diff --git a/apps/benchmarks/GranularGas/MESA_PD_GranularGas.cpp b/apps/benchmarks/GranularGas/MESA_PD_GranularGas.cpp
index 68cd3111bacb829897b20913e7d5e299e80d8566..18874901dc71d904af29d874c10b421f7aee094a 100644
--- a/apps/benchmarks/GranularGas/MESA_PD_GranularGas.cpp
+++ b/apps/benchmarks/GranularGas/MESA_PD_GranularGas.cpp
@@ -329,38 +329,34 @@ int main( int argc, char ** argv )
SNN(*ps, domain);
if (bBarrier) WALBERLA_MPI_BARRIER();
tp["SNN"].end();
-
- // if( i % 100 == 0 )
- // {
- // WALBERLA_LOG_DEVEL_ON_ROOT( "Timestep " << i << " / " << simulationSteps );
- // SNNBytesSent = SNN.getBytesSent();
- // SNNBytesReceived = SNN.getBytesReceived();
- // SNNSends = SNN.getNumberOfSends();
- // SNNReceives = SNN.getNumberOfReceives();
- // RPBytesSent = RP.getBytesSent();
- // RPBytesReceived = RP.getBytesReceived();
- // RPSends = RP.getNumberOfSends();
- // RPReceives = RP.getNumberOfReceives();
- // walberla::mpi::reduceInplace(SNNBytesSent, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(SNNBytesReceived, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(SNNSends, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(SNNReceives, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(RPBytesSent, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(RPBytesReceived, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(RPSends, walberla::mpi::SUM);
- // walberla::mpi::reduceInplace(RPReceives, walberla::mpi::SUM);
- // auto cC = walberla::mpi::reduce(contactsChecked, walberla::mpi::SUM);
- // auto cD = walberla::mpi::reduce(contactsDetected, walberla::mpi::SUM);
- // auto cT = walberla::mpi::reduce(contactsTreated, walberla::mpi::SUM);
- // WALBERLA_LOG_DEVEL_ON_ROOT( "SNN bytes communicated: " << SNNBytesSent << " / " << SNNBytesReceived );
- // WALBERLA_LOG_DEVEL_ON_ROOT( "SNN communication partners: " << SNNSends << " / " << SNNReceives );
- // WALBERLA_LOG_DEVEL_ON_ROOT( "RP bytes communicated: " << RPBytesSent << " / " << RPBytesReceived );
- // WALBERLA_LOG_DEVEL_ON_ROOT( "RP communication partners: " << RPSends << " / " << RPReceives );
- // WALBERLA_LOG_DEVEL_ON_ROOT( "contacts checked/detected/treated: " << cC << " / " << cD << " / " << cT );
- // }
}
timer.end();
+ SNNBytesSent = SNN.getBytesSent();
+ SNNBytesReceived = SNN.getBytesReceived();
+ SNNSends = SNN.getNumberOfSends();
+ SNNReceives = SNN.getNumberOfReceives();
+ RPBytesSent = RP.getBytesSent();
+ RPBytesReceived = RP.getBytesReceived();
+ RPSends = RP.getNumberOfSends();
+ RPReceives = RP.getNumberOfReceives();
+ walberla::mpi::reduceInplace(SNNBytesSent, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(SNNBytesReceived, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(SNNSends, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(SNNReceives, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(RPBytesSent, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(RPBytesReceived, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(RPSends, walberla::mpi::SUM);
+ walberla::mpi::reduceInplace(RPReceives, walberla::mpi::SUM);
+ auto cC = walberla::mpi::reduce(contactsChecked, walberla::mpi::SUM);
+ auto cD = walberla::mpi::reduce(contactsDetected, walberla::mpi::SUM);
+ auto cT = walberla::mpi::reduce(contactsTreated, walberla::mpi::SUM);
+ WALBERLA_LOG_DEVEL_ON_ROOT( "SNN bytes communicated: " << SNNBytesSent << " / " << SNNBytesReceived );
+ WALBERLA_LOG_DEVEL_ON_ROOT( "SNN communication partners: " << SNNSends << " / " << SNNReceives );
+ WALBERLA_LOG_DEVEL_ON_ROOT( "RP bytes communicated: " << RPBytesSent << " / " << RPBytesReceived );
+ WALBERLA_LOG_DEVEL_ON_ROOT( "RP communication partners: " << RPSends << " / " << RPReceives );
+ WALBERLA_LOG_DEVEL_ON_ROOT( "contacts checked/detected/treated: " << cC << " / " << cD << " / " << cT );
+
auto timer_reduced = walberla::timing::getReduced(timer, REDUCE_TOTAL, 0);
double PUpS = 0.0;
WALBERLA_ROOT_SECTION()
diff --git a/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt b/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt
index 7330aa0ddb83f973d47071a4dba1a4c03baed242..a5cd53649d6ac5c011c026346d137dbcb862a20c 100644
--- a/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt
+++ b/apps/benchmarks/ProbeVsExtraMessage/CMakeLists.txt
@@ -1,2 +1,7 @@
+waLBerla_add_executable ( NAME PackPerformance
+ FILES PackPerformance.cpp
+ DEPENDS core )
+
waLBerla_add_executable ( NAME ProbeVsExtraMessage
+ FILES ProbeVsExtraMessage.cpp
DEPENDS core postprocessing stencil )
diff --git a/apps/benchmarks/ProbeVsExtraMessage/PackPerformance.cpp b/apps/benchmarks/ProbeVsExtraMessage/PackPerformance.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..6479a3f641778ea859e126450c808febcce0d5e1
--- /dev/null
+++ b/apps/benchmarks/ProbeVsExtraMessage/PackPerformance.cpp
@@ -0,0 +1,112 @@
+//======================================================================================================================
+//
+// This file is part of waLBerla. waLBerla is free software: you can
+// redistribute it and/or modify it under the terms of the GNU General Public
+// License as published by the Free Software Foundation, either version 3 of
+// the License, or (at your option) any later version.
+//
+// waLBerla 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 General Public License
+// for more details.
+//
+// You should have received a copy of the GNU General Public License along
+// with waLBerla (see COPYING.txt). If not, see <http://www.gnu.org/licenses/>.
