[mpich-discuss] Problem with ROMIO using Windows

Jayesh Krishna jayesh at mcs.anl.gov
Wed Jul 8 09:38:00 CDT 2009


Hi,
 I have created a ticket 704
(https://trac.mcs.anl.gov/projects/mpich2/ticket/704#preview) for this
problem. We can proceed with discussions on that ticket.

Regards,
Jayesh

-----Original Message-----
From: mpich-discuss-bounces at mcs.anl.gov
[mailto:mpich-discuss-bounces at mcs.anl.gov] On Behalf Of Francisco Javier
García Blas
Sent: Wednesday, July 08, 2009 8:07 AM
To: mpich-discuss at mcs.anl.gov
Cc: Sergio Garcia
Subject: [mpich-discuss] Problem with ROMIO using Windows

Hi all,

I am using coll_perf.c in order to evaluate some optimizations at ADIO
level under Windows. Tracing ADIOI_Flatten_datatype primitive in Windows,
I got the following wrong values for flat->count and cur_idx.

C:\Users\fjblas>mpiexec -n 4 coll_perf.exe -fname myfile
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10
ADIOI_Flatten_datatype:: count 10, cur_idx = 10


I traced ADIOI_Flatten_datatype primitive and I have found that
ADIOI_Count_contiguous_blocks did not return the correct value for
flat->count.

On a Linux cluster, I tried the same application without problem, and I
have got correct values for flat->count.

I tried with different versions of MPICH2 under Windows (1.0.5, 1.0.8, and
the lasted release), and I have got the same incorrect values for
flat->count.

It is possible that is something wrong under Windows???


Thanks for all
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20090708/7ec82c8f/attachment.htm>


More information about the mpich-discuss mailing list