Bernard Leclerc (bleclerc++at++cae.ca)
Tue, 26 Sep 1995 14:01:29 -0400
> I have been reading the messages concerning the problems with
> mixing the SYS V IPC shared memory with the shared arenas that
> Performer uses. Could using Systran's SCRAMNet memory directly
> from a Performer application produce the same side-effects or
> conflicts with the shared arenas?
Concerning Systran's SCRAMNet card, we have successfully used it to interface
our visual application (based on Performer) with a simulation process running
on a different host. The mapping to SCRAMNet was done using a combination of
open("/dev/mmem",...) and mmap(). There was no conflict between Performer
arenas and SCRAMNet memory. So, regarding your problem, I don't think that
using memory mapped RAM located on a SCRAMNet board is slower than using a IRIX
shared arena. However, you have to realize that a SCRAMNet board plugs in the
VME bus. The transfer rate on the VME bus is slower than the transfer rate on
the SGI bus. Accessing SCRAMNet memory might be a little slower than accessing
the main memory. Beside this transfer rate issue, there's no difference between
the two kind of RAM.
BTW, mmap() is not what's called SysV IPC. Semaphores (semget), shared memory
segments (shmget) and message queues (msgget) form the basis for System V
Inter-Process Communication (IPC).
--
___/ | ___/ Bernard Leclerc e-mail: bleclerc++at++cae.ca
/ / | / Systems Engineer voice: +1 514 341 2000
/ / | __/ CAE Electronics Ltd. extension 2275
/ / | / 8585 Cote De Liesse fax: +1 514 340 5496
/ ____ | / P.O. Box 1800
_____/ _/ _| _____/ Saint-Laurent, Quebec, Canada, H4L-4X4
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:51:54 PDT