+//
+//! \file PackPerformance.h
+//! \author Sebastian Eibl <sebastian.eibl@fau.de>
+//
+//======================================================================================================================
+
+#include "core/DataTypes.h"
+#include "core/Environment.h"
+#include "core/math/Vector3.h"
+#include "core/mpi/BufferSystem.h"
+#include "core/mpi/MPIManager.h"
+#include "core/timing/TimingPool.h"
+
+#include <array>
+#include <iostream>
+#include <sstream>
+
+namespace walberla {
+
+int main( int /*argc*/, char ** /*argv*/ )
+{
+ const size_t numElements = 100000000;
+ mpi::SendBuffer sb0;
+ mpi::SendBuffer sb1;
+ mpi::SendBuffer sb2;
+ Vector3<real_t> v(1,2,3);
+ WcTimer timer0;
+ WcTimer timer1;
+ WcTimer timer2;
+
+ for (size_t i = 0; i < numElements; ++i)
+ {
+ sb0 << v;
+ sb1 << v;
+ sb2 << v;
+ }
+
+ WALBERLA_LOG_DEVEL_VAR(sb0.size());
+ sb0.clear();
+ sb1.clear();
+ sb2.clear();
+
+ timer0.start();
+ for (size_t i = 0; i < numElements; ++i)
+ {
+ sb0 << v;
+ }
+ timer0.end();
+
+ WALBERLA_LOG_DEVEL_VAR(sb0.size());
+ sb0.clear();
+
+ timer1.start();
+ for (size_t i = 0; i < numElements; ++i)
+ {
+ sb1 << v[0] << v[1] << v[2];
+ }
+ timer1.end();
+
+ WALBERLA_LOG_DEVEL_VAR(sb0.size());
+ sb0.clear();
+
+ timer2.start();
+ for (size_t i = 0; i < numElements; ++i)
+ {
+ auto pos = sb2.forward(sizeof(real_t) * 3);
+ memcpy(pos, v.data(), sizeof(real_t) * 3);
+ }
+ timer2.end();
+
+ WALBERLA_LOG_DEVEL_VAR(sb0.size());
+ sb0.clear();
+
+ //auto ptr0 = sb0.ptr();
+ //auto ptr1 = sb1.ptr();
+ //for (auto i = 0; i < numElements; ++i)
+ //{
+ // WALBERLA_ASSERT_EQUAL(*ptr0, *ptr1);
+ // ++ptr0;
+ // ++ptr1;
+ //}
+
+ WALBERLA_LOG_DEVEL("native: " << timer0.total());
+ WALBERLA_LOG_DEVEL("elementwise: " << timer1.total());
+ WALBERLA_LOG_DEVEL("memcpy: " << timer2.total());
+
+ return 0;
+}
+
+} // namespace walberla
+
+int main( int argc, char* argv[] )
+{
+ walberla::mpi::Environment mpiEnv( argc, argv );
+ WALBERLA_UNUSED(mpiEnv);
+
+ return walberla::main( argc, argv );
+}
diff --git a/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp b/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp
index 7eb39be162734f2daa942c53e335a2a97daca2bc..960bf1e9e7efd662d867ffe5c952cdceec9ee776 100644
--- a/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp
+++ b/apps/benchmarks/ProbeVsExtraMessage/ProbeVsExtraMessage.cpp
@@ -37,6 +37,22 @@
namespace walberla {
+class CustomBufferSystem : public mpi::BufferSystem
+{
+public:
+ explicit CustomBufferSystem( const MPI_Comm & communicator, int tag = 0 )
+ : mpi::BufferSystem(communicator, tag)
+ {}
+ auto& recvBuffer ( walberla::mpi::MPIRank rank )
+ {
+ auto it = recvInfos_.find(rank);
+ WALBERLA_CHECK_UNEQUAL(it, recvInfos_.end(), recvInfos_.size());
+ return it->second.buffer;
+ }
+
+ auto& getRecvInfos() {return recvInfos_;}
+};
+
class MPIInfo
{
public:
@@ -44,32 +60,41 @@ public:
int getNeighborRank(const stencil::Direction& dir);
private:
+ shared_ptr<mpi::MPIManager> manager_;
Vector3<uint_t> procs_;
Vector3<bool> periodicity_;
Vector3<int> pos_;
+ std::array<int, 27> ranks_;
};
MPIInfo::MPIInfo( const Vector3<uint_t>& procs, const Vector3<bool>& periodicity )
- : procs_(procs)
+ : manager_(mpi::MPIManager::instance())
+ , procs_(procs)
, periodicity_(periodicity)
{
- mpi::MPIManager::instance()->createCartesianComm(procs[0], procs[1], procs[2], periodicity[0], periodicity[1], periodicity[2]);
- mpi::MPIManager::instance()->cartesianCoord(pos_.data());
+ manager_->createCartesianComm(procs[0], procs[1], procs[2], periodicity[0], periodicity[1], periodicity[2]);
+ manager_->cartesianCoord(pos_.data());
+
+ for (auto dirIt = stencil::D3Q27::beginNoCenter(); dirIt != stencil::D3Q27::end(); ++dirIt)
+ {
+ auto neighborCoord = pos_;
+ neighborCoord[0] += stencil::cx[*dirIt];
+ neighborCoord[1] += stencil::cy[*dirIt];
+ neighborCoord[2] += stencil::cz[*dirIt];
+ if (!periodicity_[0] && (neighborCoord[0] < 0)) ranks_[*dirIt] = -1;
+ if (!periodicity_[1] && (neighborCoord[1] < 0)) ranks_[*dirIt] = -1;
+ if (!periodicity_[2] && (neighborCoord[2] < 0)) ranks_[*dirIt] = -1;
+ if (!periodicity_[0] && (neighborCoord[0] >= int_c(procs_[0]))) ranks_[*dirIt] = -1;
+ if (!periodicity_[1] && (neighborCoord[1] >= int_c(procs_[1]))) ranks_[*dirIt] = -1;
+ if (!periodicity_[2] && (neighborCoord[2] >= int_c(procs_[2]))) ranks_[*dirIt] = -1;
+ ranks_[*dirIt] = manager_->cartesianRank(uint_c(neighborCoord[0]), uint_c(neighborCoord[1]), uint_c(neighborCoord[2]));
+ }
}
+inline
int MPIInfo::getNeighborRank( const stencil::Direction& dir )
{
- auto neighborCoord = pos_;
- neighborCoord[0] += stencil::cx[dir];
- neighborCoord[1] += stencil::cy[dir];
- neighborCoord[2] += stencil::cz[dir];
- if (neighborCoord[0] < 0) return -1;
- if (neighborCoord[1] < 0) return -1;
- if (neighborCoord[2] < 0) return -1;
- if (neighborCoord[0] >= int_c(procs_[0])) return -1;
- if (neighborCoord[1] >= int_c(procs_[1])) return -1;
- if (neighborCoord[2] >= int_c(procs_[2])) return -1;
