Skip to content

LaunchMON is a software infrastructure that enables HPC run-time tools to co-locate tool daemons with a parallel job. Its API allows a tool to identify all the remote processes of a job and to scalably launch daemons into the relevant nodes.

License

ansonl/LaunchMON

 
 

Repository files navigation

2014-05-30 Dong H. Ahn <ahn1@llnl.gov>
	o LaunchMON version 1.0.1 released. Please see NEWS 

2010-06-30 Dong H. Ahn <ahn1@llnl.gov>
	o NEWS items
        o Added new Cray XT4/5/6 and OpenRTE support
        o Integrated the scalable version of PMGR, a.k.a.
          "COBO", and set it as the default bootfabric
        o Ran scaling tests and validated the scalability
          at over 100K MPI processes on BlueGene/P and
          over 200K MPI processes on Cray XT5.
        o Augmented LMON API set while maintaining
          the existing API to be backward-compatible
          - LMON_fe_regErrorCB call
          - LMON_fe_getRMInfo call
          - LMON_DONT_STOP_APP environment variable
	o Lots of bug fixes

2009-05-20 Dong H. Ahn <ahn1@llnl.gov>
	o Added LMON_fe_regErrorCB support
	o Added a fix for the "re-attach" defect on BGP. 
	  This requires IBM efix27 or higher.   
	o Added a fix for user [recv|send]UsrData functions
	  to return correct error codes. 
	  Added a new error code (LMON_ENEGCB) to lmon_api_std.h  

2009-03-11 Dong H. Ahn <ahn1@llnl.gov>
        o LaunchMON 0.7X beta release 
        o Added general BlueGene support convering both BlueGene/L 
	  and BleGene/P
   	  + --with-rm now takes bgrm instead of bglrm or bgprm
          
2008-12-12 Dong H. Ahn <ahn1@llnl.gov>
        o LaunchMON 0.6.4 service release
        o Validated the combinations of various configure options and
	  Changed the configuration policy as follows.
          + Made --with-bootfabric --with-rm --with-gcrypt mandatory options
	     - The default for the bootfabric option is --with-bootfabric=pmgr.
               --with-bootfabric=mpi has been deprecated due to a naming collision 
	       with the GCRYPT package.
	     - Merged with-slurm and with-bglrm into --with-rm[=slurm,bglrm]
 	       for conciseness. The default for with-rm is slurm. 
	o Taken together, to configure, to build and to install the package
	  On X86/X86_64/Linux/SLURM
                 ./configure --with-bootfabric=pmgr --with-rm=slurm --prefix=/myhome
                 make check
                 make install
          On BlueGene/Linux/BGLRM
                 ./configure --with-bootfabric=pmgr --with-rm=bglrm --prefix=/myhome
                 make check
                 make install
          Other configuration options can be shown by type
                 ./configure --help

2008-10-01 Dong H. Ahn <ahn1@llnl.gov>
	o LaunchMON 0.6 release
	  + Added more computer architecture and OS kernel support
	     - computer architecture: 
	       X86, X86_64, BlueGene/L with front-end POWER5 and PPC970MP.
	     - tested environments: 
               X86_64 under 2.6.18-39 and 2.6.18-52;
	       BlueGene/L front-end POWER5 under pseries64-2.6.5-7.282, 
	       front-end PPC970MP under pseries64-2.6.5-7.312, and
	       driver under V1R3M4_300_2008-080728;
	       X86 under 2.6.9-74. 
	  + Added more resource manager support
	     - SLURM (32-bit and 64-bit srun)
             - BlueGene's native RM (32-bit and 64-bit mpirun) 
	  + LaunchMON API support
	     - Full LaunchMON front-end and LaunchMON back-end 
               interface implementation 
	     - Interface implementation versioning uses 
               the libtool notation as follows.  
	         Interface Current: 0
	         Revision: 0
	         Age: 0
          + Added GNU build system support 
	     - To configure, to build and to install the package
	       On X86/X86_64/Linux/SLURM
	         ./configure --with-slurm --prefix=/myhome
	         make check
	         make install	
	       On BlueGene/Linux/BGLRM
	         ./configure --with-bglrm --prefix=/myhome
	         make check
	         make install	
             - Other configuration options can be shown by type 
                 ./configure --help

About

LaunchMON is a software infrastructure that enables HPC run-time tools to co-locate tool daemons with a parallel job. Its API allows a tool to identify all the remote processes of a job and to scalably launch daemons into the relevant nodes.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 67.7%
  • Shell 23.2%
  • Roff 4.5%
  • C 4.2%
  • Perl 0.4%