quake4-sdk/source/idlib/math/Simd_AltiVec.h

233 lines
13 KiB
C
Raw Normal View History

2007-06-15 00:00:00 +00:00
// Copyright (C) 2004 Id Software, Inc.
//
#ifndef __MATH_SIMD_ALTIVEC_H__
#define __MATH_SIMD_ALTIVEC_H__
/*
===============================================================================
AltiVec implementation of idSIMDProcessor
===============================================================================
*/
// Defines for enabling parts of the library
// Turns on/off the simple math routines (add, sub, div, etc)
#define ENABLE_SIMPLE_MATH
// Disable dot routines since they introduce artifacts with the shadows. I can't measure
// a performance improvement with sampling tools with DOT enabled, so disable for now
// Turns on/off the dot routines
//#define ENABLE_DOT
// Turns on/off the compare routines
#define ENABLE_COMPARES
// The MinMax routines introduce a couple of bugs. In the bathroom of the alphalabs2 map, the
// wrong surface appears in the mirror at times. It also introduces a noticable delay when map
// data is loaded such as going through doors.
// Turns on/off MinMax routines
//#define ENABLE_MINMAX
// Turns on/off Clamp routines
#define ENABLE_CLAMP
// Disabling because members of idAFConstraint aren't aligned properly
// Turns on/off XXX16 routines
//#define ENABLE_16ROUTINES
// Turns on/off LowerTriangularSolve, LowerTriangularSolveTranspose, and MatX_LDLTFactor
#define ENABLE_LOWER_TRIANGULAR
// Turns on/off TracePointCull, DecalPointCull, and OverlayPoint
// The Enable_Cull routines breaks the g_decals functionality, DecalPointCull is
// the likely suspect. Bullet holes do not appear on the walls when this optimization
// is enabled.
//#define ENABLE_CULL
// Turns on/off DeriveTriPlanes, DeriveTangents, DeriveUnsmoothedTangents, NormalizeTangents
#define ENABLE_DERIVE
// Turns on/off CreateTextureSpaceLightVectors, CreateShadowCache, CreateVertexProgramShadowCache
#define ENABLE_CREATE
// Turns on/off the sound routines
#define ENABLE_SOUND_ROUTINES
// Turns on/off the stuff that isn't on elsewhere
// Currently: BlendJoints, TransformJoints, UntransformJoints, ConvertJointQuatsToJointMats, and
// ConvertJointMatsToJointQuats
//#define LIVE_VICARIOUSLY
// This assumes that the dest array to the sound functions is aligned. If this is not true, we take a large
// performance hit from having to do unaligned stores
//#define SOUND_DEST_ALIGNED
// This assumes that the finalMixBuffer array to the sound functions is aligned. If this is not true, we take a large
// performance hit from having to do unaligned stores
//#define SOUND_MIX_ALIGNED
// This assumes that the vertexCache array to CreateShadowCache and CreateVertexProgramShadowCache is aligned. If it's not,
// then we take a big performance hit from unaligned stores.
//#define VERTEXCACHE_ALIGNED
// This turns on support for PPC intrinsics in the SIMD_AltiVec.cpp file. Right now it's only used for frsqrte. GCC
// supports these intrinsics but XLC does not.
#define PPC_INTRINSICS
// This assumes that the idDrawVert array that is used in DeriveUnsmoothedTangents is aligned. If its not aligned,
// then we don't get any speedup
//#define DERIVE_UNSMOOTH_DRAWVERT_ALIGNED
// The QuakeIV definition of idDrawVert is 64 bytes, so enable this optimization
#define DRAWVERT_PADDED
class idSIMD_AltiVec : public idSIMD_Generic {
#if defined(MACOS_X) && defined(__ppc__)
public:
virtual const char * VPCALL GetName( void ) const;
#ifdef ENABLE_SIMPLE_MATH
// Basic math, works for both aligned and unaligned data
virtual void VPCALL Add( float *dst, const float constant, const float *src, const int count );
virtual void VPCALL Add( float *dst, const float *src0, const float *src1, const int count );
virtual void VPCALL Sub( float *dst, const float constant, const float *src, const int count );
virtual void VPCALL Sub( float *dst, const float *src0, const float *src1, const int count );
virtual void VPCALL Mul( float *dst, const float constant, const float *src, const int count);
virtual void VPCALL Mul( float *dst, const float *src0, const float *src1, const int count );
virtual void VPCALL Div( float *dst, const float constant, const float *divisor, const int count );
virtual void VPCALL Div( float *dst, const float *src0, const float *src1, const int count );
virtual void VPCALL MulAdd( float *dst, const float constant, const float *src, const int count );
virtual void VPCALL MulAdd( float *dst, const float *src0, const float *src1, const int count );
