- announce - mailman3.cels.anl.gov

MPICH Receives 2024 ACM Software System Award
by Raffenetti, Ken 01 May '25

01 May '25

All, We are excited to share that MPICH has received the 2024 ACM Software System Award<https://urldefense.us/v3/__https://awards.acm.org/software-system__;!!G_uCf… >. This prestigious award recognizes software systems that have had a lasting impact on the computing field. We are proud to be in the company of distinguished past recipients like TCP/IP, GCC, Unix, TeX, LLVM, Java, NCSA Mosaic, and the World Wide Web. Thank you to all our collaborators and partners over the years who have been integral to MPICH’s success! Cheers, The MPICH Team

1 0

MPICH 4.3.0 released
by Raffenetti, Ken 03 Feb '25

03 Feb '25

The MPICH team is pleased to announce the availability of MPICH 4.3.0. Thanks to everyone who contributed to the release! See the CHANGES below for a detailed list of features. You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.3.0 # Support MPI memory allocation kinds side document. # Support MPI ABI Proposal. Configure with --enable-mpi-abi and build with mpicc_abi. By default, mpicc still builds and links with MPICH ABI. # Experimental API MPIX_Op_create_x. It supports user callback function with extra_state context and op destructor callback. It supports language bindings to use proxy function for language-specific user callbacks. # Experimental API MPIX_{Comm,File,Session,Win}_create_errhandler_x. They allow user error handlers to have extra_state context and corresponding destructor. This allows language bindings to implement user error handlers via proxy. # Experimental API MPIX_Request_is_complete. This is a pure request state query function that will not invoke progress, nor will free the request. This should help applications that want separate task dependency checking from progress engine to avoid progress contentions, especially in multi-threaded context. It is also useful for tools to profile non-deterministic calls such as MPI_Test. # Experimental API MPIX_Async_start. This function let applications to inject progress hooks to MPI progress. It allows application to implement custom asynchronous operations that will be progressed by MPI. It avoids having to implement separate progress mechanisms that may either take additional resource or contend with MPI progress and negatively impact performance. It also allows applications to create custom MPI operations, such as MPI nonblocking collectives, and achieve near native performance. # Added benchmark tests test/mpi/bench/p2p_{latency,bw}. # Added CMA support in CH4 IPC. # Added IPC read algorithm for intranode Allgather and Allgatherv. # Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy for inter-numa shm communication. # Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues. # ch4:ucx now supports dynamic processes. MPI_Comm_spawn{_multiple} will work. MPI_Open_port will fail due to ucx port name exceeds current MPI_MAX_PORT_NAME of 256. One can work around by use an info hint "port_name_size" and use a larger port name buffer. # PMI-1 defines PMI_MAX_PORT_NAME, which may be different from MPI_MAX_PORT_NAME. This is used by "PMI_Lookup_name". Consequently, MPI_Lookup_name accepts info hint "port_name_size" that may be larger than MPI_MAX_PORT_NAME. If the port name does not fit in "port_name_size", it will return a truncation error. # Autogen default to use -yaksa-depth=2. # Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on. # Added ch4 netmod API am_tag_send and am_tag_recv. # Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode. # Make check target will run ROMIO tests. # Add back handle conversion macros (f2c/c2f) to preserve ABI compatibility with older MPICH libraries # Fix compilation issue with g++ in -std=gnu++20 mode # Fix bug in MPI_ANY_SOURCE handling observed using the libfabric CXI provider # Add NIC information to error messages in ch4:ofi netmod

1 0

MPICH 4.3.0rc4 released
by Raffenetti, Ken 22 Jan '25

22 Jan '25

A new release candidate of MPICH, 4.3.0rc4, is now available for download. Since rc3, we have fixed a regression in static builds on macOS. See the CHANGES below for a detailed list of features. At this stage, only critical bugs will be fixed before the 4.3.0 GA release. Please give it a try! You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.3 # Support MPI memory allocation kinds side document. # Support MPI ABI Proposal. Configure with --enable-mpi-abi and build with mpicc_abi. By default, mpicc still builds and links with MPICH ABI. # Experimental API MPIX_Op_create_x. It supports user callback function with extra_state context and op destructor callback. It supports language bindings to use proxy function for language-specific user callbacks. # Experimental API MPIX_{Comm,File,Session,Win}_create_errhandler_x. They allow user error handlers to have extra_state context and corresponding destructor. This allows language bindings to implement user error handlers via proxy. # Experimental API MPIX_Request_is_complete. This is a pure request state query function that will not invoke progress, nor will free the request. This should help applications that want separate task dependency checking from progress engine to avoid progress contentions, especially in multi-threaded context. It is also useful for tools to profile non-deterministic calls such as MPI_Test. # Experimental API MPIX_Async_start. This function let applications to inject progress hooks to MPI progress. It allows application to implement custom asynchronous operations that will be progressed by MPI. It avoids having to implement separate progress mechanisms that may either take additional resource or contend with MPI progress and negatively impact performance. It also allows applications to create custom MPI operations, such as MPI nonblocking collectives, and achieve near native performance. # Added benchmark tests test/mpi/bench/p2p_{latency,bw}. # Added CMA support in CH4 IPC. # Added IPC read algorithm for intranode Allgather and Allgatherv. # Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy for inter-numa shm communication. # Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues. # ch4:ucx now supports dynamic processes. MPI_Comm_spawn{_multiple} will work. MPI_Open_port will fail due to ucx port name exceeds current MPI_MAX_PORT_NAME of 256. One can work around by use an info hint "port_name_size" and use a larger port name buffer. # PMI-1 defines PMI_MAX_PORT_NAME, which may be different from MPI_MAX_PORT_NAME. This is used by "PMI_Lookup_name". Consequently, MPI_Lookup_name accepts info hint "port_name_size" that may be larger than MPI_MAX_PORT_NAME. If the port name does not fit in "port_name_size", it will return a truncation error. # Autogen default to use -yaksa-depth=2. # Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on. # Added ch4 netmod API am_tag_send and am_tag_recv. # Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode. # Make check target will run ROMIO tests. # Add back handle conversion macros (f2c/c2f) to preserve ABI compatibility with older MPICH libraries # Fix compilation issue with g++ in -std=gnu++20 mode # Fix bug in MPI_ANY_SOURCE handling observed using the libfabric CXI provider # Add NIC information to error messages in ch4:ofi netmod

1 0

MPICH 4.3.0rc3 released
by Raffenetti, Ken 14 Jan '25

14 Jan '25

A new release candidate of MPICH, 4.3.0rc3, is now available for download. Since rc2, we have enhanced error reporting in ch4:ofi to include NIC information, fixed a bug in MPI_ANY_SOURCE handling with certain libfabric providers, and reverted a change that caused large messages with ch4:ofi to fallback to MPICH active message protocol. See the CHANGES below for a detailed list of features. At this stage, only critical bugs will be fixed before the 4.3.0 GA release. Please give it a try! You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.3 # Support MPI memory allocation kinds side document. # Support MPI ABI Proposal. Configure with --enable-mpi-abi and build with mpicc_abi. By default, mpicc still builds and links with MPICH ABI. # Experimental API MPIX_Op_create_x. It supports user callback function with extra_state context and op destructor callback. It supports language bindings to use proxy function for language-specific user callbacks. # Experimental API MPIX_{Comm,File,Session,Win}_create_errhandler_x. They allow user error handlers to have extra_state context and corresponding destructor. This allows language bindings to implement user error handlers via proxy. # Experimental API MPIX_Request_is_complete. This is a pure request state query function that will not invoke progress, nor will free the request. This should help applications that want separate task dependency checking from progress engine to avoid progress contentions, especially in multi-threaded context. It is also useful for tools to profile non-deterministic calls such as MPI_Test. # Experimental API MPIX_Async_start. This function let applications to inject progress hooks to MPI progress. It allows application to implement custom asynchronous operations that will be progressed by MPI. It avoids having to implement separate progress mechanisms that may either take additional resource or contend with MPI progress and negatively impact performance. It also allows applications to create custom MPI operations, such as MPI nonblocking collectives, and achieve near native performance. # Added benchmark tests test/mpi/bench/p2p_{latency,bw}. # Added CMA support in CH4 IPC. # Added IPC read algorithm for intranode Allgather and Allgatherv. # Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy for inter-numa shm communication. # Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues. # ch4:ucx now supports dynamic processes. MPI_Comm_spawn{_multiple} will work. MPI_Open_port will fail due to ucx port name exceeds current MPI_MAX_PORT_NAME of 256. One can work around by use an info hint "port_name_size" and use a larger port name buffer. # PMI-1 defines PMI_MAX_PORT_NAME, which may be different from MPI_MAX_PORT_NAME. This is used by "PMI_Lookup_name". Consequently, MPI_Lookup_name accepts info hint "port_name_size" that may be larger than MPI_MAX_PORT_NAME. If the port name does not fit in "port_name_size", it will return a truncation error. # Autogen default to use -yaksa-depth=2. # Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on. # Added ch4 netmod API am_tag_send and am_tag_recv. # Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode. # Make check target will run ROMIO tests. # Add back handle conversion macros (f2c/c2f) to preserve ABI compatibility with older MPICH libraries # Fix compilation issue with g++ in -std=gnu++20 mode # Fix bug in MPI_ANY_SOURCE handling observed using the libfabric CXI provider # Add NIC information to error messages in ch4:ofi netmod