- return mpi::MPIManager::instance()->cartesianRank(uint_c(neighborCoord[0]), uint_c(neighborCoord[1]), uint_c(neighborCoord[2]));
+ return ranks_[dir];
}
template <typename Stencil>
@@ -82,15 +107,13 @@ void communicate( MPIInfo& mpiInfo,
std::vector<char> sendBuf(messageSize);
std::vector<char> recvBuf(messageSize);
- WcTimer& timer = tp[iProbe ? "IProbe" : "twoMessage"];
-
- mpi::BufferSystem bs( mpi::MPIManager::instance()->comm() );
+ CustomBufferSystem bs( mpi::MPIManager::instance()->comm() );
bs.useIProbe(iProbe);
for( uint_t i =0; i < iterations; ++i )
{
- timer.start();
-
+ WALBERLA_MPI_BARRIER();
+ tp["pack"].start();
for (auto dirIt = Stencil::beginNoCenter(); dirIt != Stencil::end(); ++dirIt)
{
auto recvRank = mpiInfo.getNeighborRank( *dirIt );
@@ -98,19 +121,29 @@ void communicate( MPIInfo& mpiInfo,
bs.sendBuffer(recvRank) << sendBuf;
WALBERLA_ASSERT_EQUAL(bs.sendBuffer(recvRank).size(), messageSize + sizeof(size_t));
}
+ tp["pack"].end();
+ WALBERLA_MPI_BARRIER();
+ tp["communicate"].start();
bs.setReceiverInfoFromSendBufferState(false, true);
bs.sendAll();
-
for( auto it = bs.begin(); it != bs.end(); ++it )
{
WALBERLA_ASSERT_EQUAL(it.buffer().size(), messageSize + sizeof(size_t));
- it.buffer() >> recvBuf;
WALBERLA_ASSERT_EQUAL(recvBuf.size(), messageSize);
- WALBERLA_ASSERT(it.buffer().isEmpty());
}
+ tp["communicate"].end();
- timer.end();
+ WALBERLA_MPI_BARRIER();
+ tp["unpack"].start();
+ auto& recvInfos = bs.getRecvInfos();
+ for (auto recvIt = recvInfos.begin(); recvIt != recvInfos.end(); ++recvIt)
+ {
+ auto& rb = recvIt->second.buffer;
+ rb >> recvBuf;
+ WALBERLA_ASSERT(rb.isEmpty());
+ }
+ tp["unpack"].end();
}
}
@@ -166,25 +199,27 @@ int main( int argc, char ** argv )
MPIInfo mpiInfo(procs, periodicity);
- WcTimingPool tp;
+ WcTimingPool tp_twoMessages;
+ WcTimingPool tp_probe;
WALBERLA_MPI_BARRIER();
if (stencil == "D3Q27")
{
- communicate<stencil::D3Q27>(mpiInfo, iterations, messageSize, false, tp);
- communicate<stencil::D3Q27>(mpiInfo, iterations, messageSize, true, tp);
+ communicate<stencil::D3Q27>(mpiInfo, iterations, messageSize, false, tp_twoMessages);
+ communicate<stencil::D3Q27>(mpiInfo, iterations, messageSize, true, tp_probe);
} else if (stencil == "D3Q19")
{
- communicate<stencil::D3Q19>(mpiInfo, iterations, messageSize, false, tp);
- communicate<stencil::D3Q19>(mpiInfo, iterations, messageSize, true, tp);
+ communicate<stencil::D3Q19>(mpiInfo, iterations, messageSize, false, tp_twoMessages);
+ communicate<stencil::D3Q19>(mpiInfo, iterations, messageSize, true, tp_probe);
} else if (stencil == "D3Q7")
{
- communicate<stencil::D3Q7>(mpiInfo, iterations, messageSize, false, tp);
- communicate<stencil::D3Q7>(mpiInfo, iterations, messageSize, true, tp);
+ communicate<stencil::D3Q7>(mpiInfo, iterations, messageSize, false, tp_twoMessages);
+ communicate<stencil::D3Q7>(mpiInfo, iterations, messageSize, true, tp_probe);
} else
{
WALBERLA_ABORT("stencil not supported: " << stencil);
}
- WALBERLA_LOG_INFO_ON_ROOT(tp);
+ WALBERLA_LOG_INFO_ON_ROOT(tp_twoMessages);
+ WALBERLA_LOG_INFO_ON_ROOT(tp_probe);
WALBERLA_ROOT_SECTION()
{
@@ -216,7 +251,8 @@ int main( int argc, char ** argv )
stringProperties["SLURM_TASKS_PER_NODE"] = envToString(std::getenv( "SLURM_TASKS_PER_NODE" ));
auto runId = postprocessing::storeRunInSqliteDB( "ProbeVsTwoMessages.sqlite", integerProperties, stringProperties, realProperties );
- postprocessing::storeTimingPoolInSqliteDB( "ProbeVsTwoMessages.sqlite", runId, tp, "Timings" );
+ postprocessing::storeTimingPoolInSqliteDB( "ProbeVsTwoMessages.sqlite", runId, tp_twoMessages, "twoMessages" );
+ postprocessing::storeTimingPoolInSqliteDB( "ProbeVsTwoMessages.sqlite", runId, tp_probe, "probe" );
}
return 0;
diff --git a/apps/benchmarks/UniformGridGPU/CMakeLists.txt b/apps/benchmarks/UniformGridGPU/CMakeLists.txt
index 857df4d5f20c4be25a6dacaafd8e3c987a53531f..66497288fd68f32b56467d012b396d78de6cdb5b 100644
--- a/apps/benchmarks/UniformGridGPU/CMakeLists.txt
+++ b/apps/benchmarks/UniformGridGPU/CMakeLists.txt
@@ -10,6 +10,7 @@ waLBerla_python_file_generates(UniformGridGPU.py
UniformGridGPU_PackInfo.cu UniformGridGPU_PackInfo.h
UniformGridGPU_MacroSetter.cpp UniformGridGPU_MacroSetter.h
UniformGridGPU_MacroGetter.cpp UniformGridGPU_MacroGetter.h
+ UniformGridGPU_Defines.h
)
foreach(config srt trt mrt smagorinsky entropic )
diff --git a/apps/benchmarks/UniformGridGPU/UniformGridGPU.cpp b/apps/benchmarks/UniformGridGPU/UniformGridGPU.cpp
index a413df6934496259c62fb21b3295a5e9d64186e8..9e95c73dd6171fcf93765aa07708bae3467d742a 100644
--- a/apps/benchmarks/UniformGridGPU/UniformGridGPU.cpp
+++ b/apps/benchmarks/UniformGridGPU/UniformGridGPU.cpp
@@ -35,6 +35,7 @@
#include "UniformGridGPU_Communication.h"
#include "UniformGridGPU_MacroSetter.h"
#include "UniformGridGPU_MacroGetter.h"
+#include "UniformGridGPU_Defines.h"
using namespace walberla;
@@ -332,7 +333,10 @@ int main( int argc, char **argv )
if ( pythonCallbackResults.isCallable())
{
