[mpich-discuss] MPI I/O and Windows

[Rob Latham]

On 01/12/2016 11:17 AM, Eric Chamberland wrote:

> Otherwise, we would have to implement all of
>
> MPI_Get_address(x,y)
> MPI_File_write_at(a,b,c,d,e,f)
> MPI_File_read_at_all(a,b,c,d,e,f)
> MPI_File_write_all_begin(a,b,c,d)
> MPI_File_write_all_end(a,b,c)
> MPI_File_write(a,b,c,d,e)
> MPI_File_read_all_begin(a,b,c,d)
> MPI_File_read_all_end(a,b,c)
> MPI_FILE_delete
> MPI_Info_set
> MPI_Info_free
> MPI_Info_create

yep, those all look like routines MS-MPI and Intel MPI would support. 
nothing unusual there.

> to work in sequential, which is feasible too...
>
>> I don't know what the consistency semantics of SMB/CFS are: when are
>> writes from one client visible to another?
>>
>
> Can't tell... when I used windows for the last time I was so younger! :)

as this list is for mpich, and mpich no longer has window support, I'll 
have to encourage you to contact MS-MPI or Intel MPI folks for follow up 
questions.

==rob

>
> Thanks for the answer!
>
> Eric
>
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss