Commit befor trying to convert particle array to linked list to improve

performance.

runs/cylFlow/compCPUTime.gp

@@ -0,0 +1,54 @@
+
+set terminal qt enhanced 1 size 1600, 1000 font "Times ,10"
+
+set style line 1  pt 4              lc rgb "#B50427" #Squares   red
+set style line 2  pt 6              lc rgb "#3B4CC1" #Circles   blue
+set style line 3  pt 1              lc rgb "#2CA02C" #Crosses   green
+set style line 4  pt 2              lc rgb "#FE7F0E" #Exes      orange
+set style line 5  pt 8              lc rgb "#D6696B" #Triangles light red
+set style line 10       lt 1 lw 2.0 lc rgb "black"   #Black solid line
+
+folder1 = "2020-09-27_10.46.31/"
+folder2 = "2020-09-28_10.49.22/"
+folder3 = "2020-09-30_20.40.04/"
+folder4 = "2020-10-01_08.53.50/"
+
+set key box opaque
+set pointsize 1.5
+
+set ylabel "Time per particle (ms)"
+set xrange [0:8000]
+set xlabel "Iteration"
+
+
+set multiplot layout 2,3
+set title "Total"
+plot folder1."cpuTime.dat" u 1:(1e3*$3) t "Step 1"       ls 1, \
+     folder2."cpuTime.dat" u 1:(1e3*$3) t "Step 2"       ls 2, \
+     folder3."cpuTime.dat" u 1:(1e3*$3) t "Step 3"       ls 3, \
+     folder4."cpuTime.dat" u 1:(1e3*$3) t "Step 4"       ls 4
+	 
+set title "Push"
+plot folder1."cpuTime.dat" u 1:(1e3*$4) t "Step 1"       ls 1, \
+     folder2."cpuTime.dat" u 1:(1e3*$4) t "Step 2"       ls 2, \
+     folder3."cpuTime.dat" u 1:(1e3*$4) t "Step 3"       ls 3, \
+     folder4."cpuTime.dat" u 1:(1e3*$4) t "Step 4"       ls 4
+	 
+set title "Reset"
+plot folder1."cpuTime.dat" u 1:(1e3*$5) t "Step 1"       ls 1, \
+     folder2."cpuTime.dat" u 1:(1e3*$5) t "Step 2"       ls 2, \
+     folder3."cpuTime.dat" u 1:(1e3*$5) t "Step 3"       ls 3, \
+     folder4."cpuTime.dat" u 1:(1e3*$5) t "Step 4"       ls 4
+	 
+set title "Collisions"
+plot folder1."cpuTime.dat" u 1:(1e3*$6) t "Step 1"       ls 1, \
+     folder2."cpuTime.dat" u 1:(1e3*$6) t "Step 2"       ls 2, \
+     folder3."cpuTime.dat" u 1:(1e3*$6) t "Step 3"       ls 3, \
+     folder4."cpuTime.dat" u 1:(1e3*$6) t "Step 4"       ls 4
+	 
+set title "Weighting"
+plot folder1."cpuTime.dat" u 1:(1e3*$7) t "Step 1"       ls 1, \
+     folder2."cpuTime.dat" u 1:(1e3*$7) t "Step 2"       ls 2, \
+     folder3."cpuTime.dat" u 1:(1e3*$7) t "Step 3"       ls 3, \
+     folder4."cpuTime.dat" u 1:(1e3*$7) t "Step 4"       ls 4
+	 

runs/cylFlow/input.json

@@ -1,8 +1,8 @@
 {
   "output": {
     "path": "./runs/cylFlow/",
-	"trigger": 100,
+	"trigger": 10,
-	"cpuTime": false,
+	"cpuTime": true,
 	"numColl": false
   },
   "geometry": {
@@ -38,5 +38,10 @@
     "collisions": [
       {"species_i": "Argon", "species_j": "Argon", "crossSections": ["Ar-Ar_Elastic.dat"]}
     ]
+  },
+  "parallel": {
+    "OpenMP":{
+      "nThreads": 24
+    }
   }
 }

runs/cylFlow/plotCPUTime.gp

@@ -0,0 +1,49 @@
+reset
+set terminal qt enhanced 1 size 1800, 400 font "Times ,10"
+
+set style line 1  pt 4              lc rgb "#B50427" #Squares   red
+set style line 2  pt 6              lc rgb "#3B4CC1" #Circles   blue
+set style line 3  pt 1              lc rgb "#2CA02C" #Crosses   green
+set style line 4  pt 2              lc rgb "#FE7F0E" #Exes      orange
+set style line 5  pt 8              lc rgb "#D6696B" #Triangles light red
+set style line 10       lt 1 lw 2.0 lc rgb "black"   #Black solid line
+
+set style data histogram
+set style histogram rowstacked title offset 0,-1
+set style fill solid border -1
+set boxwidth 0.50
+
+set key box opaque
+
+set yrange [0:100]
+set ylabel "CPU Time (%)"
+set xrange [0:]
+
+
+plot "cpuTime.dat" u ($4/$3*100) t "Push"       ls 1, \
+     ""            u ($5/$3*100) t "Reset"      ls 2, \
+	 ""            u ($6/$3*100) t "Collisions" ls 3, \
+	 ""            u ($7/$3*100) t "Weighting"  ls 4
+	 
+reset
+set terminal qt enhanced 2 size 800, 600 font "Times ,10"
+
+set style line 1  pt 4              lc rgb "#B50427" #Squares   red
+set style line 2  pt 6              lc rgb "#3B4CC1" #Circles   blue
+set style line 3  pt 1              lc rgb "#2CA02C" #Crosses   green
+set style line 4  pt 2              lc rgb "#FE7F0E" #Exes      orange
+set style line 5  pt 8              lc rgb "#D6696B" #Triangles light red
+set style line 10       lt 1 lw 2.0 lc rgb "black"   #Black solid line
+
+set key box opaque
+set pointsize 1.5
+
+set ylabel "Time per particle ({/Symbol m}s)"
+set xrange [0:]
+set xlabel "Iteration"
+
+
+plot "cpuTime.dat" u 1:(1e6*$4/$2) t "Push"        ls 1, \
+     ""            u 1:(1e6*$5/$2) t "Reset"       ls 2, \
+	 ""            u 1:(1e6*$6/$2) t "Collisions"  ls 3, \
+	 ""            u 1:(1e6*$7/$2) t "Weighting"   ls 4

src/DSMC_Neutrals.f95

@@ -10,17 +10,15 @@ PROGRAM DSMC_Neutrals
   USE moduleList
   USE moduleOutput
   USE moduleCaseParam
+  USE moduleParallel
   IMPLICIT NONE
 
   ! t = time step; n = number of particle; e = number of element; i = number of inject
   INTEGER:: t, n, e, i
   CHARACTER(200):: arg1
   CHARACTER(:), ALLOCATABLE:: inputFile
-  REAL(8), EXTERNAL::omp_get_wtime, omp_get_wtick
+  REAL(8), EXTERNAL::omp_get_wtime
 
-  !Starts threads for OpenMP parallelization
-  !TODO: make this an input parameter
-  CALL OMP_SET_NUM_THREADS(8)
 
   !Gets the input file
   CALL getarg(1, arg1)
@@ -39,6 +37,10 @@ PROGRAM DSMC_Neutrals
   PRINT *, "t/tmax: ", t, "/", tmax
   PRINT *, "Particles: ", n_part_old
 
+  !Starts threads for OpenMP parallelization
+  !TODO: make this an input parameter
+  CALL OMP_SET_NUM_THREADS(openMP%nThreads)
+
   !$OMP PARALLEL PRIVATE(t) DEFAULT(SHARED) 
   DO t = 1, tmax
     tStep = omp_get_wtime()

