From a1f716d0e757ac8e33abeb7868e98757b33a0b2a Mon Sep 17 00:00:00 2001
From: Martin Bauer <martin.bauer@fau.de>
Date: Tue, 1 Oct 2019 08:25:52 +0200
Subject: [PATCH] New benchmark for generated LBM on uniform grid on CPU
---
apps/benchmarks/CMakeLists.txt | 1 +
.../simulation_setup/communication_compare.py | 196 ++++++++++++++++++
.../UniformGridGenerated/CMakeLists.txt | 12 ++
.../UniformGridGenerated.cpp | 157 ++++++++++++++
.../UniformGridGenerated.py | 133 ++++++++++++
5 files changed, 499 insertions(+)
create mode 100755 apps/benchmarks/UniformGridGPU/simulation_setup/communication_compare.py
create mode 100644 apps/benchmarks/UniformGridGenerated/CMakeLists.txt
create mode 100644 apps/benchmarks/UniformGridGenerated/UniformGridGenerated.cpp
create mode 100644 apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py
diff --git a/apps/benchmarks/CMakeLists.txt b/apps/benchmarks/CMakeLists.txt
index 8e83319a..a59d3bc1 100644
--- a/apps/benchmarks/CMakeLists.txt
+++ b/apps/benchmarks/CMakeLists.txt
@@ -14,3 +14,4 @@ add_subdirectory( ProbeVsExtraMessage )
add_subdirectory( SchaeferTurek )
add_subdirectory( UniformGrid )
add_subdirectory( UniformGridGPU )
+add_subdirectory( UniformGridGenerated )
diff --git a/apps/benchmarks/UniformGridGPU/simulation_setup/communication_compare.py b/apps/benchmarks/UniformGridGPU/simulation_setup/communication_compare.py
new file mode 100755
index 00000000..4a20b377
--- /dev/null
+++ b/apps/benchmarks/UniformGridGPU/simulation_setup/communication_compare.py
@@ -0,0 +1,196 @@
+#!/usr/bin/env python3
+
+import os
+import pandas as pd
+import waLBerla as wlb
+from waLBerla.tools.config import block_decomposition
+from waLBerla.tools.sqlitedb import sequenceValuesToScalars
+from copy import deepcopy
+import sys
+
+CSV_FILE = "new.csv"
+
+BASE_CONFIG = {
+ 'DomainSetup': {
+ 'cellsPerBlock': (256, 128, 128),
+ 'periodic': (1, 1, 1),
+ },
+ 'Parameters': {
+ 'omega': 1.8,
+ 'timesteps': 400,
+ 'cudaEnabledMPI': False,
+ 'warmupSteps': 5,
+ 'outerIterations': 3,
+ 'initShearFlow': True,
+ }
+}
+
+
+class Scenario:
+ def __init__(self, cells_per_block=(256, 128, 128), **kwargs):
+ self.config_dict = deepcopy(BASE_CONFIG)
+ self.config_dict['Parameters'].update(kwargs)
+ self.config_dict['DomainSetup']['blocks'] = block_decomposition(wlb.mpi.numProcesses())
+ self.config_dict['DomainSetup']['cellsPerBlock'] = cells_per_block
+
+ @wlb.member_callback
+ def config(self, **kwargs):
+ from pprint import pformat
+ wlb.log_info_on_root("Scenario:\n" + pformat(self.config_dict))
+ return self.config_dict
+
+ @wlb.member_callback
+ def results_callback(self, **kwargs):
+ data = {}
+ data.update(self.config_dict['Parameters'])
+ data.update(self.config_dict['DomainSetup'])
+ data.update(kwargs)
+ data['executable'] = sys.argv[0]
+ data['compile_flags'] = wlb.build_info.compiler_flags
+ data['walberla_version'] = wlb.build_info.version
+ data['build_machine'] = wlb.build_info.build_machine
+ sequenceValuesToScalars(data)
+