pythonCallbackResults.data().exposeValue( "mlupsPerProcess", mlupsPerProcess );
- pythonCallbackResults.data().exposeValue( "githash", WALBERLA_GIT_SHA1 );
+ pythonCallbackResults.data().exposeValue( "stencil", infoStencil );
+ pythonCallbackResults.data().exposeValue( "configName", infoConfigName );
+ pythonCallbackResults.data().exposeValue( "cse_global", infoCseGlobal );
+ pythonCallbackResults.data().exposeValue( "cse_pdfs", infoCsePdfs );
// Call Python function to report results
pythonCallbackResults();
}
diff --git a/apps/benchmarks/UniformGridGPU/UniformGridGPU.py b/apps/benchmarks/UniformGridGPU/UniformGridGPU.py
index b95b514b5ed4b29aeacca7f71b38440f5f7d8e0e..d3b9444f7b8c25b0ed5bdb1626bcdd218632e00c 100644
--- a/apps/benchmarks/UniformGridGPU/UniformGridGPU.py
+++ b/apps/benchmarks/UniformGridGPU/UniformGridGPU.py
@@ -55,6 +55,15 @@ options_dict = {
}
}
+info_header = """
+#include "stencil/D3Q{q}.h"\nusing Stencil_T = walberla::stencil::D3Q{q};
+const char * infoStencil = "{stencil}";
+const char * infoConfigName = "{configName}";
+const bool infoCseGlobal = {cse_global};
+const bool infoCsePdfs = {cse_pdfs};
+"""
+
+
with CodeGeneration() as ctx:
accessor = StreamPullTwoFieldsAccessor()
#accessor = StreamPushTwoFieldsAccessor()
@@ -109,3 +118,12 @@ with CodeGeneration() as ctx:
# communication
generate_pack_info_from_kernel(ctx, 'UniformGridGPU_PackInfo', update_rule, target='gpu')
+
+ infoHeaderParams = {
+ 'stencil': stencil_str,
+ 'q': q,
+ 'configName': ctx.config,
+ 'cse_global': int(options['optimization']['cse_global']),
+ 'cse_pdfs': int(options['optimization']['cse_pdfs']),
+ }
+ ctx.write_file("UniformGridGPU_Defines.h", info_header.format(**infoHeaderParams))
diff --git a/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.cpp b/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.cpp
index be38e3cb3eb6fcaadaa3caca922776b7b2b6f480..e19413a25583fe1d05e42e6ce9dfb9f61f916214 100644
--- a/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.cpp
+++ b/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.cpp
@@ -225,6 +225,7 @@ int main( int argc, char **argv )
bool useGui = parameters.getParameter<bool>( "useGui", false );
if( useGui )
{
+#ifdef WALBERLA_ENABLE_GUI
cuda::fieldCpy< PdfField_T, cuda::GPUField< real_t > >( blocks, pdfFieldCpuID, pdfFieldGpuID );
timeLoop.addFuncAfterTimeStep( cuda::fieldCpyFunctor<PdfField_T, cuda::GPUField<real_t> >( blocks, pdfFieldCpuID, pdfFieldGpuID ), "copy to CPU" );
GUI gui( timeLoop, blocks, argc, argv);
@@ -235,6 +236,9 @@ int main( int argc, char **argv )
return nullptr;
});
gui.run();
+#else
+ WALBERLA_ABORT_NO_DEBUG_INFO("Application was built without GUI. Set useGui to false or re-compile with GUI.")
+#endif
}
else
{
@@ -264,7 +268,10 @@ int main( int argc, char **argv )
if ( pythonCallbackResults.isCallable())
{
pythonCallbackResults.data().exposeValue( "mlupsPerProcess", mlupsPerProcess );
- pythonCallbackResults.data().exposeValue( "githash", WALBERLA_GIT_SHA1 );
+ pythonCallbackResults.data().exposeValue( "stencil", infoStencil );
+ pythonCallbackResults.data().exposeValue( "configName", infoConfigName );
+ pythonCallbackResults.data().exposeValue( "cse_global", infoCseGlobal );
+ pythonCallbackResults.data().exposeValue( "cse_pdfs", infoCsePdfs );
// Call Python function to report results
pythonCallbackResults();
}
diff --git a/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py b/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py
index 105cb48f4800cbc82ba16908025eb9dd120d8777..08d8b875ac232f76dbf1a065a6e7ee4810c8fc66 100644
--- a/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py
+++ b/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py
@@ -53,6 +53,16 @@ options_dict = {
}
}
+
+info_header = """
+#include "stencil/D3Q{q}.h"\nusing Stencil_T = walberla::stencil::D3Q{q};
+const char * infoStencil = "{stencil}";
+const char * infoConfigName = "{configName}";
+const bool infoCseGlobal = {cse_global};
+const bool infoCsePdfs = {cse_pdfs};
+"""
+
+
with CodeGeneration() as ctx:
accessors = {
'Even': AAEvenTimeStepAccessor(),
@@ -105,5 +115,11 @@ with CodeGeneration() as ctx:
generate_pack_info_from_kernel(ctx, 'UniformGridGPU_AA_PackInfoPull', update_rules['Odd'], kind='pull', target='gpu')
generate_pack_info_from_kernel(ctx, 'UniformGridGPU_AA_PackInfoPush', update_rules['Odd'], kind='push', target='gpu')
- ctx.write_file("UniformGridGPU_AA_Defines.h",
- '#include "stencil/D3Q{0}.h"\nusing Stencil_T = walberla::stencil::D3Q{0}; \n '.format(q))
+ infoHeaderParams = {
+ 'stencil': stencil_str,
+ 'q': q,
+ 'configName': ctx.config,
+ 'cse_global': int(options['optimization']['cse_global']),
+ 'cse_pdfs': int(options['optimization']['cse_pdfs']),
+ }
+ ctx.write_file("UniformGridGPU_AA_Defines.h", info_header.format(**infoHeaderParams))
diff --git a/apps/tutorials/mesa_pd/01_LennardJones.cpp b/apps/tutorials/mesa_pd/01_LennardJones.cpp
index eec37960f8984f3a66cec47fe8a5db57196529f8..6ed34ccabafa8d588270794ca14a13f64b85793b 100644
--- a/apps/tutorials/mesa_pd/01_LennardJones.cpp
