// ***************************************************************
bool CDisplayerVisual::getDisplayFlag(TDisplayFlags flag) const
{
//H_AUTO(R2_CDisplayerVisual_getDisplayFlag)
nlctassert(FlagCount <= 32);
nlassert((uint) flag < FlagCount);
return (_DisplayFlags & (1 << flag)) != 0;
}
// ***************************************************************
void CDisplayerVisual::setDisplayFlag(TDisplayFlags flag, bool on)
{
//H_AUTO(R2_CDisplayerVisual_setDisplayFlag)
nlctassert(FlagCount <= 32);
nlassert((uint) flag < FlagCount);
setFlags(_DisplayFlags, 1 << flag, on);
}
void CTuneMrmDlg::applySlider(UINT sliderCur, UINT sliderMax)
{
nlctassert(NL_TMD_SLIDER_SIZE*NL_TMD_SLIDER_MAX_STEP*NL_TMD_SLIDER_STEP_SIZE < ((uint)1<<31));
// compute actual max polygon
uint actualValue= sliderCur*sliderMax*NL_TMD_SLIDER_STEP_SIZE / NL_TMD_SLIDER_SIZE;
if(_Scene)
{
_Scene->setGroupLoadMaxPolygon("Skin", actualValue);
}
// refresh text views
ViewCurrentMaxPoly= NLMISC::toString(actualValue).c_str();
ViewMaxValue= NLMISC::toString(sliderMax*NL_TMD_SLIDER_STEP_SIZE).c_str();
UpdateData(FALSE);
}
//*******************************************************************************
void CWaterEnvMap::initFlattenVB()
{
_FlattenVB.setPreferredMemory(CVertexBuffer::AGPPreferred, true);
_FlattenVB.setName("Flatten VB");
_FlattenVB.clearValueEx();
_FlattenVB.addValueEx (CVertexBuffer::Position, CVertexBuffer::Float3);
_FlattenVB.addValueEx (CVertexBuffer::TexCoord0, CVertexBuffer::Float3);
_FlattenVB.initEx();
nlctassert(FVB_NUM_SIDES % 4 == 0); // number of sides must be a multiple of 4 so that sections sides will align with corners
_FlattenVB.setNumVertices(FVB_NUM_VERTS);
_FlattenIB.setFormat(NL_DEFAULT_INDEX_BUFFER_FORMAT);
_FlattenIB.setNumIndexes(3 * FVB_NUM_TRIS);
{
CVertexBufferReadWrite vbrw;
CIndexBufferReadWrite ibrw;
_FlattenVB.lock(vbrw);
_FlattenIB.lock(ibrw);
for(uint l = 0; l < FVB_NUM_SIDES; ++l)
{
double angle = NLMISC::Pi * 0.25 + 2 * NLMISC::Pi * (double) l / (double) FVB_NUM_SIDES;
for(uint k = 0; k < FVB_NUM_SECTIONS + 1; ++k)
{
double radius = (double) k / (double) (FVB_NUM_SECTIONS - 1);
float x = (float) (radius * cos(angle));
float y = (float) (radius * sin(angle));
if (k < FVB_NUM_SECTIONS)
{
ibrw.setTri(3 * 2 * (k + (l * FVB_NUM_SECTIONS)), getFVBVertex(k, l), getFVBVertex(k + 1, l + 1), getFVBVertex(k + 1, l));
ibrw.setTri(3 * (2 * (k + (l * FVB_NUM_SECTIONS)) + 1), getFVBVertex(k, l), getFVBVertex(k, l + 1), getFVBVertex(k + 1, l + 1));
}
else
{
uint side = l / (FVB_NUM_SIDES / 4);
switch(side)
{
case 0: // top
x /= y;
y = 1.f;
break;
case 1: // left
y /= -x;
x = -1.f;
break;
case 2: // bottom
x /= -y;
y = -1.f;
break;
case 3: // right
y /= x;
x = 1.f;
break;
default:
nlassert(0);
break;
}
}
CVector dir;
//dir.sphericToCartesian(1.f, (float) angle, (float) (NLMISC::Pi * 0.5 * acos(std::max(0.f, (1.f - (float) k / (FVB_NUM_SECTIONS - 1))))));
dir.sphericToCartesian(1.f, (float) angle, (float) acos(std::min(1.f, (float) k / (FVB_NUM_SECTIONS - 1))));
vbrw.setValueFloat3Ex(CVertexBuffer::Position, getFVBVertex(k, l), x, 0.5f, y);
vbrw.setValueFloat3Ex(CVertexBuffer::TexCoord0, getFVBVertex(k, l), -dir.x, dir.z, -dir.y);
}
}
}
}
// ***************************************************************************
void CMeshMRMGeom::applyArrayRawSkinNormal1(CRawVertexNormalSkin1 *src, uint8 *destVertexPtr,
CMatrix3x4 *boneMat3x4, uint nInf)
{
// must write contigously in AGP, and ASM is hardcoded...
nlctassert(NL3D_RAWSKIN_NORMAL_OFF==12);
nlctassert(NL3D_RAWSKIN_UV_OFF==24);
/*extern uint TESTYOYO_NumRawSkinVertices1;
TESTYOYO_NumRawSkinVertices1+= nInf;
H_AUTO( TestYoyo_RawSkin1 );*/
#ifdef NL3D_RAWSKIN_PRECACHE
for(;nInf>0;)
{
// number of vertices to process for this block.
uint nBlockInf= min(NumCacheVertexNormal1, nInf);
// next block.
nInf-= nBlockInf;
// cache the data in L1 cache.
