Changeset 759 for palm/trunk/SOURCE/init_pegrid.f90

Timestamp:

Sep 15, 2011 1:58:31 PM (13 years ago)

Author:

raasch

Message:

New:
---

The number of parallel I/O operations can be limited with new mrun-option -w.
(advec_particles, data_output_2d, data_output_3d, header, init_grid, init_pegrid, init_3d_model, modules, palm, parin, write_3d_binary)

Changed:

---

mrun option -T is obligatory

Errors:

---

Bugfix: No zero assignments to volume_flow_initial and volume_flow_area in
case of normal restart runs. (init_3d_model)

initialization of u_0, v_0. This is just to avoid access of uninitialized
memory in exchange_horiz_2d, which causes respective error messages
when the Intel thread checker (inspector) is used. (production_e)

Bugfix for ts limitation (prandtl_fluxes)

File:

• palm/trunk/SOURCE/init_pegrid.f90 (modified) (2 diffs)

palm/trunk/SOURCE/init_pegrid.f90

@@ -4 +4 @@
 ! Current revisions:
 ! -----------------
+! calculation of number of io_blocks and the io_group to which the respective
+! PE belongs
 ! 
 ! ATTENTION: nnz_x undefined problem still has to be solved!!!!!!!!
@@ -1237 +1239 @@
     ENDIF
 
+!
+!-- Calculate the number of groups into which parallel I/O is split.
+!-- The default for files which are opened by all PEs (or where each
+!-- PE opens his own independent file) is, that all PEs are doing input/output
+!-- in parallel at the same time. This might cause performance or even more
+!-- severe problems depending on the configuration of the underlying file
+!-- system.
+!-- First, set the default:
+    IF ( maximum_parallel_io_streams == -1  .OR. &
+         maximum_parallel_io_streams > numprocs )  THEN
+       maximum_parallel_io_streams = numprocs
+    ENDIF
+
+!
+!-- Now calculate the number of io_blocks and the io_group to which the
+!-- respective PE belongs. I/O of the groups is done in serial, but in parallel
+!-- for all PEs belonging to the same group. A preliminary setting with myid
+!-- based on MPI_COMM_WORLD has been done in parin.
+    io_blocks = numprocs / maximum_parallel_io_streams
+    io_group  = MOD( myid+1, io_blocks )
+    
+
  END SUBROUTINE init_pegrid