MPICH Release 4.3.0b1 MPICH is a high-performance and widely portable implementation of the MPI-4.1 standard from the Argonne National Laboratory. This release has all MPI 4.1 functions and features required by the standard with the exception of support for user-defined data representations for I/O. This README file should contain enough information to get you started with MPICH. More extensive installation and user guides can be found in the doc/installguide/install.pdf and doc/userguide/user.pdf files respectively. Additional information regarding the contents of the release can be found in the CHANGES file in the top-level directory, and in the RELEASE_NOTES file, where certain restrictions are detailed. Finally, the MPICH web site, http://www.mpich.org, contains information on bug fixes and new releases. 1. Getting Started 2. Reporting Installation or Usage Problems 3. Compiler Flags 4. Alternate Channels and Devices 5. Alternate Process Managers 6. Alternate Configure Options 7. Testing the MPICH installation 8. Fault Tolerance 9. Developer Builds 10. Multiple Fortran compiler support 11. ABI Compatibility 12. Capability Sets 13. Threads ------------------------------------------------------------------------- 1. Getting Started ================== Note: this guide assumes you are building MPICH from one of the MPICH release tarballs. If you are starting from a git checkout, you will need a few additional steps. Please refer to the wiki page -- https://github.com/pmodels/mpich/blob/main/doc/wiki/Index.md. The following instructions take you through a sequence of steps to get the default configuration (ch3 device, nemesis channel (with TCP and shared memory), Hydra process management) of MPICH up and running. (a) You will need the following prerequisites. - REQUIRED: This tar file mpich-4.3.0b1.tar.gz - REQUIRED: Perl - REQUIRED: A C compiler (C99 support is required. See https://github.com/pmodels/mpich/blob/main/doc/wiki/source_code/Shifting_Toward_C99.md) - OPTIONAL: A C++ compiler, if C++ applications are to be used (g++, etc.). If you do not require support for C++ applications, you can disable this support using the configure option --disable-cxx (configuring MPICH is described in step 1(d) below). - OPTIONAL: A Fortran compiler, if Fortran applications are to be used (gfortran, ifort, etc.). If you do not require support for Fortran applications, you can disable this support using --disable-fortran (configuring MPICH is described in step 1(d) below). - OPTIONAL: Python 3. Python 3 is needed to generate Fortran bindings. Also, you need to know what shell you are using since different shell has different command syntax. Command "echo $SHELL" prints out the current shell used by your terminal program. (b) Unpack the tar file and go to the top level directory: tar xzf mpich-4.3.0b1.tar.gz cd mpich-4.3.0b1 If your tar doesn't accept the z option, use gunzip mpich-4.3.0b1.tar.gz tar xf mpich-4.3.0b1.tar cd mpich-4.3.0b1 (c) Choose an installation directory, say /home//mpich-install, which is assumed to non-existent or empty. It will be most convenient if this directory is shared by all of the machines where you intend to run processes. If not, you will have to duplicate it on the other machines after installation. (d) Configure MPICH specifying the installation directory and device: for csh and tcsh: ./configure --prefix=/home//mpich-install |& tee c.txt for bash and sh: ./configure --prefix=/home//mpich-install 2>&1 | tee c.txt The configure will try to determine the best device (the internal network modules) based on system environment. You may also supply a device configuration. E.g. ./configure --prefix=... --with-device=ch4:ofi |... or: ./configure --prefix=... --with-device=ch4:ucx |... Refer to section below -- Alternate Channels and Devices -- for more details. Bourne-like shells, sh and bash, accept "2>&1 |". Csh-like shell, csh and tcsh, accept "|&". If a failure occurs, the configure command will display the error. Most errors are straight-forward to follow. For example, if the configure command fails with: "No Fortran compiler found. If you don't need to build any Fortran programs, you can disable Fortran support using --disable-fortran. If you do want to build Fortran programs, you need to install a Fortran compiler such as gfortran or ifort before you can proceed." ... it means that you don't have a Fortran compiler :-). You will need to either install one, or disable Fortran support in MPICH. If you are unable to understand what went wrong, please go to step (2) below, for reporting the issue to the MPICH developers and other users. (e) Build MPICH: for csh