//================================================================================================
bool DXShader11::tryLoadBinThenCompile(LPCWSTR pSrcFile,
LPCSTR pFuncName,
const D3D_SHADER_MACRO * pDefines)
{
//First try loading from bin
DAnsiStr strBinFile;
strBinFile.copyFromW(pSrcFile);
strBinFile = PS::FILESTRINGUTILS::ChangeFileExt(strBinFile, DAnsiStr(".bin"));
if(loadFromBinary(strBinFile.ptr()) )
return true;
//Could not find bin so try compiling the HLSL code and save bin afterwards.
HRESULT res = compile(pSrcFile, pFuncName, pDefines, true);
return SUCCEEDED(res);
}
int SaveArrayCSV(const char* lpArrayName, float* lpArray, U32 count)
{
DAnsiStr strPath = ExtractFilePath(GetExePath());
DAnsiStr strFP = printToAStr("%s/%s", strPath.ptr(), lpArrayName);
DAnsiStr strLine;
for(int i=0; i<count; i++)
{
if(i < count - 1)
strLine += printToAStr("%f, ", lpArray[i]);
else
strLine += printToAStr("%f", lpArray[i]);
}
WriteTextFile(strFP, strLine);
return count;
}
int SimdPoly::linearizeBlobTree(CBlobNode* root, int parentID, int& outIsOperator)
{
int curID = -1;
if(parentID == -1)
{
vec3f lo = root->getOctree().lower;
vec3f hi = root->getOctree().upper;
m_blobPrims.bboxLo = svec3f(lo.x, lo.y, lo.z);
m_blobPrims.bboxHi = svec3f(hi.x, hi.y, hi.z);
//Set identity matrix
float identity[] = {1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f};
m_blobPrimMatrices.count = 1;
m_blobBoxMatrices.count = 1;
for(U32 i=0; i<PRIM_MATRIX_STRIDE; i++)
m_blobPrimMatrices.matrix[i] = identity[i];
for(U32 i=0; i<BOX_MATRIX_STRIDE; i++)
m_blobBoxMatrices.matrix[i] = identity[i];
}
//Operator
outIsOperator = root->isOperator();
if(outIsOperator)
{
if(m_blobOps.ctOps >= MAX_TREE_NODES)
{
ReportError("Exceeded maximum number of allowed Operators");
FlushAllErrors();
return PS_ERROR_OPERATOR_OVERFLOW;
}
curID = m_blobOps.ctOps;
m_blobOps.ctOps++;
m_blobOps.opType[curID] = root->getNodeType();
if(root->countChildren() != 2)
{
ReportError("Not a binary Operator!");
FlushAllErrors();
return PS_ERROR_NON_BINARY_OP;
}
int isOp[2];
int kidID[2];
kidID[0] = this->linearizeBlobTree(root->getChild(0), curID, isOp[0]);
kidID[1] = this->linearizeBlobTree(root->getChild(1), curID, isOp[1]);
m_blobOps.opLeftChild[curID] = kidID[0];
m_blobOps.opRightChild[curID] = kidID[1];
m_blobOps.opChildKind[curID] = isOp[0]*2 + isOp[1];
vec3f lo = root->getOctree().lower;
vec3f hi = root->getOctree().upper;
//ReadBox from BlobTree
m_blobOps.vBoxLoX[curID] = lo.x;
m_blobOps.vBoxLoY[curID] = lo.y;
m_blobOps.vBoxLoZ[curID] = lo.z;
m_blobOps.vBoxHiX[curID] = hi.x;
m_blobOps.vBoxHiY[curID] = hi.y;
m_blobOps.vBoxHiZ[curID] = hi.z;
switch(root->getNodeType())
{
case(bntOpPCM):
{
CPcm* lpPCM = reinterpret_cast<CPcm*>(root);
m_blobOps.resX[curID] = lpPCM->getPropagateLeft();
m_blobOps.resY[curID] = lpPCM->getPropagateRight();
m_blobOps.resZ[curID] = lpPCM->getAlphaLeft();
m_blobOps.resW[curID] = lpPCM->getAlphaRight();
}
break;
case(bntOpRicciBlend):
{
CRicciBlend* ricci = reinterpret_cast<CRicciBlend*>(root);
float n = ricci->getN();
m_blobOps.resX[curID] = n;
if(n != 0.0f)
m_blobOps.resY[curID] = 1.0f / n;
}
break;
case(bntOpWarpTwist):
{
CWarpTwist* twist = reinterpret_cast<CWarpTwist*>(root);
m_blobOps.resX[curID] = twist->getWarpFactor();
m_blobOps.resY[curID] = static_cast<float>(twist->getMajorAxis());
}
break;
case(bntOpWarpTaper):
{
CWarpTaper* taper = reinterpret_cast<CWarpTaper*>(root);
m_blobOps.resX[curID] = taper->getWarpFactor();
m_blobOps.resY[curID] = static_cast<float>(taper->getAxisAlong());
