Jacek A. Klos
jakl at theochem.kun.nl
Tue Jan 22 10:15:40 GMT 2002
Check on your system hwat limits on memory usage are set with
Than try to do 'unlimit' command and than again 'limit'.
If values are the same you have to su to root and run 'dxkerneltuner'
(probably on console) to change kernel parameters.
I had to do it on DS20 with tru64 to make available 4GB of RAM for
limit was set to 1GB of RAM and there was no way to use more without
rebuilding the kernel.
unlimit command sets datasize to max_per_proc_data_size.
Probably you have set it to 1GB or 512MB.
After dxkerneltuner window pops up, you have to go to 'proc' options
and change value of max_per_proc_data_size and per_proc_data_size
to required values. per_proc_data_size is a limit of command 'limit' and
unlimit resets it to max_per_proc_data_size. So you can set
max_per_proc_data_size to your available RAM (in bytes) and
per_proc_data_size to lower value (in bytes). Than you can require whole
RAM using unlimit command before running job.
Apply changes in dxkerneltuner and reboot
Have a hope that it will boot :).
WARNING! This procedure is somewhat of dangerous sort and dxkerneltuner
will warning you too, so be careful what values are putting there and
before change write down values which were originally there.
I hope it will help.
Dr Jacek Klos
Institute of Theoretical Chemistry
University of Nijmegen
6525 ED Nijmegen, The Netherlands
Phone: +31 24-3653038
Fax: +31 24-3653041
E-mail: jakl at theochem.kun.nl
On Tue, 22 Jan 2002, Elmar Gerwalin wrote:
> Hello !
> We recently installed molpro2002.1 on a Digital Alpha, Tru64 System.
> We can only allocate 512MB (66 MW) of memory, using the -m option or
> the MEMORY keyword.
> The (first line of the) error message:
> Failure in attempting memory allocation of 100230002 words (764 Mbyte)
> It's not a hardware limitation since we have 1.5 GB of physical memory
> and molpro2000 can allocate more than 512 MB.
> An installation on a Pentium4 system doesn't show this problem.
> Is this an installation/configuration problem ?
> Any help is appreciated.
> Elmar Gerwalin
> Elmar Gerwalin , University of Kaiserslautern,Germany
> Dept. of Theoretical Chemistry
> elg at chemie.uni-kl.de 0631-205-2749
More information about the Molpro-user