h5part AT lists.psi.ch
Subject: H5Part development and discussion
List archive
[H5part] Discussion on H5Part/Block/Fed write on a subset of the MPI_COMM_WORLD space
Chronological Thread
- From: Andreas Adelmann <andreas.adelmann AT psi.ch>
- To: h5part AT lists.psi.ch
- Cc: Andreas Adelmann <Andreas.Adelmann AT psi.ch>
- Subject: [H5part] Discussion on H5Part/Block/Fed write on a subset of the MPI_COMM_WORLD space
- Date: Tue, 07 Jul 2009 13:57:58 +0200
- List-archive: <https://lists.web.psi.ch/pipermail/h5part/>
- List-id: H5Part development and discussion <h5part.lists.psi.ch>
Collegues I would like to start/continue the discussion we had over lunch.
Because of performance reasons, a partial write w.r.t. to a subset of the mpi cores
would be very attractive.
Example: the cores 1 ... 3 are writing to the H5 file
while the codes 4 ... 7 are sending the data to core 1 and so on,
as shown below.
H5-FILE --------- 0 0
0 0
----------
^ ^ ^
| | |
writing 1 2 3
sending 4 5 6 7 -> 1
8 9 10 11 -> 2
12 13 14 15 -> 3
Question: can we start H5 with a subset of MPI_COMM_WORLD (core 1,2,3 in the example above) ?
Andreas
--
Dr. sc. math. Andreas (Andy) Adelmann
Staff Scientist
Paul Scherrer Institut WBGB/132 CH-5232 Villigen PSI
Phone Office: xx41 56 310 42 33 Fax: xx41 56 310 31 91
Phone Home: xx41 62 891 91 44
-------------------------------------------------------
Wednesday: ETH CAB F 10.1 xx41 44 632 82 76
=======================================================
The more exotic, the more abstract the knowledge, the more profound will be its consequences."
Leon Lederman =======================================================
- [H5part] Discussion on H5Part/Block/Fed write on a subset of the MPI_COMM_WORLD space, Andreas Adelmann, 07/07/2009
Archive powered by MHonArc 2.6.19.