Merge branch 'master-4.3.2' into 'master'

Vendor merge MPICH 4.3.2

Closes #101 and #78

See merge request parastation/psmpi!274

Author: spickartz

Makefile.in

@@ -13,6 +13,7 @@ SRCDIR := $(shell cd @srcdir@;pwd)
 export TOP_BUILDDIR:=@TOP_BUILDDIR@
 CONFSET:=@CONFSET@
 CUDA_SUPPORT:=@CUDA_SUPPORT@
+ENABLE_MPI_ABI:=@ENABLE_MPI_ABI@
 export TOP_SRCDIR = $(SRCDIR)
 export CONFIGARGS:=@CONFIGARGS@
 exec_prefix=@exec_prefix@
@@ -52,6 +53,11 @@ endif
 endif
 endif
 
+MPI_ABI=""
+ifeq ($(ENABLE_MPI_ABI),true)
+MPI_ABI=--enable-mpi-abi ## Build tests with MPI ABI
+endif
+
 ifndef MPIEXEC
 MPIEXEC=mpiexec
 endif
@@ -121,7 +127,7 @@ else
 	( if command -v -- "$${MPIEXEC}" > /dev/null 2>&1; \
 	  then echo -n "INFO: mpiexec command: "; which $${MPIEXEC}; echo "INFO: mpiexec arguments: $${MPITEST_MPIEXECARG}";  \
 	  else echo "WARNING: $${MPIEXEC} does not seem to be valid command! Please adjust PATH or set MPIEXEC accordingly."; sleep 2; fi ); \
-	( mkdir -p mpich2/test/mpi ; cd mpich2/test/mpi ; $(SRCDIR)/mpich2/test/mpi/configure --with-mpi=$(prefix) $(CUDA_TESTS) $(LARGE_TESTS)); \
+	( mkdir -p mpich2/test/mpi ; cd mpich2/test/mpi ; $(SRCDIR)/mpich2/test/mpi/configure --with-mpi=$(prefix) $(CUDA_TESTS) $(LARGE_TESTS) $(MPI_ABI)); \
 	( hostname; rpm -qi pscom; rpm -qi psmgmt; $(SRCDIR)/scripts/vcversion -r $(SRCDIR); \
 	  $(MAKE) -C mpich2/test/mpi testing "TESTLIST=$${TESTSETLIST}" $(TESTDIRLIST) && \
 	  cp $(TESTSUMMARY).xml $${_LOGNAME}.xml && \

configure.ac

@@ -602,6 +602,16 @@ if test -z "${CHANNEL_DEVICE}" ; then
    export MPICHLIB_CFLAGS="${MPICHLIB_CFLAGS} -DMPIR_CONTEXT_DYNAMIC_PROC_WIDTH=0"
 fi
 
+AC_ARG_ENABLE(mpi-abi,
+              AS_HELP_STRING([--enable-mpi-abi],
+              [Enable building libmpi_abi.so]),
+              [
+                  ENABLE_MPI_ABI="true"
+                  CONF_OPS="$CONF_OPS --enable-mpi-abi"
+              ],[
+                  ENABLE_MPI_ABI="false"
+              ])
+
 # Check for OpenPMIx library
 openpmix_min_version=4.2.0
 openpmix_min_version_numeric="0x00040200"
@@ -756,6 +766,7 @@ test -n "$FCFLAGS" && CONF_OPS="$CONF_OPS 'FCFLAGS=$FCFLAGS'"
 
 AC_SUBST([CONFSET])
 AC_SUBST([CUDA_SUPPORT])
+AC_SUBST([ENABLE_MPI_ABI])
 AC_SUBST([MODULEFILE])
 AC_SUBST([EXTRA_VERSION_INFO])
 AC_SUBST([CUDA_HOME])
@@ -831,6 +842,10 @@ fi
 if test "$ENABLE_PMIX" = "true" ; then
    conf_echo "PMIx support."
 fi
+echo "${CONF_ECHO_PREP_PHRASE}MPICH ABI support."
+if test "$ENABLE_MPI_ABI" = "true" ; then
+	echo "${CONF_ECHO_PREP_PHRASE}MPI ABI support."
+fi
 if test "$ENABLE_STATISTICS" = "true" ; then
    conf_echo "support for statistics."
 fi

dist/Makefile.sources

@@ -11,6 +11,7 @@
 
 %{?_with_threading: %define threading yes}
 %{?_with_cuda: %define cuda 1}
+%{?_with_mpiabi: %define mpiabi 1}
 %{?_without_pmix: %define disable_pmix 1}
 
 %{?_with_manpages: %define manpages 1}
@@ -20,7 +21,7 @@
 
 Summary:   ParaStation MPI
 Vendor:    ParTec AG, Munich, Germany
-Name:      psmpi%{?subname:-%{subname}}%{?threading:-mt}%{?cuda:-cuda}%{?disable_pmix:-nopmix}
+Name:      psmpi%{?subname:-%{subname}}%{?threading:-mt}%{?cuda:-cuda}%{?mpiabi:-mpiabi}%{?disable_pmix:-nopmix}
 Version:   @VERSION_psmpi@
 Release:   @RELEASE_psmpi@
 License:   QPL and mpich2/COPYRIGHT
@@ -47,7 +48,7 @@ BuildRequires:	automake autoconf libtool
 %define _use_internal_dependency_generator 0
 %define __find_requires %_builddir/psmpi-%{version}-%{release}/dist/rpm_depgen
 