m_blobOps.resZ[curID] = static_cast<float>(taper->getAxisTaper());
}
break;
case(bntOpWarpBend):
{
CWarpBend* bend = reinterpret_cast<CWarpBend*>(root);
m_blobOps.resX[curID] = bend->getBendRate();
m_blobOps.resY[curID] = bend->getBendCenter();
m_blobOps.resZ[curID] = bend->getBendRegion().left;
m_blobOps.resW[curID] = bend->getBendRegion().right;
}
break;
case(bntOpWarpShear):
{
CWarpShear* shear = reinterpret_cast<CWarpShear*>(root);
m_blobOps.resX[curID] = shear->getWarpFactor();
m_blobOps.resY[curID] = static_cast<float>(shear->getAxisAlong());
m_blobOps.resZ[curID] = static_cast<float>(shear->getAxisDependent());
}
break;
}
return curID;
}
else
{
if(m_blobPrims.ctPrims >= MAX_TREE_NODES)
{
ReportError("Exceeded maximum number of allowed primitives");
FlushAllErrors();
return PS_ERROR_PRIM_OVERFLOW;
}
curID = m_blobPrims.ctPrims;
m_blobPrims.ctPrims++;
//
vec4f d = root->getMaterial().diffused;
m_blobPrims.colorX[curID] = d.x;
m_blobPrims.colorY[curID] = d.y;
m_blobPrims.colorZ[curID] = d.z;
vec3f lo = root->getOctree().lower;
vec3f hi = root->getOctree().upper;
//ReadBox from BlobTree
m_blobPrims.vPrimBoxLoX[curID] = lo.x;
m_blobPrims.vPrimBoxLoY[curID] = lo.y;
m_blobPrims.vPrimBoxLoZ[curID] = lo.z;
m_blobPrims.vPrimBoxHiX[curID] = hi.x;
m_blobPrims.vPrimBoxHiY[curID] = hi.y;
m_blobPrims.vPrimBoxHiZ[curID] = hi.z;
CMatrix mtxBackward = root->getTransform().getBackwardMatrix();
if(mtxBackward.isIdentity())
{
m_blobPrims.idxMatrix[curID] = 0;
}
else
{
int idxMat = m_blobPrimMatrices.count;
m_blobPrims.idxMatrix[curID] = idxMat;
idxMat *= PRIM_MATRIX_STRIDE;
float row[16];
mtxBackward.getRow(&row[0], 0);
mtxBackward.getRow(&row[4], 1);
mtxBackward.getRow(&row[8], 2);
mtxBackward.getRow(&row[12], 3);
for(int i=0; i<PRIM_MATRIX_STRIDE; i++)
m_blobPrimMatrices.matrix[idxMat + i] = row[i];
m_blobPrimMatrices.count++;
}
m_blobPrims.skeletType[curID] = root->getNodeType();
switch(root->getNodeType())
{
case(bntPrimPoint):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonPoint* skeletPoint = reinterpret_cast<CSkeletonPoint*>(sprim->getSkeleton());
vec3f pos = skeletPoint->getPosition();
m_blobPrims.posX[curID] = pos.x;
m_blobPrims.posY[curID] = pos.y;
m_blobPrims.posZ[curID] = pos.z;
}
break;
case(bntPrimLine):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonLine* skeletLine = reinterpret_cast<CSkeletonLine*>(sprim->getSkeleton());
//cfg->writeVec3f(strNodeName, "start", skeletLine->getStartPosition());
//cfg->writeVec3f(strNodeName, "end", skeletLine->getEndPosition());
vec3f s = skeletLine->getStartPosition();
vec3f e = skeletLine->getEndPosition();
m_blobPrims.posX[curID] = s.x;
m_blobPrims.posY[curID] = s.y;
m_blobPrims.posZ[curID] = s.z;
m_blobPrims.dirX[curID] = e.x;
m_blobPrims.dirY[curID] = e.y;
m_blobPrims.dirZ[curID] = e.z;
}
break;
case(bntPrimRing):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonRing* skeletRing = reinterpret_cast<CSkeletonRing*>(sprim->getSkeleton());
//cfg->writeVec3f(strNodeName, "position", skeletRing->getPosition());
//cfg->writeVec3f(strNodeName, "direction", skeletRing->getDirection());
//cfg->writeFloat(strNodeName, "radius", skeletRing->getRadius());
vec3f p = skeletRing->getPosition();
vec3f d = skeletRing->getDirection();
float r = skeletRing->getRadius();
m_blobPrims.posX[curID] = p.x;
m_blobPrims.posY[curID] = p.y;
m_blobPrims.posZ[curID] = p.z;
m_blobPrims.dirX[curID] = d.x;
m_blobPrims.dirY[curID] = d.y;
m_blobPrims.dirZ[curID] = d.z;
m_blobPrims.resX[curID] = r;
m_blobPrims.resY[curID] = r*r;
}
break;