CFastMem::precache(src, nBlockInf * sizeof(CRawVertexNormalSkin1));
#else
{
uint nBlockInf= nInf;
#endif
#ifndef NL3D_RAWSKIN_ASM
// for all InfluencedVertices only.
for(;nBlockInf>0;nBlockInf--, src++, destVertexPtr+=NL3D_RAWSKIN_VERTEX_SIZE)
{
CVector *dstVertex= (CVector*)(destVertexPtr);
CVector *dstNormal= (CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF);
// For 1 matrix, can write directly to AGP (if destVertexPtr is AGP...)
// Vertex.
boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, *(CVector*)(destVertexPtr) );
// Normal.
boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, *(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF) );
// UV copy.
*(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
}
#else
// ASM harcoded for 36
nlctassert(sizeof(CRawVertexNormalSkin1)==36);
/* 116 cycles / loop typical
58 cycles / loop in theory (no memory problem)
*/
__asm
{
mov ecx, nBlockInf
mov esi, src
mov edi, destVertexPtr
mov edx, boneMat3x4
theLoop:
// Vertex.
// **** boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, *(CVector*)(destVertexPtr) );
// eax= matrix
mov eax, [esi]src.MatrixId // uop: 0/1
lea eax, [eax*2+eax]
shl eax, 4
add eax, edx // uop: 1/0
// load x y z
fld [esi]src.Vertex.Pos.x // uop: 0/1
fld [esi]src.Vertex.Pos.y // uop: 0/1
fld [esi]src.Vertex.Pos.z // uop: 0/1
// vout.x= (a11*vin.x + a12*vin.y + a13*vin.z + a14);
fld [eax]CMatrix3x4.a11 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
fld [eax]CMatrix3x4.a12 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a13 // uop: 0/1
fmul st, st(2) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a14 // uop: 0/1
faddp st(1), st // uop: 1/0 (3)
fstp dword ptr[edi] // uop: 0/0/1/1
// vout.y= (a21*vin.x + a22*vin.y + a23*vin.z + a24);
fld [eax]CMatrix3x4.a21
fmul st, st(3)
fld [eax]CMatrix3x4.a22
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a23
fmul st, st(2)
faddp st(1), st
fld [eax]CMatrix3x4.a24
faddp st(1), st
fstp dword ptr[edi+4]
// vout.z= (a31*vin.x + a32*vin.y + a33*vin.z + a34);
fld [eax]CMatrix3x4.a31
fmul st, st(3)
fld [eax]CMatrix3x4.a32
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a33
fmul st, st(2)
faddp st(1), st
fld [eax]CMatrix3x4.a34
faddp st(1), st
fstp dword ptr[edi+8]
// free x y z
fstp st // uop: 1/0
fstp st // uop: 1/0
fstp st // uop: 1/0
// Normal
// **** boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, *(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF) );
// load x y z
fld [esi]src.Vertex.Normal.x
fld [esi]src.Vertex.Normal.y
fld [esi]src.Vertex.Normal.z
// vout.x= (a11*vin.x + a12*vin.y + a13*vin.z + a14);
fld [eax]CMatrix3x4.a11 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
fld [eax]CMatrix3x4.a12 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a13 // uop: 0/1
fmul st, st(2) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fstp dword ptr[edi+12] // uop: 0/0/1/1
// vout.y= (a21*vin.x + a22*vin.y + a23*vin.z + a24);
fld [eax]CMatrix3x4.a21
fmul st, st(3)
fld [eax]CMatrix3x4.a22
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a23
fmul st, st(2)
faddp st(1), st
fstp dword ptr[edi+16]
// vout.z= (a31*vin.x + a32*vin.y + a33*vin.z + a34);
fld [eax]CMatrix3x4.a31
fmul st, st(3)
fld [eax]CMatrix3x4.a32
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a33
fmul st, st(2)
faddp st(1), st
fstp dword ptr[edi+20]
// free x y z
fstp st
fstp st
fstp st
// UV copy.
// **** *(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
mov eax, [esi]src.Vertex.UV.U // uop: 0/1
mov dword ptr[edi+24], eax // uop: 0/0/1/1
mov eax, [esi]src.Vertex.UV.V // uop: 0/1
mov dword ptr[edi+28], eax // uop: 0/0/1/1
// **** next
add esi, 36 // uop: 1/0
add edi, NL3D_RAWSKIN_VERTEX_SIZE // uop: 1/0
dec ecx // uop: 1/0
jnz theLoop // uop: 1/1 (p1)
mov nBlockInf, ecx
mov src, esi
mov destVertexPtr, edi
}
#endif
}
}
// ***************************************************************************
void CMeshMRMGeom::applyArrayRawSkinNormal2(CRawVertexNormalSkin2 *src, uint8 *destVertexPtr,
CMatrix3x4 *boneMat3x4, uint nInf)
{
// must write contigously in AGP, and ASM is hardcoded...
nlctassert(NL3D_RAWSKIN_NORMAL_OFF==12);
nlctassert(NL3D_RAWSKIN_UV_OFF==24);
/*extern uint TESTYOYO_NumRawSkinVertices2;
TESTYOYO_NumRawSkinVertices2+= nInf;
H_AUTO( TestYoyo_RawSkin2 );*/
// Since VertexPtr may be a AGP Ram, MUST NOT read into it! (mulAdd*() do it!)
CVector tmpVert;
#ifdef NL3D_RAWSKIN_PRECACHE
for(;nInf>0;)
{
// number of vertices to process for this block.
uint nBlockInf= min(NumCacheVertexNormal2, nInf);
// next block.
nInf-= nBlockInf;
// cache the data in L1 cache.
CFastMem::precache(src, nBlockInf * sizeof(CRawVertexNormalSkin2));
#else
{
uint nBlockInf= nInf;
#endif
#ifndef NL3D_RAWSKIN_ASM
// for all InfluencedVertices only.
for(;nBlockInf>0;nBlockInf--, src++, destVertexPtr+=NL3D_RAWSKIN_VERTEX_SIZE)
{
// Vertex.
boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, src->Weights[0], tmpVert);
boneMat3x4[ src->MatrixId[1] ].mulAddPoint( src->Vertex.Pos, src->Weights[1], tmpVert);
*(CVector*)(destVertexPtr)= tmpVert;
// Normal.
boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, src->Weights[0], tmpVert);
boneMat3x4[ src->MatrixId[1] ].mulAddVector( src->Vertex.Normal, src->Weights[1], tmpVert);
*(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF)= tmpVert;
// UV copy.
*(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
}
#else
// ASM harcoded for 48
nlctassert(sizeof(CRawVertexNormalSkin2)==48);
/* 154 cycles / loop typical
124 cycles / loop in theory (no memory problem)
*/
__asm
{
mov ecx, nBlockInf
mov esi, src
mov edi, destVertexPtr
mov edx, boneMat3x4
theLoop:
// Vertex.
// **** boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, *(CVector*)(destVertexPtr) );
// eax= matrix0
mov eax, [esi+0]src.MatrixId // uop: 0/1
lea eax, [eax*2+eax]
shl eax, 4
add eax, edx // uop: 1/0
// ebx= matrix1
mov ebx, [esi+4]src.MatrixId // uop: 0/1
lea ebx, [ebx*2+ebx]
shl ebx, 4
add ebx, edx // uop: 1/0
// load x y z
fld [esi]src.Vertex.Pos.x // uop: 0/1
fld [esi]src.Vertex.Pos.y // uop: 0/1
fld [esi]src.Vertex.Pos.z // uop: 0/1
// **** vout.x= (a11*vin.x + a12*vin.y + a13*vin.z + a14);
// 1st Matrix
fld [eax]CMatrix3x4.a11 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
fld [eax]CMatrix3x4.a12 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a13 // uop: 0/1
fmul st, st(2) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a14 // uop: 0/1
faddp st(1), st // uop: 1/0 (3)
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a11
fmul st, st(4)
fld [ebx]CMatrix3x4.a12
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a13
fmul st, st(3)
faddp st(1), st
fld [ebx]CMatrix3x4.a14
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi] // uop: 0/0/1/1
// **** vout.y= (a21*vin.x + a22*vin.y + a23*vin.z + a24);
fld [eax]CMatrix3x4.a21
fmul st, st(3)
fld [eax]CMatrix3x4.a22
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a23
fmul st, st(2)
faddp st(1), st
fld [eax]CMatrix3x4.a24
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a21
fmul st, st(4)
fld [ebx]CMatrix3x4.a22
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a23
fmul st, st(3)
faddp st(1), st
fld [ebx]CMatrix3x4.a24
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+4]
// **** vout.z= (a31*vin.x + a32*vin.y + a33*vin.z + a34);
fld [eax]CMatrix3x4.a31
fmul st, st(3)
fld [eax]CMatrix3x4.a32
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a33
fmul st, st(2)
faddp st(1), st
fld [eax]CMatrix3x4.a34
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a31
fmul st, st(4)
fld [ebx]CMatrix3x4.a32
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a33
