Acoustic-main.cpp File Reference

#include "Acoustic-main.h"
#include <fenv.h>
#include "config.h"
#include "Constants.h"
#include "exahype2/UserInterface.h"
#include "observers/CreateGrid.h"
#include "observers/CreateGridAndConvergeLoadBalancing.h"
#include "observers/CreateGridButPostponeRefinement.h"
#include "observers/InitGrid.h"
#include "observers/PlotSolution.h"
#include "observers/TimeStep.h"
#include "peano4/peano.h"
#include "repositories/DataRepository.h"
#include "repositories/SolverRepository.h"
#include "repositories/StepRepository.h"
#include "tarch/accelerator/accelerator.h"
#include "tarch/accelerator/Device.h"
#include "tarch/logging/CommandLineLogger.h"
#include "tarch/logging/Log.h"
#include "tarch/logging/LogFilter.h"
#include "tarch/logging/Statistics.h"
#include "tarch/multicore/Core.h"
#include "tarch/multicore/multicore.h"
#include "tarch/NonCriticalAssertions.h"
#include "tarch/timing/Measurement.h"
#include "tarch/timing/Watch.h"
#include "tasks/AcousticEnclaveTask.h"
#include "toolbox/loadbalancing/loadbalancing.h"

Include dependency graph for Acoustic-main.cpp:

Go to the source code of this file.

Functions
void	initInputData (double *Q)
	Set input data.
void	storeOutcome (const double *Q, const double maxEigenvalue)
	Store outcome of one compute kernel.
std::tuple< double, int >	validateOutcome (double *Q, int patchIndex, const double maxEigenvalue)
	Validate data against pre-stored simulation outcome.
void	reportRuntime (const std::string &kernelIdentificator, const ::tarch::timing::Measurement &timingKernelLaunch, int patches)
template<class TempDataEnumerator , ::peano4::utils::LoopPlacement loopParallelism>
void	wrapPatchwiseHeapFunctorsHostKernel (int device, ::exahype2::CellData< double, double > &patchData)
	This is a wrapper around the kernel call with the functors.
template<class TempDataEnumerator , ::peano4::utils::LoopPlacement loopParallelism>
void	wrapBatchedHeapFunctorHostKernels (int device, ::exahype2::CellData< double, double > &patchData)
	Another wrapper.
template<class TempDataEnumerator , ::peano4::utils::LoopPlacement loopParallelism>
void	wrapVolumewiseFunctorHostKernels (int device, ::exahype2::CellData< double, double > &patchData)
	Another wrapper.
template<void(*)(::exahype2::CellData< double, double > &, ::tarch::timing::Measurement &, ::peano4::utils::LoopPlacement) Function, ::peano4::utils::LoopPlacement loopParallelism>
void	wrapStatelessHostKernel (int device, ::exahype2::CellData< double, double > &patchData)
	Wrapper around stateless kernel invocations.
template<void(*)(int, ::exahype2::CellData< double, double > &, ::tarch::timing::Measurement &) Function>
void	wrapDeviceKernel (int device, ::exahype2::CellData< double, double > &patchData)
void	runBenchmarks (int numberOfPatches, int launchingThreads)
	Run the benchmark for one particular number of patches.
int	main (int argc, char **argv)
	Main routine of the SPH code.

Variables
::tarch::logging::Log	_log ("::")
double	validMaxEigenvalue = 0.0
double *	validOutcome = nullptr
static constexpr double	TimeStamp = 0.5
static constexpr double	TimeStepSize = 1e-6
static constexpr double	CellSize = 0.1
static constexpr double	CellOffset = 4.0
static constexpr int	HaloSize = 1
::tarch::timing::Measurement	timingComputeKernel
::tarch::multicore::BooleanSemaphore	validateOutcomeSemaphore
static constexpr int	NumberOfInputEntries
static constexpr int	NumberOfOutputEntries
static constexpr int	NumberOfFiniteVolumesPerPatch

Function Documentation

initInputData()

void initInputData

(

double *

Q

)

Set input data.

We really don't care if this makes any sense. Just make the data reasonably smooth and ensure that all is positive, as there are density and energy among the unknowns. If they become negative or zero, the compute kernels do not make any sense anymore.

Definition at line 89 of file Acoustic-main.cpp.

References NumberOfInputEntries, and tarch::la::PI.

Referenced by runBenchmarks().

Here is the caller graph for this function:

main()

int main	(	int	argc,
		char **	argv )

Main routine of the SPH code.

Definition at line 2162 of file Acoustic-main.cpp.

References peano4::fillLookupTables(), tarch::mpi::Rank::getGlobalMasterRank(), tarch::accelerator::Device::getInstance(), tarch::mpi::Rank::getInstance(), tarch::multicore::Core::getInstance(), tarch::initNonCriticalAssertionEnvironment(), peano4::initParallelEnvironment(), tarch::multicore::initSmartMPI(), logInfo, runBenchmarks(), tarch::shutdownNonCriticalAssertionEnvironment(), peano4::shutdownParallelEnvironment(), tarch::multicore::shutdownSmartMPI(), toString(), and validOutcome.

