IDL on Dual processor Linux SMP box? [message #11899] |
Wed, 03 June 1998 00:00  |
krist
Messages: 6 Registered: April 1994
|
Junior Member |
|
|
Greetings:
I have been considering getting a dual processor Pentium II workstation
on which I would run Linux in SMP mode, mostly to use IDL. Does anyone
have any experience with such a configuration? I recall somewhere that
an IDL license on a x86 Linux box is tied to the Ethernet card address,
rather than the processor serial number (which perhaps Intel processors do
not report). If so, I could run IDL on both processors at the same time
under the same license, correct?
The system I may get is a Dell Precision Workstation 410 with dual
400 MHz processors, with the SCSI and video cards replaced with Buslogic
SCSI and Matrox Millenium II video cards for better Linux compatibility
(I can only get boxes from big-name companies, not from small-name Linux
specialists).
Thanks,
John Krist
krist@stsci.edu
|
|
|
Re: IDL on Dual processor Linux SMP box? [message #11966 is a reply to message #11899] |
Fri, 05 June 1998 00:00  |
Brian Guarraci
Messages: 3 Registered: June 1998
|
Junior Member |
|
|
Is the IDL server multi-threaded? If not, you could write a proxy for
IDL servers that handling all IDL requests and spawned off IDL servers
as needed - assuming you had a licensing to support that.
brian
Stein Vidar Hagfors Haugan wrote:
>
> I think the licence policy is "one licence per screen and
> user". A single user working on a single screen should thus
> only occupy a single licence, even with several IDL processes.
> Two users on a single screen (crowded office :-) might occupy
> two licences, and a single user on two screens would occupy
> two licences, AFAIK.
>
> As for the benefits of multiple processors, I sometimes
> run IDL in Remote Procedure Call (RPC) server mode to use
> it as a "plotting engine" for my C programs. This would
> (at least in theory) allow your C program to go on with
> it's business (on one processor) while IDL is plotting the
> results you just sent. The problem is though, that the RPC
> call probably waits for the plotting command to finish
> successfully...? Maybe it's possible to set the timeout
> to 0 seconds, but... It would be nice to be able to do
> RPC calls concurrently in some cleaner way.
>
> Regards,
>
> Stein Vidar
|
|
|