+ df = pd.DataFrame.from_records([data])
+ if not os.path.isfile(CSV_FILE):
+ df.to_csv(CSV_FILE, index=False)
+ else:
+ df.to_csv(CSV_FILE, index=False, mode='a', header=False)
+
+
+def overlap_benchmark():
+ scenarios = wlb.ScenarioManager()
+ inner_outer_splits = [(1, 1, 1), (4, 1, 1), (8, 1, 1), (16, 1, 1), (32, 1, 1),
+ (4, 4, 1), (8, 8, 1), (16, 16, 1), (32, 32, 1),
+ (4, 4, 4), (8, 8, 8), (16, 16, 16), (32, 32, 32)]
+
+ for comm_strategy in ['UniformGPUScheme_Baseline', 'UniformGPUScheme_Memcpy']: # 'GPUPackInfo_Baseline', 'GPUPackInfo_Streams'
+ # no overlap
+ scenarios.add(Scenario(timeStepStrategy='noOverlap', communicationScheme=comm_strategy, innerOuterSplit=(1, 1, 1)))
+
+ # overlap
+ for overlap_strategy in ['simpleOverlap', 'complexOverlap']:
+ for inner_outer_split in inner_outer_splits:
+ scenario = Scenario(timeStepStrategy=overlap_strategy,
+ communicationScheme=comm_strategy,
+ innerOuterSplit=inner_outer_split)
+ scenarios.add(scenario)
+
+
+def communication_compare():
+ scenarios = wlb.ScenarioManager()
+ for block_size in [(128, 128, 128), (32, 32, 32), (64, 64, 64), (256, 256, 256)]:
+ for comm_strategy in ['UniformGPUScheme_Baseline', 'UniformGPUScheme_Memcpy']:
+ sc = Scenario(cells_per_block=block_size,
+ gpuBlockSize=(128, 1, 1),
+ timeStepStrategy='noOverlap',
+ communicationScheme=comm_strategy)
+ scenarios.add(sc)
+ for inner_outer_split in [(4, 1, 1), (8, 1, 1), (16, 1, 1), (32, 1, 1)]:
+ if 3 * inner_outer_split[0] < block_size[0]:
+ continue
+ sc = Scenario(cells_per_block=block_size,
+ gpuBlockSize=(128, 1, 1),
+ timeStepStrategy='simpleOverlap',
+ innerOuterSplit=inner_outer_split,
+ communicationScheme=comm_strategy)
+ scenarios.add(sc)
+
+
+def single_gpu_benchmark():
+ scenarios = wlb.ScenarioManager()
+ block_sizes = [(i, i, i) for i in (64, 128, 256, 384)] + [(512, 512, 128)]
+ cuda_blocks = [(32, 1, 1), (64, 1, 1), (128, 1, 1), (256, 1, 1), (512, 1, 1),
+ (32, 2, 1), (64, 2, 1), (128, 2, 1), (256, 2, 1),
+ (32, 4, 1), (64, 4, 1), (128, 4, 1),
+ (32, 8, 1), (64, 8, 1),
+ (32, 16, 1)]
+ for block_size in block_sizes:
+ for cuda_block_size in cuda_blocks:
+ cells = block_size[0] * block_size[1] * block_size[2]
+ time_steps_for_128_cubed = 1000
+ time_steps = (128 ** 3 / cells) * time_steps_for_128_cubed
+ scenario = Scenario(cells_per_block=block_size,
+ gpuBlockSize=cuda_block_size,
+ timeStepStrategy='kernelOnly',
+ timesteps=int(time_steps))
+ scenarios.add(scenario)
+
+
+job_script_header = """
+#!/bin/bash -l
+#SBATCH --job-name=scaling
+#SBATCH --time=0:30:00
+#SBATCH --nodes={nodes}
+#SBATCH -o out_scaling_{nodes}_%j.txt
+#SBATCH -e err_scaling_{nodes}_%j.txt
+#SBATCH --ntasks-per-core=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --cpus-per-task=1
+#SBATCH --partition=normal
+#SBATCH --constraint=gpu
+#SBATCH --account=d105
+
+cd {folder}
+
+source ~/env.sh
+
+module load daint-gpu
+module load craype-accel-nvidia60
+export MPICH_RDMA_ENABLED_CUDA=1 # allow GPU-GPU data transfer
