Global Shared Memory and SCI/Dolphin

Franz Marini franz.marini at
Fri Jul 18 04:52:20 EDT 2003

On Thu, 17 Jul 2003, Greg Lindahl wrote:

> There's a terminology problem here: Some people mean cache-coherent
> shared memory, like that on an SGI Origin.

I maybe wrong but I think that all the SGI machines (including the Altix) 
implement c-c shared mem. 

> Another term for non-cache-coherent but globally addressable and
> accessible memory is SALC: Shared address, local consistency.
> And yes, all 4 of Scali/Quadrics/Myrinet/Infiniband support the
> non-cache-coherent kind of shared memory. Programming models in this
> area are:
>   * UPC: Unified Parallel C
>   * CoArray Fortran
>   * MPI-2 one-sided operations
>   * Global Arrays from PNL
>   * The Cray SHMEM library

And this should testify to the fact that the shmem programming paradigm is 
all but rarely used. As long as I can tell there is a *lot* of code out 
there that uses, e.g. the Cray SHMEM lib (btw, this is one of the things 
that makes the Scali/Dolphin solution interesting to us).

But, still, whereas, e.g. the SHMEM lib has been implemented under Scali 
(and maybe under Quadrics/Myrinet/Infiniband, not sure about it), what 
I think it'd be interesting and usefull is the support (at the OS level) 
for a GSM/single system image, providing support for POSIX threads across 
the nodes. I may be dreaming here, I know, but still... :)

Btw, on a side note, does anyone know if there is some compiler (both C 
and F90/HPF) out there supporting some kind of auto parallelization via, 
e.g. the SHMEM lib (I'm not asking for a MPI-enabled compiler, I'm not 
*so* crazy ;)) ?

Franz Marini
Sys Admin and Software Analyst,
Dept. of Physics, University of Milan, Italy.
email : franz.marini at

Beowulf mailing list, Beowulf at
To change your subscription (digest mode or unsubscribe) visit

More information about the Beowulf mailing list