaToLinear( const float3 gamma )
{
return pow( gamma, 2.2f );
}
float4 GammaToLinear( const float4 gamma )
{
return float4( pow( gamma.xyz, 2.2f ), gamma.w );
}
// sRGB gamma conversion routines
float3 SrgbGammaToLinear( float3 vSrgbGammaColor )
{
// 15 asm instructions
float3 vLinearSegment = vSrgbGammaColor.rgb / 12.92f;
float3 vExpSegment = pow( ( ( vSrgbGammaColor.rgb / 1.055f ) + ( 0.055f / 1.055f ) ), 2.4f );
float3 vLinearColor = { ( vSrgbGammaColor.r <= 0.04045f ) ? vLinearSegment.r : vExpSegment.r,
( vSrgbGammaColor.g <= 0.04045f ) ? vLinearSegment.g : vExpSegment.g,
( vSrgbGammaColor.b <= 0.04045f ) ? vLinearSegment.b : vExpSegment.b };
return vLinearColor.rgb;
}
float3 SrgbLinearToGamma( float3 vLinearColor )
{
// 15 asm instructions
float3 vLinearSegment = vLinearColor.rgb * 12.92f;
float3 vExpSegment = ( 1.055f * pow( vLinearColor.rgb, ( 1.0f / 2.4f ) ) ) - 0.055f;
float3 vGammaColor = { ( vLinearColor.r <= 0.0031308f ) ? vLinearSegment.r : vExpSegment.r,
( vLinearColor.g <= 0.0031308f ) ? vLinearSegment.g : vExpSegment.g,
( vLinearColor.b <= 0.0031308f ) ? vLinearSegment.b : vExpSegment.b };
return vGammaColor.rgb;
}
// These two functions use the XBox 360's exact piecewise linear algorithm
float3 X360GammaToLinear( float3 v360GammaColor )
{
// This code reduces the asm down to 11 instructions from the 63 instructions in the 360 XDK
float4 vTmpMul1 = { 1.0f, 2.0f, 4.0f, 8.0f };
float4 vTmpAdd1 = { 0.0f, ( -64.0f / 255.0f ), ( -96.0f / 255.0f ), ( -192.0f / 255.0f ) };
float4 vTmpAdd2 = { 0.0f, ( 64.0f / 255.0f ), ( 128.0f / 255.0f ), ( 513.0f / 255.0f ) };
float4 vRed = ( v360GammaColor.r * vTmpMul1.xyzw * 0.25f ) + ( ( ( vTmpAdd1.xyzw * vTmpMul1.xyzw ) + vTmpAdd2.xyzw ) * 0.25f );
float4 vGreen = ( v360GammaColor.g * vTmpMul1.xyzw * 0.25f ) + ( ( ( vTmpAdd1.xyzw * vTmpMul1.xyzw ) + vTmpAdd2.xyzw ) * 0.25f );
float4 vBlue = ( v360GammaColor.b * vTmpMul1.xyzw * 0.25f ) + ( ( ( vTmpAdd1.xyzw * vTmpMul1.xyzw ) + vTmpAdd2.xyzw ) * 0.25f );
float3 vMax1 = { max( vRed.x, vRed.y ), max( vGreen.x, vGreen.y ), max( vBlue.x, vBlue.y ) };
float3 vMax2 = { max( vRed.z, vRed.w ), max( vGreen.z, vGreen.w ), max( vBlue.z, vBlue.w ) };
float3 vLinearColor = max( vMax1.rgb, vMax2.rgb );
return vLinearColor.rgb;
}
float X360LinearToGamma( float flLinearValue )
{
// This needs to be optimized
float fl360GammaValue;
flLinearValue = saturate( flLinearValue );
if ( flLinearValue < ( 128.0f / 1023.0f ) )
{
if ( flLinearValue < ( 64.0f / 1023.0f ) )
{
fl360GammaValue = flLinearValue * ( 1023.0f * ( 1.0f / 255.0f ) );
}
else
{
fl360GammaValue = flLinearValue * ( ( 1023.0f / 2.0f ) * ( 1.0f / 255.0f ) ) + ( 32.0f / 255.0f );
}
}
else
{
if ( flLinearValue < ( 512.0f / 1023.0f ) )
{
fl360GammaValue = flLinearValue * ( ( 1023.0f / 4.0f ) * ( 1.0f / 255.0f ) ) + ( 64.0f / 255.0f );
}
else
{
fl360GammaValue = flLinearValue * ( ( 1023.0f /8.0f ) * ( 1.0f / 255.0f ) ) + ( 128.0f /255.0f ); // 1.0 -> 1.0034313725490196078431372549016
fl360GammaValue = saturate( fl360GammaValue );
}
}
fl360GammaValue = saturate( fl360GammaValue );
return fl360GammaValue;
}
float3 SrgbGammaTo360Gamma( float3 vSrgbGammaColor )
{
return X360LinearToGamma( SrgbGammaToLinear( vSrgbGammaColor.rgb ) );
}
// Function to do srgb read in shader code
#ifndef SHADER_SRGB_READ
#define SHADER_SRGB_READ 0
#endif
float4 tex2Dsrgb( sampler iSampler, float2 iUv )
{
// This function is named as a hint that the texture is meant to be read with
// an sRGB->linear conversion. We have to do this in shader code on the 360 sometimes.