+++ b/apps/tutorials/mesa_pd/01_LennardJones.cpp
@@ -89,15 +89,15 @@ int main( int argc, char ** argv )
WALBERLA_LOG_DEVEL(timestep);
linkedCells.clear();
storage->forEachParticle(true, kernel::SelectAll(), ac, ipilc, ac, linkedCells);
- storage->forEachParticle(true, kernel::SelectAll(), ac, vvPreForce, ac);
+ storage->forEachParticle(true, kernel::SelectLocal(), ac, vvPreForce, ac);
linkedCells.forEachParticlePairHalf(true, kernel::SelectAll(), ac, lj, ac);
const real_t coeff = real_t(0.2);
storage->forEachParticle(true,
- kernel::SelectAll(),
+ kernel::SelectLocal(),
ac,
[coeff](const size_t idx, auto& access){ access.setForce(idx, -coeff*access.getPosition(idx) + access.getForce(idx)); },
ac);
- storage->forEachParticle(true, kernel::SelectAll(), ac, vvPostForce, ac);
+ storage->forEachParticle(true, kernel::SelectLocal(), ac, vvPostForce, ac);
vtkWriter->write();
}
diff --git a/python/mesa_pd/templates/mpi/ReduceProperty.templ.h b/python/mesa_pd/templates/mpi/ReduceProperty.templ.h
index 99eab36d7cb3da7a7a9ce78c53624785b20f8951..38f7340854e17934bc5e867757968566393d658a 100644
--- a/python/mesa_pd/templates/mpi/ReduceProperty.templ.h
+++ b/python/mesa_pd/templates/mpi/ReduceProperty.templ.h
@@ -29,6 +29,7 @@
#include <mesa_pd/data/DataTypes.h>
#include <mesa_pd/data/Flags.h>
#include <mesa_pd/data/ParticleStorage.h>
+#include <mesa_pd/mpi/notifications/reset.h>
#include <core/mpi/BufferSystem.h>
#include <core/logging/Logging.h>
@@ -100,6 +101,7 @@ void ReduceProperty::operator()(data::ParticleStorage& ps) const
}
sb << Notification( p );
+ reset<Notification>( p );
} else
{
//local particles should receive the property and sum it up
diff --git a/python/mesa_pd/templates/mpi/notifications/ContactHistoryNotification.templ.h b/python/mesa_pd/templates/mpi/notifications/ContactHistoryNotification.templ.h
index 15ab081fdfb77e9102f4ff4cbf4b03323ec4c0c8..2ffca67a2aeecd4d4768ad6c4685bf426652afa7 100644
--- a/python/mesa_pd/templates/mpi/notifications/ContactHistoryNotification.templ.h
+++ b/python/mesa_pd/templates/mpi/notifications/ContactHistoryNotification.templ.h
@@ -28,6 +28,7 @@
#include <mesa_pd/data/ContactHistory.h>
#include <mesa_pd/data/DataTypes.h>
+#include <mesa_pd/mpi/notifications/reset.h>
#include <core/mpi/BufferDataTypeExtensions.h>
#include <core/mpi/Datatype.h>
@@ -56,6 +57,12 @@ public:
const data::Particle& p_;
};
+template <>
+void reset<ContactHistoryNotification>(data::Particle& p)
+{
+ p.setNewContactHistory(std::map<walberla::id_t, walberla::mesa_pd::data::ContactHistory>());
+}
+
void reduce(data::Particle&& p, const ContactHistoryNotification::Parameters& objparam)
{
auto& ch = p.getNewContactHistoryRef();
diff --git a/python/mesa_pd/templates/mpi/notifications/ForceTorqueNotification.templ.h b/python/mesa_pd/templates/mpi/notifications/ForceTorqueNotification.templ.h
index f00db9e420a65647933fc69bd5015fe144bcd873..6bfa6412b1a7d5e1d8661c36868f14b76192f69a 100644
--- a/python/mesa_pd/templates/mpi/notifications/ForceTorqueNotification.templ.h
+++ b/python/mesa_pd/templates/mpi/notifications/ForceTorqueNotification.templ.h
@@ -29,6 +29,7 @@
#include <mesa_pd/data/DataTypes.h>
#include <mesa_pd/data/ParticleStorage.h>
#include <mesa_pd/mpi/notifications/NotificationType.h>
+#include <mesa_pd/mpi/notifications/reset.h>
#include <core/mpi/Datatype.h>
#include <core/mpi/RecvBuffer.h>
@@ -55,6 +56,13 @@ public:
const data::Particle& p_;
};
+template <>
+void reset<ForceTorqueNotification>(data::Particle& p)
+{
+ p.setForce( Vec3(real_t(0)) );
+ p.setTorque( Vec3(real_t(0)) );
+}
+
void reduce(data::Particle&& p, const ForceTorqueNotification::Parameters& objparam)
{
p.getForceRef() += objparam.force_;
diff --git a/src/core/math/GenericAABB.h b/src/core/math/GenericAABB.h
index 4664a48a3e821e356488629fbf70fa2b3fbda897..9b321e738f53548dba55094722425937967bef12 100644
--- a/src/core/math/GenericAABB.h
+++ b/src/core/math/GenericAABB.h
@@ -196,7 +196,10 @@ public:
inline friend mpi::GenericRecvBuffer< ET > & operator>>( mpi::GenericRecvBuffer< ET > & buf, GenericAABB< T > & aabb )
{
buf.readDebugMarker( "bb" );
- buf >> aabb.minCorner_ >> aabb.maxCorner_;
+ static_assert ( std::is_trivially_copyable< GenericAABB< T > >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.skip(sizeof(GenericAABB< T >));
+ std::memcpy(&aabb, pos, sizeof(GenericAABB< T >));
WALBERLA_ASSERT( aabb.checkInvariant() );
return buf;
}
diff --git a/src/core/math/GenericAABB.impl.h b/src/core/math/GenericAABB.impl.h
index 7d24f164f2fbae77cbb715d3e4ac2fb35ba987ff..1fae800a88c6f7bcfb1d7a2b24cc9a30411c8066 100644
--- a/src/core/math/GenericAABB.impl.h
+++ b/src/core/math/GenericAABB.impl.h
@@ -1883,7 +1883,11 @@ template< typename T, // Element type of SendBuffer
mpi::GenericSendBuffer<T,G>& operator<<( mpi::GenericSendBuffer<T,G> & buf, const GenericAABB< VT > & aabb )
{
buf.addDebugMarker( "bb" );
- return buf << aabb.minCorner() << aabb.maxCorner();
