From: Jeff Squyres <jsquyres@XXXXXXXXX>
Date: January 18, 2008 7:23:12 PM CST
To: mpi-21@XXXXXXXXXXXXX
Subject: Re: [mpi-21] const MPI C++ handles
Reply-To: mpi-21@XXXXXXXXXXXXX

On Jan 18, 2008, at 8:00 PM, Erez Haba wrote:

>> Just trying to be cautious: would there ever be a reason to have a  
>> non-
>> const version of the Set_* functions?
>
> Yes, if the impl caches info in its own objects and change it when  
> the Set_* function is called. (but then it can use the mutable  
> keyword :))

You're talking about a pure C++ implementation, right?

If an implementation simply caches a C handle and invokes the  
underlying C MPI function, it doesn't need to be const -- perhaps  
something like this (where mpi_comm is the C handle cached in the  
MPI::Comm class):

inline void MPI::Comm::Set_name(const char* comm_name) const
{
   (void) MPI_Comm_set_name(mpi_comm, const_cast<char *>(comm_name));
}

Regardless, MPI::Comm is required to be a handle, not the actual  
underlying MPI object.  So stuff like this is ok:

    MPI::Intracomm foo = MPI::COMM_WORLD;
    foo.Send(...);
    MPI::COMM_WORLD.Send(...);

Both sends will be the same communicator.  Hence, the internal cached  
reference to the underlying object must always be immutable (allowing  
the function to be const) -- it's the underlying object that must be  
mutable.

Right?


>
> Thanks,
> .Erez
>
> -----Original Message-----
> From: owner-mpi-21@mpi-forum.org [mailto:owner-mpi-21@mpi- 
> forum.org] On Behalf Of Jeff Squyres
> Sent: Friday, January 18, 2008 11:19 AM
> To: mpi-21@mpi-forum.org
> Subject: Re: [mpi-21] const MPI C++ handles
>
> On Jan 18, 2008, at 2:11 PM, Erez Haba wrote:
>
>> I looked up this issue and did not find a C++ *standard* that spans
>> that many platforms and compilers.
>>
>> The ISO C++ 98 standard itself does have libraries, but they provide
>> the headers which are less ambiguous. (including various const
>> values (like INT_MAX)) (think ABI :))
>
> Hah!  Surely you're not implying that C++ compilers produce code that
> is ABI-compatible with other C++ compilers?  :-D
>
>> I agree that we'll need to enable the recommendation in the standard
>> interface. We'll need to agree that the semantic of changing the
>> underlying MPI handle is not changing the object (and thus can
>> define the method as const).
>>
>> I don't think you'll need an overloaded function for the Set_*
>> functions; const methods are good for non const objects.
>
> Good point -- a const method will match a corresponding non-const
> (*this).
>
> Just trying to be cautious: would there ever be a reason to have a  
> non-
> const version of the Set_* functions?
>
> --
> Jeff Squyres
> Cisco Systems
>


-- 
Jeff Squyres
Cisco Systems