virtual void VPCALL MulSub( float *dst, const float constant, const float *src, const int count );
virtual void VPCALL MulSub( float *dst, const float *src0, const float *src1, const int count );
#endif
#ifdef ENABLE_DOT
// Dot products, expects data structures in contiguous memory
virtual void VPCALL Dot( float *dst, const idVec3 &constant, const idVec3 *src, const int count );
virtual void VPCALL Dot( float *dst, const idVec3 &constant, const idPlane *src, const int count );
virtual void VPCALL Dot( float *dst, const idVec3 &constant, const idDrawVert *src, const int count );
virtual void VPCALL Dot( float *dst, const idPlane &constant,const idVec3 *src, const int count );
virtual void VPCALL Dot( float *dst, const idPlane &constant,const idPlane *src, const int count );
virtual void VPCALL Dot( float *dst, const idPlane &constant,const idDrawVert *src, const int count );
virtual void VPCALL Dot( float *dst, const idVec3 *src0, const idVec3 *src1, const int count );
virtual void VPCALL Dot( float &dot, const float *src1, const float *src2, const int count );
#endif
#ifdef ENABLE_COMPARES
// Comparisons, works for both aligned and unaligned data
virtual void VPCALL CmpGT( byte *dst, const float *src0, const float constant, const int count );
virtual void VPCALL CmpGT( byte *dst, const byte bitNum, const float *src0, const float constant, const int count );
virtual void VPCALL CmpGE( byte *dst, const float *src0, const float constant, const int count );
virtual void VPCALL CmpGE( byte *dst, const byte bitNum, const float *src0, const float constant, const int count );
virtual void VPCALL CmpLT( byte *dst, const float *src0, const float constant, const int count );
virtual void VPCALL CmpLT( byte *dst, const byte bitNum, const float *src0, const float constant, const int count );
virtual void VPCALL CmpLE( byte *dst, const float *src0, const float constant, const int count );
virtual void VPCALL CmpLE( byte *dst, const byte bitNum, const float *src0, const float constant, const int count );
#endif
#ifdef ENABLE_MINMAX
// Min/Max. Expects data structures in contiguous memory
virtual void VPCALL MinMax( float &min, float &max, const float *src, const int count );
virtual void VPCALL MinMax( idVec2 &min, idVec2 &max, const idVec2 *src, const int count );
virtual void VPCALL MinMax( idVec3 &min, idVec3 &max, const idVec3 *src, const int count );
virtual void VPCALL MinMax( idVec3 &min, idVec3 &max, const idDrawVert *src, const int count );
virtual void VPCALL MinMax( idVec3 &min, idVec3 &max, const idDrawVert *src, const int *indexes, const int count );
#endif
#ifdef ENABLE_CLAMP
// Clamp operations. Works for both aligned and unaligned data
virtual void VPCALL Clamp( float *dst, const float *src, const float min, const float max, const int count );
virtual void VPCALL ClampMin( float *dst, const float *src, const float min, const int count );
virtual void VPCALL ClampMax( float *dst, const float *src, const float max, const int count );
#endif
// These are already using memcpy and memset functions. Leaving default implementation
// virtual void VPCALL Memcpy( void *dst, const void *src, const int count );
// virtual void VPCALL Memset( void *dst, const int val, const int count );
#ifdef ENABLE_16ROUTINES
// Operations that expect 16-byte aligned data and 16-byte padded memory (with zeros), generally faster
virtual void VPCALL Zero16( float *dst, const int count );
virtual void VPCALL Negate16( float *dst, const int count );
virtual void VPCALL Copy16( float *dst, const float *src, const int count );
virtual void VPCALL Add16( float *dst, const float *src1, const float *src2, const int count );
virtual void VPCALL Sub16( float *dst, const float *src1, const float *src2, const int count );
virtual void VPCALL Mul16( float *dst, const float *src1, const float constant, const int count );
virtual void VPCALL AddAssign16( float *dst, const float *src, const int count );
virtual void VPCALL SubAssign16( float *dst, const float *src, const int count );
virtual void VPCALL MulAssign16( float *dst, const float constant, const int count );
#endif
// Most of these deal with tiny matrices or vectors, generally not worth altivec'ing since
// the scalar code is already really fast
// virtual void VPCALL MatX_MultiplyVecX( idVecX &dst, const idMatX &mat, const idVecX &vec );
// virtual void VPCALL MatX_MultiplyAddVecX( idVecX &dst, const idMatX &mat, const idVecX &vec );
// virtual void VPCALL MatX_MultiplySubVecX( idVecX &dst, const idMatX &mat, const idVecX &vec );
