From 67e53365441032070b260fc5454d648795059fd3 Mon Sep 17 00:00:00 2001
From: markus <markus.holzer@fau.de>
Date: Fri, 17 Jul 2020 11:59:03 +0200
Subject: [PATCH] Adapted flak8 formatting for apps
---
.flake8 | 2 +
apps/benchmarks/FieldCommunication/config.py | 3 +-
.../PhaseFieldAllenCahn/multiphase_codegen.py | 2 +-
.../PhaseFieldAllenCahn/profiling.py | 1 -
apps/benchmarks/UniformGrid/ecmModel.py | 87 +--
.../UniformGrid/mergeSqliteFiles.py | 114 ++--
apps/benchmarks/UniformGrid/plot.py | 524 +++++++++---------
.../UniformGridGPU/UniformGridGPU.py | 16 +-
.../UniformGridGPU/UniformGridGPU_AA.py | 8 +-
.../simulation_setup/benchmark_configs.py | 35 +-
.../UniformGridGenerated.py | 28 +-
.../benchmarks/UniformGridGenerated/params.py | 21 +-
apps/pythonmodule/setup.py | 49 +-
apps/pythonmodule/stage.py | 3 +-
.../CPU/multiphase_RTI_3D.py | 6 +-
.../CPU/multiphase_codegen.py | 6 +-
.../CPU/multiphase_rising_bubble.py | 1 +
.../GPU/multiphase_RTI_3D.py | 6 +-
.../GPU/multiphase_codegen.py | 7 +-
19 files changed, 478 insertions(+), 441 deletions(-)
diff --git a/.flake8 b/.flake8
index 8c53c3999..9bd296133 100644
--- a/.flake8
+++ b/.flake8
@@ -1,3 +1,5 @@
[flake8]
max-line-length=120
+exclude=apps/showcases/Mixer/GenerateModule.py, # contains only statements
+ apps/benchmarks/FieldCommunication/config.py # just a config file
ignore = W503 C901 E741
diff --git a/apps/benchmarks/FieldCommunication/config.py b/apps/benchmarks/FieldCommunication/config.py
index 56b480f15..154f0b90d 100644
--- a/apps/benchmarks/FieldCommunication/config.py
+++ b/apps/benchmarks/FieldCommunication/config.py
@@ -34,7 +34,8 @@ sng_network = supermuc_network_spread()
class AlreadySimulated:
- def __init__(self, db_file, properties=('processes0*processes1*processes2', 'layout', 'ghostLayers', 'cartesianCommunicator', 'stencil',
+ def __init__(self, db_file, properties=('processes0*processes1*processes2', 'layout', 'ghostLayers',
+ 'cartesianCommunicator', 'stencil',
'cellsPerBlock0', 'cellsPerBlock1', 'cellsPerBlock2',
'blocksPerProcess', 'localCommunicationMode', 'singleMessage',
'fieldsPdf', 'fieldsPdfOpt', 'fieldsVector', 'fieldsScalar',
diff --git a/apps/benchmarks/PhaseFieldAllenCahn/multiphase_codegen.py b/apps/benchmarks/PhaseFieldAllenCahn/multiphase_codegen.py
index d46056d82..5fd50cf57 100644
--- a/apps/benchmarks/PhaseFieldAllenCahn/multiphase_codegen.py
+++ b/apps/benchmarks/PhaseFieldAllenCahn/multiphase_codegen.py
@@ -147,7 +147,7 @@ sweep_block_size = (TypedSymbol("cudaBlockSize0", np.int32),
sweep_params = {'block_size': sweep_block_size}
info_header = f"""
-#include "stencil/D3Q{q_phase}.h"\nusing Stencil_phase_T = walberla::stencil::D3Q{q_phase};
+#include "stencil/D3Q{q_phase}.h"\nusing Stencil_phase_T = walberla::stencil::D3Q{q_phase};
#include "stencil/D3Q{q_hydro}.h"\nusing Stencil_hydro_T = walberla::stencil::D3Q{q_hydro};
"""
diff --git a/apps/benchmarks/PhaseFieldAllenCahn/profiling.py b/apps/benchmarks/PhaseFieldAllenCahn/profiling.py
index 4d8c9375e..71cdba3f3 100644
--- a/apps/benchmarks/PhaseFieldAllenCahn/profiling.py
+++ b/apps/benchmarks/PhaseFieldAllenCahn/profiling.py
@@ -1,4 +1,3 @@
-import os
import waLBerla as wlb
diff --git a/apps/benchmarks/UniformGrid/ecmModel.py b/apps/benchmarks/UniformGrid/ecmModel.py
index 73abc6ccf..0509ed3d0 100755
--- a/apps/benchmarks/UniformGrid/ecmModel.py
+++ b/apps/benchmarks/UniformGrid/ecmModel.py
@@ -3,75 +3,78 @@
import numpy as np
import matplotlib.pyplot as plt
-
kernels = dict()
+
class Kernel:
-
- def __init__(self,name, cyclesFirstLoop=0, cyclesSecondLoop=0, cyclesRegPerLUP =0):
+
+ def __init__(self, name, cyclesFirstLoop=0, cyclesSecondLoop=0, cyclesRegPerLUP=0):
self.name = name
if cyclesRegPerLUP <= 0:
- self.cyclesFirstLoop = cyclesFirstLoop
+ self.cyclesFirstLoop = cyclesFirstLoop
self.cyclesSecondLoop = cyclesSecondLoop
- self.cyclesRegPerLUP = cyclesFirstLoop + 9* cyclesSecondLoop
+ self.cyclesRegPerLUP = cyclesFirstLoop + 9 * cyclesSecondLoop
else:
self.cyclesRegPerLUP = cyclesRegPerLUP
-
- self.cyclesRegPerCacheLine = 8*self.cyclesRegPerLUP
- self.cyclesL1L2 = 3*19*2
- self.cyclesL2L3 = 3*19*2
-
+ self.cyclesRegPerCacheLine = 8 * self.cyclesRegPerLUP
+
+ self.cyclesL1L2 = 3 * 19 * 2
+ self.cyclesL2L3 = 3 * 19 * 2
+
self.freq = 2.7e9
self.cyclesMem = 305
- #self.cyclesMem = 191
-
+ # self.cyclesMem = 191
+
def mlups(self, processes):
singleCoreCycles = self.cyclesRegPerCacheLine + self.cyclesL1L2 + self.cyclesL2L3 + self.cyclesMem
-
+
timeSingleCore = singleCoreCycles / self.freq
-
+
mlups = 8 / timeSingleCore * 1e-6
-
- #todo
+
+ # todo
mlupsMax = 78
-
- return min ( processes * mlups, mlupsMax )
-
- def plot( self, divideByProcesses=False,processes=8, label="" ):
-
- x = np.arange( 1, processes+1, 1 )
+
+ return min(processes * mlups, mlupsMax)
+
+ def plot(self, divideByProcesses=False, processes=8, label=""):
+
+ x = np.arange(1, processes + 1, 1)
if divideByProcesses:
- y = np.array( [ self.mlups(i)/i for i in x ] )
+ y = np.array([self.mlups(i) / i for i in x])
else:
- y = np.array( [ self.mlups(i) for i in x ] )
-
- if label=="":
- label = "ecm\_" + self.name
- plt.plot( x, y, marker='^', markersize=5, label = label)
-
+ y = np.array([self.mlups(i) for i in x])
-kernels=dict()
+ if label == "":
+ label = "ecm_" + self.name
+ plt.plot(x, y, marker='^', markersize=5, label=label)
-#kernels['srt_split'] = Kernel("srt_split", 46, 12 )
-kernels['srt_pure'] = Kernel("srt_pure", 40, 8 )
-kernels['trt_split'] = Kernel("trt\_split", 41, 11 )
-kernels['srt_nonopt'] = Kernel("srt_nonopt", cyclesRegPerLUP = 1045) #SRTStreamCollide.h - pgo and lto (20cycles first loop, 35 second)
+kernels = dict()
+
+# kernels['srt_split'] = Kernel("srt_split", 46, 12 )
-#kernels['trt_pure_intelOpt'] = Kernel("trt_pure_intelOpt", 41/2, 10/2 ) # vectorized (v*pd)
+kernels['srt_pure'] = Kernel("srt_pure", 40, 8)
+kernels['trt_split'] = Kernel("trt_split", 41, 11)
+# SRTStreamCollide.h - pgo and lto (20cycles first loop, 35 second)
+kernels['srt_nonopt'] = Kernel("srt_nonopt",
+ cyclesRegPerLUP=1045)
-def plotAllKernels( divideByProcesses = False ):
+# kernels['trt_pure_intelOpt'] = Kernel("trt_pure_intelOpt", 41/2, 10/2 ) # vectorized (v*pd)
+
+
+def plotAllKernels(divideByProcesses=False):
for kernel in kernels:
- kernel.plot( divideByProcesses )
+ kernel.plot(divideByProcesses)
+
+
+def plot(kernelName, divideByProcesses=False, label=""):
+ kernels[kernelName].plot(divideByProcesses, label=label)
-def plot( kernelName, divideByProcesses = False, label = ""):
- kernels[kernelName].plot( divideByProcesses, label=label )
-
-
if __name__ == "__main__":
plotAllKernels()
plt.legend()
- plt.show()
\ No newline at end of file
+ plt.show()
diff --git a/apps/benchmarks/UniformGrid/mergeSqliteFiles.py b/apps/benchmarks/UniformGrid/mergeSqliteFiles.py
index d82a6449a..9975bb09b 100755
--- a/apps/benchmarks/UniformGrid/mergeSqliteFiles.py
+++ b/apps/benchmarks/UniformGrid/mergeSqliteFiles.py
@@ -4,85 +4,87 @@ import sqlite3
import sys
import shutil
-def getColumnNames ( db, tableName, dbName ):
+
+def getColumnNames(db, tableName, dbName):
cursor = db.cursor()
- cursor.execute("PRAGMA %s.table_info(%s)" % (dbName,tableName) )
+ cursor.execute("PRAGMA %s.table_info(%s)" % (dbName, tableName))
columns = cursor.fetchall()
-
+
res = []
for e in columns:
- res.append ( (e[1], e[2].upper()) )
-
+ res.append((e[1], e[2].upper()))
+
return res
-def mergeSqliteFiles ( targetFile, fileToMerge ):
-
- db = sqlite3.connect( targetFile )
- db.execute ('ATTACH "' + fileToMerge + '" AS toMerge')
-
- targetColumns = getColumnNames( db, "runs", "main" )
- toMergeColumns = getColumnNames( db, "runs", "toMerge" )
-
+def mergeSqliteFiles(targetFile, fileToMerge):
+ db = sqlite3.connect(targetFile)
+ db.execute('ATTACH "' + fileToMerge + '" AS toMerge')
+
+ targetColumns = getColumnNames(db, "runs", "main")
+ toMergeColumns = getColumnNames(db, "runs", "toMerge")
+
columnsToCreate = [e for e in toMergeColumns if e not in targetColumns]
-
+
for column in columnsToCreate:
- print "Adding Column %s to run table of %s " % ( column[0], targetFile )
- db.execute ( "ALTER TABLE main.runs ADD COLUMN %s %s" % ( column[0], column[1] ) )
-
+ print
+ "Adding Column %s to run table of %s " % (column[0], targetFile)
+ db.execute("ALTER TABLE main.runs ADD COLUMN %s %s" % (column[0], column[1]))
+
# Fetch all runs from toMerge,
# check if an entry with same date exists, if not add the run and the timing pool entries
# to the targetTable
c = db.cursor()
- assert( toMergeColumns[0][0] == "runId")
- columns = [ e[0] for e in toMergeColumns ]
- columnString = ",".join( columns )
- columnStringNoRunId = ",".join( columns[1:] )
-
- query = 'SELECT %s FROM toMerge.runs WHERE timestamp || " " || random NOT IN ' % (columnString,)
- query += '( SELECT timestamp || " " || random FROM main.runs )'
-
- timingPoolColumnsMain = getColumnNames ( db, "timingPool", "main" )
- timingPoolColumnsToMerge = getColumnNames ( db, "timingPool", "toMerge" )
- assert ( timingPoolColumnsMain == timingPoolColumnsToMerge )
- timingPoolColumnNames = [ e[0] for e in timingPoolColumnsMain ]
- assert ( timingPoolColumnNames[0] == "runId")
-
+ assert (toMergeColumns[0][0] == "runId")
+ columns = [e[0] for e in toMergeColumns]
+ columnString = ",".join(columns)
+ columnStringNoRunId = ",".join(columns[1:])
+
+ query = 'SELECT %s FROM toMerge.runs WHERE timestamp || " " || random NOT IN ' % (columnString,)
+ query += '( SELECT timestamp || " " || random FROM main.runs )'
+
+ timingPoolColumnsMain = getColumnNames(db, "timingPool", "main")
+ timingPoolColumnsToMerge = getColumnNames(db, "timingPool", "toMerge")
+ assert (timingPoolColumnsMain == timingPoolColumnsToMerge)
+ timingPoolColumnNames = [e[0] for e in timingPoolColumnsMain]
+ assert (timingPoolColumnNames[0] == "runId")
+
mergedRuns = 0
- for run in c.execute (query):
+ for run in c.execute(query):
# Build up insert statement for 'runs' table
- questionMarkList = ['?'] * (len(run)-1)
- questionMarkString = ",".join( questionMarkList )
- insertStatement = "INSERT INTO main.runs (%s) VALUES (%s);" % ( columnStringNoRunId, questionMarkString )
+ questionMarkList = ['?'] * (len(run) - 1)
+ questionMarkString = ",".join(questionMarkList)
+ insertStatement = "INSERT INTO main.runs (%s) VALUES (%s);" % (columnStringNoRunId, questionMarkString)
# Execute the insert
insertCursor = db.cursor()
- insertCursor.execute( insertStatement, run[1:] )
+ insertCursor.execute(insertStatement, run[1:])
# Insert the corresponding timingPool infos
insertedRunId = insertCursor.lastrowid
originalRunId = run[0]
-
- timingPoolQuery = "SELECT %s FROM toMerge.timingPool WHERE runId=?" % ( ",".join( timingPoolColumnNames[1:] ) )
+
+ timingPoolQuery = "SELECT %s FROM toMerge.timingPool WHERE runId=?" % (",".join(timingPoolColumnNames[1:]))
timingPoolInsertCursor = db.cursor()
- timingPoolQueryCursor = db.cursor()
-
- for tp in timingPoolQueryCursor.execute ( timingPoolQuery, ( originalRunId,) ):
- questionMarkList = ['?'] * len(timingPoolColumnNames)
- questionMarkString = ",".join( questionMarkList )
- insertQuery = "INSERT INTO main.timingPool (%s) VALUES (%s)" % (",".join(timingPoolColumnNames), questionMarkString)
- timingPoolInsertCursor.execute ( insertQuery, (insertedRunId,) + tp )
-
- mergedRuns = mergedRuns +1
-
- print "Merged %s runs from %s to %s " % ( mergedRuns, fileToMerge, targetFile )
+ timingPoolQueryCursor = db.cursor()
+
+ for tp in timingPoolQueryCursor.execute(timingPoolQuery, (originalRunId,)):
+ questionMarkList = ['?'] * len(timingPoolColumnNames)
+ questionMarkString = ",".join(questionMarkList)
+ insertQuery = "INSERT INTO main.timingPool (%s) VALUES (%s)" % (",".join(timingPoolColumnNames),
+ questionMarkString)
+ timingPoolInsertCursor.execute(insertQuery, (insertedRunId,) + tp)
+
+ mergedRuns = mergedRuns + 1
+
+ print("Merged %s runs from %s to %s " % (mergedRuns, fileToMerge, targetFile))
db.commit()
db.close()
-if ( len(sys.argv) < 3 ):
- print "Usage: mergeSqliteFiles resultFile <filesToMerge>"
+if len(sys.argv) < 3:
+ print("Usage: mergeSqliteFiles resultFile <filesToMerge>")
else:
- print "Copying " + sys.argv[2] + " to " + sys.argv[1]
- shutil.copy( sys.argv[2], sys.argv[1] )
- for i in range ( 3, len(sys.argv) ):
- print "Merging " + sys.argv[i]
- mergeSqliteFiles ( sys.argv[1], sys.argv[i] )
+ print("Copying " + sys.argv[2] + " to " + sys.argv[1])
+ shutil.copy(sys.argv[2], sys.argv[1])
+ for i in range(3, len(sys.argv)):
+ print("Merging " + sys.argv[i])
+ mergeSqliteFiles(sys.argv[1], sys.argv[i])
diff --git a/apps/benchmarks/UniformGrid/plot.py b/apps/benchmarks/UniformGrid/plot.py
index fcdedfd0a..877dc3635 100755
--- a/apps/benchmarks/UniformGrid/plot.py
+++ b/apps/benchmarks/UniformGrid/plot.py
@@ -1,7 +1,6 @@
#!/usr/bin/python
import sqlite3
-import numpy as np
import sys
import re
import math
@@ -11,270 +10,283 @@ from optparse import OptionParser
import ecmModel
+# Defining graph types
-####################################################################################################################################################
-############################ Defining graph types ##############################################
-####################################################################################################################################################
+graphs = dict(nonOpt=["cores,MAX(MLUPS)",
+ "SPLIT IS NULL AND PURE IS NULL AND CELL_OP IS NULL AND "
+ + "COMPRESSED IS NULL AND D3Q19_OPT IS NULL"],
+ split=["cores,MAX(MLUPS)", "SPLIT=1"],
+ pure=["cores,MAX(MLUPS)", "PURE=1"],
+ normal=["cores,MAX(MLUPS)", "CELL_OP=1", ],
+ compressed=["cores,MAX(MLUPS)", "COMPRESSED=1"],
+ opt=["cores,MAX(MLUPS)", "D3Q19_OPT=1"],
+ split_lbm=["cores,MAX(percentage)", "SPLIT=1 AND name=\"Timeloop\" AND sweep LIKE \"LBM%\""],
+ split_comm=["cores,MAX(percentage)", "SPLIT=1 AND name=\"Communication\" AND sweep LIKE \"%MPI%\""],
+ split_time=["cores,MAX(average) ", "SPLIT=1 AND name=\"Timeloop\" AND sweep LIKE \"LBM%\""]
+ )
+# Defining all options
-graphs = dict( nonOpt = [ "cores,MAX(MLUPS)", "SPLIT IS NULL AND PURE IS NULL AND CELL_OP IS NULL AND COMPRESSED IS NULL AND D3Q19_OPT IS NULL" ],
- split = [ "cores,MAX(MLUPS)", "SPLIT=1" ],
- pure = [ "cores,MAX(MLUPS)", "PURE=1" ],
- normal = [ "cores,MAX(MLUPS)", "CELL_OP=1", ],
- compressed = [ "cores,MAX(MLUPS)", "COMPRESSED=1"],
- opt = [ "cores,MAX(MLUPS)", "D3Q19_OPT=1"],
- split_lbm = [ "cores,MAX(percentage)", "SPLIT=1 AND name=\"Timeloop\" AND sweep LIKE \"LBM%\"" ],
- split_comm = [ "cores,MAX(percentage)", "SPLIT=1 AND name=\"Communication\" AND sweep LIKE \"%MPI%\"" ],