fmul st, st(3)
faddp st(1), st
fld [ebx]CMatrix3x4.a34
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+8]
// free x y z
fstp st // uop: 1/0
fstp st // uop: 1/0
fstp st // uop: 1/0
// Normal
// **** boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, *(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF) );
// load x y z
fld [esi]src.Vertex.Normal.x
fld [esi]src.Vertex.Normal.y
fld [esi]src.Vertex.Normal.z
// **** vout.x= (a11*vin.x + a12*vin.y + a13*vin.z + a14);
fld [eax]CMatrix3x4.a11 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
fld [eax]CMatrix3x4.a12 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a13 // uop: 0/1
fmul st, st(2) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a11
fmul st, st(4)
fld [ebx]CMatrix3x4.a12
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a13
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+12] // uop: 0/0/1/1
// **** vout.y= (a21*vin.x + a22*vin.y + a23*vin.z + a24);
fld [eax]CMatrix3x4.a21
fmul st, st(3)
fld [eax]CMatrix3x4.a22
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a23
fmul st, st(2)
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a21
fmul st, st(4)
fld [ebx]CMatrix3x4.a22
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a23
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+16]
// **** vout.z= (a31*vin.x + a32*vin.y + a33*vin.z + a34);
fld [eax]CMatrix3x4.a31
fmul st, st(3)
fld [eax]CMatrix3x4.a32
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a33
fmul st, st(2)
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a31
fmul st, st(4)
fld [ebx]CMatrix3x4.a32
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a33
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+20]
// free x y z
fstp st
fstp st
fstp st
// UV copy.
// **** *(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
mov eax, [esi]src.Vertex.UV.U // uop: 0/1
mov dword ptr[edi+24], eax // uop: 0/0/1/1
mov eax, [esi]src.Vertex.UV.V // uop: 0/1
mov dword ptr[edi+28], eax // uop: 0/0/1/1
// **** next
add esi, 48 // uop: 1/0
add edi, NL3D_RAWSKIN_VERTEX_SIZE // uop: 1/0
dec ecx // uop: 1/0
jnz theLoop // uop: 1/1 (p1)
mov nBlockInf, ecx
mov src, esi
mov destVertexPtr, edi
}
#endif
}
}
// ***************************************************************************
void CMeshMRMGeom::applyArrayRawSkinNormal3(CRawVertexNormalSkin3 *src, uint8 *destVertexPtr,
CMatrix3x4 *boneMat3x4, uint nInf)
{
// must write contigously in AGP, and ASM is hardcoded...
nlctassert(NL3D_RAWSKIN_NORMAL_OFF==12);
nlctassert(NL3D_RAWSKIN_UV_OFF==24);
/*extern uint TESTYOYO_NumRawSkinVertices3;
TESTYOYO_NumRawSkinVertices3+= nInf;
H_AUTO( TestYoyo_RawSkin3 );*/
// Since VertexPtr may be a AGP Ram, MUST NOT read into it! (mulAdd*() do it!)
CVector tmpVert;
#ifdef NL3D_RAWSKIN_PRECACHE
for(;nInf>0;)
{
// number of vertices to process for this block.
uint nBlockInf= min(NumCacheVertexNormal3, nInf);
// next block.
nInf-= nBlockInf;
// cache the data in L1 cache.
CFastMem::precache(src, nBlockInf * sizeof(CRawVertexNormalSkin3));
#else
{
uint nBlockInf= nInf;
#endif
#ifndef NL3D_RAWSKIN_ASM
// for all InfluencedVertices only.
for(;nBlockInf>0;nBlockInf--, src++, destVertexPtr+=NL3D_RAWSKIN_VERTEX_SIZE)
{
// Vertex.
boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, src->Weights[0], tmpVert);
boneMat3x4[ src->MatrixId[1] ].mulAddPoint( src->Vertex.Pos, src->Weights[1], tmpVert);
boneMat3x4[ src->MatrixId[2] ].mulAddPoint( src->Vertex.Pos, src->Weights[2], tmpVert);
*(CVector*)(destVertexPtr)= tmpVert;
// Normal.
boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, src->Weights[0], tmpVert);
boneMat3x4[ src->MatrixId[1] ].mulAddVector( src->Vertex.Normal, src->Weights[1], tmpVert);
boneMat3x4[ src->MatrixId[2] ].mulAddVector( src->Vertex.Normal, src->Weights[2], tmpVert);
*(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF)= tmpVert;
// UV copy.
*(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
}
#else
// ASM harcoded for 56
nlctassert(sizeof(CRawVertexNormalSkin3)==56);
/* 226 cycles / loop typical
192 cycles / loop in theory (no memory problem)
148 optimal
*/
__asm
{
mov ecx, nBlockInf
mov esi, src
mov edi, destVertexPtr
theLoop:
// Vertex.
