[mpich-discuss] How expensive is MPI_Win_create() compared to memcpy()?

[Pavan Balaji]

On 12/27/2011 09:49 PM, Jed Brown wrote:
>     Hmm... On second thought, it might be possible to make it not
>     synchronize by passing an info argument (though it's not implemented
>     yet).  In this case, the base address and other information can be
>     queried on-demand when the next PUT/GET/ACCUMULATE operation occurs.
>
> Or in the next Win_post/Win_start sequence?

Sure.

>     The part I don't understand here is what benefit does wrapping an
>     existing window object around new memory give you?  Why not create a
>     new window object around your new memory?
>
> Nothing in terms of semantics, but if it is less expensive to wrap new
> memory than to create a new window, then we can do that.

I see.  In that case, the dynamically attaching memory to windows option 
in MPI-3 is what you need.  In this case, the window object is created 
collectively once, and then memory can be attached to it as a local 
operation without requiring collective synchronization.

  -- Pavan

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji