[R] Segmentation Fault on Unix box with nloptr, works on Windows

[Corak, Robert (US - Newton)]

I have an R script that is giving me a Segmentation Fault depending on the size of the dataset.  It is only happening on our Unix installation of R Server.  I am able to run it against a Windows server with the exact same data and script successfully.

The Segmentation Fault occurs when I call nloptr.  The data I am passing in only has about 1350 records.  I have a print level of 2 set for the nloptr call.  I have been trying different record counts and was able to get to a point where I would either:


1)      1395 Recs: Fault immediately after calling the nloptr function

2)      1394 Recs: Start the optimization iteration inside nloptr for 1 iteration then fault

3)      1393 Recs: Get the optimization to iterate around 130 times before faulting

4)      1392 Recs: Getting the optimization to succeed

I am running the script from the command line using "Rscript myscript.R"

I have a tryCatch around the call but it just crashes with no additional info and is never caught.

I would assume that this is memory related but it looks like there is plenty of memory resources available (at least within the JVM).  I also tried to call:

options(java.parameters = "-Xmx8192m")

but it didn't seem to help.

These are the libraries I am installing:

*         library(rJava)

*         library(RJDBC)

*         library(RCurl)

*         library(stringr)

*         library(nloptr)

*         library(gsubfn)

Below is the setup snippet for the nloptr call:

# 4) Pick Algorithm to be used
                local_opts <- list (
                                "algorithm"   = "NLOPT_LD_MMA",
                                "xtol_rel"    = 1.0e-7
                )
                opts <- list (
                                "algorithm"   = "NLOPT_LD_AUGLAG",
                                "xtol_rel"    = 1.0e-7,
                                "maxeval"     = 1000,
                                "print_level" = 2,
                                "local_opts"  = local_opts
                )

# 5) Do optimization
                optRes <- nloptr (
                                x0          = x0,
                                eval_f      = eval_f,
                                lb          = lb,
                                ub          = ub,
                                eval_g_ineq = eval_g_ineq,
                                opts        = opts
                )

Does anyone have any thoughts or ideas on what might be happening here?

Thanks

Rob





This message (including any attachments) contains confid...{{dropped:18}}