****************************************************************************** GPI-2 http://www.gpi-site.com Version: 1.3.0 Copyright (C) 2013-2019 Fraunhofer ITWM ****************************************************************************** 1. INTRODUCTION =============== GPI-2 is the second generation of GPI (www.gpi-site.com). GPI-2 implements the GASPI specification (www.gaspi.de), an API specification which originates from the ideas and concepts of GPI. GPI-2 is an API for asynchronous communication. It provides a flexible, scalable and fault tolerant interface for parallel applications. 2. INSTALLATION =============== Requirements: ------------ The current version of GPI-2 has the following requirements. Software: - libibverbs v1.1.6 (Verbs library from OFED) if running on Infiniband. - ssh server running on compute nodes (requiring no password). - autotools utilities (autoconf>=2.63,libtool>=2.2,automake>=1.11) - gawk (GNU Awk) and sed utilities. Hardware: - Infiniband/RoCE device or Ethernet device. Basic configuration: ------------------- To install GPI-2, it needs to be configured using the script `./configure`. The available options and the relevant environment variables are printed by `./configure --help`. The basic configuration: `./configure --prefix=$HOME/local` use the compilers defined by the environment variables CC and FC for the general checking procedure and sets up `$HOME/local` as the installation directory. By default, the script: - checks for the Infiniband header and library files, and fall backs to the Ethernet device in case they are not available or usable, - targets to the production, debugging and statistic libraries, as well as, the Fortran modules (if the Fortran compilers are found), - configure GPI for use PBS as the batch system, - checks the existence of `doxygen` and `dot` for the documentation target. Compilation, testing and cleaning: ----------------------- The compilation step: `make -j$NPROC` builds in parallel the GPI libraries, the Fortran modules, and the binary tests and microbenchmarks. After successful completion, the user can define the working hosts in `tests/machines` and run the predefined tests by: `make check` Clean of the configuration/compilation files can be done as usual with the commands `make distclean` and `make clean`. Documentation and tutorial: -------------------------- If required, the (doxygen) documentation and the tutorial code are built through `make docs` and `make tutorial`, respectively. Installation and uninstallation: ------------------------------- Finally, `make install` installs: - the running scripts in the `$HOME/local/bin` directory, - the libraries in `$HOME/local/lib64`, - the headers and Fortran modules in `$HOME/local/include`, - the full tests directory in `$HOME/local/tests` If required the package can be removed from the target directory by using `make uninstall`. Custom configurations: ---------------------- Specific configurations can be setup by predefined flags. ### DEVICES ### GPI-2 is intended to be linked to the libibverbs from the OFED stack. In case the configure script is not able to find it in the default paths of the host system, the user can pass the path of the OFED installation: `./configure --with-infiniband<=full_path_to_ofed>` On the other hand, GPI-2 can be installed on a system without Infiniband, using the standard TCP sockets: `./configure --with-ethernet` Such support is, however, primarily targetted at the development of GPI-2 applications without the need to access a system with Infiniband, with less focus on performance. ### BATCH SYSTEM ### PBS is the default batch system of GPI-2, however, the user can request for Slurm `./configure --with-slurm` or LoadLeveler: `./configure --with-loadleveler` ### MPI Interoperability ### If the plan is to use GPI-2 with MPI to, for instance, start an incremental port of a large application or to use some libraries that require MPI, the user can enable MPI interoperability in several ways: - check for MPI in the standard path: `./configure --with-mpi` - check for MPI in a specific path, e.g.: `./configure --with-mpi=<=path_to_mpi_installation>` - specify the MPI compilers, e.g.: `CC=mpiicc FC=mpiifort ./configure` For this MPI+GPI2 mixed mode, the only constraint is that MPI_Init() must be invoked before gaspi_proc_init() and it is assumed that the application starts with mpirun (or mpiexec, etc.). Also, note that this option will require that the GPI-2 application is linked to the MPI library (even if MPI is not used). Therefore, if the interest is to use GPI-2 only, GPI-2 must not be build with with this option. Furthermore fine control of MPI can be done through the `--with-mpi-extra-flags` option (used, for example, to link an specific version of the Intel MPI libraries). 