Ask ROTANG

The MM5 MPP Answer Beast

This is an archive of notes plus email from users with questions on aspects of the distributed-memory version of MM5 and responses. Since this is an archive, information is not usually removed. Therefore, please be aware of the date of a link you access on this page. Old information may have been superseded.

Address questions for ROTANG care of:

John Michalakes
michalak@ucar.edu


New:

  • 2005 03 11 Portland Group Compiler Setting Needed for MPP MM5v3.7 on Linux

  • Performance:

  • Ongoing MM5 MPP Performance Plots

  • Other links:

  • For general information on the DM-parallel MM5, please click here.
  • For machine-specific hints and fixes, please click here.
  • For the Weather Models on Parallel Computers web page at Argonnne NL, click here.

    Version-specific information: MM5 Version 3


    (plus some carry-over from V2 when still relevant)

  • 2005 03 11 Portland Group Compiler Setting Needed for MPP MM5v3.7 on Linux

  • (v3.7 released December 23, 2004)
  • 2004 03 08 Bug Fix for FLIC: Handle % Characters in Input
  • 2004 03 04 MM5 MPP Performance Plots
  • 2004 01 28 mm5etime, stats.awk, and information for interpreting MM5 T3A performance
  • 2003 10 01 Configuration information for Linux Itanium (IA-64) Compiles (thanks John Taylor)
  • 2003 07 25 Link problem involving reference to errno on Linux
  • 2003 03 06 Compiling MM5 for Double (64-bit) Precision (With Linux Addendum)

  • (v3.6 released December, 2002)
  • 2002 10 18 Fix: MPP_IO_NODE=1 causes problems with the MMOUT_DOMAIN files
  • 2002 09 06 Linux: CPP -traditional glitch with some versions of Linux

  • (v3.5.3 released August 28, 2002)
  • 2002 02 08 Linux compile note: settings in 7g of configure.user and RSL makefile
  • 2002 01 29 Fujitsu VPP invalid communicator MPI error (fix)

  • (v3.5 released December 5, 2001)
  • 2001 10 22 Numerical results and PGF compiler flags on Linux systems
  • 2001 10 22 Attn: Solaris users using Sun HPC tools
  • 2001 10 11 Update to README.MPP file
  • 2001 08 02 Problem with IOVERW=1 (Linux; nest initialization from nest input file)
  • 2001 08 02 Two patches for the FDDA obs-nudging option
  • 2001 08 02 Notes on compiler optimization levels for Sun and IBM
  • 2001 08 01 Linux: error compiling cup.F
  • 2001 08 20 Problem with timing of restart output (Thanks Norm Henry)
  • 2001 08 02 Problem with IOVERW=1 (nest initialization from nest input file)
  • 2001 08 02 Two patches for the FDDA obs-nudging option
  • 2001 08 02 Notes on compiler optimization levels for Sun and IBM
  • 2001 08 01 Linux: error compiling cup.F
  • 2001 05 18 Draft MM5 MPP Design and Implementation Document
  • 2001 05 04 Linux Compile Notes: /lib/cpp, MPICH, etc. (thanks B. Brashers)
  • 2001 04 25 Sun/Solaris compilation: workaround for 'sort' bug (thanks M. Walters)
  • 2001 03 23 MPICH/P4 environment on Linux PC cluster for large domain(thanks N. Henry)
  • 2001 02 13 Using DBX with the parallel MM5 under MPICH/CH-P4
  • 2001 01 23 Problem modifying MPP I/O routines for MM5v3 (advanced)
  • 2001 01 16 Parallel Computing in Atmospheric Modeling (Bibliography)

  • (v3.4 released November 7, 2000)
  • 2000 08 28 Running out of thread space on Compaq w/OpenMP. (thanks M. Kelly, M. Michelson)
  • 2000 07 25 Bug Fixes: Two bugs in MPP LSM code for V3.3.
  • 2000 07 05 Running very large problems on very large numbers of processors
  • 2000 06 30 Problem with Compaq F77 5.3 compiler and workaround.
  • 2000 06 26 Bug Fix: Serious bug in MPP code for V3.3.
  • 2000 06 25 Bug Fix: FDDA on T3E.
  • 2000 06 21 ORIGIN 2000 USERS NOTE: Setting for MPI resources.
  • 2000 05 12 Fix to make ISOIL=1 more stable.
  • 2000 05 11 Advisory: MPHYSTBL=1 is not safe for operational use.
  • 2000 04 12 Cray T3E USERS NOTE: Fix for aborts doing I/O on large numbers of processors.
  • 2000 04 11 FUJITSU USERS NOTE: Fix for bad boundaries on lowest level of vert. winds on nests.
  • 2000 03 28 Cray T3E USERS NOTE: Fix for output bug involving XTIME not advancing in file headers. (thanks Erik Berge)
  • 2000 03 06 Fix for floating point errors with Linux and pgf77
  • 2000 01 25 FIX: CCM2 Radiation bug in MM5v3.

