/* * Advanced Simulation Library * * Copyright 2015 Avtech Scientific * * * This file is part of Advanced Simulation Library (ASL). * * ASL is free software: you can redistribute it and/or modify it * under the terms of the GNU Affero General Public License as * published by the Free Software Foundation, version 3 of the License. * * ASL is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Affero General Public License for more details. * * You should have received a copy of the GNU Affero General Public License * along with ASL. If not, see . * */ #include "aslCrystalGrowthBC.h" #include "acl/aclMath/aclVectorOfElements.h" #include "acl/acl.h" #include "acl/aclHardware.h" #include "acl/aclGenerators.h" #include #include "acl/Kernels/aclKernelConfigurationTemplates.h" #include #include #include "acl/aclMath/aclMathAlg.h" #include #include #include namespace asl { BCLinearGrowthMap::BCLinearGrowthMap(Data d, const acl::VectorOfElements & ceq, const acl::VectorOfElements & b, Data map, const VectorTemplate *const t): BCondWithMap(map, t), kernel(new acl::Kernel(acl::KERNEL_SIMDUA)), data(d), cEq(ceq), beta(b) { } BCLinearGrowthMap::BCLinearGrowthMap(Data d, const acl::VectorOfElements & ceq, const acl::VectorOfElements & b, Data map, Data computationalDomain, const VectorTemplate *const t): BCondWithMap(map, computationalDomain, t), kernel(new acl::Kernel(acl::KERNEL_SIMDUA)), data(d), cEq(ceq), beta(b) { } BCLinearGrowthMap::~BCLinearGrowthMap() { } void BCLinearGrowthMap::init() { acl::ExpressionContainer kk; unsigned int nd(nD(data->getBlock())); unsigned int nc(data->getEContainer().size()); auto dataX(generateDCFullSafe(data)); initMapInfrastructure(kk); acl::TypeID type(getElementType(data->getEContainer())); vector dataTVE(nc); for(unsigned int ii(0); ii < nc; ++ii) { dataTVE[ii].init(*dataX, *templ, ii); kk << dataTVE[ii].initValues; } auto normal(generateVEPrivateVariable(nd,type)); auto vLocal(generateVEPrivateVariable(nc,type)); auto counter(generateVEPrivateVariable(1,type)); auto dirCoef(generateVEPrivateVariable(1,type)); unsigned int nDir(templ->vectors.size()); kk << (vLocal = acl::generateVEConstantN(nc,0.)); kk << (counter = acl::generateVEConstant(0.)); kk << (normal = gradient(*mapTVE)); kk << (gcNormalize(normal)); for(unsigned int i(1); i < nDir; ++i) { kk << (dirCoef = (normal * normalize(templ->vectors[i])) + 1.); kk << (dirCoef *= dirCoef ); auto xx(exBoundaryX(*mapTVE,i)); for(unsigned int ii(0); ii < nc; ++ii) { auto vBulk(subVE(dataTVE[ii].values,i)); auto anv(subVE(beta,ii)*(normal * templ->vectors[i])); auto c(subVE(cEq,ii)); kk << (subVE(vLocal,ii) += select(acl::generateVEConstant(0.), (vBulk * (1.+ anv * xx) - anv*c) / (1. - anv*(1. - xx)), isComputationNode(i), type) * dirCoef); } kk << (counter+= select(acl::generateVEConstant( 0.), dirCoef, isComputationNode(i), type)); } kk << (vLocal /= max(counter,acl::generateVEConstant(.01))); for(unsigned int ii(0); ii < nc; ++ii) { kk << (acl::assignmentSafe(subVE(data->getEContainer(),ii), select(subVE(dataTVE[ii].values,0), subVE(vLocal,ii), isGhostNode(0), type))); } kernel->addExpression(acl::elementOperators:: ifElse(acl::elementOperators::any((map->getEContainer() <= 0 && map->getEContainer() > -.9999)[0]), kk.expression, {} )); kernel->setup(); } void BCLinearGrowthMap::execute() { kernel->compute(); } BCLinearGrowthMap1::BCLinearGrowthMap1(Data d, const acl::VectorOfElements & ceq, const acl::VectorOfElements & b, Data map, const VectorTemplate *const t): BCondWithMap(map, t), // kernel(new acl::Kernel(acl::KERNEL_SIMDUA)), kernel(new acl::Kernel(acl::KERNEL_BASIC)), data(d), cEq(ceq), beta(b) { } BCLinearGrowthMap1::BCLinearGrowthMap1(Data d, const acl::VectorOfElements & ceq, const acl::VectorOfElements & b, Data map, Data computationalDomain, const VectorTemplate *const t): BCondWithMap(map, computationalDomain, t), // kernel(new acl::Kernel(acl::KERNEL_SIMDUA)), kernel(new