+export CRAY_CUDA_MPS=1 # allow GPU sharing
+export MPICH_G2G_PIPELINE=256 # adapt maximum number of concurrent in-flight messages
+
+export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
+export CRAY_CUDA_MPS=1
+
+export MPICH_RANK_REORDER_METHOD=3
+export PMI_MMAP_SYNC_WAIT_TIME=300
+
+
+# grid_order -R -H -c 1,1,8 -g 16,16,8
+
+ulimit -c 0
+"""
+
+job_script_exe_part = """
+
+export WALBERLA_SCENARIO_IDX=0
+while srun -n {nodes} ./{app} {config}
+do
+ ((WALBERLA_SCENARIO_IDX++))
+done
+"""
+
+
+all_executables = ('UniformGridBenchmarkGPU_mrt_d3q27',
+ 'UniformGridBenchmarkGPU_smagorinsky_d3q27',
+ 'UniformGridBenchmarkGPU_cumulant'
+ 'UniformGridBenchmarkGPU_cumulant_d3q27',
+ )
+
+
+def generate_jobscripts(exe_names=all_executables):
+ for node_count in [1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 2400]:
+ folder_name = "scaling_{:04d}".format(node_count)
+ os.makedirs(folder_name, exist_ok=True)
+
+ # run grid_order
+ import subprocess
+ decomposition = block_decomposition(node_count)
+ decomposition_str = ",".join(str(e) for e in decomposition)
+ subprocess.check_call(['grid_order', '-R', '-H', '-g', decomposition_str])
+
+ job_script = job_script_header.format(nodes=node_count, folder=os.path.join(os.getcwd(), folder_name))
+ for exe in exe_names:
+ job_script += job_script_exe_part.format(app="../" + exe, nodes=node_count, config='../communication_compare.py')
+
+ with open(os.path.join(folder_name, 'job.sh'), 'w') as f:
+ f.write(job_script)
+
+
+if __name__ == '__main__':
+ print("Called without waLBerla - generating job scripts for PizDaint")
+ generate_jobscripts()
+else:
+ single_gpu_benchmark()
diff --git a/apps/benchmarks/UniformGridGenerated/CMakeLists.txt b/apps/benchmarks/UniformGridGenerated/CMakeLists.txt
new file mode 100644
index 00000000..def7d93f
--- /dev/null
+++ b/apps/benchmarks/UniformGridGenerated/CMakeLists.txt
@@ -0,0 +1,12 @@
+waLBerla_python_file_generates(UniformGridGenerated.py
+ UniformGridGenerated_LatticeModel.cpp
+ UniformGridGenerated_Defines.h)
+
+
+foreach(config trt )
+ waLBerla_add_executable ( NAME UniformGridBenchmarkGenerated_${config}
+ FILES UniformGridGenerated.cpp UniformGridGenerated.py
+ DEPENDS blockforest boundary core domain_decomposition field geometry timeloop vtk gui
+ CODEGEN_CFG ${config})
+endforeach()
+
diff --git a/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.cpp b/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.cpp
new file mode 100644
index 00000000..f48b0f75
--- /dev/null
+++ b/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.cpp
@@ -0,0 +1,157 @@
+#include "core/Environment.h"
+#include "core/logging/Initialization.h"
+#include "core/math/Random.h"
+#include "python_coupling/CreateConfig.h"
+#include "python_coupling/PythonCallback.h"
+#include "python_coupling/DictWrapper.h"
+#include "blockforest/Initialization.h"
+#include "field/vtk/VTKWriter.h"
+#include "field/communication/PackInfo.h"
+#include "blockforest/communication/UniformBufferedScheme.h"