1 0

MPICH 4.3.0rc2 released
by Raffenetti, Ken 02 Jan '25

02 Jan '25

A new release candidate of MPICH, 4.3.0rc2, is now available for download. Since rc1, we have fixed a compilation issue with the C++ bindings and added back the handle conversion macros (f2c/c2f) for ABI compatibility. See the CHANGES below for a detailed list of features. At this stage, only critical bugs will be fixed before the 4.3.0 GA release. Please give it a try! You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.3 # Support MPI memory allocation kinds side document. # Support MPI ABI Proposal. Configure with --enable-mpi-abi and build with mpicc_abi. By default, mpicc still builds and links with MPICH ABI. # Experimental API MPIX_Op_create_x. It supports user callback function with extra_state context and op destructor callback. It supports language bindings to use proxy function for language-specific user callbacks. # Experimental API MPIX_{Comm,File,Session,Win}_create_errhandler_x. They allow user error handlers to have extra_state context and corresponding destructor. This allows language bindings to implement user error handlers via proxy. # Experimental API MPIX_Request_is_complete. This is a pure request state query function that will not invoke progress, nor will free the request. This should help applications that want separate task dependency checking from progress engine to avoid progress contentions, especially in multi-threaded context. It is also useful for tools to profile non-deterministic calls such as MPI_Test. # Experimental API MPIX_Async_start. This function let applications to inject progress hooks to MPI progress. It allows application to implement custom asynchronous operations that will be progressed by MPI. It avoids having to implement separate progress mechanisms that may either take additional resource or contend with MPI progress and negatively impact performance. It also allows applications to create custom MPI operations, such as MPI nonblocking collectives, and achieve near native performance. # Added benchmark tests test/mpi/bench/p2p_{latency,bw}. # Added CMA support in CH4 IPC. # Added IPC read algorithm for intranode Allgather and Allgatherv. # Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy for inter-numa shm communication. # Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues. # ch4:ucx now supports dynamic processes. MPI_Comm_spawn{_multiple} will work. MPI_Open_port will fail due to ucx port name exceeds current MPI_MAX_PORT_NAME of 256. One can work around by use an info hint "port_name_size" and use a larger port name buffer. # PMI-1 defines PMI_MAX_PORT_NAME, which may be different from MPI_MAX_PORT_NAME. This is used by "PMI_Lookup_name". Consequently, MPI_Lookup_name accepts info hint "port_name_size" that may be larger than MPI_MAX_PORT_NAME. If the port name does not fit in "port_name_size", it will return a truncation error. # Autogen default to use -yaksa-depth=2. # Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on. # Added ch4 netmod API am_tag_send and am_tag_recv. # Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode. # Make check target will run ROMIO tests. # Add back handle conversion macros (f2c/c2f) to preserve ABI compatibility with older MPICH libraries # Fix compilation issue with g++ in -std=gnu++20 mode