acl::Kernel(acl::KERNEL_BASIC)), data(d), cEq(ceq), beta(b) { } BCLinearGrowthMap1::~BCLinearGrowthMap1() { } void BCLinearGrowthMap1::init() { acl::ExpressionContainer kk; unsigned int nd(nD(data->getBlock())); //< number of dimmentions unsigned int nc(data->getEContainer().size()); // number of components auto dataX(generateDCFullSafe(data)); initMapInfrastructure(kk); acl::TypeID type(getElementType(data->getEContainer())); acl::TypeID typeI(acl::TYPE_SELECT[type]); TemplateVE dataTVE(*dataX, *templ); kk << dataTVE.initValues; auto normal(generateVEPrivateVariable(nd,type)); auto area(generateVEPrivateVariable(1,type)); auto sw(generateVEPrivateVariable(1,type)); auto swc(generateVEPrivateVariable(1,type)); auto sna(generateVEPrivateVariable(1,type)); auto sna1mx(generateVEPrivateVariable(1,type)); auto snaxc(generateVEPrivateVariable(nc,type)); auto isBoundary(generateVEPrivateVariable(1,typeI)); unsigned int nDir(templ->vectors.size()); kk << (sw = acl::generateVEConstant(0.)); kk << (swc = acl::generateVEConstant(0.)); kk << (sna = acl::generateVEConstant(0.)); kk << (sna1mx = acl::generateVEConstant(0.)); kk << (snaxc = acl::generateVEConstant(0.)); kk << (normal = gradient(*mapTVE)); kk << (gcNormalize(normal)); kk << gcBoundaryAreaPerGhostPoint(*mapTVE, area); for(unsigned int i(1); i < nDir; ++i) { kk << (isBoundary = isComputationNode(i)); kk << (sw += select(acl::generateVEConstant(templ->laplasCoefs[i]), isBoundary,type)); kk << (swc += select(dataTVE.getValue(i) * templ->laplasCoefs[i], isBoundary,type)); auto a(normalize(templ->vectors[i])); kk << (sna += select(normal * a, isBoundary,type)); kk << (sna1mx += select((normal * a)*(1.-exBoundaryX(*mapTVE,i)), isBoundary,type)); kk << (snaxc += select((normal * a)*exBoundaryX(*mapTVE,i)*dataTVE.getValue(i), isBoundary,type)); } auto vLocal((swc*sna-area*beta*snaxc+area*beta*cEq*sna)/(area*beta*sna1mx+sw*sna)); kk << (acl::assignmentSafe(data->getEContainer(), select(dataTVE.getValue(0), vLocal, isGhostNode(0), type))); kernel->addExpression(acl::elementOperators:: ifElse(acl::elementOperators::any((isGhostNode() && map->getEContainer() > -.9999)[0]), kk.expression, {} )); kernel->setup(); } void BCLinearGrowthMap1::execute() { kernel->compute(); } BCLinearGrowthMap2::BCLinearGrowthMap2(Data d, const acl::VectorOfElements & ceq, const acl::VectorOfElements & b, Data map, const VectorTemplate *const t): BCondWithMap(map, t), // kernelCN(new acl::Kernel(acl::KERNEL_SIMDUA)), // kernelGN(new acl::Kernel(acl::KERNEL_SIMDUA)), kernelCN(new acl::Kernel(acl::KERNEL_BASIC)), kernelGN(new acl::Kernel(acl::KERNEL_BASIC)), data(d), cEq(ceq), beta(b) { } BCLinearGrowthMap2::BCLinearGrowthMap2(Data d, const acl::VectorOfElements & ceq, const acl::VectorOfElements & b, Data map, Data computationalDomain, const VectorTemplate *const t): BCondWithMap(map, computationalDomain, t), // kernelCN(new acl::Kernel(acl::KERNEL_SIMDUA)), // kernelGN(new acl::Kernel(acl::KERNEL_SIMDUA)), kernelCN(new acl::Kernel(acl::KERNEL_BASIC)), kernelGN(new acl::Kernel(acl::KERNEL_BASIC)), data(d), cEq(ceq), beta(b) { } BCLinearGrowthMap2::~BCLinearGrowthMap2() { } void BCLinearGrowthMap2::init() { unsigned int nd(nD(data->getBlock())); unsigned int nc(data->getEContainer().size()); unsigned int nDir(templ->vectors.size()); acl::TypeID type(getElementType(data->getEContainer())); auto normal(generateVEPrivateVariable(nd,type)); auto vLocal(generateVEPrivateVariable(nc,type)); auto counter(generateVEPrivateVariable(1,type)); auto dirCoef(generateVEPrivateVariable(1,type)); auto xx(generateVEPrivateVariable(1,type)); auto dataX(generateDCFullSafe(data)); TemplateVE dataTVE; acl::ExpressionContainer kkCN; initMapInfrastructure(kkCN); kkCN << (vLocal = acl::generateVEConstantN(nc,0.)); kkCN << (counter = acl::generateVEConstant(0.)); kkCN << (normal = gradient(*mapTVE)); kkCN << (gcNormalize(normal)); kkCN << (xx = acl::generateVEConstant(0.)); for(unsigned int i(1); i < nDir; ++i) { kkCN << (dirCoef = (normal * normalize(templ->vectors[i])) + 