and tcsh: make |& tee m.txt for bash and sh: make 2>&1 | tee m.txt This step should succeed if there were no problems with the preceding step. Check file m.txt. If there were problems, do a "make clean" and then run make again with V=1. make V=1 |& tee m.txt (for csh and tcsh) OR make V=1 2>&1 | tee m.txt (for bash and sh) Then go to step (2) below, for reporting the issue to the MPICH developers and other users. (f) Install the MPICH commands: for csh and tcsh: make install |& tee mi.txt for bash and sh: make install 2>&1 | tee mi.txt This step collects all required executables and scripts in the bin subdirectory of the directory specified by the prefix argument to configure. (g) Add the bin subdirectory of the installation directory to your path in your startup script (.bashrc for bash, .cshrc for csh, etc.): for csh and tcsh: setenv PATH /home//mpich-install/bin:$PATH for bash and sh: PATH=/home//mpich-install/bin:$PATH ; export PATH Check that everything is in order at this point by doing: which mpicc which mpiexec These commands should display the path to your bin subdirectory of your install directory. IMPORTANT NOTE: The install directory has to be visible at exactly the same path on all machines you want to run your applications on. This is typically achieved by installing MPICH on a shared NFS file-system. If you do not have a shared NFS directory, you will need to manually copy the install directory to all machines at exactly the same location. (h) MPICH uses a process manager for starting MPI applications. The process manager provides the "mpiexec" executable, together with other utility executables. MPICH comes packaged with multiple process managers; the default is called Hydra. Now we will run an MPI job, using the mpiexec command as specified in the MPI standard. There are some examples in the install directory, which you have already put in your path, as well as in the directory mpich-4.3.0b1/examples. One of them is the classic CPI example, which computes the value of pi by numerical integration in parallel. To run the CPI example with 'n' processes on your local machine, you can use: mpiexec -n ./examples/cpi Test that you can run an 'n' process CPI job on multiple nodes: mpiexec -f machinefile -n ./examples/cpi The 'machinefile' is of the form: host1 host2:2 host3:4 # Random comments host4:1 'host1', 'host2', 'host3' and 'host4' are the hostnames of the machines you want to run the job on. The ':2', ':4', ':1' segments depict the number of processes you want to run on each node. If nothing is specified, ':1' is assumed. More details on interacting with Hydra can be found at https://github.com/pmodels/mpich/blob/main/doc/wiki/how_to/Using_the_Hydra_Process_Manager.md If you have completed all of the above steps, you have successfully installed MPICH and run an MPI example. ------------------------------------------------------------------------- 2. Reporting Installation or Usage Problems =========================================== [VERY IMPORTANT: PLEASE COMPRESS ALL FILES BEFORE SENDING THEM TO US. DO NOT SPAM THE MAILING LIST WITH LARGE ATTACHMENTS.] The distribution has been tested by us on a variety of machines in our environments as well as our partner institutes. If you have problems with the installation or usage of MPICH, please follow these steps: 1. First see the Frequently Asked Questions (FAQ) page at https://github.com/pmodels/mpich/blob/main/doc/wiki/faq/Frequently_Asked_Questions.md to see if the problem you are facing has a simple solution. Many common problems and their solutions are listed here. 2. If you cannot find an answer on the FAQ page, look through previous email threads on the discuss@mpich.org mailing list archive (https://lists.mpich.org/mailman/listinfo/discuss). It is likely someone else had a similar problem, which has already been resolved before. 3. If neither of the above steps work, please send an email to discuss@mpich.org. You need to subscribe to this list (https://lists.mpich.org/mailman/listinfo/discuss) before sending an email. Your email should contain the following files. ONCE AGAIN, PLEASE COMPRESS BEFORE SENDING, AS THE FILES CAN BE LARGE. Note that, depending on which step the build failed, some of the files might not exist. mpich-4.3.0b1/c.txt (generated in step 1(d) above) mpich-4.3.0b1/m.txt (generated in step 1(e) above) mpich-4.3.0b1/mi.txt (generated in step 1(f) above) mpich-4.3.0b1/config.log (generated in step 1(d) above) mpich-4.3.0b1/src/mpl/config.log (generated in step 1(d) above) mpich-4.3.0b1/src/pm/hydra/config.log (generated in step 1(d) above) DID WE MENTION? DO NOT FORGET TO COMPRESS THESE FILES! If you have compiled MPICH and are having trouble running an application, please provide the output of the following command in your email. mpiexec -info Finally, please include the actual error you are seeing when running the application, including the mpiexec command used, and the host file. If possible, please try to reproduce the error with a smaller application or benchmark and send that along in your bug report. 