Skip to content

Error: cannot allocate vector of size 3.0 Gb

6 messages · Liaw, Andy, Daniel Nordlund, Uwe Ligges +1 more

John
# 
Hello,

I have read recent posts on this topic (Dr. Ronnen Levinson's Monday 02:39:55 
pm), but before I install a 64 bit system, and purchase more RAM, I want to 
make sure I understand this interesting issue.

I was attempting to plot a data frame containing Dow Jones stock information:
[1] 19957
[1] 19957
Date     Open     High      Low    Close     Volume Adj.Close
1 03/26/08 12531.79 12531.79 12309.62 12422.86 4055670000  12422.86
2 03/25/08 12547.34 12639.82 12397.62 12532.60 4145120000  12532.60
3 03/24/08 12361.97 12687.61 12346.17 12548.64 4499000000  12548.64
4 03/20/08 12102.43 12434.34 12024.68 12361.32 2078450000  12361.32
5 03/19/08 12391.52 12525.19 12077.27 12099.66 1203830000  12099.66
6 03/18/08 11975.92 12411.63 11975.92 12392.66 1263390000  12392.66
Date   Open   High    Low  Close  Volume Adj.Close
19952 10/08/28 240.17 243.33 237.72 239.55 3940000    239.55
19953 10/05/28 240.00 243.08 238.22 240.44 4360000    240.44
19954 10/04/28 237.75 242.53 237.72 240.00 4330000    240.00
19955 10/03/28 238.14 239.14 233.60 237.75 4060000    237.75
19956 10/02/28 240.01 241.54 235.42 238.14 3850000    238.14
19957 10/01/28 239.43 242.46 238.24 240.01 3500000    240.01
Error: cannot allocate vector of size 3.0 Gb
I currently have installed R 2.60, from source, onto Linux SuSE 10.2. I 
believe it is a 32 bit install:
I have 1GB of RAM, with circa 883.61 not allocated to hardware after boot. 
I naively thought 19957 rows of data would be easily handled by a fairly new 
computer. R was able to cease execution, and place control back onto its 
interpreter, while a Chart attempt in OpenOffice locks up OpenOffice.

So,  will installation of a 64 bit Operating System, and at least an 
additional 1GB of RAM  suffice to generate memory space for a 3.0 Gb vector, 
in R?

Much appreciated, time to buy a new computer?

John
Liaw, Andy
# 
Looks like you attach() the data frame before you try t/o plot.  Note
that in the "Details" section of ?attach, it says:

"... The database is not actually attached. Rather, a new environment is
created on the search path and the elements of a list (including columns
of a data frame) or objects in a save file or an environment are copied
into the new environment. ..."

so it's probably a bad idea to do the attach().  Have you tried:

with(DowJones, plot(Date, Close))

?

About 4 years ago I've posted an example of creating a matrix that took
up >6GB, on a dual Opteron w/ 16GB ram running 64-bit SuSE and 64-bit R.
So yes, it's possible.  Nowadays we do it nearly on a daily basis.

Andy

From: John
------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachme...{{dropped:15}}
Daniel Nordlund
# 
John,

Am I missing something here?  You have one 1 GB of RAM, you want to get a second 1 GB of RAM, and you want to know if you will be able to create a 3 GB vector?  With or without a 64bit OS, I don't think you will be successful.  In addition to total memory available, I believe you need to have sufficient contiguous memory to allocate an object.  If memory is fragmented, you won't be able to create the vector, even if the total free space is sufficient.  If I am incorrect, I'm sure someone will be along to correct me shortly.

Hope this is helpful,

Dan

Daniel Nordlund
Bothell, WA
Uwe Ligges
# 
Can you make the data.frame available somewhere. Actually, I am 
surprised it need that huge amount of memory to do the plot.

Best,
Uwe Ligges
John wrote:
John
#
On Friday 28 March 2008 14:28, Daniel Nordlund wrote:
information:
Hi Daniel,

Thank you for the information. It has saved me from needlessly installing a 
64-bit linux instance. In the short term, I'll try and generate multiple 
graphs, side by side. And purchase more RAM. A lot more.

Enjoying the notion of multiple plots in R,

John
Uwe Ligges
# 
Dear John,

note that you "Date" is a factor rather than some date object.
If you convert it to some date object, just a few megabytes will suffice!

Best wishes,
Uwe
John wrote: