Rod wrote:
> On Jan 8, 2008 3:40 PM, Duncan Murdoch <[EMAIL PROTECTED]> wrote:
>
>> On 1/8/2008 8:49 AM, Rod wrote:
>>
>>> On Jan 8, 2008 12:41 PM, Duncan Murdoch <[EMAIL PROTECTED]> wrote:
>>>
>>>> Rod wrote:
>>>>
>>>>> Hello,
>>>>>
>>>>> I have a memory problem when I run package WinBUGS with R (2.6.1).
>>>>> Until now I have been using a Pentium IV 3.2Ghz computer with 512Mb of
>>>>> RAM memory (with Windows XP Pro SP2), and I hadn't had any problem.
>>>>> Now I have a new computer with the following characteristics: Intel
>>>>> Core 2 duo 2.33Ghz and 3Gb of RAM memory (with Windows XP Pro SP2) and
>>>>> when I am running the same R syntax I receive the following error
>>>>> message:
>>>>>
>>>>> "Error: cannot allocate vector of size 216.0 Mb".
>>>>>
>>>>> This is an extract of my R syntax:
>>>>>
>>>>>
>>>>>
>>>>>> memory.limit(4000)
>>>>>>
>>>>>>
>>>>> NULL
>>>>>
>>>>>
>>>>>> simu.BYM<-bugs(data, inits, parameters, model.file="Modelo.txt",
>>>>>> n.chains=3, n.iter=200000, n.burnin=20000, n.thin=180, debug=FALSE,
>>>>>> DIC=FALSE, digits=5, codaPkg=FALSE, bugs.directory="c:/Archivos de
>>>>>> programa/WinBUGS14", working.directory=NULL)
>>>>>>
>>>>>>
>>>>> Error: cannot allocate vector of size 216.0 Mb
>>>>>
>>>>> I have tried:
>>>>> -Change size of virtual memory inside Windows XP
>>>>> -change the R memory settings
>>>>> -run from DOS (R CMD BATCH...)
>>>>> -reviewed the R Windows FAQ and R-Help.
>>>>>
>>>>> but I haven't be able to find a solution !
>>>>>
>>>> Could you try R-patched? I recently fixed a minor bug that could have
>>>> symptoms like this. Under certain conditions it would miscalculate the
>>>> amount of available memory. You can download a copy from
>>>> cran.r-project.org/bin/windows/base/rpatched.html.
>>>>
>>>> Duncan Murdoch
>>>>
>>>>
>>> Dear Duncan,
>>>
>>> Thank for your advice. But, the patched version produce the same error.
>>>
>> In that case, you are probably really running out of memory. It's hard
>> to see why this would happen on the larger machine and not the smaller
>> one, but perhaps the code tries to adapt itself to available resources.
>> You are likely to have a bit of work diagnosing the problem.
>>
>> The tools to use are memory.size(), memory.limit() (which report on what
>> is available); memory.profile() (which gives more detail on usage).
>> Rprofmem() would also be nice, but it's not enabled in the standard
>> Windows build.
>>
>> Duncan Murdoch
>>
>>
>
> Dear Duncan,
>
> I have tried to run the same syntax in another pc (with exactly the
> same mark and model as mine) and I have obtained the same error. Then
> we can rule out that it is a problem with the hardware of my pc.
>
> I have tried to change the memory with command --max-mem-size=4000M
> ("c:\...\Rgui.exe" --max-mem-size=4000Mb) but I obtain this alert:
>
> WARNING: --max-mem-size=4000M: too large and taken as 2047
>
Windows normally reserves 2 GB for itself. You can reduce this to 1 GB
(the method is described in the R Windows FAQ 2.9), but there's no way
to allocate more than 3 GB to a process in 32 bit Windows.
More below...
> Here I show you the parameters of memory that I obtain before and
> after I run my syntax, and the gc() results obtained during the
> process. Perhaps this values can help you.
>
> thank you,
>
> Rodrigo
>
>
> ############# Memory values before I run the syntax###############
>
>
>> memory.limit()
>>
> [1] 1535.875
>
>> memory.size()
>>
> [1] 29.42623
>
>> memory.profile()
>>
> NULL symbol pairlist closure environment promise
> 1 9368 407965 12045 2013 7091
> language special builtin char logical integer
> 107052 269 2397 11503 13032 8429
> double complex character ... any list
> 11122 17 122950 1 0 7535
> expression bytecode externalptr weakref raw
> 1 0 1341 359 1
>
>> gc()
>>
> used (Mb) gc trigger (Mb) max used (Mb)
> Ncells 727690 19.5 1166886 31.2 984024 26.3
> Vcells 282062 2.2 786432 6.0 519695 4.0
>
> ############# Memory values after I run the syntax###############
>
>
>> memory.limit()
>>
> [1] 4000
>
>> memory.size()
>>
> [1] 1417.948
>
>> memory.profile()
>>
> NULL symbol pairlist closure environment promise
> 1 9921 466736 12211 2013 7091
> language special builtin char logical integer
> 114491 269 2397 14534 13341 46991
> double complex character ... any list
> 35508 17 131349 1 0 14663
> expression bytecode externalptr weakref raw
> 1 0 1324 342 1
>
>> gc()
>>
> used (Mb) gc trigger (Mb) max used (Mb)
> Ncells 876371 23.5 1835812 49.1 1835812 49.1
> Vcells 1247398 9.6 121622414 928.0 180932899 1380.5
>
>> memory.size(TRUE)
>>
> [1] 1541.688
>
>
> ############# gc() values during the process###############
>
>
> Garbage collection 330 = 271+30+29 (level 2) ...
> 23.6 Mbytes of cons cells used (48%)
> 9.9 Mbytes of vectors used (8%)
> Garbage collection 331 = 271+30+30 (level 2) ...
> 23.6 Mbytes of cons cells used (48%)
> 117.8 Mbytes of vectors used (49%)
> Garbage collection 332 = 271+30+31 (level 2) ...
> 23.6 Mbytes of cons cells used (48%)
> 225.8 Mbytes of vectors used (64%)
> Garbage collection 333 = 271+30+32 (level 2) ...
> 23.6 Mbytes of cons cells used (48%)
> 333.8 Mbytes of vectors used (72%)
> Garbage collection 334 = 272+30+32 (level 0) ...
> 23.6 Mbytes of cons cells used (48%)
> 441.9 Mbytes of vectors used (95%)
... deletions
> Garbage collection 454 = 369+38+47 (level 2) ...
> 24.2 Mbytes of cons cells used (49%)
> 1217.2 Mbytes of vectors used (91%)
> Garbage collection 455 = 369+38+48 (level 2) ...
> 24.2 Mbytes of cons cells used (49%)
> 1325.1 Mbytes of vectors used (91%)
> Garbage collection 456 = 369+38+49 (level 2) ...
> 24.4 Mbytes of cons cells used (50%)
> 1380.4 Mbytes of vectors used (82%)
> Garbage collection 457 = 370+38+49 (level 0) ...
> 24.4 Mbytes of cons cells used (50%)
> 1380.4 Mbytes of vectors used (82%)
> Error: cannot allocate vector of size 216.0 Mb
>
It looks as though your memory use really is growing over the run, up to
1.4 GB. At this point the memory manager was unable to find a 216 MB
block.
This looks like a problem in your code, or in the package: you seem to
have a memory leak. You can't increase memory indefinitely, eventually
you'll run out.
Duncan Murdoch
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
