On 12.01.2010 21:07, Alexander Shenkin wrote:
Hi Alessia,
Note that, while your physical limit might be 6 GB, Windows memory
management allows more memory than that to be allocated (aka Virtual
Memory, or at least that's what they called it in XP). ?Windows swaps
out memory from RAM to the hard disk and back when necessary (please
excuse the explanation if you already know all this). ?For processing
large vectors, this swapping might bring your system to a standstill.
Regardless, the "maximum" memory for a windows process is larger than
the physical RAM you have available.
allie
In this case 6Gb was the default (as physical maximum in the particular
machine) and there was bug in the *experimental* version of R that did not
allow to increase memory size from within R using memory.limit() which
already has been fixed thanks to Brian Ripley.
Uwe Ligges
On 1/12/2010 6:27 AM, alessia matano wrote:
Fine, it worked. I will try in this way.
Just the last question and I won't bother you further today. My
machine right now has just 6 giga of RAM (it will be increased to 16
in a few days), and I see that with this experimental version
memory.limit is 6135.
How is the command to increase the memory usage until the maximum I
can (5 giga?). If I am writing memory.limit(5000) it still gives me
the error:
don't be silly! Your machine has a 4Gb address limit
which is quite odd.
Many thanks
Best
A.
2010/1/12 alessia matano<alexis.rtd at gmail.com>:
ok, perfect!
I will try with it...many many thanks. Have you got there also the
quantreg package, which has actually the same problem of sparseM
(32bit version)?
best
alessia
2010/1/12 Uwe Ligges<ligges at statistik.tu-dortmund.de>:
On 12.01.2010 12:09, alessia matano wrote:
I am sorry, I know it is an experimental version, and I have been
misleading saying a new version.
Therefore, I will wait for when they will be available officially,
since it is just a few days.
Or just use today my private repository I indicated in the other mail.
Uwe Ligges
However, I tried also to go to the cran pages and download them and
insert into the library. For quantreg it worked, for sparseM it did
not probably because it's a win32 version, as you said.
2010/1/12 Prof Brian Ripley<ripley at stats.ox.ac.uk>:
On Tue, 12 Jan 2010, alessia matano wrote:
Dear all,
I just download and set this new version of R. I am now trying to
download the packages I need which are sperseM and quantreg. I
downloaded and insert into the library file the quantreg pacjkage
and
it seems to work. However, when I try to do the same with sparseM I
get the following error message:
Loading required package: SparseM
Error in inDL(x, as.logical(local), as.logical(now), ...) :
?unable to load shared library
'C:/PROGRA~1/R/R-211~1.0DE/library/SparseM/libs/SparseM.dll':
?LoadLibrary failure: ?%1 non ? un'applicazione di Win32 valida.
Any help for it?
Please do refer to the posting referred to in that thread (and
Henrique,
please do not post just the URL without the explanations).
https://stat.ethz.ch/pipermail/r-devel/2010-January/056301.html
You cannot mix 32-bit Windows binary packages with this experimental
port
(it is not a 'new version'): you need to install from the package
sources.
?If that is too difficult for you, please do not try to use
unsupported
experimental builds (and Uwe Ligges may have some binary packages
available
for test in a few days).
Thanks a lot
alessia
2010/1/11 Henrique Dallazuanna<wwwhsd at gmail.com>:
Dear all,
do you know if there is any particular version of R to implement
with
windows 64 bit, in such a way to increase the amount of memory it
can
use?
How should I increase the memory, and more importantly to set a
higher
max vector size? It still stops me saying "Could not allocate
vector
of size 145"
thanks to all
alessia