case(bntPrimDisc):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonDisc* skeletDisc = reinterpret_cast<CSkeletonDisc*>(sprim->getSkeleton());
//cfg->writeVec3f(strNodeName, "position", skeletDisc->getPosition());
//cfg->writeVec3f(strNodeName, "direction", skeletDisc->getDirection());
//cfg->writeFloat(strNodeName, "radius", skeletDisc->getRadius());
vec3f p = skeletDisc->getPosition();
vec3f d = skeletDisc->getDirection();
float r = skeletDisc->getRadius();
m_blobPrims.posX[curID] = p.x;
m_blobPrims.posY[curID] = p.y;
m_blobPrims.posZ[curID] = p.z;
m_blobPrims.dirX[curID] = d.x;
m_blobPrims.dirY[curID] = d.y;
m_blobPrims.dirZ[curID] = d.z;
m_blobPrims.resX[curID] = r;
m_blobPrims.resY[curID] = r*r;
}
break;
case(bntPrimCylinder):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonCylinder* skeletCyl = reinterpret_cast<CSkeletonCylinder*>(sprim->getSkeleton());
//cfg->writeVec3f(strNodeName, "position", skeletCyl->getPosition());
//cfg->writeVec3f(strNodeName, "direction", skeletCyl->getDirection());
//cfg->writeFloat(strNodeName, "radius", skeletCyl->getRadius());
//cfg->writeFloat(strNodeName, "height", skeletCyl->getHeight());
vec3f p = skeletCyl->getPosition();
vec3f d = skeletCyl->getDirection();
m_blobPrims.posX[curID] = p.x;
m_blobPrims.posY[curID] = p.y;
m_blobPrims.posZ[curID] = p.z;
m_blobPrims.dirX[curID] = d.x;
m_blobPrims.dirY[curID] = d.y;
m_blobPrims.dirZ[curID] = d.z;
m_blobPrims.resX[curID] = skeletCyl->getRadius();
m_blobPrims.resY[curID] = skeletCyl->getHeight();
}
break;
case(bntPrimCube):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonCube* skeletCube = reinterpret_cast<CSkeletonCube*>(sprim->getSkeleton());
//cfg->writeVec3f(strNodeName, "position", skeletCube->getPosition());
//cfg->writeFloat(strNodeName, "side", skeletCube->getSide());
vec3f p = skeletCube->getPosition();
float side = skeletCube->getSide();
m_blobPrims.posX[curID] = p.x;
m_blobPrims.posY[curID] = p.y;
m_blobPrims.posZ[curID] = p.z;
m_blobPrims.resX[curID] = side;
}
break;
case(bntPrimTriangle):
{
CSkeletonPrimitive *sprim = reinterpret_cast<CSkeletonPrimitive*>(root);
CSkeletonTriangle* skeletTriangle = reinterpret_cast<CSkeletonTriangle*>(sprim->getSkeleton());
//cfg->writeVec3f(strNodeName, "corner0", skeletTriangle->getTriangleCorner(0));
//cfg->writeVec3f(strNodeName, "corner1", skeletTriangle->getTriangleCorner(1));
//cfg->writeVec3f(strNodeName, "corner2", skeletTriangle->getTriangleCorner(2));
vec3f p0 = skeletTriangle->getTriangleCorner(0);
vec3f p1 = skeletTriangle->getTriangleCorner(1);
vec3f p2 = skeletTriangle->getTriangleCorner(2);
m_blobPrims.posX[curID] = p0.x;
m_blobPrims.posY[curID] = p0.y;
m_blobPrims.posZ[curID] = p0.z;
m_blobPrims.dirX[curID] = p1.x;
m_blobPrims.dirY[curID] = p1.y;
m_blobPrims.dirZ[curID] = p1.z;
m_blobPrims.resX[curID] = p2.x;
m_blobPrims.resY[curID] = p2.y;
m_blobPrims.resX[curID] = p2.z;
}
break;
case(bntPrimNull):
{
m_blobPrims.posX[curID] = 0.0f;
m_blobPrims.posY[curID] = 0.0f;
m_blobPrims.posZ[curID] = 0.0f;
}
break;
default:
{
DAnsiStr strMsg = printToAStr("Primitive %s has not been implemented in compact mode yet!", root->getName().c_str());
ReportError(strMsg.ptr());
FlushAllErrors();
}
}
return curID;
}
}
HRESULT DXShader11::compile( LPCWSTR pSrcFile,
LPCSTR pFuncName,
const D3D_SHADER_MACRO * pDefines,
bool bSaveByteCode)
{
HRESULT hr;
if(!PS::FILESTRINGUTILS::FileExists( pSrcFile))
return E_FAIL;
DWORD dwShaderFlags = D3DCOMPILE_ENABLE_STRICTNESS;
#if defined( DEBUG ) || defined( _DEBUG )
// Set the D3DCOMPILE_DEBUG flag to embed debug information in the shaders.