// virtual void VPCALL MatX_TransposeMultiplyVecX( idVecX &dst, const idMatX &mat, const idVecX &vec );
// virtual void VPCALL MatX_TransposeMultiplyAddVecX( idVecX &dst, const idMatX &mat, const idVecX &vec );
// virtual void VPCALL MatX_TransposeMultiplySubVecX( idVecX &dst, const idMatX &mat, const idVecX &vec );
// virtual void VPCALL MatX_MultiplyMatX( idMatX &dst, const idMatX &m1, const idMatX &m2 );
// virtual void VPCALL MatX_TransposeMultiplyMatX( idMatX &dst, const idMatX &m1, const idMatX &m2 );
#ifdef ENABLE_LOWER_TRIANGULAR
virtual void VPCALL MatX_LowerTriangularSolve( const idMatX &L, float *x, const float *b, const int n, int skip = 0 );
virtual void VPCALL MatX_LowerTriangularSolveTranspose( const idMatX &L, float *x, const float *b, const int n );
#if __GNUC__ >= 4
virtual bool VPCALL MatX_LDLTFactor( idMatX &mat, idVecX &invDiag, const int n );
#else
virtual unsigned char VPCALL MatX_LDLTFactor( idMatX &mat, idVecX &invDiag, const int n );
#endif
#endif
#ifdef LIVE_VICARIOUSLY
virtual void VPCALL BlendJoints( idJointQuat *joints, const idJointQuat *blendJoints, const float lerp, const int *index, const int numJoints );
virtual void VPCALL ConvertJointQuatsToJointMats( idJointMat *jointMats, const idJointQuat *jointQuats, const int numJoints );
virtual void VPCALL ConvertJointMatsToJointQuats( idJointQuat *jointQuats, const idJointMat *jointMats, const int numJoints );
#endif
#ifdef LIVE_VICARIOUSLY
virtual void VPCALL TransformJoints( idJointMat *jointMats, const int *parents, const int firstJoint, const int lastJoint );
virtual void VPCALL UntransformJoints( idJointMat *jointMats, const int *parents, const int firstJoint, const int lastJoint );
#endif
#ifdef ENABLE_CULL
virtual void VPCALL TracePointCull( byte *cullBits, byte &totalOr, const float radius, const idPlane *planes, const idDrawVert *verts, const int numVerts );
virtual void VPCALL DecalPointCull( byte *cullBits, const idPlane *planes, const idDrawVert *verts, const int numVerts );
virtual void VPCALL OverlayPointCull( byte *cullBits, idVec2 *texCoords, const idPlane *planes, const idDrawVert *verts, const int numVerts );
#endif
#ifdef ENABLE_DERIVE
virtual void VPCALL DeriveTriPlanes( idPlane *planes, const idDrawVert *verts, const int numVerts, const int *indexes, const int numIndexes );
virtual void VPCALL DeriveTangents( idPlane *planes, idDrawVert *verts, const int numVerts, const int *indexes, const int numIndexes );
virtual void VPCALL DeriveUnsmoothedTangents( idDrawVert *verts, const dominantTri_s *dominantTris, const int numVerts );
virtual void VPCALL NormalizeTangents( idDrawVert *verts, const int numVerts );
#endif
#ifdef ENABLE_CREATE
virtual void VPCALL CreateTextureSpaceLightVectors( idVec3 *lightVectors, const idVec3 &lightOrigin, const idDrawVert *verts, const int numVerts, const int *indexes, const int numIndexes );
virtual void VPCALL CreateSpecularTextureCoords( idVec4 *texCoords, const idVec3 &lightOrigin, const idVec3 &viewOrigin, const idDrawVert *verts, const int numVerts, const int *indexes, const int numIndexes );
virtual int VPCALL CreateShadowCache( idVec4 *vertexCache, int *vertRemap, const idVec3 &lightOrigin, const idDrawVert *verts, const int numVerts );
virtual int VPCALL CreateVertexProgramShadowCache( idVec4 *vertexCache, const idDrawVert *verts, const int numVerts );
#endif
#ifdef ENABLE_SOUND_ROUTINES
// Sound upsampling and mixing routines, works for aligned and unaligned data
virtual void VPCALL UpSamplePCMTo44kHz( float *dest, const short *pcm, const int numSamples, const int kHz, const int numChannels );
virtual void VPCALL UpSampleOGGTo44kHz( float *dest, const float * const *ogg, const int numSamples, const int kHz, const int numChannels );
virtual void VPCALL MixSoundTwoSpeakerMono( float *mixBuffer, const float *samples, const int numSamples, const float lastV[2], const float currentV[2] );
virtual void VPCALL MixSoundTwoSpeakerStereo( float *mixBuffer, const float *samples, const int numSamples, const float lastV[2], const float currentV[2] );
virtual void VPCALL MixSoundSixSpeakerMono( float *mixBuffer, const float *samples, const int numSamples, const float lastV[6], const float currentV[6] );
virtual void VPCALL MixSoundSixSpeakerStereo( float *mixBuffer, const float *samples, const int numSamples, const float lastV[6], const float currentV[6] );
virtual void VPCALL MixedSoundToSamples( short *samples, const float *mixBuffer, const int numSamples );
#endif
#endif
};
#endif /* !__MATH_SIMD_ALTIVEC_H__ */