4. If you have found a bug in MPICH, you can report it on our Github page (https://github.com/pmodels/mpich/issues). ------------------------------------------------------------------------- 3. Compiler Flags ================= MPICH allows several sets of compiler flags to be used. The first three sets are configure-time options for MPICH, while the fourth is only relevant when compiling applications with mpicc and friends. (a) CFLAGS, CPPFLAGS, CXXFLAGS, FFLAGS, FCFLAGS, LDFLAGS and LIBS (abbreviated as xFLAGS): Setting these flags would result in the MPICH library being compiled/linked with these flags and the flags internally being used in mpicc and friends. (b) MPICHLIB_CFLAGS, MPICHLIB_CPPFLAGS, MPICHLIB_CXXFLAGS, MPICHLIB_FFLAGS, MPICHLIB_FCFLAGS, MPICHLIB_LDFLAGS and MPICHLIB_LIBS (abbreviated as MPICHLIB_xFLAGS): Setting these flags would result in the MPICH library being compiled/linked with these flags. However, these flags will *not* be used by mpicc and friends. (c) MPICH_MPICC_CFLAGS, MPICH_MPICC_CPPFLAGS, MPICH_MPICC_LDFLAGS, MPICH_MPICC_LIBS, and so on for MPICXX, MPIF77 and MPIFORT (abbreviated as MPICH_MPIX_FLAGS): These flags do *not* affect the compilation of the MPICH library itself, but will be internally used by mpicc and friends. +--------------------------------------------------------------------+ | | | | | | MPICH library | mpicc and friends | | | | | +--------------------+----------------------+------------------------+ | | | | | xFLAGS | Yes | Yes | | | | | +--------------------+----------------------+------------------------+ | | | | | MPICHLIB_xFLAGS | Yes | No | | | | | +--------------------+----------------------+------------------------+ | | | | | MPICH_MPIX_FLAGS | No | Yes | | | | | +--------------------+----------------------+------------------------+ All these flags can be set as part of configure command or through environment variables. Default flags -------------- By default, MPICH automatically adds certain compiler optimizations to MPICHLIB_CFLAGS. The currently used optimization level is -O2. ** IMPORTANT NOTE: Remember that this only affects the compilation of the MPICH library and is not used in the wrappers (mpicc and friends) that are used to compile your applications or other libraries. This optimization level can be changed with the --enable-fast option passed to configure. For example, to build an MPICH environment with -O3 for all language bindings, one can simply do: ./configure --enable-fast=O3 Or to disable all compiler optimizations, one can do: ./configure --disable-fast For more details of --enable-fast, see the output of "configure --help". For performance testing, we recommend the following flags: ./configure --enable-fast=O3,ndebug --disable-error-checking --without-timing \ --without-mpit-pvars Examples -------- Example 1: ./configure --disable-fast MPICHLIB_CFLAGS=-O3 MPICHLIB_FFLAGS=-O3 \ MPICHLIB_CXXFLAGS=-O3 MPICHLIB_FCFLAGS=-O3 This will cause the MPICH libraries to be built with -O3, and -O3 will *not* be included in the mpicc and other MPI wrapper script. Example 2: ./configure --disable-fast CFLAGS=-O3 FFLAGS=-O3 CXXFLAGS=-O3 FCFLAGS=-O3 This will cause the MPICH libraries to be built with -O3, and -O3 will be included in the mpicc and other MPI wrapper script. ------------------------------------------------------------------------- 4. Alternate Channels and Devices ================================= The communication mechanisms in MPICH are called "devices". MPICH supports ch3 and ch4 (default), as well as many third-party devices that are released and maintained by other institutes. ************************************* ch3 device ********** The ch3 device contains different internal communication options called "channels". We currently support nemesis (default) and sock channels. nemesis channel --------------- Nemesis provides communication using different networks (tcp, mx) as well as various shared-memory optimizations. To configure MPICH with nemesis, you can use the following configure option: --with-device=ch3:nemesis Shared-memory optimizations are enabled by default to improve performance for multi-processor/multi-core platforms. They can be disabled (at the cost of performance) either by setting the environment variable MPICH_NO_LOCAL to 1, or using the following configure option: --enable-nemesis-dbg-nolocal The --with-shared-memory= configure option allows you to choose how Nemesis allocates shared memory. The options are "auto", "sysv", and "mmap". Using "sysv" will allocate shared memory using the System V shmget(), shmat(), etc. functions. Using "mmap" will allocate shared memory by creating a file (in /dev/shm if it exists, otherwise /tmp), then mmap() the file. The default is "auto". Note that System V shared memory has limits on the size of shared memory segments so using this for Nemesis may limit the number of processes that can be started on a single node. ofi network module ``````````````````` The ofi netmod provides support for the OFI network programming interface. To enable, configure with the following option: --with-device=ch3:nemesis:ofi If the OFI include files and libraries are not in the normal search paths, you can specify them with the following options: --with-ofi-include= and --with-ofi-lib= ... or the if lib/ and include/ are in the same directory, you can use the following option: --with-ofi= If the OFI libraries are shared libraries, they need to be in the shared library search path. This can be done by adding the path to /etc/ld.so.conf, or by setting the LD_LIBRARY_PATH variable in your environment. It's also possible to set the shared library search path in the binary. If you're using gcc, you can do this by adding LD_LIBRARY_PATH=/path/to/lib (and) LDFLAGS="-Wl,-rpath -Wl,/path/to/lib" ... as arguments to configure. sock channel ------------ sock is the traditional TCP sockets based communication channel. It uses TCP/IP sockets for all communication including intra-node communication. So, though the performance of this channel is worse than that of nemesis, it should work on almost every platform. This channel can be configured using the following option: --with-device=ch3:sock ch4 device ********** The ch4 device contains different network and shared memory modules for communication. We currently support the ofi and ucx network modules, and posix shared memory module. ofi network module ``````````````````` The ofi netmod provides support for the OFI network programming interface. To enable, configure with the following option: --with-device=ch4:ofi[:provider] If the OFI include files and libraries are not in the normal search paths, you can specify them with the following options: --with-libfabric-include= and --with-libfabric-lib= ... or the if lib/ and include/ are in the same directory, you can use the following option: --with-libfabric= If specifying the provider, the MPICH library will be optimized specifically for the requested provider by removing runtime branches to determine provider capabilities. Note that using this feature with a version of the libfabric library older than that recommended with this version of MPICH is unsupported and may result in unexpected behavior. This is also true when using the environment variable FI_PROVIDER. The currently expected version of libfabric is: . ucx network module `````````````````` The ucx netmod provides support for the Unified Communication X library. It can be built with the following configure option: --with-device=ch4:ucx If the UCX include files and libraries are not in the normal search paths, you can specify them with the following options: --with-ucx-include= and --with-ucx-lib= ... or the if lib/ and include/ are in the same directory, you can use the following option: --with-ucx= By default, the UCX library throws warnings when the system does not enable certain features that might hurt performance. These are important warnings that might cause performance degradation on your system. But you might need root privileges to fix some of them. If you would like to disable such warnings, you can set the UCX log level to "error" instead of the default "warn" by using: UCX_LOG_LEVEL=error export UCX_LOG_LEVEL GPU support *********** GPU support is automatically enabled if CUDA, ZE, or HIP runtime is detected during configure. To specify where your GPU runtime is installed, use: --with-cuda= or --with-ze= or --with-hip= If the lib/ and include/ are not in the same path, both can be specified separately, for example: --with-cuda-include= and --with-cuda-lib= In addition, GPU support can be explicitly disabled by using: --without-cuda or --without-ze or --without-hip If desirable, GPU support can be disabled during runtime by setting environment variable MPIR_CVAR_ENABLE_GPU=0. This may help avoid the GPU initialization and detection overhead for non-GPU