[MITgcm-support] crash with a new processor / grid size setup
Jean-Michel Campin
jmc at ocean.mit.edu
Sat Mar 28 09:56:28 EDT 2015
Hi Neil and Dimitris,
The problem is not nPy <> 1 (in fact there are cases, e.g. coupled
set-up, where we do need to keep Ny = cube-sphere face size), but
rather the fact that your tile size (sNx,sNy) does not divide the
face size (510x510).
I guess one (at least) of the STDERR files should contain some explicit
error message.
Cheers,
Jean-Michel
On Fri, Mar 27, 2015 at 05:12:01PM -0700, Dimitris Menemenlis wrote:
> pretty sure that nPy must be equal to 1 for cube sphere topologies
> try nPx=68*18, nPy=1
>
> > On Mar 27, 2015, at 4:58 PM, Neil Patel <nigellius at gmail.com> wrote:
> >
> > I recompiled my MITgcm setup (cube face 510 x 6 x 510) for a higher number of processors (1224 cores). I used these parameters in my SIZE.h:
> >
> > (PID.TID 0000.0001) nPx = 68 ; /* No. processes in X */
> > (PID.TID 0000.0001) nPy = 18 ; /* No. processes in Y */
> > (PID.TID 0000.0001) nSx = 1 ; /* No. tiles in X per process */
> > (PID.TID 0000.0001) nSy = 1 ; /* No. tiles in Y per process */
> > (PID.TID 0000.0001) sNx = 75 ; /* Tile size in X */
> > (PID.TID 0000.0001) sNy = 17 ; /* Tile size in Y */
> > (PID.TID 0000.0001) OLx = 4 ; /* Tile overlap distance in X */
> > (PID.TID 0000.0001) OLy = 4 ; /* Tile overlap distance in Y */
> > (PID.TID 0000.0001) nTx = 1 ; /* No. threads in X per process */
> > (PID.TID 0000.0001) nTy = 1 ; /* No. threads in Y per process */
> > (PID.TID 0000.0001) Nr = 50 ; /* No. levels in the vertical */
> >
> > and the program crashed with this error message:
> >
> > ABNORMAL END: S/R W2_SET_MAP_TILES
> >
> > Haven???t seen this one before. Anyone have any clue what to check? Number of grid cells = cube face, told the computer I was running mitgcm with 1224 cores, can???t think of what else I would need to change. Checking STDOUT.0000, these are the last line outputted:
> >
> > (PID.TID 0000.0001) // ======================================================
> > (PID.TID 0000.0001) // Mapping of tiles to threads
> > (PID.TID 0000.0001) // ======================================================
> > (PID.TID 0000.0001) // -o- Thread 1, tiles ( 1: 1, 1: 1)
> > (PID.TID 0000.0001)
> > (PID.TID 0000.0001) W2_READPARMS: file data.exch2 not found
> > (PID.TID 0000.0001) => use W2_EXCH2 default: regular 6-facets Cube
> > (PID.TID 0000.0001) W2_useE2ioLayOut= T ;/* T: use Exch2 glob IO map; F: use model default */
> > (PID.TID 0000.0001) W2_mapIO = -1 ; /* select option for Exch2 global-IO map */
> > (PID.TID 0000.0001) W2_printMsg = -1 ; /* select option for printing information */
> > (PID.TID 0000.0001) ===== Start setting W2 TOPOLOGY:
> > (PID.TID 0000.0001) write to log-file: w2_tile_topology.0000.log
> >
> > the tiles log files exist and this is each one contains:
> >
> > (PID.TID 0003.0001) W2_E2SETUP: number of Active Tiles = 1224
> > (PID.TID 0003.0001) W2_E2SETUP: number of Blank Tiles = 0
> > (PID.TID 0003.0001) W2_E2SETUP: Total number of Tiles = 1224
> > (PID.TID 0003.0001) W2_SET_CS6_FACETS: preDefTopol= 3 selected
> > (PID.TID 0003.0001) W2_SET_CS6_FACETS: facet-dims Unset; assume nRd=nGr=nBl= 510
> > (PID.TID 0003.0001) W2_E2SETUP: Total number of Facets = 6
> > (PID.TID 0003.0001) W2_SET_F2F_INDEX: index matrix for connected Facet-Edges:
> > (PID.TID 0003.0001) W2_SET_MAP_TILES: tile mapping within facet and global Map:
> >
> > Thanks,
> >
> > Neil
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-support
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
More information about the MITgcm-support
mailing list