Parallel Examples

Author:	Mitch Richling
Updated:	2022-06-04 16:17:46

1. Metadata

Files related to this document may be found on github: https://github.com/richmit/ex-R

Directory contents:

`src`	-	The org-mode file that generated this HTML document
`docs`	-	This html document
`data`	-	Data files
`tangled`	-	Tangled R code from this document

library(parallel)

instancesInMyComputeCluster <- detectCores()
instancesInMyComputeCluster

[1] 8

myComputeCluster <- makeCluster(instancesInMyComputeCluster)
myComputeCluster

socket cluster with 8 nodes on host 'localhost'

unlist(clusterCall(myComputeCluster, 'Sys.getpid'))

[1] 25704 17520 15828 18808 21660 21896 26024 12048

unlist(clusterEvalQ(myComputeCluster, 2*Sys.getpid()))

[1] 51408 35040 31656 37616 43320 43792 52048 24096

Notice aVar need not be exported because it is the second argument of parSapply!

aVar <- 1:10
parSapply(myComputeCluster, aVar, sin)
#parLapply(myComputeCluster, aVar, sin)

[1]  0.8414710  0.9092974  0.1411200 -0.7568025 -0.9589243 -0.2794155  0.6569866  0.9893582  0.4121185 -0.5440211

While aVar need not be exported, bVar must be exported – because it is not the second argument of parSapply.

bVar <- 10
clusterExport(myComputeCluster, 'bVar')
parSapply(myComputeCluster, aVar, function (x) bVar*x)

[1]  10  20  30  40  50  60  70  80  90 100

cVar <- rnorm(instancesInMyComputeCluster*2^14)

clusterExport(myComputeCluster, "cVar")

system.time(b<-sapply(cVar, function (x) for(i in 1:500) sin(x)))

user  system elapsed 
4.81    0.00    4.82

system.time(b<-parSapply(myComputeCluster, cVar, function (x) for(i in 1:500) sin(x)))

user  system elapsed 
0.07    0.00    1.12

stopCluster(myComputeCluster)