Here is the call graph for this function:

reportRuntime()

void reportRuntime	(	const std::string &	kernelIdentificator,
		const ::tarch::timing::Measurement &	timingKernelLaunch,
		int	patches )

Definition at line 161 of file Acoustic-main.cpp.

References tarch::timing::Measurement::getValue(), logInfo, NumberOfFiniteVolumesPerPatch, timingComputeKernel, and tarch::timing::Measurement::toString().

Referenced by runBenchmarks().

Here is the call graph for this function:

Here is the caller graph for this function:

runBenchmarks()

void runBenchmarks	(	int	numberOfPatches,
		int	launchingThreads )

Run the benchmark for one particular number of patches.

This operation benchmarks exclusively the host performance. This happens in two steps: We first assess the baseline performance, i.e., the serial kernel call, and then we look if the multithreaded stateless optimisation pays off and what speedup we obtain.

These two steps may run on multiple threads to mimic the spacetree tasking and domain decomposition. In OpenMP for example, there is one parallel loop around the two steps which invokes the same benchmark per loop entry. However, only the first thread always will report its runtime. If you are not bandwidth-bound, the data you obtain should thus be independent of the number of threads used.

Parameters

numberOfPatches

Number of patches to study

Definition at line 482 of file Acoustic-main.cpp.

References assessKernel(), exahype2::CellData< inType, outType >::cellCentre, CellOffset, CellSize, exahype2::CellData< inType, outType >::cellSize, exahype2::CellData< inType, outType >::dt, endParallelFor, tarch::timing::Measurement::erase(), tarch::freeMemory(), tarch::timing::Watch::getCalendarTime(), tarch::mpi::Rank::getGlobalMasterRank(), tarch::mpi::Rank::getInstance(), tarch::mpi::Rank::getNumberOfRanks(), tarch::mpi::Rank::getRank(), HaloSize, tarch::accelerator::Device::HostDevice, initInputData(), logError, logInfo, tarch::ManagedSharedAcceleratorDeviceMemory, exahype2::CellData< inType, outType >::maxEigenvalue, peano4::utils::Nested, NumberOfInputEntries, NumberOfOutputEntries, parallelFor, exahype2::CellData< inType, outType >::QIn, exahype2::CellData< inType, outType >::QOut, reportRuntime(), peano4::utils::Serial, peano4::grid::GridTraversalEvent::setH(), tarch::timing::Measurement::setValue(), peano4::grid::GridTraversalEvent::setX(), peano4::utils::SpreadOut, tarch::timing::Watch::stop(), storeOutcome(), exahype2::CellData< inType, outType >::t, TimeStamp, TimeStepSize, exahype2::fv::rusanov::timeStepWithRusanovBatchedHeapStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovBatchedHeapStateless(), exahype2::fv::rusanov::omp::timeStepWithRusanovBatchedHeapStateless(), exahype2::fv::rusanov::timeStepWithRusanovBatchedInsituStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovBatchedManagedStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovBatchedUSMStateless(), exahype2::fv::rusanov::cpp::timeStepWithRusanovBatchedUSMStateless(), exahype2::fv::rusanov::omp::timeStepWithRusanovBatchedUSMStateless(), exahype2::fv::rusanov::timeStepWithRusanovPatchwiseHeapStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovPatchwiseHeapStateless(), exahype2::fv::rusanov::omp::timeStepWithRusanovPatchwiseHeapStateless(), exahype2::fv::rusanov::timeStepWithRusanovPatchwiseInsituStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovPatchwiseUSMStateless(), exahype2::fv::rusanov::cpp::timeStepWithRusanovPatchwiseUSMStateless(), exahype2::fv::rusanov::omp::timeStepWithRusanovPatchwiseUSMStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovTaskgraphCopyStateless(), exahype2::fv::rusanov::sycl::timeStepWithRusanovTaskgraphUSMStateless(), exahype2::fv::rusanov::timeStepWithRusanovVolumewiseStateless(), exahype2::fv::rusanov::omp::timeStepWithRusanovVolumewiseStateless(), timingComputeKernel, validateOutcome(), void, wrapBatchedHeapFunctorHostKernels(), wrapDeviceKernel(), wrapPatchwiseHeapFunctorsHostKernel(), wrapStatelessHostKernel(), and wrapVolumewiseFunctorHostKernels().

Referenced by main().

Here is the call graph for this function:

Here is the caller graph for this function:

storeOutcome()

void storeOutcome	(	const double *	Q,
		const double	maxEigenvalue )

Store outcome of one compute kernel.

Make a persistent snapshot of a solution and assume, from hereon, that this snapshot is the valid data. You can call this routine as often as you want. Only the very first call will trigger a snapshot.

Definition at line 103 of file Acoustic-main.cpp.

References logInfo, applications::exahype2::acoustic::maxEigenvalue(), NumberOfOutputEntries, validateOutcomeSemaphore, validMaxEigenvalue, and validOutcome.

Referenced by runBenchmarks().

Here is the call graph for this function:

Here is the caller graph for this function:

validateOutcome()

std::tuple< double, int > validateOutcome	(	double *	Q,
		int	patchIndex,
		const double	maxEigenvalue )

Validate data against pre-stored simulation outcome.

Works if and only if storeOutcome has been performed before.

Returns

Tuple of maximum difference and total number of errors.

Definition at line 121 of file Acoustic-main.cpp.

References std::abs(), tarch::la::equals(), logError, applications::exahype2::acoustic::maxEigenvalue(), NumberOfOutputEntries, validateOutcomeSemaphore, validMaxEigenvalue, and validOutcome.

Referenced by runBenchmarks().

Here is the call graph for this function:

Here is the caller graph for this function:

wrapBatchedHeapFunctorHostKernels()

template<class TempDataEnumerator , ::peano4::utils::LoopPlacement loopParallelism>

void wrapBatchedHeapFunctorHostKernels	(	int	device,
		::exahype2::CellData< double, double > &	patchData )

Another wrapper.

See also

wrapPatchwiseHeapFunctorsHostKernel()

Definition at line 285 of file Acoustic-main.cpp.

References assertionEquals, HaloSize, tarch::accelerator::Device::HostDevice, exahype2::CellData< inType, outType >::maxEigenvalue, exahype2::fv::rusanov::timeStepWithRusanovBatchedHeapFunctors(), and timingComputeKernel.

Referenced by runBenchmarks().

Here is the call graph for this function:

Here is the caller graph for this function:

wrapDeviceKernel()

template<void(*)(int, ::exahype2::CellData< double, double > &, ::tarch::timing::Measurement &) Function>

void wrapDeviceKernel	(	int	device,
		::exahype2::CellData< double, double > &	patchData )

Definition at line 459 of file Acoustic-main.cpp.

References assertion, tarch::accelerator::Device::HostDevice, and timingComputeKernel.

Referenced by runBenchmarks().

Here is the caller graph for this function:

wrapPatchwiseHeapFunctorsHostKernel()

template<class TempDataEnumerator , ::peano4::utils::LoopPlacement loopParallelism>

void wrapPatchwiseHeapFunctorsHostKernel	(	int	device,
		::exahype2::CellData< double, double > &	patchData )

This is a wrapper around the kernel call with the functors.

We want to use all kernels exactly the same way. However, the various kernels all have slightly different signatures. So we write small helper functions (wrappers) which map the generic test signature onto the specific kernel.

To make this possible, all parameters which are not part of the generic interface, i.e., which are not patch data or the boolean, have to be mapped onto template arguments.

Here, we use the functor-based generic kernel implementation.

Definition at line 204 of file Acoustic-main.cpp.

References assertionEquals, HaloSize, tarch::accelerator::Device::HostDevice, exahype2::CellData< inType, outType >::maxEigenvalue, exahype2::fv::rusanov::timeStepWithRusanovPatchwiseHeapFunctors(), and timingComputeKernel.

Referenced by runBenchmarks().

Here is the call graph for this function:

Here is the caller graph for this function:

wrapStatelessHostKernel()

template<void(*)(::exahype2::CellData< double, double > &, ::tarch::timing::Measurement &, ::peano4::utils::LoopPlacement) Function, ::peano4::utils::LoopPlacement loopParallelism>

void wrapStatelessHostKernel	(	int	device,
		::exahype2::CellData< double, double > &	patchData )

Wrapper around stateless kernel invocations.

See also

wrapPatchwiseHeapFunctorsHostKernel()

Definition at line 451 of file Acoustic-main.cpp.

References assertionEquals, tarch::accelerator::Device::HostDevice, and timingComputeKernel.

Referenced by runBenchmarks().

Here is the caller graph for this function:

wrapVolumewiseFunctorHostKernels()

template<class TempDataEnumerator , ::peano4::utils::LoopPlacement loopParallelism>

void wrapVolumewiseFunctorHostKernels	(	int	device,
		::exahype2::CellData< double, double > &	patchData )

Another wrapper.

See also

wrapPatchwiseHeapFunctorsHostKernel()

Definition at line 366 of file Acoustic-main.cpp.

References assertionEquals, HaloSize, tarch::accelerator::Device::HostDevice, exahype2::CellData< inType, outType >::maxEigenvalue, exahype2::fv::rusanov::timeStepWithRusanovVolumewiseFunctors(), and timingComputeKernel.