+#include "timeloop/all.h"
+#include "core/timing/TimingPool.h"
+#include "core/timing/RemainingTimeLogger.h"
+#include "domain_decomposition/SharedSweep.h"
+#include "lbm/communication/PdfFieldPackInfo.h"
+#include "lbm/field/AddToStorage.h"
+#include "lbm/vtk/VTKOutput.h"
+#include "lbm/gui/Connection.h"
+#include "lbm/vtk/Velocity.h"
+#include "gui/Gui.h"
+
+#include "UniformGridGenerated_LatticeModel.h"
+#include "UniformGridGenerated_Defines.h"
+
+
+using namespace walberla;
+
+typedef lbm::UniformGridGenerated_LatticeModel LatticeModel_T;
+typedef LatticeModel_T::Stencil Stencil_T;
+typedef LatticeModel_T::CommunicationStencil CommunicationStencil_T;
+typedef lbm::PdfField< LatticeModel_T > PdfField_T;
+
+
+void initShearVelocity(const shared_ptr<StructuredBlockStorage> & blocks, BlockDataID pdfFieldId,
+ const real_t xMagnitude=0.1, const real_t fluctuationMagnitude=0.05 )
+{
+ math::seedRandomGenerator(0);
+ auto halfZ = blocks->getDomainCellBB().zMax() / 2;
+ for( auto & block: *blocks)
+ {
+ auto pdfField = block.getData<PdfField_T>( pdfFieldId );
+ WALBERLA_FOR_ALL_CELLS_INCLUDING_GHOST_LAYER_XYZ(pdfField,
+ Cell globalCell;
+ blocks->transformBlockLocalToGlobalCell(globalCell, block, Cell(x, y, z));
+ real_t randomReal = xMagnitude * math::realRandom<real_t>(-fluctuationMagnitude, fluctuationMagnitude);
+
+ if( globalCell[2] >= halfZ ) {
+ pdfField->setDensityAndVelocity(x, y, z, Vector3<real_t>(xMagnitude, 0, randomReal), real_t(1.0));
+ } else {
+ pdfField->setDensityAndVelocity(x, y, z, Vector3<real_t>(-xMagnitude, 0, randomReal), real_t(1.0));
+ }
+ );
+ }
+}
+
+
+int main( int argc, char **argv )
+{
+ mpi::Environment env( argc, argv );
+
+ for( auto cfg = python_coupling::configBegin( argc, argv ); cfg != python_coupling::configEnd(); ++cfg )
+ {
+ WALBERLA_MPI_WORLD_BARRIER();
+
+ auto config = *cfg;
+ logging::configureLogging( config );
+ auto blocks = blockforest::createUniformBlockGridFromConfig( config );
+
+ Vector3<uint_t> cellsPerBlock = config->getBlock( "DomainSetup" ).getParameter<Vector3<uint_t> >( "cellsPerBlock" );
+ // Reading parameters
+ auto parameters = config->getOneBlock( "Parameters" );
+ const std::string timeStepStrategy = parameters.getParameter<std::string>( "timeStepStrategy", "normal");
+ const real_t omega = parameters.getParameter<real_t>( "omega", real_c( 1.4 ));
+ const uint_t timesteps = parameters.getParameter<uint_t>( "timesteps", uint_c( 50 ));
+ const bool initShearFlow = parameters.getParameter<bool>("initShearFlow", false);
+
+ // Creating fields
+ LatticeModel_T latticeModel = LatticeModel_T( omega );
+ BlockDataID pdfFieldId = lbm::addPdfFieldToStorage( blocks, "pdf field", latticeModel);
+
+ if( initShearFlow ) {
+ initShearVelocity(blocks, pdfFieldId);
+ }
+
+ SweepTimeloop timeLoop( blocks->getBlockStorage(), timesteps );
+ blockforest::communication::UniformBufferedScheme< CommunicationStencil_T > communication( blocks );
+ communication.addPackInfo( make_shared< lbm::PdfFieldPackInfo< LatticeModel_T > >( pdfFieldId ) );
