On Feb 1, 2011, at 9:00 PM, Dario Strbenac wrote: > Hello, > > I'm trying to track down the cause of some extreme memory usage and I've been > using Dirk Eddelbuettel's lsos() function he posted on stack overflow. There > is a large difference between R's RAM usage : > > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > 6637 darstr 20 0 30.0g 29g 4712 S 0 63.2 10:34.43 R > > and what objects I have loaded in memory : > >> lsos() > Type Size PrettySize Rows Columns > A list 552387720 526.8 Mb 2 NA > B GRangesList 552376408 526.8 Mb 4 NA > C SimpleRleList 353421896 337 Mb 24 NA > D GRanges 236410608 225.5 Mb 15272853 NA > E data.frame 6981952 6.7 Mb 24966 14 > F data.frame 6782136 6.5 Mb 24966 13 > G list 4393704 4.2 Mb 24964 NA > H matrix 3195760 3 Mb 24964 16 > I list 1798752 1.7 Mb 24964 NA > J GRanges 312656 305.3 Kb 24964 NA > > (The total looks like about 1.5 GB) > > I haven't got any calls to external C code in my R script, although the > Bioconductor packages I am using do. How can I regain those missing Gigabytes > ? >
There is no reason why the two numbers should have anything in common. The OS (assuming the above is supposed to be ps output) reports memory that the OS reserved (it is not necessarily what the application is currently using) and lsos (assuming you meant to point to this: http://stackoverflow.com/questions/1358003/tricks-to-manage-the-available-memory-in-an-r-session ) simply shows approximate size estimates of objects in your workspace. So you can expect the former to be larger than the latter (often but not necessarily always) but that's about all you can say. If you want to know what R objects are allocated, just look at gc() - it has the actual numbers. However, that only includes transient allocations using the R GC, any memory reserved directly from the OS will not be included. Furthermore, many OSes don't release memory from processes for performance reasons so even if you had exact allocation numbers they would not match what you see in ps. I you want to lean more, you'll have to grab the tools available for your platform to find out who's allocating what. Since you didn't even mention your platform I can't really help you with the specifics, though. Cheers, Simon ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