applications. ------------------------------------------------------------------------- 5. Alternate Process Managers ============================= hydra ----- Hydra is the default process management framework that uses existing daemons on nodes (e.g., ssh, pbs, slurm, sge) to start MPI processes. More information on Hydra can be found at https://github.com/pmodels/mpich/blob/main/doc/wiki/how_to/Using_the_Hydra_Process_Manager.md gforker ------- gforker is a process manager that creates processes on a single machine, by having mpiexec directly fork and exec them. gforker is mostly meant as a research platform and for debugging purposes, as it is only meant for single-node systems. slurm ----- Slurm is an external process manager not distributed with MPICH. MPICH's default process manager, hydra, has native support for Slurm and you can directly use it in Slurm environments (it will automatically detect Slurm and use Slurm capabilities). However, if you want to use the Slurm-provided "srun" process manager, you can use the "--with-pmi=slurm --with-pm=no" option with configure. Note that the "srun" process manager that comes with Slurm uses an older PMI standard which does not have some of the performance enhancements that hydra provides in Slurm environments. ------------------------------------------------------------------------- 6. Alternate Configure Options ============================== MPICH has a number of other features. If you are exploring MPICH as part of a development project, you might want to tweak the MPICH build with the following configure options. A complete list of configuration options can be found using: ./configure --help ------------------------------------------------------------------------- 7. Testing the MPICH installation ================================== To test MPICH, we package the MPICH test suite in the MPICH distribution. You can run the test suite after "make install" using: make testing The results summary will be placed in test/summary.xml. The test suite can be used independently to test any installed MPI implementations: cd test/mpi ./configure --with-mpi=/path/to/mpi make testing ------------------------------------------------------------------------- 8. Fault Tolerance ================== MPICH has some tolerance to process failures, and supports checkpointing and restart. Tolerance to Process Failures ----------------------------- The features described in this section should be considered experimental. Which means that they have not been fully tested, and the behavior may change in future releases. The below notes are some guidelines on what can be expected in this feature: - ERROR RETURNS: Communication failures in MPICH are not fatal errors. This means that if the user sets the error handler to MPI_ERRORS_RETURN, MPICH will return an appropriate error code in the event of a communication failure. When a process detects a failure when communicating with another process, it will consider the other process as having failed and will no longer attempt to communicate with that process. The user can, however, continue making communication calls to other processes. Any outstanding send or receive operations to a failed process, or wildcard receives (i.e., with MPI_ANY_SOURCE) posted to communicators with a failed process, will be immediately completed with an appropriate error code. - COLLECTIVES: For collective operations performed on communicators with a failed process, the collective would return an error on some, but not necessarily all processes. A collective call returning MPI_SUCCESS on a given process means that the part of the collective performed by that process has been successful. - PROCESS MANAGER: If used with the hydra process manager, hydra will detect failed processes and notify the MPICH library. Users can query the list of failed processes using MPIX_Comm_group_failed(). This functions returns a group consisting of the failed processes in the communicator. The function MPIX_Comm_remote_group_failed() is provided for querying failed processes in the remote processes of an intercommunicator. Note that hydra by default will abort the entire application when any process terminates before calling MPI_Finalize. In order to allow an application to continue running despite failed processes, you will need to pass the -disable-auto-cleanup option to mpiexec. - FAILURE NOTIFICATION: THIS IS AN UNSUPPORTED FEATURE AND WILL ALMOST CERTAINLY CHANGE IN THE FUTURE! In the current release, hydra notifies the MPICH library of failed processes by sending a SIGUSR1 signal. The application can catch this signal to be notified of failed processes. If the