1 0

MPICH 4.3.0rc1 released
by Raffenetti, Ken 17 Dec '24

17 Dec '24

A new release candidate of MPICH, 4.3.0rc1, is now available for download. See the CHANGES below for a detailed list of features. At this stage, only critical bugs will be fixed before the 4.3.0 GA release. Please give it a try! You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.3 # Support MPI memory allocation kinds side document. # Support MPI ABI Proposal. Configure with --enable-mpi-abi and build with mpicc_abi. By default, mpicc still builds and links with MPICH ABI. # Experimental API MPIX_Op_create_x. It supports user callback function with extra_state context and op destructor callback. It supports language bindings to use proxy function for language-specific user callbacks. # Experimental API MPIX_{Comm,File,Session,Win}_create_errhandler_x. They allow user error handlers to have extra_state context and corresponding destructor. This allows language bindings to implement user error handlers via proxy. # Experimental API MPIX_Request_is_complete. This is a pure request state query function that will not invoke progress, nor will free the request. This should help applications that want separate task dependency checking from progress engine to avoid progress contentions, especially in multi-threaded context. It is also useful for tools to profile non-deterministic calls such as MPI_Test. # Experimental API MPIX_Async_start. This function let applications to inject progress hooks to MPI progress. It allows application to implement custom asynchronous operations that will be progressed by MPI. It avoids having to implement separate progress mechanisms that may either take additional resource or contend with MPI progress and negatively impact performance. It also allows applications to create custom MPI operations, such as MPI nonblocking collectives, and achieve near native performance. # Added benchmark tests test/mpi/bench/p2p_{latency,bw}. # Added CMA support in CH4 IPC. # Added IPC read algorithm for intranode Allgather and Allgatherv. # Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy for inter-numa shm communication. # Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues. # ch4:ucx now supports dynamic processes. MPI_Comm_spawn{_multiple} will work. MPI_Open_port will fail due to ucx port name exceeds current MPI_MAX_PORT_NAME of 256. One can work around by use an info hint "port_name_size" and use a larger port name buffer. # PMI-1 defines PMI_MAX_PORT_NAME, which may be different from MPI_MAX_PORT_NAME. This is used by "PMI_Lookup_name". Consequently, MPI_Lookup_name accepts info hint "port_name_size" that may be larger than MPI_MAX_PORT_NAME. If the port name does not fit in "port_name_size", it will return a truncation error. # Autogen default to use -yaksa-depth=2. # Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on. # Added ch4 netmod API am_tag_send and am_tag_recv. # Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode. # Make check target will run ROMIO tests.

1 0

Join us for the MPICH BoF on Wednesday evening at SC24
by Raffenetti, Ken 18 Nov '24

18 Nov '24

All, The MPICH team will host a Birds-of-a-Feather session Wednesday evening as part of the SC24 proceedings. We hope you will join us to discuss the latest MPICH developments and hear from the community. We will again feature an MPICH user presentation alongside the traditional partner updates this year. See below for more info. Date and Time: Wednesday, November 20th, 5:15pm - 6:45pm EST Location: B308 Agenda: * Welcome and ANL Update * Welcome – Ken Raffenetti * ANL Update – Yanfei Guo & Hui Zhou * Partner Presentations * Maria Garzaran, Intel * Hari Subramoni, MVAPICH * Simon Pickartz, ParTex AG * Naveen Ravi, HPE * Jim Dinan, NVIDIA * Charles Shereda, Cornelis Networks * User Presentations * Jeff Hammond, NVIDIA * Q&A Cheers, The MPICH team