// **** boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, *(CVector*)(destVertexPtr) );
// eax= matrix0
mov eax, [esi+0]src.MatrixId // uop: 0/1
lea eax, [eax*2+eax]
shl eax, 4
add eax, boneMat3x4 // uop: 1/0
// ebx= matrix1
mov ebx, [esi+4]src.MatrixId // uop: 0/1
lea ebx, [ebx*2+ebx]
shl ebx, 4
add ebx, boneMat3x4 // uop: 1/0
// edx= matrix2
mov edx, [esi+8]src.MatrixId // uop: 0/1
lea edx, [edx*2+edx]
shl edx, 4
add edx, boneMat3x4 // uop: 1/0
// load x y z
fld [esi]src.Vertex.Pos.x // uop: 0/1
fld [esi]src.Vertex.Pos.y // uop: 0/1
fld [esi]src.Vertex.Pos.z // uop: 0/1
// **** vout.x= (a11*vin.x + a12*vin.y + a13*vin.z + a14);
// 1st Matrix
fld [eax]CMatrix3x4.a11 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
fld [eax]CMatrix3x4.a12 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a13 // uop: 0/1
fmul st, st(2) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a14 // uop: 0/1
faddp st(1), st // uop: 1/0 (3)
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a11
fmul st, st(4)
fld [ebx]CMatrix3x4.a12
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a13
fmul st, st(3)
faddp st(1), st
fld [ebx]CMatrix3x4.a14
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// 3rd matrix
fld [edx]CMatrix3x4.a11
fmul st, st(4)
fld [edx]CMatrix3x4.a12
fmul st, st(4)
faddp st(1), st
fld [edx]CMatrix3x4.a13
fmul st, st(3)
faddp st(1), st
fld [edx]CMatrix3x4.a14
faddp st(1), st
// mul by scale, and append
fmul [esi+8]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi] // uop: 0/0/1/1
// **** vout.y= (a21*vin.x + a22*vin.y + a23*vin.z + a24);
fld [eax]CMatrix3x4.a21
fmul st, st(3)
fld [eax]CMatrix3x4.a22
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a23
fmul st, st(2)
faddp st(1), st
fld [eax]CMatrix3x4.a24
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a21
fmul st, st(4)
fld [ebx]CMatrix3x4.a22
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a23
fmul st, st(3)
faddp st(1), st
fld [ebx]CMatrix3x4.a24
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// 3rd matrix
fld [edx]CMatrix3x4.a21
fmul st, st(4)
fld [edx]CMatrix3x4.a22
fmul st, st(4)
faddp st(1), st
fld [edx]CMatrix3x4.a23
fmul st, st(3)
faddp st(1), st
fld [edx]CMatrix3x4.a24
faddp st(1), st
// mul by scale, and append
fmul [esi+8]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+4]
// **** vout.z= (a31*vin.x + a32*vin.y + a33*vin.z + a34);
fld [eax]CMatrix3x4.a31
fmul st, st(3)
fld [eax]CMatrix3x4.a32
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a33
fmul st, st(2)
faddp st(1), st
fld [eax]CMatrix3x4.a34
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a31
fmul st, st(4)
fld [ebx]CMatrix3x4.a32
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a33
fmul st, st(3)
faddp st(1), st
fld [ebx]CMatrix3x4.a34
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// 3rd matrix
fld [edx]CMatrix3x4.a31
fmul st, st(4)
fld [edx]CMatrix3x4.a32
fmul st, st(4)
faddp st(1), st
fld [edx]CMatrix3x4.a33
fmul st, st(3)
faddp st(1), st
fld [edx]CMatrix3x4.a34
faddp st(1), st
// mul by scale, and append
fmul [esi+8]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+8]
// free x y z
fstp st // uop: 1/0
fstp st // uop: 1/0
fstp st // uop: 1/0
// Normal
// **** boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, *(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF) );
// load x y z
fld [esi]src.Vertex.Normal.x
fld [esi]src.Vertex.Normal.y
fld [esi]src.Vertex.Normal.z
// **** vout.x= (a11*vin.x + a12*vin.y + a13*vin.z + a14);
fld [eax]CMatrix3x4.a11 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
fld [eax]CMatrix3x4.a12 // uop: 0/1
fmul st, st(3) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
fld [eax]CMatrix3x4.a13 // uop: 0/1
fmul st, st(2) // uop: 1/0 (5)
faddp st(1), st // uop: 1/0 (3)
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a11
fmul st, st(4)
fld [ebx]CMatrix3x4.a12
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a13
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// 3rd matrix
fld [edx]CMatrix3x4.a11
fmul st, st(4)
fld [edx]CMatrix3x4.a12
fmul st, st(4)
faddp st(1), st
fld [edx]CMatrix3x4.a13
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+8]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+12] // uop: 0/0/1/1
// **** vout.y= (a21*vin.x + a22*vin.y + a23*vin.z + a24);
fld [eax]CMatrix3x4.a21
fmul st, st(3)
fld [eax]CMatrix3x4.a22
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a23
fmul st, st(2)
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a21
fmul st, st(4)
fld [ebx]CMatrix3x4.a22
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a23
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// 3rd matrix
fld [edx]CMatrix3x4.a21
fmul st, st(4)
fld [edx]CMatrix3x4.a22
fmul st, st(4)
faddp st(1), st
fld [edx]CMatrix3x4.a23
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+8]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+16]
// **** vout.z= (a31*vin.x + a32*vin.y + a33*vin.z + a34);
fld [eax]CMatrix3x4.a31
fmul st, st(3)
fld [eax]CMatrix3x4.a32
fmul st, st(3)
faddp st(1), st
fld [eax]CMatrix3x4.a33
fmul st, st(2)
faddp st(1), st
// mul by scale
fmul [esi+0]src.Weights
// 2nd matrix
fld [ebx]CMatrix3x4.a31
fmul st, st(4)
fld [ebx]CMatrix3x4.a32
fmul st, st(4)
faddp st(1), st
fld [ebx]CMatrix3x4.a33
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+4]src.Weights
faddp st(1), st
// 3rd matrix
fld [edx]CMatrix3x4.a31
fmul st, st(4)
fld [edx]CMatrix3x4.a32
fmul st, st(4)
faddp st(1), st
fld [edx]CMatrix3x4.a33
fmul st, st(3)
faddp st(1), st
// mul by scale, and append
fmul [esi+8]src.Weights
faddp st(1), st
// store
fstp dword ptr[edi+20]
// free x y z
fstp st
fstp st
fstp st
// UV copy.
// **** *(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
mov eax, [esi]src.Vertex.UV.U // uop: 0/1
mov dword ptr[edi+24], eax // uop: 0/0/1/1
mov eax, [esi]src.Vertex.UV.V // uop: 0/1
mov dword ptr[edi+28], eax // uop: 0/0/1/1
// **** next
add esi, 56 // uop: 1/0
add edi, NL3D_RAWSKIN_VERTEX_SIZE // uop: 1/0
dec ecx // uop: 1/0
jnz theLoop // uop: 1/1 (p1)
mov nBlockInf, ecx
mov src, esi
mov destVertexPtr, edi
}
#endif
}
}
// ***************************************************************************
void CMeshMRMGeom::applyArrayRawSkinNormal4(CRawVertexNormalSkin4 *src, uint8 *destVertexPtr,
CMatrix3x4 *boneMat3x4, uint nInf)
{
// must write contigously in AGP, and ASM is hardcoded...
nlctassert(NL3D_RAWSKIN_NORMAL_OFF==12);
nlctassert(NL3D_RAWSKIN_UV_OFF==24);
/*extern uint TESTYOYO_NumRawSkinVertices4;
TESTYOYO_NumRawSkinVertices4+= nInf;
H_AUTO( TestYoyo_RawSkin4 );*/
// Since VertexPtr may be a AGP Ram, MUST NOT read into it! (mulAdd*() do it!)
CVector tmpVert;
#ifdef NL3D_RAWSKIN_PRECACHE
for(;nInf>0;)
{
// number of vertices to process for this block.
uint nBlockInf= min(NumCacheVertexNormal4, nInf);
// next block.
nInf-= nBlockInf;
// cache the data in L1 cache.
CFastMem::precache(src, nBlockInf * sizeof(CRawVertexNormalSkin4));
#else
{
uint nBlockInf= nInf;
#endif
// for all InfluencedVertices only.
for(;nBlockInf>0;nBlockInf--, src++, destVertexPtr+=NL3D_RAWSKIN_VERTEX_SIZE)
{
// Vertex.
boneMat3x4[ src->MatrixId[0] ].mulSetPoint( src->Vertex.Pos, src->Weights[0], tmpVert);
boneMat3x4[ src->MatrixId[1] ].mulAddPoint( src->Vertex.Pos, src->Weights[1], tmpVert);
boneMat3x4[ src->MatrixId[2] ].mulAddPoint( src->Vertex.Pos, src->Weights[2], tmpVert);
boneMat3x4[ src->MatrixId[3] ].mulAddPoint( src->Vertex.Pos, src->Weights[3], tmpVert);
*(CVector*)(destVertexPtr)= tmpVert;
// Normal.
boneMat3x4[ src->MatrixId[0] ].mulSetVector( src->Vertex.Normal, src->Weights[0], tmpVert);
boneMat3x4[ src->MatrixId[1] ].mulAddVector( src->Vertex.Normal, src->Weights[1], tmpVert);
boneMat3x4[ src->MatrixId[2] ].mulAddVector( src->Vertex.Normal, src->Weights[2], tmpVert);
boneMat3x4[ src->MatrixId[3] ].mulAddVector( src->Vertex.Normal, src->Weights[3], tmpVert);
*(CVector*)(destVertexPtr + NL3D_RAWSKIN_NORMAL_OFF)= tmpVert;
// UV copy.
*(CUV*)(destVertexPtr + NL3D_RAWSKIN_UV_OFF)= src->Vertex.UV;
}
// NB: ASM not done for 4 vertices, cause very rare and negligeable ...
}
}
// ***************************************************************************
void CMeshMRMGeom::applyRawSkinWithNormal(CLod &lod, CRawSkinNormalCache &rawSkinLod, const CSkeletonModel *skeleton, uint8 *vbHard, float alphaLod)
{
nlassert(_Skinned);
if(_SkinWeights.size()==0)
return;
// Some assert
//===========================
// must have XYZ, Normal and UV only
nlassert( _VBufferFinal.getVertexFormat() == (CVertexBuffer::PositionFlag | CVertexBuffer::NormalFlag | CVertexBuffer::TexCoord0Flag) );
nlassert( _VBufferFinal.getValueType(CVertexBuffer::TexCoord0) == CVertexBuffer::Float2 );
nlassert( _VBufferFinal.getVertexSize() ==NL3D_RAWSKIN_VERTEX_SIZE);
// HardCoded for normalOff==12 (see applyArrayRawSkinNormal*)
nlassert( _VBufferFinal.getNormalOff()==NL3D_RAWSKIN_NORMAL_OFF );
nlassert( _VBufferFinal.getTexCoordOff()==NL3D_RAWSKIN_UV_OFF );
// assert, code below is written especially for 4 per vertex.
nlassert( NL3D_MESH_SKINNING_MAX_MATRIX==4 );
// Compute useful Matrix for this lod.
//===========================
// Those arrays map the array of bones in skeleton.
static vector<CMatrix3x4> boneMat3x4;
computeBoneMatrixes3x4(boneMat3x4, lod.MatrixInfluences, skeleton);
// TestYoyo
/*extern uint TESTYOYO_NumRawSkinVertices;
TESTYOYO_NumRawSkinVertices+= rawSkinLod.Vertices1.size();
TESTYOYO_NumRawSkinVertices+= rawSkinLod.Vertices2.size();
TESTYOYO_NumRawSkinVertices+= rawSkinLod.Vertices3.size();
TESTYOYO_NumRawSkinVertices+= rawSkinLod.Vertices4.size();*/
uint nInf;
// Manage "SoftVertices"
if(rawSkinLod.TotalSoftVertices)
{
// apply skinning into Temp RAM for vertices that are Src of Geomorph
//===========================
static vector<uint8> tempSkin;
uint tempVbSize= rawSkinLod.TotalSoftVertices*NL3D_RAWSKIN_VERTEX_SIZE;
if(tempSkin.size() < tempVbSize)
tempSkin.resize(tempVbSize);
uint8 *destVertexPtr= &tempSkin[0];
// 1 Matrix
nInf= rawSkinLod.SoftVertices[0];
if(nInf>0)
{
applyArrayRawSkinNormal1(&rawSkinLod.Vertices1[0], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// 2 Matrix
nInf= rawSkinLod.SoftVertices[1];
if(nInf>0)
{
applyArrayRawSkinNormal2(&rawSkinLod.Vertices2[0], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// 3 Matrix
nInf= rawSkinLod.SoftVertices[2];
if(nInf>0)
{
applyArrayRawSkinNormal3(&rawSkinLod.Vertices3[0], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// 4 Matrix
nInf= rawSkinLod.SoftVertices[3];
if(nInf>0)
{
applyArrayRawSkinNormal4(&rawSkinLod.Vertices4[0], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// Fast Copy this into AGP Ram. NB: done before Geomorphs, because ensure some precaching this way!!
//===========================
// Skin geomorphs.
uint8 *vbHardStart= vbHard + rawSkinLod.Geomorphs.size()*NL3D_RAWSKIN_VERTEX_SIZE;
// fast copy
CFastMem::memcpy(vbHardStart, &tempSkin[0], tempVbSize);
// Geomorphs directly into AGP Ram
//===========================
clamp(alphaLod, 0.f, 1.f);
float a= alphaLod;
float a1= 1 - alphaLod;
// Fast Geomorph
applyGeomorphPosNormalUV0(rawSkinLod.Geomorphs, &tempSkin[0], vbHard, NL3D_RAWSKIN_VERTEX_SIZE, a, a1);
}
// Manage HardVertices
if(rawSkinLod.TotalHardVertices)
{
// apply skinning directly into AGP RAM for vertices that are not Src of Geomorph
//===========================
uint startId;
// Skip Geomorphs and SoftVertices.