- split_time = [ "cores,MAX(average) " , "SPLIT=1 AND name=\"Timeloop\" AND sweep LIKE \"LBM%\"" ]
- )
+def addCommonOptions(parser):
+ parser.add_option("", "--smIntel", action="store_true", default=False, help="On SuperMUC using IntelMPI")
+ parser.add_option("", "--smIbm", action="store_true", default=False, help="On SuperMUC using IBM MPI")
+ parser.add_option("", "--zyxf", action="store_true", default=False, help="Only Layout zyxf")
+ parser.add_option("", "--fzyx", action="store_true", default=False, help="Only Layout fzyx")
+ parser.add_option("", "--trt", action="store_true", default=False, help="Only TRT")
+ parser.add_option("", "--srt", action="store_true", default=False, help="Only SRT")
+
+ parser.add_option("", "--lto", action="store_true", default=False, help="Only where link time optimization was on")
+ parser.add_option("", "--noLto", action="store_true", default=False,
+ help="Only where link time optimization was off")
+
+ parser.add_option("", "--pgo", action="store_true", default=False,
+ help="Only where profile guided optimization was on")
+ parser.add_option("", "--noPgo", action="store_true", default=False,
+ help="Only where profile guided optimization was off")
+
+ parser.add_option("", "--intelOpt", action="store_true", default=False, help="Only where intel pragmas were used")
+ parser.add_option("", "--noIntelOpt", action="store_true", default=False, help="Only where intel pragmas not used")
+
+ parser.add_option("", "--pinCore", action="store_true", default=False, help="First fill up socket.")
+ parser.add_option("", "--pinMCM", action="store_true", default=False, help="Socket round robin")
+
+ parser.add_option("-t", "--totalMLUPS", action="store_true", default=False,
+ help="Show total MLUPS instead of MLUPS/cores")
+
+ parser.add_option("-w", "--where", help="SQL Where or Group By clause", default="")
+
+ parser.add_option("-p", "--printDataset", action="store_true", default=False,
+ help="Prints the dataset that is plotted")
-####################################################################################################################################################
-############################ Defining all options ##############################################
-####################################################################################################################################################
-
-
-def addCommonOptions( parser ):
- parser.add_option("", "--smIntel", action="store_true", default=False, help="On SuperMUC using IntelMPI" )
- parser.add_option("", "--smIbm", action="store_true", default=False, help="On SuperMUC using IBM MPI" )
- parser.add_option("", "--zyxf", action="store_true", default=False, help="Only Layout zyxf" )
- parser.add_option("", "--fzyx", action="store_true", default=False, help="Only Layout fzyx" )
- parser.add_option("", "--trt", action="store_true", default=False, help="Only TRT" )
- parser.add_option("", "--srt", action="store_true", default=False, help="Only SRT" )
-
- parser.add_option("", "--lto", action="store_true", default=False, help="Only where link time optimization was on")
- parser.add_option("", "--noLto", action="store_true", default=False, help="Only where link time optimization was off")
-
- parser.add_option("", "--pgo", action="store_true", default=False, help="Only where profile guided optimization was on")
- parser.add_option("", "--noPgo", action="store_true", default=False, help="Only where profile guided optimization was off")
-
- parser.add_option("", "--intelOpt", action="store_true", default=False, help="Only where intel pragmas were used")
- parser.add_option("", "--noIntelOpt",action="store_true", default=False, help="Only where intel pragmas not used" )
-
-
- parser.add_option("", "--pinCore", action="store_true", default=False, help="First fill up socket." )
- parser.add_option("", "--pinMCM", action="store_true", default=False, help="Socket round robin" )
-
- parser.add_option("-t", "--totalMLUPS", action="store_true", default=False, help="Show total MLUPS instead of MLUPS/cores" )
-
- parser.add_option("-w", "--where", help="SQL Where or Group By clause", default = "")
-
- parser.add_option("-p", "--printDataset", action="store_true", default=False, help="Prints the dataset that is plotted")
-
- parser.add_option("", "--legend", default="", help="Legend entry for the graph" )
-
-
-def addOuterOptions( parser ):
- parser.add_option("-f", "--file", default="timing.sqlite", help="Sqlite3 File with timing data")
- parser.add_option("", "--title", default="", help="Title of Graph" )
- parser.add_option("-s", "--save", default="", help="Print graph to file")
- parser.add_option("", "--xlabel",default="", help="Label of x Axis")
- parser.add_option("", "--ylabel",default="", help="Label of y Axis")
- parser.add_option("", "--figureWidth", default="", help="When using --save, the width has to be specified in pts")
- parser.add_option("", "--figureHeight", default="", help="When using --save, the height has to be specified in pts")
- parser.add_option("", "--legendPos", default="", help="Position of legend (matplotlib syntax)")
-
-def addInnerOptions( parser ):
- parser.add_option("-g", "--graph", default="", help="Graph from database (measured)" )
- parser.add_option("-e", "--ecm", default="", help="Plot ECM model for given kernel name")
-
-
-def stringFromOptions( opt ):
+ parser.add_option("", "--legend", default="", help="Legend entry for the graph")
+
+
+def addOuterOptions(parser):
+ parser.add_option("-f", "--file", default="timing.sqlite", help="Sqlite3 File with timing data")
+ parser.add_option("", "--title", default="", help="Title of Graph")
+ parser.add_option("-s", "--save", default="", help="Print graph to file")
+ parser.add_option("", "--xlabel", default="", help="Label of x Axis")
+ parser.add_option("", "--ylabel", default="", help="Label of y Axis")
+ parser.add_option("", "--figureWidth", default="", help="When using --save, the width has to be specified in pts")
+ parser.add_option("", "--figureHeight", default="", help="When using --save, the height has to be specified in pts")
+ parser.add_option("", "--legendPos", default="", help="Position of legend (matplotlib syntax)")
+
+
+def addInnerOptions(parser):
+ parser.add_option("-g", "--graph", default="", help="Graph from database (measured)")
+ parser.add_option("-e", "--ecm", default="", help="Plot ECM model for given kernel name")
+
+
+def stringFromOptions(opt):
c = ""
- if ( opt.zyxf ):
+ if opt.zyxf:
c += "FZYX "
- elif ( opt.fzyx ):
+ elif opt.fzyx:
c += "ZYXF"
else:
- c+="ALL"
-
- c+= "|"
-
- if ( opt.trt ):
+ c += "ALL"
+
+ c += "|"
+
+ if opt.trt:
c += "TRT"
- elif ( opt.srt ):
+ elif opt.srt:
c += "SRT"
else:
- c+= "ALL"
-
- c+="|"
-
- if ( opt.smIntel ):
- c += "SmIntel|"
- if ( opt.smIbm ):
- c += "SmIBM"
-
-
- if opt.where and len( opt.where ) > 0:
- c += opt.where + "|"
-
- if opt.pinCore: c+="pinCore|"
- if opt.pinMCM: c+="pinMCM|"
-
- if opt.lto: c+="lto|"
- if opt.noLto: c+="noLto|"
-
- if opt.pgo: c+="pgo|"
- if opt.noPgo: c+="noPgo|"
-
- if opt.intelOpt: c+="intelOpt|"
- if opt.noIntelOpt: c+="noIntelOpt|"
+ c += "ALL"
+
+ c += "|"
+
+ if opt.smIntel:
+ c += "SmIntel|"
+ if opt.smIbm:
+ c += "SmIBM"
+
+ if opt.where and len(opt.where) > 0:
+ c += opt.where + "|"
+
+ if opt.pinCore:
+ c += "pinCore|"
+ if opt.pinMCM:
+ c += "pinMCM|"
+
