DA/raaql-paper-experiments Changeset - 6da277f993c5 · Centrum Wiskunde & Informatica (CWI)

Changeset - 6da277f993c5

Parent rev.

Child rev.

[Not reviewed]

0 20 0

Hannes Muehleisen - 9 years ago 2015-06-25 10:01:41
hannes@muehleisen.org

mc3

20 files changed with 33 insertions and 33 deletions:

defer.pdf

bin+mod

identity-noopt.pdf

bin+mod

identity-opt.pdf

bin+mod

identity.pdf

bin+mod

makedot.sh

makeplots.R

operators.R

operators.pdf

bin+mod

parallel-tree.pdf

bin+mod

parallel.pdf

bin+mod

pushdown-noopt.pdf

bin+mod

pushdown-opt.pdf

bin+mod

pushdown.pdf

bin+mod

recycling-noopt.pdf

bin+mod

recycling-opt.pdf

bin+mod

recycling.pdf

bin+mod

sqlsurvey.R

survey-noopt.pdf

bin+mod

survey-opt.pdf

bin+mod

survey.pdf

bin+mod

0 comments (0 inline, 0 general)

defer.pdf

➞

Show inline comments

binary diff not shown

identity-noopt.pdf

➞

Show inline comments

binary diff not shown

identity-opt.pdf

➞

Show inline comments

binary diff not shown

identity.pdf

➞

Show inline comments

binary diff not shown

makedot.sh

➞

Show inline comments

@@ @@ -2,13 +2,13 @@ @@
 DOTOPTS="-Nfontsize=15 -Tpdf"
 dot recycling-noopt.dot $DOTOPTS > recycling-noopt.pdf
 dot recycling-opt.dot $DOTOPTS > recycling-opt.pdf
 dot identity-noopt.dot $DOTOPTS > identity-noopt.pdf
+dot identity-noopt.dot -Granksep=0.3 $DOTOPTS > identity-noopt.pdf
 dot identity-opt.dot $DOTOPTS > identity-opt.pdf
 dot pushdown-noopt.dot $DOTOPTS > pushdown-noopt.pdf
 dot pushdown-opt.dot $DOTOPTS > pushdown-opt.pdf
 dot parallel.dot $DOTOPTS > parallel-tree.pdf

makeplots.R

➞

Show inline comments

@@ @@ -24,12 +24,39 @@ theme(axis.title.y=element_text(vjust=0.9), @@
 se <- function(x) sqrt(var(x)/length(x))
 # survey
 d <- all %>% filter(exp=="survey") %>% group_by(sys, s, conf) %>%
 	summarize(meant=mean(timesec), se=se(timesec)) %>%
 	mutate(tool=paste(sys,conf),
 		datasize=sapply(s, switch, alabama=47512, california=1060060, acs3yr=9093077)) %>%
 	filter(tool == "Renjin jitopt" | sys != "Renjin")
 print(d %>% select(sys,s,meant) %>% arrange(s))
 limits <- aes(ymax = meant + se, ymin=meant - se, width=2)
 pdf("survey.pdf",width=10,height=7)
 ggplot(d,aes(datasize,meant,group=tool)) +
   geom_point(size=4) + geom_line(size=1.5, aes(group=tool, linetype=tool)) +
   geom_pointrange(limits) +
   #scale_y_log10(breaks=ybreaks, labels=ylabels) +
     scale_x_log10(breaks=c(47512,1060060,9093077)) +
 	xlab("Dataset Size (elements, log scale)") + ylab("Execution Time (s)") + theme +
 	annotate("text", x=10^6.6, y=45, label="GNU R", family="serif", size=10)+
 	annotate("text", x=10^6.4, y=100, label="sqlsurvey", family="serif", size=10)+
 	annotate("text", x=10^6.8, y=15, label="Renjin", family="serif", size=10)
 	#scale_color_brewer(palette=cBrwPl) +
  #guides(colour=guide_legend(keywidth=3.5))
 dev.off()
 # parallel
@@ @@ -76,13 +103,12 @@ dev.off() @@
 # 	# annotate("text", x=3.5, y=20, label="Renjin ", family="serif", size=10)
 # print(p)
 # dev.off()
 stop()
 # recycling
 d <- all %>% filter(exp=="recycling", s > 5) %>% group_by(exp, sys, s, conf) %>%
 	summarize(meant=mean(timesec), se=se(timesec)) %>%
 	mutate(datasize=10^as.integer(s), tool=paste(sys, conf))
 print(d)
@@ @@ -102,39 +128,12 @@ p <- ggplot(d,aes(datasize,meant,group=tool)) + @@
 	annotate("text", x=10^7.8, y=21, label="GNU R", family="serif", size=10)
 print(p)
 dev.off()
 # survey
 d <- all %>% filter(exp=="survey") %>% group_by(sys, s, conf) %>%
 	summarize(meant=mean(timesec), se=se(timesec)) %>%
 	mutate(tool=paste(sys,conf),
 		datasize=sapply(s, switch, alabama=47512, california=1060060, acs3yr=9093077)) %>%
 	filter(tool == "Renjin jitopt" | sys != "Renjin")
 print(d)
 limits <- aes(ymax = meant + se, ymin=meant - se, width=2)
 pdf("survey.pdf",width=10,height=7)
 ggplot(d,aes(datasize,meant,group=tool)) +
   geom_point(size=4) + geom_line(size=1.5, aes(group=tool, linetype=tool)) +
   geom_pointrange(limits) +
   #scale_y_log10(breaks=ybreaks, labels=ylabels) +
     scale_x_log10(breaks=c(47512,1060060,9093077)) +
 	xlab("Dataset Size (elements, log scale)") + ylab("Execution Time (s)") + theme +
 	annotate("text", x=10^6.6, y=45, label="GNU R", family="serif", size=10)+
 	annotate("text", x=10^6.4, y=100, label="sqlsurvey", family="serif", size=10)+
 	annotate("text", x=10^6.8, y=15, label="Renjin", family="serif", size=10)
 	#scale_color_brewer(palette=cBrwPl) +
  #guides(colour=guide_legend(keywidth=3.5))
 dev.off()
 # # identity
 d <- all %>% filter(exp=="identity", s > 5) %>% group_by(exp, sys, conf, s) %>%
 	summarize(meant=mean(timesec), se=se(timesec)) %>%
 	mutate(tool=ifelse(conf=="none", sys, paste(sys,conf)), datasize=10^as.integer(s))
@@ @@ -160,13 +159,13 @@ ggplot(d,aes(datasize,meant,group=tool)) + @@
 dev.off()
 # pushdown
-d <- all %>% filter(exp=="pushdown", s > 4) %>% group_by(exp, sys, s) %>%
+d <- all %>% filter(exp=="pushdown", s > 5) %>% group_by(exp, sys, s) %>%
 	summarize(meant=mean(timesec), se=se(timesec)) %>%
 	mutate(tool=sys, datasize=10^as.integer(s))
 print(d)
 limits <- aes(ymax = meant + se, ymin=meant - se, width=2)
@@ @@ -176,13 +175,13 @@ p <- ggplot(d,aes(datasize,meant,group=tool)) + @@
   geom_point(size=4) + geom_line(size=1.5, aes(group=tool, linetype=tool)) +
   geom_pointrange(limits) +
  # scale_y_log10(breaks=ybreaks, labels=ylabels) +
     scale_x_log10(breaks=xbreaks, labels=xlabels) +
 	xlab("Dataset Size (elements, log scale)") + ylab("Execution Time (s)") + theme +
-	annotate("text", x=10^7.5, y=5, label="GNU R", family="serif", size=10)+
+	annotate("text", x=10^7.6, y=5, label="GNU R", family="serif", size=10)+
 	annotate("text", x=10^7.6, y=.4, label="Renjin ", family="serif", size=10)
 print(p)
 dev.off()

operators.R

➞

Show inline comments

@@ @@ -17,7 +17,8 @@ for (s in 4:sm) { @@
+			}
 			print(m)
 		})[[3]]
 		log.result("operators", sys, conf, s, r, timing)
 		clearResultRecycler()
+	}
+}
@@ \ No newline at end of file @@
+}

operators.pdf

➞

Show inline comments

binary diff not shown

parallel-tree.pdf

➞

Show inline comments

binary diff not shown

parallel.pdf

➞

Show inline comments

binary diff not shown

pushdown-noopt.pdf

➞

Show inline comments

binary diff not shown

pushdown-opt.pdf

➞

Show inline comments

binary diff not shown

pushdown.pdf

➞

Show inline comments

binary diff not shown

recycling-noopt.pdf

➞

Show inline comments

binary diff not shown

recycling-opt.pdf

➞

Show inline comments

binary diff not shown

recycling.pdf

➞

Show inline comments

binary diff not shown

sqlsurvey.R

➞

Show inline comments

     database="monetdb://localhost/acs2", driver=MonetDB.R(),
     key="idkey",user="monetdb",password="monetdb",table.name=s, check.factors=data.frame())
   for (r in 1:5) {
     timing <- system.time({
       print(svymean(~agep, svydsgn, se=TRUE))
-      print(svymean(~relp, svydsgn, se=TRUE))
+      print(svymean(~adjinc, svydsgn, se=TRUE))
     })[[3]]
     log.result("survey", "sqlsurvey", conf, s, r, timing)
+  }
+}
 # TODO: test with Oct14 branch, probably faster

survey-noopt.pdf

➞

Show inline comments

binary diff not shown

survey-opt.pdf

➞

Show inline comments

binary diff not shown

survey.pdf

➞

Show inline comments

binary diff not shown

0 comments (0 inline, 0 general)