#if ( SHADER_SRGB_READ == 0 )
{
// Don't fake the srgb read in shader code
return tex2D( iSampler, iUv.xy );
}
#else
{
if ( IsX360() )
{
float4 vTextureValue = tex2D( iSampler, iUv.xy );
vTextureValue.rgb = X360GammaToLinear( vTextureValue.rgb );
return vTextureValue.rgba;
}
else
{
float4 vTextureValue = tex2D( iSampler, iUv.xy );
vTextureValue.rgb = SrgbGammaToLinear( vTextureValue.rgb );
return vTextureValue.rgba;
}
}
#endif
}
// Tangent transform helper functions
float3 Vec3WorldToTangent( float3 iWorldVector, float3 iWorldNormal, float3 iWorldTangent, float3 iWorldBinormal )
{
float3 vTangentVector;
vTangentVector.x = dot( iWorldVector.xyz, iWorldTangent.xyz );
vTangentVector.y = dot( iWorldVector.xyz, iWorldBinormal.xyz );
vTangentVector.z = dot( iWorldVector.xyz, iWorldNormal.xyz );
return vTangentVector.xyz; // Return without normalizing
}
float3 Vec3WorldToTangentNormalized( float3 iWorldVector, float3 iWorldNormal, float3 iWorldTangent, float3 iWorldBinormal )
{
return normalize( Vec3WorldToTangent( iWorldVector, iWorldNormal, iWorldTangent, iWorldBinormal ) );
}
float3 Vec3TangentToWorld( float3 iTangentVector, float3 iWorldNormal, float3 iWorldTangent, float3 iWorldBinormal )
{
float3 vWorldVector;
vWorldVector.xyz = iTangentVector.x * iWorldTangent.xyz;
vWorldVector.xyz += iTangentVector.y * iWorldBinormal.xyz;
vWorldVector.xyz += iTangentVector.z * iWorldNormal.xyz;
return vWorldVector.xyz; // Return without normalizing
}
float3 Vec3TangentToWorldNormalized( float3 iTangentVector, float3 iWorldNormal, float3 iWorldTangent, float3 iWorldBinormal )
{
return normalize( Vec3TangentToWorld( iTangentVector, iWorldNormal, iWorldTangent, iWorldBinormal ) );
}
// returns 1.0f for no fog, 0.0f for fully fogged
float CalcRangeFogFactorFixedFunction( float3 worldPos, float3 eyePos, float flFogMaxDensity, float flFogEndOverRange, float flFogOORange )
{
float dist = distance( eyePos.xyz, worldPos.xyz );
return max( flFogMaxDensity, ( -dist * flFogOORange ) + flFogEndOverRange );
}
// returns 0.0f for no fog, 1.0f for fully fogged which is opposite of what fixed function fog expects so that we don't have to do a "1-x" in the pixel shader.
float CalcRangeFogFactorNonFixedFunction( float3 worldPos, float3 eyePos, float flFogMaxDensity, float flFogEndOverRange, float flFogOORange )
{
float dist = distance( eyePos.xyz, worldPos.xyz );
return min( flFogMaxDensity, saturate( flFogEndOverRange + ( dist * flFogOORange ) ) );
}
#endif //#ifndef COMMON_FXC_H_