+ static_assert ( std::is_trivially_copyable< GenericAABB< VT > >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.forward(sizeof(GenericAABB< VT >));
+ std::memcpy(pos, &aabb, sizeof(GenericAABB< VT >));
+ return buf;
}
diff --git a/src/core/math/Matrix2.h b/src/core/math/Matrix2.h
index 8e0504359bb722f912d159557034c5bdab9773e6..e3d8893333b3e8ba27f59b3565d551379b8fd774 100644
--- a/src/core/math/Matrix2.h
+++ b/src/core/math/Matrix2.h
@@ -958,9 +958,11 @@ template< typename T, // Element type of SendBuffer
typename MT > // Element type of matrix
mpi::GenericSendBuffer<T,G>& operator<<( mpi::GenericSendBuffer<T,G> & buf, const Matrix2<MT> & m )
{
- for(unsigned int i=0; i<4; ++i)
- buf << m[i];
-
+ buf.addDebugMarker( "m2" );
+ static_assert ( std::is_trivially_copyable< Matrix2<MT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.forward(sizeof(Matrix2<MT>));
+ std::memcpy(pos, &m, sizeof(Matrix2<MT>));
return buf;
}
@@ -968,9 +970,12 @@ template< typename T, // Element type of RecvBuffer
typename MT > // Element type of matrix
mpi::GenericRecvBuffer<T>& operator>>( mpi::GenericRecvBuffer<T> & buf, Matrix2<MT> & m )
{
- for(unsigned int i=0; i<4; ++i)
- buf >> m[i];
-
+ buf.readDebugMarker( "m2" );
+ static_assert ( std::is_trivially_copyable< Matrix2<MT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.skip(sizeof(Matrix2<MT>));
+ //suppress https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
+ std::memcpy(static_cast<void*>(&m), pos, sizeof(Matrix2<MT>));
return buf;
}
diff --git a/src/core/math/Matrix3.h b/src/core/math/Matrix3.h
index 311a041200f49c08bb21c4f456c382dc2d0fb75b..88c7818b7926b05747cdaa09578aa8bbc32c6cb1 100644
--- a/src/core/math/Matrix3.h
+++ b/src/core/math/Matrix3.h
@@ -35,6 +35,7 @@
#include <type_traits>
#include <algorithm>
+#include <array>
#include <cmath>
#include <iostream>
#include <limits>
@@ -174,6 +175,7 @@ public:
template< typename Other > inline const Vector3<HIGH> solve( const Vector3<Other> &rhs ) const;
inline Type trace() const;
inline Type* data() {return v_;}
+ inline Type const * data() const {return v_;}
//@}
//*******************************************************************************************************************
@@ -246,9 +248,9 @@ private:
* 6 & 7 & 8 \\
* \end{array}\right)\f]
**/
- Type v_[9] = {Type(1), Type(0), Type(0),
- Type(0), Type(1), Type(0),
- Type(0), Type(0), Type(1)};
+ std::array<Type, 9> v_ = {{Type(1), Type(0), Type(0),
+ Type(0), Type(1), Type(0),
+ Type(0), Type(0), Type(1)}};
//@}
//*******************************************************************************************************************
};
@@ -1769,9 +1771,11 @@ namespace mpi {
typename MT > // Element type of matrix
mpi::GenericSendBuffer<T,G>& operator<<( mpi::GenericSendBuffer<T,G> & buf, const Matrix3<MT> & m )
{
- for(unsigned int i=0; i<9; ++i)
- buf << m[i];
-
+ buf.addDebugMarker( "m3" );
+ static_assert ( std::is_trivially_copyable< Matrix3<MT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.forward(sizeof(Matrix3<MT>));
+ std::memcpy(pos, &m, sizeof(Matrix3<MT>));
return buf;
}
@@ -1779,9 +1783,12 @@ namespace mpi {
typename MT > // Element type of matrix
mpi::GenericRecvBuffer<T>& operator>>( mpi::GenericRecvBuffer<T> & buf, Matrix3<MT> & m )
{
- for(unsigned int i=0; i<9; ++i)
- buf >> m[i];
-
+ buf.readDebugMarker( "m3" );
+ static_assert ( std::is_trivially_copyable< Matrix3<MT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.skip(sizeof(Matrix3<MT>));
+ //suppress https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
+ std::memcpy(static_cast<void*>(&m), pos, sizeof(Matrix3<MT>));
return buf;
}
diff --git a/src/core/math/Quaternion.h b/src/core/math/Quaternion.h
index dd3a4298ab22a0fc814ab65a7e27b90321a5a888..90fd652de353026cab2612aed1be1570d754ea66 100644
--- a/src/core/math/Quaternion.h
+++ b/src/core/math/Quaternion.h
@@ -161,6 +161,8 @@ public:
inline void rotateZ( Type angle );
inline void swap( Quaternion& q ) /* throw() */;
inline const Vector3<Type> getEulerAnglesXYZ() const;
+ inline Type* data() {return v_;}
+ inline Type const * data() const {return v_;}
//@}
//**********************************************************************************************
@@ -1097,7 +1099,10 @@ namespace mpi {
mpi::GenericSendBuffer<T,G>& operator<<( mpi::GenericSendBuffer<T,G> & buf, const math::Quaternion<VT> & quat )
{
buf.addDebugMarker( "q4" );
- buf << quat[0] << quat[1] << quat[2] << quat[3];
+ static_assert ( std::is_trivially_copyable< math::Quaternion<VT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.forward(sizeof(math::Quaternion<VT>));
+ std::memcpy(pos, &quat, sizeof(math::Quaternion<VT>));
return buf;
}
@@ -1106,9 +1111,11 @@ namespace mpi {
mpi::GenericRecvBuffer<T>& operator>>( mpi::GenericRecvBuffer<T> & buf, math::Quaternion<VT> & quat )
{
buf.readDebugMarker( "q4" );
- VT tmp1, tmp2, tmp3, tmp4;
- buf >> tmp1 >> tmp2 >> tmp3 >> tmp4;
- quat.set(tmp1, tmp2, tmp3, tmp4);
+ static_assert ( std::is_trivially_copyable< math::Quaternion<VT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.skip(sizeof(math::Quaternion<VT>));