  • (v3.3 released January 25, 2000)
  • 1999 12 20 LINUX and FUJITSU USERS NOTE: Fix for CFL problem with MM5V3.2 (Thanks Yong Hee Lee)
  • 1999 12 04 RE: MPI deadlock problem on Linux Cluster (thanks Gorka Perez Land)
  • 1999 12 03 LINUX USERS NOTE: New version of MPICH, release 1.2.0
  • 1999 11 12 FIX Model hang when using Surface FDDA analysis nudging
  • 1999 11 09 FIX parallel bug in Schultz (IMPHYS=8) Microphysics
  • 1999 10 14 MM5v3 output bug on Linux PC Beowulf clusters (with FIX)
  • 1999 10 01 SERIOUS BUG in MPP V3.2 code. Patch file with FIX.
  • 1999 09 28 MM5 on dual-processed linux boxes (thanks Todd Hutchinson)
  • 1999 09 27 Corrupt tar file: please re-download MPP.TAR.gz.
  • 1999 09 27 SMP performance mod: parallelize, reorder loop in solve.F
  • 1999 09 26 An update on release 3.2

  • (v3.2 released Sept. 24, 1999)
  • 1999 09 22 FIX for PGF77 3.0 problems on Red Hat Linux 6 (Item 1999 09 14 )
  • 1999 09 22 Assign command on T3E for MM5v3
  • 1999 09 14 User-reported problem: PGF77 3.0 compiler problem on Red Hat Linux 6
  • 1999 09 08 Code UPDATE: Pre-release MM5v3.2 for MPP
  • 1999 08 13 FIX with patch file: fatal T3E error; non-fatal Compaq output problem
  • 1999 08 10 F77 compiler error on bdyin, outtap
  • 1999 08 10 FIX for FLIC (compiling on T3E)
  • 1999 05 27 How to compile on PC's using Linux/pgf77/MPICH (thanks Steve Webb)
  • 1999 05 28 Fix for hard-wired limit of 128 processors in RSL
  • 1999 04 28 Where does the MPP version of MM5 compute the decomposition over procs?
  • 1999 04 25 What is this PROCMIN_NS and PROCMIN_EW business?
  • 1999 04 23 How to print values from model arrays within physics for debugging.
  • 1999 04 05 Testing bit-for-bit agreement between MPP and non-MPP versions of the model.
  • 1999 01 22 Adding new modules; how does MPP code build itself?

  • Version-specific information: MM5 Version 2

  • 1999 12 14 FIX for MM5V2 with One-Way nesting on Cray T3E (thanks David Wong, EPA)
  • 1999 12 04 RE: MPI deadlock problem on Linux Cluster (thanks Gorka Perez Land)
  • 1999 12 03 LINUX USERS NOTE: New version of MPICH, release 1.2.0
  • 1999 11 09 FIX parallel bug in Schultz (IMPHYS=8) Microphysics (v2.12 only)
  • 1999 07 31 FIX for nesting on Fujitsu-VPP
  • 1999 06 24 FIXED at least part of the MPI deadlock problem on U. Hong Kong Linux Cluster
  • 1999 06 15 More on MPI deadlock problem on Linux Cluster...
  • 1999 05 28 Fix for hard-wired limit of 128 processors in RSL
  • 1999 05 27 How to compile on PC's using Linux/pgf77/MPICH (thanks Steve Webb)
  • 1999 05 11 MPI_Send deadlock on Linux Cluster?
  • 1999 05 11 Problem compiling GSFC Microphysics
  • 1999 04 28 Where does the MPP version of MM5 compute the decomposition over procs?
  • 1999 04 27 Bug in MPP/RSL/Makefile.RSL hinders bit-for-bit agreement T3E.
  • 1999 04 27 Fujitsu VPP performance and scaling issues.
  • 1999 04 25 What is this PROCMIN_NS and PROCMIN_EW business?
  • 1999 04 23 How to print values from model arrays within physics for debugging.
  • 1999 04 19 Core dump on Cray T3E when writing fort.41.
  • 1999 04 05 How to add a physics module to the MPP framework.
  • 1999 04 05 Testing bit-for-bit agreement between MPP and non-MPP versions of the model.
  • 1999 01 22 Adding new modules; how does MPP code build itself?