uint8 *destVertexPtr= vbHard + (rawSkinLod.Geomorphs.size()+rawSkinLod.TotalSoftVertices)*NL3D_RAWSKIN_VERTEX_SIZE;
// 1 Matrix
nInf= rawSkinLod.HardVertices[0];
startId= rawSkinLod.SoftVertices[0];
if(nInf>0)
{
applyArrayRawSkinNormal1(&rawSkinLod.Vertices1[startId], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// 2 Matrix
nInf= rawSkinLod.HardVertices[1];
startId= rawSkinLod.SoftVertices[1];
if(nInf>0)
{
applyArrayRawSkinNormal2(&rawSkinLod.Vertices2[startId], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// 3 Matrix
nInf= rawSkinLod.HardVertices[2];
startId= rawSkinLod.SoftVertices[2];
if(nInf>0)
{
applyArrayRawSkinNormal3(&rawSkinLod.Vertices3[startId], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
// 4 Matrix
nInf= rawSkinLod.HardVertices[3];
startId= rawSkinLod.SoftVertices[3];
if(nInf>0)
{
applyArrayRawSkinNormal4(&rawSkinLod.Vertices4[startId], destVertexPtr, &boneMat3x4[0], nInf);
destVertexPtr+= nInf * NL3D_RAWSKIN_VERTEX_SIZE;
}
}
}
// Remplit les différents informations concernant le craft par un item
void FillCraftData( char craft, int eco, int level, bool creature, int bestStat,
int worstStat1, int worstStat2, CSString& outStr )
{
CSString data;
char buf[10];
int index;
static CSString carac[] = { "Durability", "Weight", "SapLoad", "DMG", "Speed",
"Range", "DodgeModifier", "ParryModifier",
"AdversaryDodgeModifier", "AdversaryParryModifier",
"ProtectionFactor", "MaxSlashingProtection",
"MaxBluntProtection", "MaxPiercingProtection",
"ElementalCastingTimeFactor", "ElementalPowerFactor",
"OffensiveAfflictionCastingTimeFactor",
"OffensiveAfflictionPowerFactor", "HealCastingTimeFactor",
"HealPowerFactor", "DefensiveAfflictionCastingTimeFactor",
"DefensiveAfflictionPowerFactor",
"AcidProtection",
"ColdProtection",
"FireProtection",
"RotProtection",
"ShockWaveProtection",
"PoisonProtection",
"ElectricityProtection",
"DesertResistance",
"ForestResistance",
"LacustreResistance",
"JungleResistance",
"PrimaryRootResistance",
};
nlctassert((sizeof(carac)/sizeof(carac[0]))==NumMPStats);
static int mediumStatsByStatQuality[] = { 20, 35, 50, 65, 80 };
int stat, remaining, nbToRaise, ajout;
int stats[NumMPStats];
index = craft - 'A';
outStr = " <STRUCT Name=\"";
outStr += craftParts[index].Desc;
outStr += "\">\n";
nbToRaise = 0;
remaining = 0;
ajout = 0;
currentDocItem.push( DtCraftSlotName, craftParts[index].Desc.splitFrom( "(" ).splitTo( ")" ) );
// enregistrements des stats de chaque
// caractèristique
for ( int i=0; i<NumMPStats; i++ )
{
if ( craftParts[index].Carac[i] )
{
if ( !creature )
stat = mediumStatsByStatQuality[ level-1 ];
else
stat = mediumStatsByStatQuality[ level ];
// gestion des points forts/faibles de la MP
if ( i == bestStat )
{
if ( worstStat2 == -1 )
stat += 20;
else
stat += 40;
if ( stat > 100 )
{
// si une des stats dépasse 100, les autres
// stats sont augmentées
remaining = stat - 100;
stat = 100;
}
}
else if ( ( i == worstStat1 ) || ( i == worstStat2 ) )
{
stat -= 20;
// durabilité minimum
if ( ( i == 0 ) && ( stat < 1 ) )
stat = 1;
}
else
nbToRaise++;
stats[i] = stat;
}
else
stats[i] = -1;
}
if ( nbToRaise != 0 )
ajout = remaining/nbToRaise;
// ajout des informations pour chaque carac
for ( int i=0; i<NumMPStats; i++ )
{
if ( stats[i] != -1 )
{
if ( ( i != bestStat ) && ( i != worstStat1 ) && ( i != worstStat2 ) )
stats[i] += ajout;
outStr += " <ATOM Name=\"";
outStr += carac[i];
outStr += "\" Value=\"";
sprintf( buf, "%d", stats[i] );
outStr += buf;
outStr += "\"/>\n";
}
}
// CraftCivSpec en fonction de l'écosystème
outStr += " <ATOM Name=\"CraftCivSpec\" Value=\"";
CSString craftCiv;
switch ( eco )
{
case 'c' :
outStr += "common";
craftCiv = "All";
break;
case 'd' :
outStr += "fyros";
craftCiv = "Fyros";
break;
case 'f' :
outStr += "matis";
craftCiv = "Matis";
break;
case 'j' :
outStr += "zorai";
craftCiv = "Zorai";
break;
case 'l' :
outStr += "tryker";
craftCiv = "Tryker";
break;
case 'p' :
outStr += "common";
craftCiv = "All";
break;
}
currentDocItem.push( DtCraftCivSpec, craftCiv );
outStr += "\"/>\n </STRUCT>\n";
}