application replaces the library's signal handler with its own, the application must be sure to call the library's handler from it's own handler. Note that you cannot call any MPI function from inside a signal handler. Checkpoint and Restart ---------------------- MPICH supports checkpointing and restart fault-tolerance using BLCR. CONFIGURATION First, you need to have BLCR version 0.8.2 or later installed on your machine. If it's installed in the default system location, you don't need to do anything. If BLCR is not installed in the default system location, you'll need to tell MPICH's configure where to find it. You might also need to set the LD_LIBRARY_PATH environment variable so that BLCR's shared libraries can be found. In this case add the following options to your configure command: --with-blcr= LD_LIBRARY_PATH=/lib where is the directory where BLCR has been installed (whatever was specified in --prefix when BLCR was configured). After it's configured compile as usual (e.g., make; make install). Note, checkpointing is only supported with the Hydra process manager. VERIFYING CHECKPOINTING SUPPORT Make sure MPICH is correctly configured with BLCR. You can do this using: mpiexec -info This should display 'BLCR' under 'Checkpointing libraries available'. CHECKPOINTING THE APPLICATION There are two ways to cause the application to checkpoint. You can ask mpiexec to periodically checkpoint the application using the mpiexec option -ckpoint-interval (seconds): mpiexec -ckpointlib blcr -ckpoint-prefix /tmp/app.ckpoint \ -ckpoint-interval 3600 -f hosts -n 4 ./app Alternatively, you can also manually force checkpointing by sending a SIGUSR1 signal to mpiexec. The checkpoint/restart parameters can also be controlled with the environment variables HYDRA_CKPOINTLIB, HYDRA_CKPOINT_PREFIX and HYDRA_CKPOINT_INTERVAL. To restart a process: mpiexec -ckpointlib blcr -ckpoint-prefix /tmp/app.ckpoint -f hosts -n 4 -ckpoint-num where is the checkpoint number you want to restart from. These instructions can also be found on the MPICH wiki: https://github.com/pmodels/mpich/blob/main/doc/wiki/design/Checkpointing.md ------------------------------------------------------------------------- 9. Developer Builds =================== For MPICH developers who want to directly work on the primary version control system, there are a few additional steps involved (people using the release tarballs do not have to follow these steps). Details about these steps can be found here: https://github.com/pmodels/mpich/blob/main/doc/wiki/source_code/Github.md ------------------------------------------------------------------------- 10. Multiple Fortran compiler support ===================================== If the C compiler that is used to build MPICH libraries supports both multiple weak symbols and multiple aliases of common symbols, the Fortran binding can support multiple Fortran compilers. The multiple weak symbols support allow MPICH to provide different name mangling scheme (of subroutine names) required by different Fortran compilers. The multiple aliases of common symbols support enables MPICH to equal different common block symbols of the MPI Fortran constant, e.g. MPI_IN_PLACE, MPI_STATUS_IGNORE. So they are understood by different Fortran compilers. Since the support of multiple aliases of common symbols is new/experimental, users can disable the feature by using configure option --disable-multi-aliases if it causes any undesirable effect, e.g. linker warnings of different sizes of common symbols, MPIFCMB* (the warning should be harmless). We have only tested this support on a limited set of platforms/compilers. On linux, if the C compiler that builds MPICH is either gcc or icc, the above support will be enabled by configure. At the time of this writing, pgcc does not seem to have this multiple aliases of common symbols, so configure will detect the deficiency and disable the feature automatically. The tested Fortran compilers include GNU Fortran compilers (gfortan), Intel Fortran compiler (ifort), Portland Group Fortran compilers (pgfortran), Absoft Fortran compilers (af90), and IBM XL fortran compiler (xlf). What this means is that if mpich is built by gcc/gfortran, the resulting mpich library can be used to link a Fortran program compiled/linked by another fortran compiler, say pgf90, say through mpifort -fc=pgf90. As long as the Fortran program is linked without any errors by one of these compilers, the program shall be running fine. ------------------------------------------------------------------------- 11. ABI Compatibility ===================== The MPICH ABI compatibility initiative was announced at