+
+ timeLoop.add() << BeforeFunction( communication, "communication" )
+ << Sweep( LatticeModel_T::Sweep( pdfFieldId ), "LB stream & collide" );
+
+ int warmupSteps = parameters.getParameter<int>( "warmupSteps", 2 );
+ int outerIterations = parameters.getParameter<int>( "outerIterations", 1 );
+ for(int i=0; i < warmupSteps; ++i )
+ timeLoop.singleStep();
+
+ auto remainingTimeLoggerFrequency = parameters.getParameter< double >( "remainingTimeLoggerFrequency", -1.0 ); // in seconds
+ if (remainingTimeLoggerFrequency > 0) {
+ auto logger = timing::RemainingTimeLogger( timeLoop.getNrOfTimeSteps() * outerIterations, remainingTimeLoggerFrequency );
+ timeLoop.addFuncAfterTimeStep( logger, "remaining time logger" );
+ }
+
+ // VTK
+ uint_t vtkWriteFrequency = parameters.getParameter<uint_t>( "vtkWriteFrequency", 0 );
+ if( vtkWriteFrequency > 0 )
+ {
+ auto vtkOutput = vtk::createVTKOutput_BlockData( *blocks, "vtk", vtkWriteFrequency, 0, false, "vtk_out",
+ "simulation_step", false, true, true, false, 0 );
+ auto velWriter = make_shared< lbm::VelocityVTKWriter<LatticeModel_T> >(pdfFieldId, "vel");
+ vtkOutput->addCellDataWriter(velWriter);
+ timeLoop.addFuncAfterTimeStep( vtk::writeFiles( vtkOutput ), "VTK Output" );
+ }
+
+
+ bool useGui = parameters.getParameter<bool>( "useGui", false );
+ if( useGui )
+ {
+ GUI gui( timeLoop, blocks, argc, argv);
+ lbm::connectToGui<LatticeModel_T>(gui);
+ gui.run();
+ }
+ else
+ {
+ for ( int outerIteration = 0; outerIteration < outerIterations; ++outerIteration )
+ {
+ timeLoop.setCurrentTimeStepToZero();
+ WcTimer simTimer;
+ WALBERLA_LOG_INFO_ON_ROOT( "Starting simulation with " << timesteps << " time steps" );
+ simTimer.start();
+ timeLoop.run();
+ simTimer.end();
+ WALBERLA_LOG_INFO_ON_ROOT( "Simulation finished" );
+ auto time = simTimer.last();
+ auto nrOfCells = real_c( cellsPerBlock[0] * cellsPerBlock[1] * cellsPerBlock[2] );
+ auto mlupsPerProcess = nrOfCells * real_c( timesteps ) / time * 1e-6;
+ WALBERLA_LOG_RESULT_ON_ROOT( "MLUPS per process " << mlupsPerProcess );
+ WALBERLA_LOG_RESULT_ON_ROOT( "Time per time step " << time / real_c( timesteps ));
+ WALBERLA_ROOT_SECTION()
+ {
+ python_coupling::PythonCallback pythonCallbackResults( "results_callback" );
+ if ( pythonCallbackResults.isCallable())
+ {
+ pythonCallbackResults.data().exposeValue( "mlupsPerProcess", mlupsPerProcess );
+ 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();
+ }
+ }
+ }
+ }
+ }
+
+ return 0;
+}
diff --git a/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py b/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py
new file mode 100644
index 00000000..8f7bf791
--- /dev/null
+++ b/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py
@@ -0,0 +1,133 @@
+import sympy as sp
+import pystencils as ps
+from lbmpy.creationfunctions import create_lb_collision_rule
+from lbmpy.fieldaccess import StreamPullTwoFieldsAccessor
+from lbmpy_walberla import generate_lattice_model
