[BioC] RIPSeeker multicore option not working fully?

[Davis, Wade]

I just wanted to report that, as expected, the suggestion of using options(mc.cores=) to increase the number of cores did work on Linux.

Wade

-----Original Message-----
From: Yue Li [mailto:yueli at cs.toronto.edu] 
Sent: Wednesday, July 31, 2013 3:59 PM
To: Dan Tenenbaum
Cc: Davis, Wade; bioconductor at r-project.org
Subject: Re: [BioC] RIPSeeker multicore option not working fully?

Hi Wade,

As Dan mentioned, that's most likely the cost. Other than that, it would be able to use however many cores you have on the machine when multicore set TRUE.

Thanks Dan.
Yue


On 2013-07-31, at 4:51 PM, Dan Tenenbaum <dtenenba at fhcrc.org> wrote:

> On Wed, Jul 31, 2013 at 1:28 PM, Davis, Wade 
> <davisjwa at health.missouri.edu> wrote:
>> Dear Yue,
>> I am using your RIPSeeker package and trying to use the option multicore=TRUE. Trying it on both  windows and Linux machines, I did not observe multiple cores being used until I first called library(parallel) before running the code below. Also, it appears that only 2 cores were being used on the Linux machine, and only 1 on the windows, even though detectCores() indicated 32 cores on the Linux and 8 on the windows.
>> 
>> Are there any additional parameters that I need to set to increase the number of cores being used?
>> 
> 
> I don't know about RIPSeeker in particular but it seems that it's 
> using parallel::mclapply under the hood, so looking at ?mclapply You 
> can see that mc.cores defaults to:
> 
> mc.cores = getOption("mc.cores", 2L)
> 
> Therefore you could set it as follows:
> 
> options(mc.cores=detectCores())
> 
> See if that gives you different results in linux.
> 
> In windows, mclapply will only work if mc.cores is set to 1. That is, 
> it doesn't matter how many cores you have on Windows, your task will 
> run serially.
> 
> Dan
> 
> 
> 
> 
>> Thanks,
>> Wade
>> 
>> #####################################################################
>> ###
>> 
>> library(ShortRead)
>> library(RIPSeeker)
>> library(parallel)
>> 
>> filedir<-"~/bulkdata/PI/RIP"
>> setwd(filedir)
>> bamFiles <- list.files( pattern="_2\\.bam", recursive=TRUE, 
>> full.names=TRUE) cNAME <- "IgG"
>> outDir <- file.path(filedir, "RIPSeeker_output_2") # Parameters 
>> setting
>> binSize <- NULL      # set to NULL to automatically determine bin size
>> multicore <- TRUE                            # use multicore
>> strandType <- NULL                       # set strand type to minus strand
>> biomart <- "ENSEMBL_MART_ENSEMBL"           # use archive to get ensembl 65
>> biomaRt_dataset <- "mmusculus_gene_ensembl" # mouse dataset id name
>> goAnno <- "org.Mm.eg.db"                    # GO annotation database
>> ################ run main function ripSeek to predict RIP 
>> ################ seekOut.HuR <- ripSeek(bamPath = bamFiles, cNAME = cNAME,
>>                    reverseComplement = FALSE, #genomeBuild = "mm10", not relevant for BAM
>>                    strandType = strandType,
>>                    uniqueHit = TRUE,
>>                    assignMultihits = TRUE,
>>                    rerunWithDisambiguatedMultihits = TRUE,
>>                    binSize=binSize,
>>                    #minBinSize = minBinSize,
>>                    #maxBinSize = maxBinSize,
>>                    biomart=biomart,
>>                    host=host,
>>                    biomaRt_dataset = biomaRt_dataset,
>>                    goAnno = goAnno,
>>                    multicore=multicore,
>>                    logOddCutoff = 2,
>>                    pvalCutoff = 1,
>>                                pvalAdjCutoff = 0.2,
>>                    eFDRCutoff = 0.2,
>>                    outDir=outDir)
>> 
>> 
>> 
>>        [[alternative HTML version deleted]]
>> 
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives: 
>> http://news.gmane.org/gmane.science.biology.informatics.conductor