// Setting this flag improves the shader debugging experience, but still allows
// the shaders to be optimized and to run exactly the way they will run in
// the release configuration of this program.
dwShaderFlags |= D3DCOMPILE_DEBUG;
#endif
// We generally prefer to use the higher CS shader profile when possible as CS 5.0 is better performance on 11-class hardware
LPCSTR pProfile = ( m_pDXDevice->getDevice()->GetFeatureLevel() >= D3D_FEATURE_LEVEL_11_0 ) ? "cs_5_0" : "cs_4_0";
ID3DBlob* pErrorBlob = NULL;
ID3DBlob* pBlob = NULL;
hr = D3DX11CompileFromFile(pSrcFile, pDefines, NULL, pFuncName, pProfile,
dwShaderFlags, NULL, NULL, &pBlob, &pErrorBlob, NULL );
if ( FAILED(hr) )
{
if ( pErrorBlob )
{
OutputDebugStringA( (char*)pErrorBlob->GetBufferPointer() );
ReportError((const char*)pErrorBlob->GetBufferPointer());
}
SAFE_RELEASE( pErrorBlob );
SAFE_RELEASE( pBlob );
return hr;
}
hr = m_pDXDevice->getDevice()->CreateComputeShader( pBlob->GetBufferPointer(), pBlob->GetBufferSize(), NULL, &m_pComputeShader );
if(SUCCEEDED(hr))
{
m_bCompiled = true;
if(bSaveByteCode)
{
DAnsiStr strBinFile;
strBinFile.copyFromW(pSrcFile);
strBinFile = PS::FILESTRINGUTILS::ChangeFileExt(strBinFile, DAnsiStr(".bin"));
ofstream ofs(strBinFile.ptr(), ios::out | ios::trunc | ios::binary);
if(!ofs.is_open())
return false;
ofs.write( reinterpret_cast<const char*>(pBlob->GetBufferPointer()), pBlob->GetBufferSize());
ofs.close();
}
}
#if defined(DEBUG) || defined(PROFILE)
if ( m_pComputeShader )
(m_pComputeShader)->SetPrivateData( WKPDID_D3DDebugObjectName, lstrlenA(pFuncName), pFuncName);
#endif
SAFE_RELEASE( pErrorBlob );
SAFE_RELEASE( pBlob );
return hr;
}
void CPerfTest::runTests()
{
if(this->FOnPerformTest == NULL)
return;
//count csv file events
size_t rowoffset = countFileEvents();
//Some variables
TestConfig* pConfig = NULL;
CPerfLogger* perf = PS::CPerfLogger::getInstance();
//Count
int ctConfigs = (int)m_lstTestConfigs.size();
int ctFiles = (int)m_lstTestFiles.size();
//Get TASK Manager
TaskManager::JobResult arrResult;
RenderTaskParams param;
//TaskManager* pTaskManager = TaskManager::getTaskManager();
DAnsiStr strTestFile;
DAnsiStr temp;
for(int i=0; i < ctConfigs; i++)
{
pConfig = dynamic_cast<TestConfig*>(m_lstTestConfigs[i]);
//Run all files with pConfig
for(int j=0; j < ctFiles; j++)
{
strTestFile = m_lstTestFiles[j];
temp = PS::FILESTRINGUTILS::ExtractFileName(strTestFile);
//Output some progress status
if(FOnOverallProgress)
FOnOverallProgress(temp.ptr(), static_cast<float>(i * ctFiles + j) / static_cast<float>(ctConfigs * ctFiles) );
//Control all performance outputs
if(FOnPerformTest)
{
param.strPathFileName = strTestFile;
param.lpConfig = pConfig;
FOnPerformTest(¶m);
//pTaskManager->dispatch(&arrResult, TaskManager::JobFunction(FOnPerformTest), ¶m);
}
//FOnPerformTest(strTestFile.ptr(), *pConfig);
//pTaskManager->wait(&arrResult);
//Store test result
storeTestResult(strTestFile, pConfig, perf, rowoffset);
//Clear Performance
perf->cleanup();
}
}
//Write Final CSV file
this->writeCSV();
}