diff --git a/apps/gpu/LBM/BoundaryLayer/BoundaryLayer.cpp b/apps/gpu/LBM/BoundaryLayer/BoundaryLayer.cpp
index c57b474d9c4805528b9c2b9e925638a89d4467dd..aaefbadd67b531ebe4438d1ec790a4bc52a82d46 100644
--- a/apps/gpu/LBM/BoundaryLayer/BoundaryLayer.cpp
+++ b/apps/gpu/LBM/BoundaryLayer/BoundaryLayer.cpp
@@ -178,7 +178,7 @@ void multipleLevel(const std::string& configPath)
para->setDensityRatio( 1.0 );
bool useStreams = (nProcs > 1 ? true: false);
- useStreams=true;
+ // useStreams=true;
para->setUseStreams(useStreams);
para->setMainKernel("CumulantK17Almighty");
para->setIsBodyForce( config.getValue<bool>("bodyForce") );
diff --git a/src/gpu/VirtualFluids_GPU/Calculation/CollisisionStrategy.cpp b/src/gpu/VirtualFluids_GPU/Calculation/CollisisionStrategy.cpp
index 819119176f131d50a569dbf63c863d346149585c..c06be75d9f24175b1d69e391e7dfb5dbc0885988 100644
--- a/src/gpu/VirtualFluids_GPU/Calculation/CollisisionStrategy.cpp
+++ b/src/gpu/VirtualFluids_GPU/Calculation/CollisisionStrategy.cpp
@@ -41,8 +41,8 @@ void CollisionAndExchange_noStreams_indexKernel::operator()(UpdateGrid27 *update
//!
updateGrid->collisionUsingIndices( level, t,
para->getParD(level)->taggedFluidNodeIndices[CollisionTemplate::Default],
- para->getParD(level)->numberOfTaggedFluidNodes[CollisionTemplate::Default],
- CollisionTemplate::Default,
+ para->getParD(level)->numberOfTaggedFluidNodes[CollisionTemplate::Default],
+ CollisionTemplate::Default,
CudaStreamIndex::Legacy);
//! 2. exchange information between GPUs
diff --git a/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.cpp b/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.cpp
index ae72af41f1fd7367f51c9f0b5059c26fe01b09e8..2b56160bca003fca66aba4399d0adb28bcb569fd 100644
--- a/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.cpp
+++ b/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.cpp
@@ -71,10 +71,10 @@ void UpdateGrid27::collisionAllNodes(int level, unsigned int t)
collisionAdvectionDiffusion(level);
}
-void UpdateGrid27::collisionUsingIndices(int level, unsigned int t, uint *indices, uint numberOfIndices, CollisionTemplate collisionTemplate, CudaStreamIndex stream)
+void UpdateGrid27::collisionUsingIndices(int level, unsigned int t, uint *taggedFluidNodeIndices, uint numberOfTaggedFluidNodes, CollisionTemplate collisionTemplate, CudaStreamIndex stream)
{
- if (indices != nullptr && numberOfIndices != 0)
- kernels.at(level)->runOnIndices(indices, numberOfIndices, collisionTemplate, stream);
+ if (taggedFluidNodeIndices != nullptr && numberOfTaggedFluidNodes != 0)
+ kernels.at(level)->runOnIndices(taggedFluidNodeIndices, numberOfTaggedFluidNodes, collisionTemplate, stream);
else
std::cout << "In collision: fluidNodeIndices or numberOfFluidNodes not definded"
<< std::endl;
diff --git a/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.h b/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.h
index 6f5db02aee211021994fea4257e4b19feaa56c48..4a25821672a86e9368ebacc3c8012e69267a9afe 100644
--- a/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.h
+++ b/src/gpu/VirtualFluids_GPU/Calculation/UpdateGrid27.h
@@ -31,7 +31,7 @@ public:
private:
void collisionAllNodes(int level, unsigned int t);
- void collisionUsingIndices(int level, unsigned int t, uint *indices, uint numberOfIndices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIndex=CudaStreamIndex::Legacy);
+ void collisionUsingIndices(int level, unsigned int t, uint *taggedFluidNodeIndices = nullptr, uint numberOfTaggedFluidNodes = 0, CollisionTemplate collisionTemplate = CollisionTemplate::Default, CudaStreamIndex streamIndex=CudaStreamIndex::Legacy);
void collisionAdvectionDiffusion(int level);
void postCollisionBC(int level);
diff --git a/src/gpu/VirtualFluids_GPU/Kernel/Kernel.h b/src/gpu/VirtualFluids_GPU/Kernel/Kernel.h
index 94b42728ba6fb6a942653bd056a48a3fa24319c1..7eaf0050b41a0aa67de0a4e592b8a123d9ef3e9f 100644
--- a/src/gpu/VirtualFluids_GPU/Kernel/Kernel.h
+++ b/src/gpu/VirtualFluids_GPU/Kernel/Kernel.h
@@ -16,7 +16,7 @@ class Kernel
public:
virtual ~Kernel() = default;
virtual void run() = 0;