3. BUILDING GPI-2 APPLICATIONS ============================== GPI-2 provides two libraries: libGPI2.a and libGPI2-dbg.a. The libGPI2.a aims at high-performance and is to be used in production whereas the libGPI2-dbg.a provides a debug version, with extra parameter checking and debug messages and is to be used to debug and during development. 4. RUNNING GPI-2 APPLICATIONS ============================= The gaspi_run utility is used to start and run GPI-2 applications. A machine file with the hostnames of nodes where the application will run, must be provided. For example, to start 1 process per node (on 4 nodes), the machine file looks like: node01 node02 node03 node04 Similarly, to start 2 processes per node (on 4 nodes): node01 node01 node02 node02 node03 node03 node04 node04 The gaspi_run utility is invoked as follows: gaspi_run -m [OPTIONS] IMPORTANT: The path to the program must exist on all nodes where the program should be started. The gaspi_run utility has the following further options [OPTIONS]: -b Use a different binary for first node (master). The master (first entry in the machine file) is started with a different application than the rest of the nodes (workers). -N Enable NUMA for processes on same node. With this option it is only possible to start the same number of processes as NUMA nodes present on the system. The processes running on same node will be set with affinity to the proper NUMA node. -n Start as many from machine file. This option is used to start less processes than those listed in the machine file. -d Run with GDB (debugger) on master node. With this option, GDB is started in the master node, to allow debugging the application. -p Ping hosts before starting the binary to make sure they are available. -h Show help. 5. THE GASPI_LOGGER =================== The gaspi_logger utility is used to view and separate the output from all nodes when the function gaspi_printf is called. The gaspi_logger is started, on another session, on the master node. The output of the application, when using gaspi_printf, will be redirected to the gaspi_logger. Other I/O routines (e.g. printf) will not. A further separation of output (useful for debugging) can be achieved by using the routine gaspi_printf_to which sends the output to the gaspi_logger started on a particular node. For example, gaspi_printf_to(1, "Hello 1\n"); will display the string "Hello 1" in the gaspi_logger started on rank 1. 6. TROUBLESHOOTING AND KNOWN ISSUES =================================== If there are troubles when building GPI-2 with support for Infiniband, make sure the OFED stack is correctly installed and running. As above mentioned, it is possible to specify the OFED path in the actual host system. When installing GPI-2 with MPI mixed-mode support (using the options `--with-mpi` or `--with-mpi<=path_to_mpi_installation>`) and the installation is failing when trying to build the tests due to missing libraries, try to setup directly the MPI compilers (wrappers) through the environment variables CC and FC. Environment variables --------------------- You might have some trouble when your application requires some dynamically set environment setting (e.g. the LD_LIBRARY_PATH), for instance, through the module system of your jobs batch system. Currently, neither the gaspi_run or the GPI-2 library take care of such environment settings. To this situation there are 2 workarounds: i) you set the required environment variables in your shell initialization file (e.g. ~/.bashrc). ii) you create an executable shell script which sets the required environment variables and then starts the application. Then you can use gaspi_run to start the application, providing the shell script as the application to execute. gaspi_run -m machinefile ./my_wrapper_script.sh where my_wrapper_script.sh contains: #!/bin/sh LD_LIBRARY_PATH=$LD_LIBRARY_PATH: /my_application exit $? If you're running in MPI mixed-mode, starting your application with mpirun/mpiexec, this should not be an issue. 7. UP COMING FEATURES ===================== GPI-2 is on-going work and more features are still to come. Here are some that are in our roadmap: - support to add spare nodes (fault tolerance) - better debugging possibilities 8. MORE INFORMATION ==================== For more information, check the GPI-2 website ( www.gpi-site.com ) and don't forget to subscribe to the GPI-2 mailing list. You subscribe it at https://listserv.itwm.fraunhofer.de/mailman/listinfo/gpi2-users