1.); kkCN << (dirCoef *= dirCoef ); kkCN << (counter+= select(dirCoef, isGhostNode(templ->invertVectors[i]), type)); kkCN << (xx+= select(dirCoef * exBoundaryX(*mapTVE,templ->invertVectors[i]), isGhostNode(templ->invertVectors[i]), type)); } kkCN << (xx=select(acl::generateVEConstant(1.),xx/counter,counter>0.001,type)); kkCN << (counter = max(counter,acl::generateVEConstant(.001), type)); for(unsigned int ii(0); ii < nc; ++ii) { dataTVE.init(*dataX, *templ, ii,false); kkCN << (vLocal = acl::generateVEConstant(0.)); for(unsigned int i(1); i < nDir; ++i) { kkCN << (dirCoef = (normal * normalize(templ->vectors[i])) + 1.); kkCN << (dirCoef *= dirCoef); auto anv(subVE(beta,ii)*(-normal * templ->vectors[i])); kkCN << (vLocal += select((dataTVE.getValue(i)*(1.+anv*xx) + anv*subVE(cEq,ii)) / (1.+ anv*(1.+xx)) * dirCoef, isGhostNode(templ->invertVectors[i]), type)); } kkCN << (acl::assignmentSafe(subVE(data->getEContainer(),ii), select(subVE(dataTVE.values,0), xx*dataTVE.getValue(0) + (1.-xx)*vLocal/counter, isComputationNode(0), type))); } kernelCN->addExpression(acl::elementOperators:: ifElse(acl::elementOperators::any((isGhostNode() && map->getEContainer() > -.9999)[0]), kkCN.expression, {} )); kernelCN->setup(); acl::ExpressionContainer kkGN; initMapInfrastructure(kkGN); kkGN << (vLocal = acl::generateVEConstantN(nc,0.)); kkGN << (counter = acl::generateVEConstant(0.)); kkGN << (normal = gradient(*mapTVE)); kkGN << (gcNormalize(normal)); kkGN << (xx = acl::generateVEConstant(0.)); for(unsigned int i(1); i < nDir; ++i) { kkGN << (dirCoef = (normal * normalize(templ->vectors[i])) + 1.); kkGN << (dirCoef *= dirCoef ); kkGN << (counter+= select(dirCoef, isComputationNode(i), type)); kkGN << (xx+= select(dirCoef * exBoundaryX(*mapTVE,i), isComputationNode(i), type)); } kkGN << (xx=select(acl::generateVEConstant(1.),xx/counter,counter>0.001,type)); kkGN << (counter = max(counter,acl::generateVEConstant(.001), type)); for(unsigned int ii(0); ii < nc; ++ii) { dataTVE.init(*dataX, *templ, ii,false); kkGN << (vLocal = acl::generateVEConstant(0.)); for(unsigned int i(1); i < nDir; ++i) { kkGN << (dirCoef = (normal * normalize(templ->vectors[i])) + 1.); kkGN << (dirCoef *= dirCoef ); auto anv(subVE(beta,ii)*(-normal * templ->vectors[i])); kkGN << (vLocal += select((dataTVE.getValue(i)*(1.-anv*xx) + anv*subVE(cEq,ii)) / (1.+ anv*(1.-xx)) * dirCoef, isComputationNode(i), type)); } kkGN << (acl::assignmentSafe(subVE(data->getEContainer(),ii), select(subVE(dataTVE.values,0), vLocal/counter, isGhostNode(0), type))); } kernelGN->addExpression(acl::elementOperators:: ifElse(acl::elementOperators::any((isGhostNode() && map->getEContainer() > -.9999)[0]), kkGN.expression, {} )); kernelGN->setup(); } void BCLinearGrowthMap2::execute() { kernelCN->compute(); kernelGN->compute(); } SPNumMethod generateBCLinearGrowth(SPAbstractDataWithGhostNodes d, double cEq, double beta, SPAbstractDataWithGhostNodes map, const VectorTemplate *const t) { return make_shared (d, acl::generateVEConstant(cEq), acl::generateVEConstant(beta), map, t); } SPNumMethod generateBCLinearGrowth(SPAbstractDataWithGhostNodes d, double cEq, double beta, SPAbstractDataWithGhostNodes map, SPAbstractDataWithGhostNodes computationalDomain, const VectorTemplate *const t) { return make_shared (d, acl::generateVEConstant(cEq), acl::generateVEConstant(beta), map, computationalDomain, t); } SPNumMethod generateBCLinearGrowth2(SPAbstractDataWithGhostNodes d, double cEq, double beta, SPAbstractDataWithGhostNodes map, const VectorTemplate *const t) { return make_shared (d, acl::generateVEConstant(cEq), acl::generateVEConstant(beta), map, t); } SPNumMethod generateBCLinearGrowth2(SPAbstractDataWithGhostNodes d, double cEq, double beta, SPAbstractDataWithGhostNodes map, SPAbstractDataWithGhostNodes computationalDomain, const VectorTemplate *const t) { return make_shared (d, acl::generateVEConstant(cEq), acl::generateVEConstant(beta), map, computationalDomain, t); } } // asl