SC 2014 (http://www.mpich.org/abi). As a part of this initiative, Argonne, Intel, IBM and Cray have committed to maintaining ABI compatibility with each other. As a first step in this initiative, starting with version 3.1, MPICH is binary (ABI) compatible with Intel MPI 5.0. This means you can build your program with one MPI implementation and run with the other. Specifically, binary-only applications that were built and distributed with one of these MPI implementations can now be executed with the other MPI implementation. Some setup is required to achieve this. Suppose you have MPICH installed in /path/to/mpich and Intel MPI installed in /path/to/impi. You can run your application with mpich using: % export LD_LIBRARY_PATH=/path/to/mpich/lib:$LD_LIBRARY_PATH % mpiexec -np 100 ./foo or using Intel MPI using: % export LD_LIBRARY_PATH=/path/to/impi/lib:$LD_LIBRARY_PATH % mpiexec -np 100 ./foo This works irrespective of which MPI implementation your application was compiled with, as long as you use one of the MPI implementations in the ABI compatibility initiative. ------------------------------------------------------------------------- 12. Capability Sets ===================== The CH4 device contains a feature called "capability sets" to simplify configuration of MPICH on systems using the OFI netmod. This feature configures MPICH to use a predetermined set of OFI features based on the provider being used. Capability sets can be configured at compile time or runtime. Compile time configuration provides better performance by reducing unnecessary code branches, but at the cost of flexibility. To configure at compile time, the device string should be amended to include the OFI provider with the following option: --with-device=ch4:ofi:sockets This will setup the OFI netmod to use the optimal configuration for the sockets provider, and will set various compile time constants. These settings cannot be changed at runtime. If runtime configuration is needed, use: --with-device=ch4:ofi i.e. without the OFI provider extension, and set various environment variables to achieve a similar configuration. To select the desired provider: % export FI_PROVIDER=sockets This will select the OFI provider and the associated MPICH capability set. To change the preset configuration, there exists an extended set of environment variables. As an example, native provider RMA atomics can be disabled by using the environment variable: % export MPIR_CVAR_CH4_OFI_ENABLE_ATOMICS=0 For some configuration options (in particular, MPIR_CVAR_CH4_OFI_ENABLE_TAGGED and MPIR_CVAR_CH4_OFI_ENABLE_RMA), if disabled, some functionality may fallback to generic implementations. A full list of capability set configuration variables can be found in the environment variables README.envvar. ------------------------------------------------------------------------- 13. Threads =========== The supported thread level are configured by option: --enable-threads={single,funneled,serialized,multiple} The default depends on the configured device. With ch4, "multiple" is the default. Set thread level to "single" provides best performance when application does not use multiple threads. Use "multiple" to allow application to access MPI from multiple threads concurrently. With "multiple" thread level, there are a few choices for the internal critical section models. This is controlled by configure option: --enable-thread-cs={global,per-vci} Current default is to use "global" cs. Applications that do heavy concurrent MPI communications may experience slow down due to this global cs. The "per-vci" cs internally will use multiple VCI (virtual communication interface) critical sections, thus can provide much better performance. To achieve the best performance, applications should try to expose as much parallel information to MPI as possible. For example, if each threads use separate communicators, MPICH may be able to assign separate VCI for each thread, thus achieving the maximum performance. The multiple VCI support may increase the resource allocation and overheads during initialization. By default, only a single vci is used. Set MPIR_CVAR_CH4_NUM_VCIS= to enable multiple vcis at runtime. For best performance, match number of VCIs to the number threads application is using. MPICH supports multiple threading packages. The default is posix threads (pthreads), but solaris threads, windows threads, argobots and qthreads are also supported. To configure mpich to work with argobots or qthreads, use the following configure options: --with-thread-package=argobots \ CFLAGS="-I" \ LDFLAGS="-L" --with-thread-package=qthreads \ CFLAGS="-I" \ LDFLAGS="-L"