+ if opt.lto:
+ c += "lto|"
+ if opt.noLto:
+ c += "noLto|"
+
+ if opt.pgo:
+ c += "pgo|"
+ if opt.noPgo:
+ c += "noPgo|"
+
+ if opt.intelOpt:
+ c += "intelOpt|"
+ if opt.noIntelOpt:
+ c += "noIntelOpt|"
if hasattr(opt, 'graph') and opt.graph:
- c+= opt.graph
-
-
+ c += opt.graph
+
return c
-def whereClauseFromOptions ( opt ):
+def whereClauseFromOptions(opt):
c = ""
- if ( opt.zyxf ) : c += "FZYX IS NULL AND "
- if ( opt.fzyx ) : c += "FZYX = 1 AND "
-
- if ( opt.trt ) : c += "TRT=1 AND "
- if ( opt.srt ) : c += "TRT IS NULL AND "
-
- if ( opt.smIntel ): c += "buildMachine LIKE \"supermuc_intel%\" AND machine LIKE \"i__r__a__\" AND "
-
- if ( opt.smIbm ) : c += "buildMachine LIKE \"supermuc_ibm%\" AND machine LIKE \"i__r__a__\" AND "
-
- if opt.pinCore: c+="MP_TASK_AFFINITY='CORE' AND "
- if opt.pinMCM: c+="MP_TASK_AFFINITY='MCM' AND "
-
- if opt.lto: c+="compilerFlags LIKE '%-ipo%' AND "
- if opt.noLto: c+="compilerFlags NOT LIKE '%-ipo%' AND "
-
- if opt.pgo: c+="compilerFlags LIKE '%-prof-use%' AND "
- if opt.noPgo: c+="compilerFlags NOT LIKE '%-prof-use%' AND "
-
- if opt.intelOpt: c+="intelCompilerOpt=1 AND "
- if opt.noIntelOpt: c+="intelCompilerOpt IS NULL AND "
-
-
- if opt.where and len( opt.where ) > 0:
- c += opt.where + " AND "
-
- return c
+ if opt.zyxf:
+ c += "FZYX IS NULL AND "
+ if opt.fzyx:
+ c += "FZYX = 1 AND "
+
+ if opt.trt:
+ c += "TRT=1 AND "
+ if opt.srt:
+ c += "TRT IS NULL AND "
+
+ if opt.smIntel:
+ c += "buildMachine LIKE \"supermuc_intel%\" AND machine LIKE \"i__r__a__\" AND "
+
+ if opt.smIbm:
+ c += "buildMachine LIKE \"supermuc_ibm%\" AND machine LIKE \"i__r__a__\" AND "
+ if opt.pinCore:
+ c += "MP_TASK_AFFINITY='CORE' AND "
+ if opt.pinMCM:
+ c += "MP_TASK_AFFINITY='MCM' AND "
+ if opt.lto:
+ c += "compilerFlags LIKE '%-ipo%' AND "
+ if opt.noLto:
+ c += "compilerFlags NOT LIKE '%-ipo%' AND "
+ if opt.pgo:
+ c += "compilerFlags LIKE '%-prof-use%' AND "
+ if opt.noPgo:
+ c += "compilerFlags NOT LIKE '%-prof-use%' AND "
-####################################################################################################################################################
-############################ Database and plotting ##############################################
-####################################################################################################################################################
+ if opt.intelOpt:
+ c += "intelCompilerOpt=1 AND "
+ if opt.noIntelOpt:
+ c += "intelCompilerOpt IS NULL AND "
+ if opt.where and len(opt.where) > 0:
+ c += opt.where + " AND "
+ return c
+
+# Database and plotting
-def getListFromDatabase( databaseFile, graphKeyword, whereClause ):
- db = sqlite3.connect( databaseFile )
+
+def getListFromDatabase(databaseFile, graphKeyword, whereClause):
+ db = sqlite3.connect(databaseFile)
db.row_factory = sqlite3.Row
- query = "SELECT " + graphs[graphKeyword][0]
+ query = "SELECT " + graphs[graphKeyword][0]
query += " FROM runs JOIN timingPool ON runs.runId = timingPool.runId "
if not whereClause or len(whereClause) == 0:
whereClause = " 1 "
-
- combinedWhere = " WHERE " + graphs[graphKeyword][1] + " AND " + whereClause + "GROUP BY cores"
+
+ combinedWhere = " WHERE " + graphs[graphKeyword][1] + " AND " + whereClause + "GROUP BY cores"
query += combinedWhere
- print combinedWhere;
+ print(combinedWhere)
c = db.cursor()
- c.execute( query )
+ c.execute(query)
return c.fetchall()
-
-
-def numericPlot( dataset, legendLabel = "" , divideByProcesses=True):
- assert ( len(dataset) > 0 )
- assert ( isinstance(dataset[0][0], (int, long, float) ) )
- columns = len( dataset[0] )
+
+
+def numericPlot(dataset, legendLabel="", divideByProcesses=True):
+ assert (len(dataset) > 0)
+ assert (isinstance(dataset[0][0], (int, float)))
+ columns = len(dataset[0])
columnNames = dataset[0].keys()
-
- xVals = [ e[0] for e in dataset ]
-
+
+ xVals = [e[0] for e in dataset]
+
plots = []
- for i in range(1,columns):
- yVals = [ e[i] for e in dataset ]
+ for i in range(1, columns):
+ yVals = [e[i] for e in dataset]
if divideByProcesses:
- yVals = [ yVals[i] / xVals[i] for i in range(0,len(yVals)) ]
-
+ yVals = [yVals[i] / xVals[i] for i in range(0, len(yVals))]
+
if legendLabel == "":
legendLabel = columnNames[i]
-
- p, = plt.plot( xVals, yVals, marker='^', markersize=5, label = legendLabel )
- plots.append( p )
-
- plt.xlabel ( "Cores" )
- plt.ylabel ( "MLUPS" )
- #plt.gca().yaxis.grid(color='gray', linestyle='dashed')
- #plt.gca().xaxis.grid(color='gray', linestyle='dashed')
+ p, = plt.plot(xVals, yVals, marker='^', markersize=5, label=legendLabel)
+ plots.append(p)
-def printDataset( dataset ):
+ plt.xlabel("Cores")
+ plt.ylabel("MLUPS")
+ # plt.gca().yaxis.grid(color='gray', linestyle='dashed')
+ # plt.gca().xaxis.grid(color='gray', linestyle='dashed')
+
+
+def printDataset(dataset):
header = dataset[0].keys()
- row_format ="{:<30}" * (len(header) )
- print row_format.format( *header).upper()
-
- for row in dataset:
- print row_format.format( *row)
+ row_format = "{:<30}" * (len(header))
+ print(row_format.format(*header).upper())
+ for row in dataset:
+ print(row_format.format(*row))
-def matplotlibLatexSetup( figureWidth, figureHeight ):
+
+def matplotlibLatexSetup(figureWidth, figureHeight):
ieee = False
if ieee:
- params ={\
+ params = {
'backend': 'GTKAgg',
-
+
'font.family': 'serif',
'font.serif': ['Times', 'Palatino', 'New Century Schoolbook', 'Bookman', 'Computer Modern Roman'],
- 'font.sans-serif' : ['Helvetica', 'Avant Garde', 'Computer Modern Sans serif'],
- #font.cursive : Zapf Chancery
- #font.monospace : Courier, Computer Modern Typewriter
+ 'font.sans-serif': ['Helvetica', 'Avant Garde', 'Computer Modern Sans serif'],
+ # font.cursive : Zapf Chancery
+ # font.monospace : Courier, Computer Modern Typewriter
'text.usetex': True,
-
+
'axes.labelsize': 9,
'axes.linewidth': .75,
-
+
'figure.figsize': (3.5, 2.5),
- 'figure.subplot.left' : 0.175,
+ 'figure.subplot.left': 0.175,
'figure.subplot.right': 0.95,
'figure.subplot.bottom': 0.15,
'figure.subplot.top': .95,
-
- 'figure.dpi':150,
-
+
+ 'figure.dpi': 150,
+
'text.fontsize': 9,
'legend.fontsize': 8,
'xtick.labelsize': 8,
'ytick.labelsize': 8,
-
- 'lines.linewidth':.75,
- 'savefig.dpi':600,
- }
+
+ 'lines.linewidth': .75,
+ 'savefig.dpi': 600,
+ }
else:
# Input comes usually directly from latex, and looks like "124.0pt"
# first the "pt" is removed, then it is converted to a floating point number
- maxFigHeight = float( figureWidth .replace("pt","") )
- maxFigWidth = float( figureHeight.replace("pt","") )
-
+ maxFigHeight = float(figureWidth.replace("pt", ""))
+ maxFigWidth = float(figureHeight.replace("pt", ""))
+
# Convert pt to inch
- inches_per_pt = 1.0/72.27
- maxFigHeight *= inches_per_pt
- maxFigWidth *= inches_per_pt
-
- golden_mean = (math.sqrt(5)-1.0)/2.0 # Aesthetic ratio
-
+ inches_per_pt = 1.0 / 72.27
+ maxFigHeight *= inches_per_pt
+ maxFigWidth *= inches_per_pt
+
+ golden_mean = (math.sqrt(5) - 1.0) / 2.0 # Aesthetic ratio
+
if maxFigWidth * golden_mean < maxFigHeight:
height = maxFigWidth * golden_mean
width = maxFigWidth
else:
height = maxFigHeight
width = maxFigHeight / golden_mean
-
- fig_size = [width,height]
+
+ fig_size = [width, height]
params = {'backend': 'ps',
'font.family': 'serif',
'font.serif': ['Times', 'Palatino', 'New Century Schoolbook', 'Bookman', 'Computer Modern Roman'],
- 'font.sans-serif' : ['Helvetica', 'Avant Garde', 'Computer Modern Sans serif'],
+ 'font.sans-serif': ['Helvetica', 'Avant Garde', 'Computer Modern Sans serif'],
'axes.labelsize': 10,
'text.fontsize': 10,
'legend.fontsize': 8,
@@ -282,116 +294,116 @@ def matplotlibLatexSetup( figureWidth, figureHeight ):
'ytick.labelsize': 8,
'text.usetex': True,
'figure.figsize': fig_size,
- 'savefig.dpi':600,
- 'antialiased': True,}
+ 'savefig.dpi': 600,
+ 'antialiased': True, }
plt.rcParams.update(params)
- plt.rc('lines', aa = True)
-####################################################################################################################################################
-############################ Option parsing ###############################################
-####################################################################################################################################################
+ plt.rc('lines', aa=True)
+
+# Option parsing
def bracketsAsExtraElements(list):
""" Transforms the list [ '[abc', 'def]' ] to [ '[', 'abc', 'def', ']' ] """
processedList = []
for e in list:
- splitted = re.split('(\[|\])', e)
- while '' in splitted: splitted.remove('')
- processedList.extend( splitted )
-
+ # TODO: write better
+ splitted = re.split('(\[|\])', e) # noqa: W605
+ while '' in splitted:
+ splitted.remove('')
+ processedList.extend(splitted)
+
return processedList
-def parseList( list ):
- list = bracketsAsExtraElements( list )
- outerList = []
+def parseList(list):
+ list = bracketsAsExtraElements(list)
+ outerList = []
subLists = []
-
+
curSubList = None
for element in list:
-
+
if element == "[":
- if not curSubList == None:
- raise ValueError( "Two opening brackets" )
- curSubList = []
+ if curSubList is not None:
+ raise ValueError("Two opening brackets")
+ curSubList = []
elif element == "]":
- if curSubList == None:
+ if curSubList is None:
raise ValueError("Closing bracket without opening bracket")
- subLists.append( curSubList )
+ subLists.append(curSubList)
curSubList = None
else:
- if curSubList == None:
- outerList.append( element )
+ if curSubList is None:
+ outerList.append(element)
else:
- curSubList.append( element )
-
- if not curSubList == None:
- raise ValueError( "Missing closing bracket")
-
- return ( outerList, subLists )
-
-
+ curSubList.append(element)
+
+ if curSubList is not None:
+ raise ValueError("Missing closing bracket")
+
+ return outerList, subLists
+
outerParser = OptionParser()
-addCommonOptions( outerParser )
-addOuterOptions( outerParser )
+addCommonOptions(outerParser)
+addOuterOptions(outerParser)
innerParser = OptionParser()
-addCommonOptions( innerParser )
-addInnerOptions( innerParser )
+addCommonOptions(innerParser)
+addInnerOptions(innerParser)
-(outerList,innerLists) = parseList( sys.argv )
+(outerList, innerLists) = parseList(sys.argv)
-
-(outerOptions, outerArgs) = outerParser.parse_args( outerList )
-outerWhereClause = whereClauseFromOptions( outerOptions )
+(outerOptions, outerArgs) = outerParser.parse_args(outerList)
+outerWhereClause = whereClauseFromOptions(outerOptions)
# Matplotlib setup
if outerOptions.save != "":
- matplotlibLatexSetup( outerOptions.figureWidth, outerOptions.figureHeight )
+ matplotlibLatexSetup(outerOptions.figureWidth, outerOptions.figureHeight)
legend = []
for innerList in innerLists:
- ( innerOptions,innerArgs ) = innerParser.parse_args( innerList )
-
+ (innerOptions, innerArgs) = innerParser.parse_args(innerList)
+
if innerOptions.ecm != "":
- ecmModel.plot( innerOptions.ecm, not outerOptions.totalMLUPS, innerOptions.legend )
+ ecmModel.plot(innerOptions.ecm, not outerOptions.totalMLUPS, innerOptions.legend)
else:
- innerWhereClause = whereClauseFromOptions( innerOptions )
- dataset = getListFromDatabase ( outerOptions.file, innerOptions.graph, outerWhereClause + innerWhereClause + " 1 ")
-
+ innerWhereClause = whereClauseFromOptions(innerOptions)
+ dataset = getListFromDatabase(outerOptions.file, innerOptions.graph,
+ outerWhereClause + innerWhereClause + " 1 ")
+
if len(dataset) == 0:
- print "Empty"
+ print("Empty")
continue
-
- if ( innerOptions.printDataset or outerOptions.printDataset ):
- printDataset( dataset )
-
-
- if innerOptions.legend == "" : innerOptions.legend = stringFromOptions( innerOptions )
- numericPlot( dataset, innerOptions.legend, not (outerOptions.totalMLUPS or innerOptions.totalMLUPS) )
+
+ if innerOptions.printDataset or outerOptions.printDataset:
+ printDataset(dataset)
+
+ if innerOptions.legend == "":
+ innerOptions.legend = stringFromOptions(innerOptions)
+ numericPlot(dataset, innerOptions.legend, not (outerOptions.totalMLUPS or innerOptions.totalMLUPS))
# Title
if outerOptions.title == "auto":
outerOptions.title = stringFromOptions(outerOptions)
-if outerOptions.title != "":
- plt.title( outerOptions.title )
-
+if outerOptions.title != "":
+ plt.title(outerOptions.title)
+
# Axis labels
-if outerOptions.xlabel != "": plt.xlabel( outerOptions.xlabel )
-if outerOptions.ylabel != "": plt.ylabel( outerOptions.ylabel )
+if outerOptions.xlabel != "":
+ plt.xlabel(outerOptions.xlabel)
+if outerOptions.ylabel != "":
+ plt.ylabel(outerOptions.ylabel)
# Legend
if outerOptions.legendPos != "":
- plt.legend( loc = outerOptions.legendPos )
+ plt.legend(loc=outerOptions.legendPos)
else:
- plt.legend( loc = 'upper left' )
-
+ plt.legend(loc='upper left')
if outerOptions.save != "":
- plt.savefig( outerOptions.save + ".pdf" )
+ plt.savefig(outerOptions.save + ".pdf")
else:
plt.show()
-
diff --git a/apps/benchmarks/UniformGridGPU/UniformGridGPU.py b/apps/benchmarks/UniformGridGPU/UniformGridGPU.py
index 98eed9e8e..c1371278a 100644
--- a/apps/benchmarks/UniformGridGPU/UniformGridGPU.py
+++ b/apps/benchmarks/UniformGridGPU/UniformGridGPU.py
@@ -3,7 +3,7 @@ import numpy as np
import pystencils as ps
from lbmpy.creationfunctions import create_lb_method, create_lb_update_rule, create_lb_collision_rule
from lbmpy.boundaries import NoSlip, UBB
-from lbmpy.fieldaccess import StreamPullTwoFieldsAccessor, StreamPushTwoFieldsAccessor
+from lbmpy.fieldaccess import StreamPullTwoFieldsAccessor
from pystencils_walberla import generate_pack_info_from_kernel
from lbmpy_walberla import generate_lattice_model, generate_boundary
from pystencils_walberla import CodeGeneration, generate_sweep
@@ -45,7 +45,8 @@ options_dict = {
'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]],
+ 'relaxation_rates': [omega_fill[0], omega, omega_fill[1], omega_fill[2],
+ omega_fill[3], omega_fill[4], omega_fill[5]],
},
'entropic': {
'method': 'mrt',
@@ -77,7 +78,7 @@ options_dict = {
}
info_header = """
-#include "stencil/D3Q{q}.h"\nusing Stencil_T = walberla::stencil::D3Q{q};
+#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};
@@ -87,7 +88,7 @@ const bool infoCsePdfs = {cse_pdfs};
with CodeGeneration() as ctx:
accessor = StreamPullTwoFieldsAccessor()
- #accessor = StreamPushTwoFieldsAccessor()
+ # accessor = StreamPushTwoFieldsAccessor()
assert not accessor.is_inplace, "This app does not work for inplace accessors"
common_options = {
@@ -118,7 +119,7 @@ with CodeGeneration() as ctx:
options['stencil'] = 'D3Q27'
stencil_str = options['stencil']
- q = int(stencil_str[stencil_str.find('Q')+1:])
+ 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
@@ -143,7 +144,8 @@ with CodeGeneration() as ctx:
# CPU lattice model - required for macroscopic value computation, VTK output etc.
options_without_opt = options.copy()
del options_without_opt['optimization']
- generate_lattice_model(ctx, 'UniformGridGPU_LatticeModel', create_lb_collision_rule(lb_method=lb_method, **options_without_opt))
+ generate_lattice_model(ctx, 'UniformGridGPU_LatticeModel', create_lb_collision_rule(lb_method=lb_method,
+ **options_without_opt))
# gpu LB sweep & boundaries
generate_sweep(ctx, 'UniformGridGPU_LbKernel', update_rule,
@@ -158,7 +160,7 @@ with CodeGeneration() as ctx:
setter_assignments = macroscopic_values_setter(lb_method, velocity=velocity_field.center_vector,
pdfs=pdfs.center_vector, density=1)
getter_assignments = macroscopic_values_getter(lb_method, velocity=velocity_field.center_vector,
- pdfs=pdfs.center_vector, density=None)
+ pdfs=pdfs.center_vector, density=None)
generate_sweep(ctx, 'UniformGridGPU_MacroSetter', setter_assignments)
generate_sweep(ctx, 'UniformGridGPU_MacroGetter', getter_assignments)
diff --git a/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py b/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py
index 64ccfa6ac..1fb6261d3 100644
--- a/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py
+++ b/apps/benchmarks/UniformGridGPU/UniformGridGPU_AA.py
@@ -56,7 +56,7 @@ options_dict = {
info_header = """
-#include "stencil/D3Q{q}.h"\nusing Stencil_T = walberla::stencil::D3Q{q};
+#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};
@@ -113,8 +113,10 @@ with CodeGeneration() as ctx:
generate_sweep(ctx, 'UniformGridGPU_AA_MacroGetter', getter_assignments)
# communication
- 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')
+ 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')
infoHeaderParams = {
'stencil': stencil_str,
diff --git a/apps/benchmarks/UniformGridGPU/simulation_setup/benchmark_configs.py b/apps/benchmarks/UniformGridGPU/simulation_setup/benchmark_configs.py
index ca66b364c..5fbf36a94 100755
--- a/apps/benchmarks/UniformGridGPU/simulation_setup/benchmark_configs.py
+++ b/apps/benchmarks/UniformGridGPU/simulation_setup/benchmark_configs.py
@@ -9,8 +9,8 @@ Look at the end of the file to select the benchmark to run
import os
import waLBerla as wlb
-from waLBerla.tools.config import block_decomposition, toPrm
-from waLBerla.tools.sqlitedb import *
+from waLBerla.tools.config import block_decomposition
+from waLBerla.tools.sqlitedb import sequenceValuesToScalars, checkAndUpdateSchema, storeSingle
from copy import deepcopy
import sys
import sqlite3
@@ -53,7 +53,7 @@ class Scenario:
from pprint import pformat
wlb.log_info_on_root("Scenario:\n" + pformat(self.config_dict))
# Write out the configuration as text-based prm:
- #print(toPrm(self.config_dict))
+ # print(toPrm(self.config_dict))
return self.config_dict
@wlb.member_callback
@@ -71,16 +71,17 @@ class Scenario:
result = data
sequenceValuesToScalars(result)
num_tries = 4
- for num_try in range(num_tries): # check multiple times e.g. may fail when multiple benchmark processes are running
+ # check multiple times e.g. may fail when multiple benchmark processes are running
+ for num_try in range(num_tries):
try:
checkAndUpdateSchema(result, "runs", DB_FILE)
storeSingle(result, "runs", DB_FILE)
break
except sqlite3.OperationalError as e:
- wlb.log_warning("Sqlite DB writing failed: try {}/{} {}".format(num_try+1, num_tries, str(e)))
+ wlb.log_warning("Sqlite DB writing failed: try {}/{} {}".format(num_try + 1, num_tries, str(e)))
-# -------------------------------------- Functions trying different parameter sets -------------------------------------------------------------------
+# -------------------------------------- Functions trying different parameter sets -----------------------------------
def overlap_benchmark():
@@ -93,9 +94,11 @@ def overlap_benchmark():
(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'
+ # 'GPUPackInfo_Baseline', 'GPUPackInfo_Streams'
+ for comm_strategy in ['UniformGPUScheme_Baseline', 'UniformGPUScheme_Memcpy']:
# no overlap
- scenarios.add(Scenario(timeStepStrategy='noOverlap', communicationScheme=comm_strategy, innerOuterSplit=(1, 1, 1)))
+ scenarios.add(Scenario(timeStepStrategy='noOverlap', communicationScheme=comm_strategy,
+ innerOuterSplit=(1, 1, 1)))
# overlap
for overlap_strategy in ['simpleOverlap', 'complexOverlap']:
@@ -123,7 +126,8 @@ def communication_compare():
timesteps=num_time_steps(block_size))
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]: # ensure that the inner part of the domain is still large enough
+ # ensure that the inner part of the domain is still large enough
+ if 3 * inner_outer_split[0] > block_size[0]:
continue
sc = Scenario(cells_per_block=block_size,
gpuBlockSize=(128, 1, 1),
@@ -155,7 +159,7 @@ def single_gpu_benchmark():
scenarios.add(scenario)
-# -------------------------------------- Optional job script generation for PizDaint -------------------------------------------------------------------
+# -------------------------------------- Optional job script generation for PizDaint ---------------------------------
job_script_header = """
@@ -223,7 +227,8 @@ def generate_jobscripts(exe_names=all_executables):
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')
+ 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)
@@ -235,6 +240,8 @@ if __name__ == '__main__':
else:
wlb.log_info_on_root("Batch run of benchmark scenarios, saving result to {}".format(DB_FILE))
# Select the benchmark you want to run
- single_gpu_benchmark() # benchmarks different CUDA block sizes and domain sizes and measures single GPU performance of compute kernel (no communication)
- #communication_compare() # benchmarks different communication routines, with and without overlap
- #overlap_benchmark() # benchmarks different communication overlap options
+ single_gpu_benchmark()
+ # benchmarks different CUDA block sizes and domain sizes and measures single
+ # GPU performance of compute kernel (no communication)
+ # communication_compare(): benchmarks different communication routines, with and without overlap
+ # overlap_benchmark(): benchmarks different communication overlap options
diff --git a/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py b/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py
index 6a4d6c452..b9ca24ba1 100644
--- a/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py
+++ b/apps/benchmarks/UniformGridGenerated/UniformGridGenerated.py
@@ -1,7 +1,8 @@
import sympy as sp
import pystencils as ps
from lbmpy.creationfunctions import create_lb_update_rule, create_lb_collision_rule
-from pystencils_walberla import CodeGeneration, generate_pack_info_from_kernel, generate_sweep, generate_mpidtype_info_from_kernel
+from pystencils_walberla import CodeGeneration, generate_pack_info_from_kernel, generate_sweep,\
+ generate_mpidtype_info_from_kernel
from lbmpy.macroscopic_value_kernels import macroscopic_values_getter, macroscopic_values_setter
from lbmpy.fieldaccess import AAEvenTimeStepAccessor, AAOddTimeStepAccessor
@@ -30,7 +31,8 @@ options_dict = {
'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]],
+ 'relaxation_rates': [omega_fill[0], omega, omega_fill[1], omega_fill[2],
+ omega_fill[3], omega_fill[4], omega_fill[5]],
},
'entropic': {
'method': 'mrt',
@@ -62,7 +64,7 @@ options_dict = {
}
info_header = """
-#include "stencil/D3Q{q}.h"\nusing Stencil_T = walberla::stencil::D3Q{q};
+#include "stencil/D3Q{q}.h"\nusing Stencil_T = walberla::stencil::D3Q{q};
const char * infoStencil = "{stencil}";
const char * infoConfigName = "{configName}";
const char * optimizationDict = "{optimizationDict}";
@@ -108,7 +110,7 @@ with CodeGeneration() as ctx:
options['stencil'] = 'D3Q27'
stencil_str = options['stencil']
- q = int(stencil_str[stencil_str.find('Q')+1:])
+ 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')
update_rule_two_field = create_lb_update_rule(optimization={'symbolic_field': pdfs,
@@ -128,12 +130,16 @@ with CodeGeneration() as ctx:
((0, 0, 1), UBB([0.05, 0, 0])),
((0, 0, -1), NoSlip()),
))
- cr_even = create_lb_collision_rule(stencil="D3Q19", compressible=False, optimization={'cse_global': opts['aa_even_cse_global'],
- 'cse_pdfs': opts['aa_even_cse_pdfs']})
- cr_odd = create_lb_collision_rule(stencil="D3Q19", compressible=False, optimization={'cse_global': opts['aa_odd_cse_global'],
- 'cse_pdfs': opts['aa_odd_cse_pdfs']})
- update_rule_aa_even = update_rule_with_push_boundaries(cr_even, pdfs, boundaries, AAEvenTimeStepAccessor, AAOddTimeStepAccessor.read)
- update_rule_aa_odd = update_rule_with_push_boundaries(cr_odd, pdfs, boundaries, AAOddTimeStepAccessor, AAEvenTimeStepAccessor.read)
+ cr_even = create_lb_collision_rule(stencil="D3Q19", compressible=False,
+ optimization={'cse_global': opts['aa_even_cse_global'],
+ 'cse_pdfs': opts['aa_even_cse_pdfs']})
+ cr_odd = create_lb_collision_rule(stencil="D3Q19", compressible=False,
+ optimization={'cse_global': opts['aa_odd_cse_global'],
+ 'cse_pdfs': opts['aa_odd_cse_pdfs']})
+ update_rule_aa_even = update_rule_with_push_boundaries(cr_even, pdfs, boundaries,
+ AAEvenTimeStepAccessor, AAOddTimeStepAccessor.read)
+ update_rule_aa_odd = update_rule_with_push_boundaries(cr_odd, pdfs, boundaries,
+ AAOddTimeStepAccessor, AAEvenTimeStepAccessor.read)
else:
update_rule_aa_even = create_lb_update_rule(kernel_type=AAEvenTimeStepAccessor(),
optimization={'symbolic_field': pdfs,
@@ -146,7 +152,7 @@ with CodeGeneration() as ctx:
'cse_global': opts['aa_odd_cse_global'],
'cse_pdfs': opts['aa_odd_cse_pdfs']}, **options)
- vec = { 'assume_aligned': True, 'assume_inner_stride_one': True}
+ vec = {'assume_aligned': True, 'assume_inner_stride_one': True}
# check if openmp is enabled in cmake
if ctx.openmp:
diff --git a/apps/benchmarks/UniformGridGenerated/params.py b/apps/benchmarks/UniformGridGenerated/params.py
index 724238f4b..598855b63 100644
--- a/apps/benchmarks/UniformGridGenerated/params.py
+++ b/apps/benchmarks/UniformGridGenerated/params.py
@@ -2,7 +2,7 @@ import math
import os
import operator
import waLBerla as wlb
-from waLBerla.tools.sqlitedb import *
+from waLBerla.tools.sqlitedb import sequenceValuesToScalars, checkAndUpdateSchema, storeSingle
from waLBerla.tools.config import block_decomposition
from functools import reduce
import sqlite3
@@ -137,18 +137,19 @@ def single_node_benchmark():
continue
scenarios.add(sc)
else:
- sc = BenchmarkScenario(block_size=block_size, direct_comm=direct_comm,
- domain_decomposition_func=domain_decomposition_func_z,
- time_step_mode=time_step_mode)
- if not block_size_ok(sc):
- continue
- scenarios.add(sc)
+ sc = BenchmarkScenario(block_size=block_size, direct_comm=direct_comm,
+ domain_decomposition_func=domain_decomposition_func_z,
+ time_step_mode=time_step_mode)
+ if not block_size_ok(sc):
+ continue
+ # scenarios.add(sc)
def single_node_benchmark_small():
scenarios = wlb.ScenarioManager()
for block_size in [(128, 128, 128), (128, 64, 64), (64, 64, 128), (64, 128, 64), (64, 64, 64),
- (1024, 64, 32), (2048, 64, 16), (64, 32, 32), (32, 32, 32), (16, 16, 16), (256, 128, 64), (512, 128, 32)]:
+ (1024, 64, 32), (2048, 64, 16), (64, 32, 32), (32, 32, 32), (16, 16, 16),
+ (256, 128, 64), (512, 128, 32)]:
for direct_comm in (True, False):
for time_step_mode in ['aa', 'aaKernelOnly', 'twoField']:
sc = BenchmarkScenario(block_size=block_size, direct_comm=direct_comm, time_step_mode=time_step_mode)
@@ -160,7 +161,8 @@ def single_node_benchmark_small():
def weak_scaling():
scenarios = wlb.ScenarioManager()
for block_size in [(128, 128, 128), (128, 64, 64), (64, 64, 128), (64, 128, 64), (64, 64, 64),
- (1024, 64, 32), (2048, 64, 16), (64, 32, 32), (32, 32, 32), (16, 16, 16), (256, 128, 64), (512, 128, 32)]:
+ (1024, 64, 32), (2048, 64, 16), (64, 32, 32), (32, 32, 32), (16, 16, 16),
+ (256, 128, 64), (512, 128, 32)]:
for direct_comm in (True, False):
sc = BenchmarkScenario(block_size=block_size, direct_comm=direct_comm,
domain_decomposition_func=domain_decomposition_func_full)
@@ -168,4 +170,5 @@ def weak_scaling():
continue
scenarios.add(sc)
+
single_node_benchmark()
diff --git a/apps/pythonmodule/setup.py b/apps/pythonmodule/setup.py
index 74532db4f..5929b0fb9 100644
--- a/apps/pythonmodule/setup.py
+++ b/apps/pythonmodule/setup.py
@@ -5,37 +5,36 @@ import platform
import sys
# The following variables are configure by CMake
-walberla_source_dir = "${walberla_SOURCE_DIR}"
+walberla_source_dir = "${walberla_SOURCE_DIR}"
walberla_binary_dir = "${CMAKE_CURRENT_BINARY_DIR}"
if platform.system() == 'Windows':
- extension = ( 'dll', 'pyd' )
- configuration = 'Release'
+ extension = ('dll', 'pyd')
+ configuration = 'Release'
else:
- extension = ( 'so', 'so' )
- configuration = ''
-
+ extension = ('so', 'so')
+ configuration = ''
+
def collectFiles():
- src_shared_lib = join(walberla_binary_dir, configuration, 'walberla_cpp.' + extension[0] )
- dst_shared_lib = join(walberla_binary_dir, 'waLBerla', 'walberla_cpp.' + extension[1] )
+ src_shared_lib = join(walberla_binary_dir, configuration, 'walberla_cpp.' + extension[0])
+ dst_shared_lib = join(walberla_binary_dir, 'waLBerla', 'walberla_cpp.' + extension[1])
# copy everything inplace
-
- print( src_shared_lib )
-
- if not exists( src_shared_lib ):
+
+ print(src_shared_lib)
+
+ if not exists(src_shared_lib):
print("Python Module was not built yet - run 'make walberla_cpp'")
exit(1)
-
+ shutil.rmtree(join(walberla_binary_dir, 'waLBerla'), ignore_errors=True)
- shutil.rmtree( join(walberla_binary_dir, 'waLBerla'), ignore_errors=True )
+ shutil.copytree(join(walberla_source_dir, 'python', 'waLBerla'),
+ join(walberla_binary_dir, 'waLBerla'))
- shutil.copytree( join(walberla_source_dir, 'python', 'waLBerla'),
- join(walberla_binary_dir, 'waLBerla') )
+ shutil.copy(src_shared_lib,