1 0

MPICH 4.3.0b1 Released!
by Zhou, Hui 15 Nov '24

15 Nov '24

MPICH team is happy to announce the release of version 4.3.0b1. It is now available for download at https://urldefense.us/v3/__https://www.mpich.org/downloads__;!!G_uCfscf7eWS… . This is the first beta release for the MPICH 4.3.x series. Regards, The MPICH team =============================================================================== Changes in 4.3 =============================================================================== # Support MPI memory allocation kinds side document. # Support MPI ABI Proposal. Configure with --enable-mpi-abi and build with mpicc_abi. By default, mpicc still builds and links with MPICH ABI. # Experimental API MPIX_Op_create_x. It supports user callback function with extra_state context and op destructor callback. It supports language bindings to use proxy function for language-specific user callbacks. # Experimental API MPIX_{Comm,File,Session,Win}_create_errhandler_x. They allow user error handlers to have extra_state context and corresponding destructor. This allows language bindings to implement user error handlers via proxy. # Experimental API MPIX_Request_is_complete. This is a pure request state query function that will not invoke progress, nor will free the request. This should help applications that want separate task dependency checking from progress engine to avoid progress contentions, especially in multi-threaded context. It is also useful for tools to profile non-deterministic calls such as MPI_Test. # Experimental API MPIX_Async_start. This function let applications to inject progress hooks to MPI progress. It allows application to implement custom asynchronous operations that will be progressed by MPI. It avoids having to implement separate progress mechanisms that may either take additional resource or contend with MPI progress and negatively impact performance. It also allows applications to create custom MPI operations, such as MPI nonblocking collectives, and achieve near native performance. # Added benchmark tests test/mpi/bench/p2p_{latency,bw}. # Added CMA support in CH4 IPC. # Added IPC read algorithm for intranode Allgather and Allgatherv. # Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy for inter-numa shm communication. # Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues. # ch4:ucx now supports dynamic processes. MPI_Comm_spawn{_multiple} will work. MPI_Open_port will fail due to ucx port name exceeds current MPI_MAX_PORT_NAME of 256. One can work around by use an info hint "port_name_size" and use a larger port name buffer. # PMI-1 defines PMI_MAX_PORT_NAME, which may be different from MPI_MAX_PORT_NAME. This is used by "PMI_Lookup_name". Consequently, MPI_Lookup_name accepts info hint "port_name_size" that may be larger than MPI_MAX_PORT_NAME. If the port name does not fit in "port_name_size", it will return a truncation error. # Autogen default to use -yaksa-depth=2. # Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on. # Added ch4 netmod API am_tag_send and am_tag_recv. # Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode. # Make check target will run ROMIO tests.

1 0

MPICH 4.2.3 released
by Raffenetti, Ken 02 Oct '24

02 Oct '24

All, A new stable release of MPICH, 4.2.3, is now available for download. This is the third bug-fix release in the 4.2.x series. See the CHANGES below for a detailed list of fixes. For the full lists of commits, see https://urldefense.us/v3/__https://github.com/pmodels/mpich/compare/v4.2.2_… …v4.2.3/ You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.2.3 =============================================================================== # Update embedded libfabric to fix a build issue on FreeBSD # Fix HIP support for use with AMD GPUs # Fix potential invalid context issue in CUDA memory hooks # Fix GPU fallback path in ch4/ofi for Intel GPU buffers # Fix IPC handle destruction with Level Zero API (Intel GPU) # Fix potential crash in MPI_ISENDRECV with derived datatypes # Fix bug in persistent MPI_GATHER that incorrectly cached buffer contents at init time # Fix memory allocation bug in ROMIO GPFS driver # Fix missing error names in ch4/ofi netmod # Fix potential hang in multi-VCI active message RMA # Fix bug in ch3 large count support with derived datatypes # Fix manpage generation to provide aliases for large-count versions # Fix potential crash in Hydra with long PMI command messages # Fix bug in exit status capture in Hydra when there are multiple processes with non-zero exit # Fix implementation of C/Fortran status conversion functions # Fix implementation of MPI_Type_create_f90_xxx functions

1 0

MPICH 4.2.3rc1 released
by Raffenetti, Ken 25 Sep '24

25 Sep '24

A new stable release candidate of MPICH, 4.2.3rc1, is now available for download. See the CHANGES below for a detailed list of fixes. At this stage, only critical bugs will be fixed before the 4.2.3 GA release. Please give it a try! You can find the release on our downloads page (https://urldefense.us/v3/__https://www.mpich.org/downloads/__;!!G_uCfscf7eW… ) Regards, The MPICH team =============================================================================== Changes in 4.2.3 =============================================================================== # Update embedded libfabric to fix a build issue on FreeBSD # Fix HIP support for use with AMD GPUs # Fix potential invalid context issue in CUDA memory hooks # Fix GPU fallback path in ch4/ofi for Intel GPU buffers # Fix IPC handle destruction with Level Zero API (Intel GPU) # Fix potential crash in MPI_ISENDRECV with derived datatypes # Fix bug in persistent MPI_GATHER that incorrectly cached buffer contents at init time # Fix memory allocation bug in ROMIO GPFS driver # Fix missing error names in ch4/ofi netmod # Fix potential hang in multi-VCI active message RMA # Fix bug in ch3 large count support with derived datatypes # Fix manpage generation to provide aliases for large-count versions # Fix potential crash in Hydra with long PMI command messages # Fix bug in exit status capture in Hydra when there are multiple processes with non-zero exit # Fix implementation of C/Fortran status conversion functions # Fix implementation of MPI_Type_create_f90_xxx functions

1 0