paralellize plots

fuzzers/FRET/benchmark/plot_multi.r

@@ -1,10 +1,18 @@
 library("mosaic")
 library("dplyr")
+library("foreach")
+library("doParallel")
+
+#setup parallel backend to use many processors
+cores=detectCores()
+cl <- makeCluster(cores[1]-1) #not to overload your computer
+registerDoParallel(cl)
+
 args = commandArgs(trailingOnly=TRUE)
 
 if (length(args)==0) {
-  runtype="timedump_253048_1873f6/timedump"
+  runtype="timedump_253048_1873f6_full/timedump"
-  target="waters"
+  target="watersv2"
   outputpath="~/code/FRET/LibAFL/fuzzers/FRET/benchmark/"
   #MY_SELECTION <- c('state', 'afl', 'graph', 'random')
   SAVE_FILE=TRUE
@@ -85,57 +93,94 @@ frame2maxlines <- function(tr) {
   return(tr)
 }
 
-all_maxlines = c()
+trace2maxpoints <- function(tr) {
-for (bn in BASENAMES) {
+  minval = tr[1,1]
+  collect = tr[1,]
+  for (i in seq_len(dim(tr)[1])) {
+    if (minval < tr[i,1]) {
+      collect = rbind(collect,tr[i,])
+      minval = tr[i,1]
+    }
+  }
+  return(collect)
+}
+
+sample_maxpoints <- function(tr,po) {
+  index = 1
+  collect=NULL
+  for (p in po) {
+    done = FALSE
+    while (!done) {
+      if (p<=tr[1,2] || (index < dim(tr)[1] && tr[index,2] <= p && p < tr[index+1,2])) {
+        tmp = tr[index,]
+        tmp[2] = p
+        collect = rbind(collect, tmp)
+        done = TRUE
+      } else { if ( p >= tr[dim(tr)[1],2] ) {
+        tmp = tr[dim(tr)[1],]
+        tmp[2] = p
+        collect = rbind(collect, tmp)
+        done = TRUE
+      } else {
+        index = index + 1
+      } }
+    }
+  }
+  return(collect)
+}
+
+#https://www.r-bloggers.com/2012/01/parallel-r-loops-for-windows-and-linux/
+all_runtypetables <- foreach (bn=BASENAMES) %do% {
   runtypefiles <- list.files(file.path(BENCHDIR,bn),pattern=sprintf(PATTERNS,target),full.names = TRUE)
   if (length(runtypefiles) > 0) {
-    runtypetables <- lapply(seq_len(length(runtypefiles)),
+    runtypetables_reduced <- foreach(i=seq_len(length(runtypefiles))) %dopar% {
-        function(i)read.csv(runtypefiles[[i]], col.names=c(sprintf("%s%d",bn,i),"times")))
+      rtable = read.csv(runtypefiles[[i]], col.names=c(sprintf("%s%d",bn,i),sprintf("times%d",i)))
-    runtypetables = trim_data(runtypetables)
+      trace2maxpoints(rtable)
-    runtypetables = lapply(runtypetables, function(i) i[1])
+    }
-    list_of_maxlines = data2maxlines(runtypetables)
+    #runtypetables <- lapply(seq_len(length(runtypefiles)),
-    tmp_frame <- Reduce(bind_cols, list_of_maxlines)
+    #    function(i)read.csv(runtypefiles[[i]], col.names=c(sprintf("%s%d",bn,i),sprintf("times%d",i))))
-    statframe <- bind_cols(rowMeans(tmp_frame),apply(tmp_frame, 1, sd),apply(tmp_frame, 1, min),apply(tmp_frame, 1, max))
+    #runtypetables_reduced <- lapply(runtypetables, trace2maxpoints)
+    runtypetables_reduced
+    #all_runtypetables = c(all_runtypetables, list(runtypetables_reduced))
+  }
+}
+all_runtypetables = all_runtypetables[lapply(all_runtypetables, length) > 0]
+all_points = sort(unique(Reduce(c, lapply(all_runtypetables, function(v) Reduce(c, lapply(v, function(w) w[[2]]))))))
+all_maxlines <- foreach (rtt=all_runtypetables) %dopar% {
+  bn = substr(names(rtt[[1]])[1],1,nchar(names(rtt[[1]])[1])-1)
+  runtypetables_sampled = lapply(rtt, function(v) sample_maxpoints(v, all_points)[1])
+  tmp_frame <- Reduce(cbind, runtypetables_sampled)
+  statframe <- data.frame(rowMeans(tmp_frame),apply(tmp_frame, 1, sd),apply(tmp_frame, 1, min),apply(tmp_frame, 1, max))
   names(statframe) <- c(bn, sprintf("%s_sd",bn), sprintf("%s_min",bn), sprintf("%s_max",bn))
-    all_maxlines = c(all_maxlines, list(round(statframe)))
+  #statframe[sprintf("%s_times",bn)] = all_points
-    #all_maxlines = append(all_maxlines, list_of_maxlines)
+  round(statframe)
-    #mean_maxline<-Reduce(function(a,b) a+b,list_of_maxlines,0)/length(runtypetables)
+  #all_maxlines = c(all_maxlines, list(round(statframe))) 
-    #all_maxlines=append(all_maxlines,mean_maxline)
 }
-}
-min_length <- min(sapply(all_maxlines, function(x) dim(x)[1]))
-all_maxlines=lapply(all_maxlines, function(v) v[1:min_length,])
 one_frame<-data.frame(all_maxlines)
-one_frame[length(one_frame)+1] <- seq_len(length(one_frame[[1]]))
+one_frame[length(one_frame)+1] <- all_points/(3600 * 1000)
-names(one_frame)[length(one_frame)] <- 'iters'
+names(one_frame)[length(one_frame)] <- 'time'
 
-typenames = names(one_frame)[which(names(one_frame) != 'iters')]
+typenames = names(one_frame)[which(names(one_frame) != 'time')]
+typenames = typenames[which(!endsWith(typenames, "_sd"))]
 ylow=min(one_frame[typenames])
 yhigh=max(one_frame[typenames],worst_case)
-typenames = typenames[which(!endsWith(typenames, "_sd"))]
 typenames = typenames[which(!endsWith(typenames, "_min"))]
 typenames = typenames[which(!endsWith(typenames, "_max"))]
-#yhigh=3400000
-#yhigh=max(one_frame[typenames],405669)
 
 ml2lines <- function(ml) {
   lines = NULL
-  last = -1
+  last = 0
-  for (i in seq_len(length(ml))) {
+  for (i in seq_len(dim(ml)[1])) {
-    if (ml[[i]] != last || (i >= length(ml))) {
+    lines = rbind(lines, cbind(X=last, Y=ml[i,1]))
-      if (i != 1) {
+    lines = rbind(lines, cbind(X=ml[i,2], Y=ml[i,1]))
-        lines = rbind(lines, cbind(X=i, Y=last)) 
+    last = ml[i,2]
-      }
-      lines = rbind(lines, cbind(X=i, Y=ml[[i]]))
-      last=ml[[i]]
-    }
   }
   return(lines)
 }
 
 plotting <- function(selection, filename, MY_COLORS_) {
 # filter out names of iters and sd cols
-typenames = names(one_frame)[which(names(one_frame) != 'iters')]
+typenames = names(one_frame)[which(names(one_frame) != 'times')]
 typenames = typenames[which(!endsWith(typenames, "_sd"))]
 typenames = typenames[which(!endsWith(typenames, "_min"))]
 typenames = typenames[which(!endsWith(typenames, "_max"))]
@@ -149,16 +194,16 @@ if (SAVE_FILE) {png(file=sprintf("%s%s_%s.png",outputpath,target,filename), widt
 par(mar=c(4,4,1,1))
 par(oma=c(0,0,0,0))
 
-plot(c(1,length(one_frame[[1]])),c(ylow,yhigh), col='white', xlab="Iters", ylab="WORT", pch='.')
+plot(c(1,max(one_frame['time'])),c(ylow,yhigh), col='white', xlab="Time [h]", ylab="WORT [insn]", pch='.')
 
 for (t in seq_len(length(typenames))) {
   proj = one_frame[seq(1, dim(one_frame)[1], by=max(1, length(one_frame[[1]])/(10*w_))),]
   #points(proj[c('iters',typenames[t])], col=MY_COLORS_[t], pch='.')
-  avglines = ml2lines(one_frame[[typenames[t]]])
+  avglines = ml2lines(one_frame[c(typenames[t],'time')])
   lines(avglines, col=MY_COLORS_[t])
   if (exists("RIBBON") && ( RIBBON=='both' || RIBBON=='span')) {
-    milines = ml2lines(one_frame[[sprintf("%s_min",typenames[t])]])
+    milines = ml2lines(one_frame[c(sprintf("%s_min",typenames[t]),'time')])
-    malines = ml2lines(one_frame[[sprintf("%s_max",typenames[t])]])
+    malines = ml2lines(one_frame[c(sprintf("%s_max",typenames[t]),'time')])
     lines(milines, col=MY_COLORS_[t], lty='dashed')
     lines(malines, col=MY_COLORS_[t], lty='dashed')
     #points(proj[c('iters',sprintf("%s_min",typenames[t]))], col=MY_COLORS_[t], pch='.')
@@ -167,7 +212,7 @@ for (t in seq_len(length(typenames))) {
   if (exists("RIBBON") && RIBBON != '') {
   for (i in seq_len(dim(proj)[1])) {
     row = proj[i,]
-    x_ <- row['iters'][[1]]
+    x_ <- row['time'][[1]]
     y_ <- row[typenames[t]][[1]]
     sd_ <- row[sprintf("%s_sd",typenames[t])][[1]]
     min_ <- row[sprintf("%s_min",typenames[t])][[1]]
@@ -203,9 +248,9 @@ if (exists("MY_SELECTION")) {
   #MY_SELECTION=c('state', 'afl', 'random', 'feedlongest', 'feedgeneration', 'feedgeneration10')
   #MY_SELECTION=c('state_int', 'afl_int', 'random_int', 'feedlongest_int', 'feedgeneration_int', 'feedgeneration10_int')
   #MY_SELECTION=c('state', 'frAFL', 'statenohash', 'feedgeneration10')
-  #MY_SELECTION=c('state_int', 'frAFL_int', 'statenohash_int', 'feedgeneration10_int')
+  MY_SELECTION=c('state_int', 'frAFL_int', 'statenohash_int', 'feedgeneration10_int')
   MY_SELECTION=typenames
-  RIBBON='both'
+  RIBBON='span'
   for (i in seq_len(length(MY_SELECTION))) {
     n <- MY_SELECTION[i]
     plotting(c(n), n, c(MY_COLORS[i]))