+ //suppress https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
+ std::memcpy(static_cast<void*>(&quat), pos, sizeof(math::Quaternion<VT>));
return buf;
}
diff --git a/src/core/math/Vector2.h b/src/core/math/Vector2.h
index c9b9dba4be0d8c5e38fc9296ed037e47ced1559c..b62d58f4651a743e33a0b9c6f0e4f7c860d174dd 100644
--- a/src/core/math/Vector2.h
+++ b/src/core/math/Vector2.h
@@ -160,6 +160,7 @@ public:
inline Type sqrLength() const;
inline Vector2<Length> getNormalized() const;
inline Type* data() {return v_;}
+ inline Type const * data() const {return v_;}
//@}
//*******************************************************************************************************************
@@ -1609,7 +1610,10 @@ namespace mpi {
mpi::GenericSendBuffer<T,G>& operator<<( mpi::GenericSendBuffer<T,G> & buf, const Vector2<VT> & vec )
{
buf.addDebugMarker( "v2" );
- buf << vec[0] << vec[1];
+ static_assert ( std::is_trivially_copyable< Vector2<VT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.forward(sizeof(Vector2<VT>));
+ std::memcpy(pos, &vec, sizeof(Vector2<VT>));
return buf;
}
@@ -1618,7 +1622,11 @@ namespace mpi {
mpi::GenericRecvBuffer<T>& operator>>( mpi::GenericRecvBuffer<T> & buf, Vector2<VT> & vec )
{
buf.readDebugMarker( "v2" );
- buf >> vec[0] >> vec[1] ;
+ static_assert ( std::is_trivially_copyable< Vector2<VT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.skip(sizeof(Vector2<VT>));
+ //suppress https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
+ std::memcpy(static_cast<void*>(&vec), pos, sizeof(Vector2<VT>));
return buf;
}
diff --git a/src/core/math/Vector3.h b/src/core/math/Vector3.h
index c5ec3e7b8ccb7840fc59905c7d6eaa081661f4f2..267e90c38f7b9752820a8397d7729c36134e3c96 100644
--- a/src/core/math/Vector3.h
+++ b/src/core/math/Vector3.h
@@ -166,6 +166,7 @@ public:
inline Vector3<Length> getNormalizedOrZero() const;
inline void reset();
inline Type* data() {return v_;}
+ inline Type const * data() const {return v_;}
//@}
//*******************************************************************************************************************
@@ -1864,7 +1865,10 @@ namespace mpi {
mpi::GenericSendBuffer<T,G>& operator<<( mpi::GenericSendBuffer<T,G> & buf, const Vector3<VT> & vec )
{
buf.addDebugMarker( "v3" );
- buf << vec[0] << vec[1] << vec[2];
+ static_assert ( std::is_trivially_copyable< Vector3<VT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.forward(sizeof(Vector3<VT>));
+ std::memcpy(pos, &vec, sizeof(Vector3<VT>));
return buf;
}
@@ -1873,7 +1877,11 @@ namespace mpi {
mpi::GenericRecvBuffer<T>& operator>>( mpi::GenericRecvBuffer<T> & buf, Vector3<VT> & vec )
{
buf.readDebugMarker( "v3" );
- buf >> vec[0] >> vec[1] >> vec[2];
+ static_assert ( std::is_trivially_copyable< Vector3<VT> >::value,
+ "type has to be trivially copyable for the memcpy to work correctly" );
+ auto pos = buf.skip(sizeof(Vector3<VT>));
+ //suppress https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
+ std::memcpy(static_cast<void*>(&vec), pos, sizeof(Vector3<VT>));
return buf;
}
diff --git a/src/core/mpi/BufferSystem.impl.h b/src/core/mpi/BufferSystem.impl.h
index 8a4185dd5de991eb85137ed5855ea53104773c1c..f3d7539865e2d8894cda86492f67463006fe0513 100644
--- a/src/core/mpi/BufferSystem.impl.h
+++ b/src/core/mpi/BufferSystem.impl.h
@@ -417,6 +417,11 @@ void GenericBufferSystem<Rb, Sb>::send( MPIRank rank )
template< typename Rb, typename Sb>
void GenericBufferSystem<Rb, Sb>::startCommunication()
{
+ // Clear receive buffers
+ for( auto iter = recvInfos_.begin(); iter != recvInfos_.end(); ++iter ) {
+ iter->second.buffer.clear();
+ }
+
const auto tag = currentComm_->getTag();
WALBERLA_CHECK_EQUAL(activeTags_.find(tag), activeTags_.end(),
"Another communication with the same MPI tag is currently in progress.");
@@ -457,11 +462,6 @@ void GenericBufferSystem<Rb, Sb>::endCommunication()
iter->second.buffer.clear();
}
- // Clear receive buffers
- for( auto iter = recvInfos_.begin(); iter != recvInfos_.end(); ++iter ) {
- iter->second.buffer.clear();
- }
-
if( !sizeChangesEverytime_ )
setCommunicationType( true );
diff --git a/src/mesa_pd/mpi/ReduceProperty.h b/src/mesa_pd/mpi/ReduceProperty.h
index 22ca078f13da4dbf19e0ad6b12b1c556b532326a..7f55c334973729778ca7dd87ada38ed175a522e7 100644
--- a/src/mesa_pd/mpi/ReduceProperty.h
+++ b/src/mesa_pd/mpi/ReduceProperty.h
@@ -29,6 +29,7 @@
#include <mesa_pd/data/DataTypes.h>
#include <mesa_pd/data/Flags.h>
#include <mesa_pd/data/ParticleStorage.h>
+#include <mesa_pd/mpi/notifications/reset.h>
#include <core/mpi/BufferSystem.h>
#include <core/logging/Logging.h>
@@ -100,6 +101,7 @@ void ReduceProperty::operator()(data::ParticleStorage& ps) const
}
sb << Notification( p );
+ reset<Notification>( p );
} else
{
//local particles should receive the property and sum it up
diff --git a/src/mesa_pd/mpi/notifications/ContactHistoryNotification.h b/src/mesa_pd/mpi/notifications/ContactHistoryNotification.h