- virtual void runOnIndices(const unsigned int *indices, unsigned int size_indices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIdx=CudaStreamIndex::Legacy) = 0;
+ virtual void runOnIndices(const unsigned int *indices, unsigned int size_indices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIdx=CudaStreamIndex::Legacy) = 0; //if stream == -1: run on default stream
virtual bool checkParameter() = 0;
virtual std::vector<PreProcessorType> getPreProcessorTypes() = 0;
diff --git a/src/gpu/VirtualFluids_GPU/Kernel/KernelImp.h b/src/gpu/VirtualFluids_GPU/Kernel/KernelImp.h
index d7ebb3eee35678cca64cf55a34524740a0ca7e7f..a96c2c123472ca33f635273e06a5bf36a745654d 100644
--- a/src/gpu/VirtualFluids_GPU/Kernel/KernelImp.h
+++ b/src/gpu/VirtualFluids_GPU/Kernel/KernelImp.h
@@ -16,7 +16,7 @@ class KernelImp : public Kernel
{
public:
virtual void run() = 0;
- virtual void runOnIndices(const unsigned int *indices, unsigned int size_indices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIndex);
+ virtual void runOnIndices(const unsigned int *indices, unsigned int size_indices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIndex=CudaStreamIndex::Legacy);
bool checkParameter();
std::vector<PreProcessorType> getPreProcessorTypes();
diff --git a/src/gpu/VirtualFluids_GPU/Kernel/Kernels/BasicKernels/FluidFlow/Compressible/CumulantK17Almighty/CumulantK17Almighty.cu b/src/gpu/VirtualFluids_GPU/Kernel/Kernels/BasicKernels/FluidFlow/Compressible/CumulantK17Almighty/CumulantK17Almighty.cu
index 8ddaa6381d65707936b580509bfd277b3f990bda..bb1b36ad8e14129a379fdce505345904f7b1f8ce 100644
--- a/src/gpu/VirtualFluids_GPU/Kernel/Kernels/BasicKernels/FluidFlow/Compressible/CumulantK17Almighty/CumulantK17Almighty.cu
+++ b/src/gpu/VirtualFluids_GPU/Kernel/Kernels/BasicKernels/FluidFlow/Compressible/CumulantK17Almighty/CumulantK17Almighty.cu
@@ -37,8 +37,10 @@ void CumulantK17Almighty<turbulenceModel>::run()
}
template<TurbulenceModel turbulenceModel>
-void CumulantK17Almighty<turbulenceModel>::runOnIndices( const unsigned int *indices, unsigned int size_indices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIndex)
+void CumulantK17Almighty<turbulenceModel>::runOnIndices( const unsigned int *indices, unsigned int size_indices, CollisionTemplate collisionTemplate, CudaStreamIndex streamIndex )
{
+ cudaStream_t stream = para->getStreamManager()->getStream(streamIndex);
+
switch (collisionTemplate)
{
case CollisionTemplate::Default:
diff --git a/src/gpu/VirtualFluids_GPU/LBM/Simulation.cpp b/src/gpu/VirtualFluids_GPU/LBM/Simulation.cpp
index a98d290997a1c20afc30b66eec68988eb10fa995..68c4555f57b97133e05b2bd0e7692ff95ea70068 100644
--- a/src/gpu/VirtualFluids_GPU/LBM/Simulation.cpp
+++ b/src/gpu/VirtualFluids_GPU/LBM/Simulation.cpp
@@ -117,6 +117,10 @@ void Simulation::init(GridProvider &gridProvider, BoundaryConditionFactory *bcFa
para->getStreamManager()->launchStreams();
para->getStreamManager()->createCudaEvents();
}
+ std::cout << "OKDIDOKI 1" << std::endl;
+ auto test = para->getStreamManager()->getStream(CudaStreamIndex::Legacy);
+ if(test == CU_STREAM_LEGACY)
+ std::cout << "OKDIDOKI 2" << std::endl;
//////////////////////////////////////////////////////////////////////////
VF_LOG_INFO("LB_Modell: D3Q{}", para->getD3Qxx());
VF_LOG_INFO("Re: {}", para->getRe());
diff --git a/src/gpu/VirtualFluids_GPU/Parameter/Parameter.cpp b/src/gpu/VirtualFluids_GPU/Parameter/Parameter.cpp
index eea23b991a529da60cd4dea021c7b0fbb3d41eb0..b8d5064bbe59ad8bbc37d65fa47e553c92b3fbfa 100644
--- a/src/gpu/VirtualFluids_GPU/Parameter/Parameter.cpp
+++ b/src/gpu/VirtualFluids_GPU/Parameter/Parameter.cpp
@@ -65,6 +65,8 @@ Parameter::Parameter(int numberOfProcesses, int myId, std::optional<const vf::ba
initGridPaths();
initGridBasePoints();
initDefaultLBMkernelAllLevels();
+
+ this->cudaStreamManager = std::make_unique<CudaStreamManager>();
}
Parameter::~Parameter() = default;
@@ -2655,7 +2657,6 @@ void Parameter::setUseStreams(bool useStreams)
if (useStreams) {
if (this->getNumprocs() != 1) {
this->useStreams = useStreams;
- this->cudaStreamManager = std::make_unique<CudaStreamManager>();
return;
} else {
std::cout << "Can't use streams with only one process!" << std::endl;