src/makefile

@@ -4,7 +4,7 @@ OBJECTS = $(OBJDIR)/$(OUTPUT).o \
 			  	$(OBJDIR)/moduleErrors.o $(OBJDIR)/moduleList.o $(OBJDIR)/moduleOutput.o \
 					$(OBJDIR)/moduleMeshCyl.o	$(OBJDIR)/moduleMeshCylRead.o	$(OBJDIR)/moduleMeshCylBoundary.o \
 					$(OBJDIR)/moduleBoundary.o $(OBJDIR)/moduleCaseParam.o $(OBJDIR)/moduleRefParam.o \
-					$(OBJDIR)/moduleCollisions.o $(OBJDIR)/moduleTable.o
+					$(OBJDIR)/moduleCollisions.o $(OBJDIR)/moduleTable.o $(OBJDIR)/moduleParallel.o
 
 
 all: $(OUTPUT)

src/modules/makefile

@@ -2,14 +2,15 @@
 OBJS = moduleCaseParam.o moduleCompTime.o moduleList.o\
 			 moduleMesh.o moduleMeshCyl.o moduleMeshCylBoundary.o\
 			 moduleMeshCylRead.o moduleOutput.o moduleInput.o \
-			 moduleSolver.o moduleCollisions.o moduleTable.o
+			 moduleSolver.o moduleCollisions.o moduleTable.o \
+			 moduleParallel.o
 
 all: $(OBJS)
 
 moduleCollisions.o: moduleTable.o moduleSpecies.o moduleRefParam.o moduleConstParam.o moduleMeshCyl.f95
 	$(FC) $(FCFLAGS) -c $(subst .o,.f95, $@) -o $(OBJDIR)/$@
 
-moduleInput.o: moduleRefParam.o moduleCaseParam.o moduleSolver.o moduleInject.o moduleBoundary.o moduleMesh.o moduleMeshCylRead.o moduleErrors.o moduleSpecies.o moduleInput.f95
+moduleInput.o: moduleParallel.o moduleRefParam.o moduleCaseParam.o moduleSolver.o moduleInject.o moduleBoundary.o moduleMesh.o moduleMeshCylRead.o moduleErrors.o moduleSpecies.o moduleInput.f95
 	$(FC) $(FCFLAGS) -c $(subst .o,.f95, $@) -o $(OBJDIR)/$@
 
 moduleInject.o: moduleSpecies.o moduleSolver.o moduleMesh.o moduleMeshCyl.o moduleInject.f95

src/modules/moduleInput.f95

@@ -45,6 +45,9 @@ MODULE moduleInput
       !Read injection of particles
       CALL readInject(config)
 
+      !Read parallel parameters
+      CALL readParallel(config)
+
     END SUBROUTINE readConfig
 
     !Reads the reference parameters
@@ -325,9 +328,9 @@ MODULE moduleInput
 
       TYPE(json_file), INTENT(inout):: config
       INTEGER:: i
-      character(2):: istring
+      CHARACTER(2):: istring
-      character(:), allocatable:: object
+      CHARACTER(:), ALLOCATABLE:: object
-      logical:: found
+      LOGICAL:: found
       CHARACTER(:), ALLOCATABLE:: speciesName
       CHARACTER(:), ALLOCATABLE:: name
       REAL(8):: v
@@ -366,4 +369,28 @@ MODULE moduleInput
       
     END SUBROUTINE readInject
 
+    SUBROUTINE readParallel(config)
+      USE moduleParallel
+      USE moduleErrors
+      USE json_module
+      IMPLICIT NONE
+
+      TYPE(json_file), INTENT(inout):: config
+      CHARACTER(:), ALLOCATABLE:: object
+      LOGICAL:: found
+
+      !Reads OpenMP parameters
+      object = 'parallel.OpenMP'
+
+      CALL config%get(object // '.nThreads', openMP%nThreads, found)
+
+      IF (.NOT. found) THEN
+        openMP%nthreads = 8
+        CALL warningError('No OpenMP configuration detected, using 8 threads as default.')
+
+      END IF
+
+    END SUBROUTINE readParallel
+
+
 END MODULE moduleInput

src/modules/moduleParallel.f95

@@ -0,0 +1,10 @@
+MODULE moduleParallel
+  
+  TYPE openMP_type
+    INTEGER:: nThreads
+
+  END TYPE
+
+  TYPE(openMP_type):: openMP
+
+END MODULE moduleParallel