Referenced by runBenchmarks().

Here is the call graph for this function:

Here is the caller graph for this function:

Variable Documentation

_log

::tarch::logging::Log _log("::")

(

"::"

)

CellOffset

constexpr double CellOffset = 4.0

staticconstexpr

Definition at line 43 of file Acoustic-main.cpp.

Referenced by runBenchmarks().

CellSize

constexpr double CellSize = 0.1

staticconstexpr

Definition at line 42 of file Acoustic-main.cpp.

Referenced by runBenchmarks().

HaloSize

constexpr int HaloSize = 1

staticconstexpr

Definition at line 44 of file Acoustic-main.cpp.

Referenced by runBenchmarks(), exahype2::fv::rusanov::tests::CopyPatchTest::testCopyPatch(), exahype2::fv::rusanov::internal::timeStepWithRusanovBatchedFunctors(), exahype2::fv::rusanov::internal::timeStepWithRusanovBatchedStateless(), exahype2::fv::rusanov::cpp::internal::timeStepWithRusanovBatchedStateless(), exahype2::fv::rusanov::omp::internal::timeStepWithRusanovBatchedStateless(), exahype2::fv::rusanov::internal::timeStepWithRusanovPatchwiseFunctors(), exahype2::fv::rusanov::internal::timeStepWithRusanovPatchwiseStateless(), exahype2::fv::rusanov::cpp::internal::timeStepWithRusanovPatchwiseStateless(), exahype2::fv::rusanov::omp::internal::timeStepWithRusanovPatchwiseStateless(), wrapBatchedHeapFunctorHostKernels(), wrapPatchwiseHeapFunctorsHostKernel(), and wrapVolumewiseFunctorHostKernels().

NumberOfFiniteVolumesPerPatch

constexpr int NumberOfFiniteVolumesPerPatch

staticconstexpr

Initial value:

= Acoustic::NumberOfFiniteVolumesPerAxisPerPatch
    * Acoustic::NumberOfFiniteVolumesPerAxisPerPatch
    * Acoustic::NumberOfFiniteVolumesPerAxisPerPatch

Definition at line 74 of file Acoustic-main.cpp.

Referenced by reportRuntime().

NumberOfInputEntries

constexpr int NumberOfInputEntries

staticconstexpr

Initial value:

= (Acoustic::NumberOfFiniteVolumesPerAxisPerPatch + 2 * HaloSize)
    * (Acoustic::NumberOfFiniteVolumesPerAxisPerPatch + 2 * HaloSize)
    * (Acoustic::NumberOfFiniteVolumesPerAxisPerPatch + 2 * HaloSize)
    * (Acoustic::NumberOfUnknowns + Acoustic::NumberOfAuxiliaryVariables)

Definition at line 64 of file Acoustic-main.cpp.

Referenced by initInputData(), and runBenchmarks().

NumberOfOutputEntries

constexpr int NumberOfOutputEntries

staticconstexpr

Initial value:

= (Acoustic::NumberOfFiniteVolumesPerAxisPerPatch + 0)
    * (Acoustic::NumberOfFiniteVolumesPerAxisPerPatch + 0)
    * (Acoustic::NumberOfFiniteVolumesPerAxisPerPatch + 0)
    * (Acoustic::NumberOfUnknowns + Acoustic::NumberOfAuxiliaryVariables)

Definition at line 69 of file Acoustic-main.cpp.

Referenced by runBenchmarks(), storeOutcome(), and validateOutcome().

TimeStamp

constexpr double TimeStamp = 0.5

staticconstexpr

Definition at line 40 of file Acoustic-main.cpp.

Referenced by runBenchmarks().

TimeStepSize

constexpr double TimeStepSize = 1e-6

staticconstexpr

Definition at line 41 of file Acoustic-main.cpp.

Referenced by runBenchmarks().

timingComputeKernel

::tarch::timing::Measurement timingComputeKernel

Definition at line 45 of file Acoustic-main.cpp.

Referenced by reportRuntime(), runBenchmarks(), wrapBatchedHeapFunctorHostKernels(), wrapDeviceKernel(), wrapPatchwiseHeapFunctorsHostKernel(), wrapStatelessHostKernel(), and wrapVolumewiseFunctorHostKernels().

validateOutcomeSemaphore

::tarch::multicore::BooleanSemaphore validateOutcomeSemaphore

Definition at line 46 of file Acoustic-main.cpp.

Referenced by storeOutcome(), and validateOutcome().

validMaxEigenvalue

double validMaxEigenvalue = 0.0

Definition at line 38 of file Acoustic-main.cpp.

Referenced by storeOutcome(), and validateOutcome().

validOutcome

double* validOutcome = nullptr

Definition at line 39 of file Acoustic-main.cpp.

Referenced by main(), storeOutcome(), and validateOutcome().