[ofa-general] Re: New proposal for memory management

[Jeff Squyres]

On Apr 30, 2009, at 8:49 AM, Steven Truelove wrote:

> I have seen this problem in our own MPI application.  The 'simple'
> solution I have seen used in at least one MPI implementation for this
> problem is to prevent the malloc/free implementation being used from
> ever returning memory to the OS.  The virtual/physical mapping can  
> only
> become invalid if virtual addresses are given back to the OS, then
> returned with different physical pages.  Under Linux with at least, it
> is quite easy to tell libc to never return memory to the OS.
>

Unfortunately, this is a false assumption.  There are definitely code  
paths in glibc where, even if you use the mallopt() hints, memory  
*can* (will) be returned to the OS.  This led Open MPI to change its  
memory allocation / intercept scheme in 1.3.2.  See:

     http://www.open-mpi.org/community/lists/announce/2009/03/0029.php

-- 
Jeff Squyres
Cisco Systems