index c306f6c8977d174203aaea6f1042731ccdf561c1..b177f313c61d1c522a116c83781109275f2c0c07 100644
--- a/src/mesa_pd/mpi/notifications/ContactHistoryNotification.h
+++ b/src/mesa_pd/mpi/notifications/ContactHistoryNotification.h
@@ -28,6 +28,7 @@
#include <mesa_pd/data/ContactHistory.h>
#include <mesa_pd/data/DataTypes.h>
+#include <mesa_pd/mpi/notifications/reset.h>
#include <core/mpi/BufferDataTypeExtensions.h>
#include <core/mpi/Datatype.h>
@@ -56,6 +57,12 @@ public:
const data::Particle& p_;
};
+template <>
+void reset<ContactHistoryNotification>(data::Particle& p)
+{
+ p.setNewContactHistory(std::map<walberla::id_t, walberla::mesa_pd::data::ContactHistory>());
+}
+
void reduce(data::Particle&& p, const ContactHistoryNotification::Parameters& objparam)
{
auto& ch = p.getNewContactHistoryRef();
diff --git a/src/mesa_pd/mpi/notifications/ForceTorqueNotification.h b/src/mesa_pd/mpi/notifications/ForceTorqueNotification.h
index c46f58724790f804b51b44fe966f5616efdaac76..d19d0d084635528194d048820d9d70785d03940a 100644
--- a/src/mesa_pd/mpi/notifications/ForceTorqueNotification.h
+++ b/src/mesa_pd/mpi/notifications/ForceTorqueNotification.h
@@ -29,6 +29,7 @@
#include <mesa_pd/data/DataTypes.h>
#include <mesa_pd/data/ParticleStorage.h>
#include <mesa_pd/mpi/notifications/NotificationType.h>
+#include <mesa_pd/mpi/notifications/reset.h>
#include <core/mpi/Datatype.h>
#include <core/mpi/RecvBuffer.h>
@@ -55,6 +56,13 @@ public:
const data::Particle& p_;
};
+template <>
+void reset<ForceTorqueNotification>(data::Particle& p)
+{
+ p.setForce( Vec3(real_t(0)) );
+ p.setTorque( Vec3(real_t(0)) );
+}
+
void reduce(data::Particle&& p, const ForceTorqueNotification::Parameters& objparam)
{
p.getForceRef() += objparam.force_;
diff --git a/src/mesa_pd/mpi/notifications/reset.h b/src/mesa_pd/mpi/notifications/reset.h
new file mode 100644
index 0000000000000000000000000000000000000000..24bab4533bacbfba9695fb4927136112da31b47e
--- /dev/null
+++ b/src/mesa_pd/mpi/notifications/reset.h
@@ -0,0 +1,35 @@
+//======================================================================================================================
+//
+// This file is part of waLBerla. waLBerla is free software: you can
+// redistribute it and/or modify it under the terms of the GNU General Public
+// License as published by the Free Software Foundation, either version 3 of
+// the License, or (at your option) any later version.
+//
+// waLBerla 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 General Public License
+// for more details.
+//
+// You should have received a copy of the GNU General Public License along
+// with waLBerla (see COPYING.txt). If not, see <http://www.gnu.org/licenses/>.
+//
+//! \file reset.h
+//! \author Sebastian Eibl <sebastian.eibl@fau.de>
+//
+//======================================================================================================================
+
+#pragma once
+
+#include <mesa_pd/data/ParticleStorage.h>
+
+namespace walberla {
+namespace mesa_pd {
+
+template <class Notification>
+void reset(data::Particle& /*p*/)
+{
+ WALBERLA_ABORT("not implemented!");
+}
+
+} // mpi
+} // walberla
diff --git a/tests/core/math/GenericAABBTest.cpp b/tests/core/math/GenericAABBTest.cpp
index ee6db55766e9528b00de3a7a6bb323ca0677ee3b..f1d386ade803ce23909bf47a5df4f6b1307618d2 100644
--- a/tests/core/math/GenericAABBTest.cpp
+++ b/tests/core/math/GenericAABBTest.cpp
@@ -102,7 +102,7 @@ void testNonEmptyAABB( const GenericAABB< T > & aabb )
WALBERLA_CHECK_EQUAL( tmpAABB, aabb.getIntersection( intersectingBox ) );
WALBERLA_CHECK_EQUAL( aabb.getIntersection( intersectingBox ), tmpAABB );
WALBERLA_CHECK_FLOAT_EQUAL( tmpAABB.volume(), aabb.volume() / T(8) );
- WALBERLA_CHECK_IDENTICAL( tmpAABB.volume(), aabb.intersectionVolume( intersectingBox ) );
+ WALBERLA_CHECK_FLOAT_EQUAL( tmpAABB.volume(), aabb.intersectionVolume( intersectingBox ) );
intersectingBox.init( aabb.minCorner() + aabb.sizes(), aabb.maxCorner() + aabb.sizes() );
tmpAABB = aabb;
@@ -231,7 +231,7 @@ void testAnyAABB( const GenericAABB< T > & aabb )
WALBERLA_CHECK( aabb.intersectsClosedInterval( aabb ) );
WALBERLA_CHECK_EQUAL( aabb.getIntersection( aabb ), aabb );
- WALBERLA_CHECK_IDENTICAL( aabb.intersectionVolume( aabb ), aabb.volume() );
+ WALBERLA_CHECK_FLOAT_EQUAL( aabb.intersectionVolume( aabb ), aabb.volume() );
WALBERLA_CHECK( aabb.isIdentical( aabb ) );
WALBERLA_CHECK( aabb.isEqual( aabb ) );
diff --git a/tests/mesa_pd/mpi/ReduceProperty.cpp b/tests/mesa_pd/mpi/ReduceProperty.cpp
index 300fc5dc106f938b619b810ddba72e8a7738820f..50c5b77836e842f6ba88cb4695c0bccf87c5389d 100644
--- a/tests/mesa_pd/mpi/ReduceProperty.cpp
+++ b/tests/mesa_pd/mpi/ReduceProperty.cpp
@@ -95,7 +95,7 @@ void main( int argc, char ** argv )
WALBERLA_CHECK_FLOAT_EQUAL( pIt->getForce(), Vec3(real_t(28)) );
} else
{
- WALBERLA_CHECK_FLOAT_EQUAL( pIt->getForce(), Vec3(real_t(walberla::mpi::MPIManager::instance()->rank())) );
+ WALBERLA_CHECK_FLOAT_EQUAL( pIt->getForce(), Vec3(0) );
}
}