+from pystencils_walberla import CodeGeneration
+
+omega = sp.symbols("omega")
+omega_fill = sp.symbols("omega_:10")
+
+options_dict = {
+ 'srt': {
+ 'method': 'srt',
+ 'stencil': 'D3Q19',
+ 'relaxation_rate': omega,
+ 'compressible': False,
+ },
+ 'trt': {
+ 'method': 'trt',
+ 'stencil': 'D3Q19',
+ 'relaxation_rate': omega,
+ },
+ 'mrt': {
+ 'method': 'mrt',
+ 'stencil': 'D3Q19',
+ 'relaxation_rates': [0, omega, 1.3, 1.4, omega, 1.2, 1.1, 1.15, 1.234, 1.4235, 1.242, 1.2567, 0.9, 0.7],
+ },
+ 'mrt_full': {
+ 'method': 'mrt',
+ 'stencil': 'D3Q19',
+ 'relaxation_rates': [omega_fill[0], omega, omega_fill[1], omega_fill[2], omega_fill[3], omega_fill[4], omega_fill[5]],
+ },
+ 'mrt3': {
+ 'method': 'mrt3',
+ 'stencil': 'D3Q19',
+ 'relaxation_rates': [omega, 1.1, 1.2],
+ },
+ 'entropic': {
+ 'method': 'mrt3',
+ 'stencil': 'D3Q19',
+ 'compressible': True,
+ 'relaxation_rates': [omega, omega, sp.Symbol("omega_free")],
+ 'entropic': True,
+ },
+ 'entropic_kbc_n4': {
+ 'method': 'trt-kbc-n4',
+ 'stencil': 'D3Q27',
+ 'compressible': True,
+ 'relaxation_rates': [omega, sp.Symbol("omega_free")],
+ 'entropic': True,
+ },
+ 'smagorinsky': {
+ 'method': 'srt',
+ 'stencil': 'D3Q19',
+ 'smagorinsky': True,
+ 'relaxation_rate': omega,
+ },
+ 'cumulant': {
+ 'stencil': 'D3Q19',
+ 'compressible': True,
+ 'method': 'mrt',
+ 'cumulant': True,
+ 'relaxation_rates': [0, omega, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
+ },
+}
+
+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()
+ assert not accessor.is_inplace, "This app does not work for inplace accessors"
+
+ common_options = {
+ 'field_name': 'pdfs',
+ 'temporary_field_name': 'pdfs_tmp',
+ 'kernel_type': accessor,
+ 'optimization': {'cse_global': True,
+ 'cse_pdfs': False}
+ }
+ config_name = ctx.config
+ noopt = False
+ d3q27 = False
+ if config_name.endswith("_noopt"):
+ noopt = True
+ config_name = config_name[:-len("_noopt")]
+ if config_name.endswith("_d3q27"):
+ d3q27 = True
+ config_name = config_name[:-len("_d3q27")]
+
+ options = options_dict[config_name]
+ options.update(common_options)
+ options = options.copy()
+
+ if noopt:
+ options['optimization']['cse_global'] = False
+ options['optimization']['cse_pdfs'] = False
+ if d3q27:
+ options['stencil'] = 'D3Q27'
+
+ stencil_str = options['stencil']
+ q = int(stencil_str[stencil_str.find('Q')+1:])
+ pdfs, velocity_field = ps.fields("pdfs({q}), velocity(3) : double[3D]".format(q=q), layout='fzyx')
+ options['optimization']['symbolic_field'] = pdfs
+
+ vp = [
+ ('double', 'omega_0'),
+ ('double', 'omega_1'),
+ ('double', 'omega_2'),
+ ('double', 'omega_3'),
+ ('double', 'omega_4'),
+ ('double', 'omega_5'),
+ ('double', 'omega_6'),
+ ('int32_t', 'cudaBlockSize0'),
+ ('int32_t', 'cudaBlockSize1'),
+ ]
+ update_rule = create_lb_collision_rule(**options)
+ generate_lattice_model(ctx, 'UniformGridGenerated_LatticeModel', update_rule)
+
+ 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("UniformGridGenerated_Defines.h", info_header.format(**infoHeaderParams))
--
GitLab