[MPICH] MPI_GET_ADDRESS
William Gropp
gropp at mcs.anl.gov
Sun Apr 23 13:03:28 CDT 2006
At 01:51 PM 4/21/2006, Kenneth Rempe wrote:
>Bill,
>
>Okay, what am I looking for in the configure output. Is there a why to
>override what configure is setting?
>Also, is this a problem because in C++ you can use an unsigned int for
>addresses up to 4GB but in fortran it is a signed integer
>which only can go up to 2GB?
The first thing to look for is the size of void *. The line should look like
checking size of void *...8
If the size if 4, then you have 32 bit pointers. There may then be a
problem, as you suspect, with the signed integer in Fortran.
Bill
>Thanks
>
>Ken
>
>William Gropp wrote:
>
>>At 08:02 AM 4/21/2006, Kenneth Rempe wrote:
>>
>>>...
>>>On both the Linux and IBM the value of MPI_ADDRESS_KIND is set to a 4
>>>byte integer.
>>
>>
>>That's the problem. If the sizeof a pointer is 8 bytes, then
>>MPI_ADDRESS_KIND should be an 8 byte integer. What is the output of
>>configure for MPICH2? Note that MPICH2 must be configured with
>>compatible compiler options for systems that support multiple datatype
>>sizes (such as 32 and 64bit pointers).
>>
>>Bill
>>
>>
>>>Linux Red Hat Enterprise 3
>>> Lahey Fortran Compiler version 6.2c
>>> gcc version 3.2.3
>>>
>>>IBM Aix 5.2
>>> xlf compiler version 8.1
>>> xlc compiler version 6.0
>>>
>>>
>>>--
>>>*** NOTE: soa.com will no longer be used after February 1, 2006, please
>>>change your address book ***
>>>
>>>Kenneth R. Rempe * email:
>>>kenneth.rempe at studsvikscandpower.com
>>>Studsvik Scandpower, Inc. * 504 Shoup Avenue Suite #201 *
>>>voice: 208-522-4630
>>>Idaho Falls, ID USA 83402-3502 * fax: 208-522-1187
>>
>>
>>William Gropp
>>http://www.mcs.anl.gov/~gropp
>
>
>
>--
>*** NOTE: soa.com will no longer be used after February 1, 2006, please
>change your address book ***
>
>Kenneth R. Rempe * email: kenneth.rempe at studsvikscandpower.com
>Studsvik Scandpower, Inc. * 504 Shoup Avenue Suite #201 *
>voice: 208-522-4630
>Idaho Falls, ID USA 83402-3502 * fax: 208-522-1187
William Gropp
http://www.mcs.anl.gov/~gropp
More information about the mpich-discuss
mailing list