CSCD$543$Project$Proposal$
$
Darren$Altenhofen$
$
Brief$Summary:$$A$comparison$of$sequential$and$parallel$versions$of$radix$sort.$
$
Description:$$Various$sizes$of$random$integer$arrays$will$be$run$through$sequential$
and$parallel$versions$of$the$algorithm.$$The$goal$is$to$show$which$conditions$the$
algorithm$perform$better$and$worse$to$make$some$conclusions$into$why.$Various$
sizes$of$integer$arrays$will$be$used$and$multiple$runs$will$be$performed$to$average$
results.$$I$will$start$with$random$data,$and$then$experiment$with$pre$sorted$and$
other$data$patterns.$
$
Information:$$Radix$sort$is$not$a$comparison$sort$and$has$potential$to$be$quicker$
than$other$parallel$sorting$methods$such$as$merge$sort.$$For$the$project,$I$will$be$
using$the$most$significant$digit$(left$most$digit)$to$split$the$array$into$different$
“buckets”$based$on$that$digit.$$These$buckets$of$numbers$are$then$sent$out$to$
available$nodes$to$be$sorted$by$the$sequential$version$of$radix$sort$and$returned.$$At$
the$end,$the$buckets$are$concatenated$together$and$we$have$a$sorted$array$of$
integers.$
$ One$challenge$presented$is$accounting$for$the$possible$very$unbalanced$
bucket$sizes.$$I$will$look$into$how$to$create$some$load$balancing$sort$conditions.$
$
Results:$$Graphs$of$the$timing$results$will$present$data$findings.$$Some$analysis$will$
be$done$into$why$I$get$the$results$I$do.$
$
Possible$Expansion:$$If$time$allows,$I$will$possibly$add$more$test$conditions.$
$
$
$
$
$
$
$
Sources$for$picture:$
$
http://en.wikipedia.org/wiki/Radix_sort$