+ dst_shared_lib)
- shutil.copy( src_shared_lib,
- dst_shared_lib )
packages = ['waLBerla',
'waLBerla.evaluation',
@@ -48,13 +47,13 @@ packages = ['waLBerla',
collectFiles()
-setup( name='waLBerla',
- version='1.0',
- author='Martin Bauer',
- author_email='martin.bauer@fau.de',
- url='http://www.walberla.net',
- packages=packages,
- package_data = {'' : ['walberla_cpp.' + extension[1]] }
+setup(name='waLBerla',
+ version='1.0',
+ author='Martin Bauer',
+ author_email='martin.bauer@fau.de',
+ url='http://www.walberla.net',
+ packages=packages,
+ package_data={'': ['walberla_cpp.' + extension[1]]}
)
if sys.argv[1] == 'build':
diff --git a/apps/pythonmodule/stage.py b/apps/pythonmodule/stage.py
index 50cfd4362..2072a7e86 100644
--- a/apps/pythonmodule/stage.py
+++ b/apps/pythonmodule/stage.py
@@ -1,3 +1,2 @@
-
if __name__ == "__main__":
- install()
\ No newline at end of file
+ install() # noqa: F821
diff --git a/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_RTI_3D.py b/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_RTI_3D.py
index 1050b08e9..54f656cf9 100755
--- a/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_RTI_3D.py
+++ b/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_RTI_3D.py
@@ -89,17 +89,17 @@ class Scenario:
spike_data = wlb.field.gather(blocks, 'phase', makeSlice[self.size[0] // 2, :, self.size[2] // 2])
if spike_data:
spike_field = np.asarray(spike_data.buffer()).squeeze()
- location_of_spike = (np.argmax(spike_field > 0.5) - ny//2)/l0
+ location_of_spike = (np.argmax(spike_field > 0.5) - ny // 2) / l0
bubble_data = wlb.field.gather(blocks, 'phase', makeSlice[0, :, 0])
if bubble_data:
bubble_field = np.asarray(bubble_data.buffer()).squeeze()
- location_of_bubble = (np.argmax(bubble_field > 0.5) - ny//2)/l0
+ location_of_bubble = (np.argmax(bubble_field > 0.5) - ny // 2) / l0
saddle_data = wlb.field.gather(blocks, 'phase', makeSlice[0, :, self.size[2] // 2])
if saddle_data:
saddle_field = np.asarray(saddle_data.buffer()).squeeze()
- location_of_saddle = (np.argmax(saddle_field > 0.5) - ny//2)/l0
+ location_of_saddle = (np.argmax(saddle_field > 0.5) - ny // 2) / l0
phase = wlb.field.gather(blocks, 'phase', makeSlice[:, :, :])
if phase:
diff --git a/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_codegen.py b/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_codegen.py
index 9010cdcf8..42e990dc3 100644
--- a/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_codegen.py
+++ b/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_codegen.py
@@ -1,4 +1,4 @@
-from pystencils import fields, TypedSymbol
+from pystencils import fields
from pystencils.simp import sympy_cse
from pystencils import AssignmentCollection
@@ -6,7 +6,7 @@ from lbmpy.boundaries import NoSlip
from lbmpy.creationfunctions import create_lb_method, create_lb_update_rule
from lbmpy.stencils import get_stencil
-from pystencils_walberla import CodeGeneration, generate_sweep, generate_pack_info_from_kernel
+from pystencils_walberla import CodeGeneration, generate_sweep
from lbmpy_walberla import generate_boundary
from lbmpy.phasefield_allen_cahn.kernel_equations import initializer_kernel_phase_field_lb, \
@@ -155,7 +155,7 @@ vp = [('int32_t', 'cudaBlockSize0'),
('int32_t', 'cudaBlockSize1')]
info_header = f"""
-#include "stencil/D3Q{q_phase}.h"\nusing Stencil_phase_T = walberla::stencil::D3Q{q_phase};
+#include "stencil/D3Q{q_phase}.h"\nusing Stencil_phase_T = walberla::stencil::D3Q{q_phase};
#include "stencil/D3Q{q_hydro}.h"\nusing Stencil_hydro_T = walberla::stencil::D3Q{q_hydro};
"""
diff --git a/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_rising_bubble.py b/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_rising_bubble.py
index e1ca7bdda..9a8948868 100755
--- a/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_rising_bubble.py
+++ b/apps/showcases/PhaseFieldAllenCahn/CPU/multiphase_rising_bubble.py
@@ -5,6 +5,7 @@ import numpy as np
from waLBerla.core_extension import makeSlice
from lbmpy.phasefield_allen_cahn.parameter_calculation import calculate_dimensionless_rising_bubble
+
class Scenario:
def __init__(self):
# output frequencies
diff --git a/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_RTI_3D.py b/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_RTI_3D.py
index a7c5ed1c4..d862b1a48 100755
--- a/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_RTI_3D.py
+++ b/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_RTI_3D.py
@@ -89,17 +89,17 @@ class Scenario:
spike_data = wlb.field.gather(blocks, 'phase', makeSlice[self.size[0] // 2, :, self.size[2] // 2])
if spike_data:
spike_field = np.asarray(spike_data.buffer()).squeeze()
- location_of_spike = (np.argmax(spike_field > 0.5) - ny//2)/l0
+ location_of_spike = (np.argmax(spike_field > 0.5) - ny // 2) / l0
bubble_data = wlb.field.gather(blocks, 'phase', makeSlice[0, :, 0])
if bubble_data:
bubble_field = np.asarray(bubble_data.buffer()).squeeze()
- location_of_bubble = (np.argmax(bubble_field > 0.5) - ny//2)/l0
+ location_of_bubble = (np.argmax(bubble_field > 0.5) - ny // 2) / l0
saddle_data = wlb.field.gather(blocks, 'phase', makeSlice[0, :, self.size[2] // 2])
if saddle_data:
saddle_field = np.asarray(saddle_data.buffer()).squeeze()
- location_of_saddle = (np.argmax(saddle_field > 0.5) - ny//2)/l0
+ location_of_saddle = (np.argmax(saddle_field > 0.5) - ny // 2) / l0
phase = wlb.field.gather(blocks, 'phase', makeSlice[:, :, :])
if phase:
diff --git a/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_codegen.py b/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_codegen.py
index b98348ab4..16bdcfb5a 100644
--- a/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_codegen.py
+++ b/apps/showcases/PhaseFieldAllenCahn/GPU/multiphase_codegen.py
@@ -9,9 +9,8 @@ from lbmpy.stencils import get_stencil
from pystencils_walberla import CodeGeneration, generate_sweep, generate_pack_info_from_kernel
from lbmpy_walberla import generate_boundary
-from lbmpy.phasefield_allen_cahn.kernel_equations import initializer_kernel_phase_field_lb, \
- initializer_kernel_hydro_lb, interface_tracking_force, \
- hydrodynamic_force, get_collision_assignments_hydro
+from lbmpy.phasefield_allen_cahn.kernel_equations import initializer_kernel_phase_field_lb,\
+ initializer_kernel_hydro_lb, interface_tracking_force, hydrodynamic_force, get_collision_assignments_hydro
from lbmpy.phasefield_allen_cahn.force_model import MultiphaseForceModel
@@ -159,7 +158,7 @@ sweep_block_size = (TypedSymbol("cudaBlockSize0", np.int32),
sweep_params = {'block_size': sweep_block_size}
info_header = f"""
-#include "stencil/D3Q{q_phase}.h"\nusing Stencil_phase_T = walberla::stencil::D3Q{q_phase};
+#include "stencil/D3Q{q_phase}.h"\nusing Stencil_phase_T = walberla::stencil::D3Q{q_phase};
#include "stencil/D3Q{q_hydro}.h"\nusing Stencil_hydro_T = walberla::stencil::D3Q{q_hydro};
"""
--
GitLab