-%define prefix /opt/parastation/mpi%{?subname:-%{subname}}%{?threading:-mt}%{?cuda:-cuda}
+%define prefix /opt/parastation/mpi%{?subname:-%{subname}}%{?threading:-mt}%{?cuda:-cuda}%{?mpiabi:-mpiabi}
 Prefix: %{prefix}
 
 
@@ -61,6 +62,7 @@ ParaStation MPI. Based on MPICH using the ParaStation communication library psco
 %{?threading: Threading support: MPI_THREAD_MULTIPLE}
 %{?cuda: CUDA awareness   : YES}%{!?cuda: CUDA awareness   : NO}
 %{!?disable_pmix: PMIx support     : YES}%{?disable_pmix: PMIx support     : NO}
+%{?mpiabi: MPICH ABI and MPI ABI support}%{!?mpiabi: MPICH ABI support  }
 MPICH is a freely available, portable implementation of MPI, the Standard for message-passing libraries.
 http://www-unix.mcs.anl.gov/mpi/mpich/
 
@@ -86,13 +88,14 @@ Build options (confset):
 additive options:
 "rpmbuild --with threading" Support for MPI_THREAD_MULTIPLE
 "rpmbuild --with cuda"      Enable CUDA awareness
+"rpmbuild --with mpiabi"    MPI ABI support in addition to MPICH ABI
 "rpmbuild --without pmix"   Disable PMIx support
 "rpmbuild --with manpages"  Include MPICH's man pages in doc package (build needs Sowing tools, see http://wgropp.cs.illinois.edu/projects/software/sowing/)
 
 Optional overwrite subname (default to compiler confset):
 "rpmbuild --define='subname _any_subname_'"
 
-Now using --with %{?confset:%{confset}}%{!?confset:default} %{?threading:--enable-threading} %{?cuda:--with-cuda} --with-pmix%{?disable_pmix:=no}
+Now using --with %{?confset:%{confset}}%{!?confset:default} %{?threading:--enable-threading} %{?cuda:--with-cuda} %{?mpiabi:--enable-mpi-abi} --with-pmix%{?disable_pmix:=no}
 subname: %{?subname:%{subname}}%{!?subname:<none>}
 =========================================================
 DOK
@@ -102,7 +105,7 @@ sleep 1
 
 %build
 ./autogen.sh
-%configure --with-confset=%{?confset:%{confset}}%{!?confset:default} %{?threading:--enable-threading} %{?cuda:--with-cuda} --with-pmix%{?disable_pmix:=no}
+%configure --with-confset=%{?confset:%{confset}}%{!?confset:default} %{?threading:--enable-threading} %{?cuda:--with-cuda} %{?mpiabi:--enable-mpi-abi} --with-pmix%{?disable_pmix:=no}
 
 %make_build
 
@@ -128,6 +131,7 @@ make mandoc
 %{_bindir}/mpif90
 %{_bindir}/mpifort
 %{_bindir}/mpivars
+%{_sysconfdir}/mpixxx_opts.conf
 %{_includedir}/mpi-ext.h
 %{_includedir}/mpi.h
 %{_includedir}/mpi.mod

dist/psmpi.spec.templ

@@ -96,6 +96,10 @@ We recommend developers to use the `devel` confset for more error checking and d
 | disabled
 | Enable multi-thread support
 
+| `--enable-mpi-abi`
+| disabled
+| Enable MPI ABI support (provides `libmpi_abi.so` and `mpicc_abi`/ `mpicxx_abi`)
+
 | `--enable-hydra`
 | disabled
 | Enable build and installation of MPICH's process manager Hydra

doc/antora/modules/ROOT/pages/installation.adoc

@@ -147,6 +147,10 @@ Makefile.am-stamp
 # random stuff
 /maint/gcovmerge
 /src/binding/cxx/buildiface-stamp
+/src/binding/abi/mpi_abi_internal.h
+/src/binding/abi/mpi_abi_util.c
+/src/binding/abi/c_binding_abi.c
+/src/mpi/romio/include/romio_abi_internal.h
 
 # script-generated f90 test files
 
@@ -583,7 +587,7 @@ Makefile.am-stamp
 
 # /test/mpi/
 /test/mpi/libtool
-/test/mpi/coll/testlist.cvar
+/test/mpi/coll/testlist.collalgo
 /test/mpi/attr/testlist.dtp
 /test/mpi/pt2pt/testlist.dtp
 /test/mpi/coll/testlist.dtp

mpich2/.gitignore

@@ -1,98 +1,169 @@
 ===============================================================================
-                               Changes in 4.2.3
+                               Changes in 4.3.2
 ===============================================================================
+# Improve libfabric provider selection when available providers have
+  negative internal score
 
-# Update embedded libfabric to fix a build issue on FreeBSD
+# Improve error messages when Level Zero failures are detected
 
-# Fix HIP support for use with AMD GPUs
+# Improve localhost detection in Hydra
 
-# Fix potential invalid context issue in CUDA memory hooks
+# Update libfabric usage to silence deprecation warnings
 
-# Fix GPU fallback path in ch4/ofi for Intel GPU buffers
+# Update yaksa for improved reproducibility in code generation
 
-# Fix IPC handle destruction with Level Zero API (Intel GPU)
+# Update embedded UCX to v1.19.0
 
-# Fix potential crash in MPI_ISENDRECV with derived datatypes
+# Update embedded libfabric to fix build issue with GCC 15
 
-# Fix bug in persistent MPI_GATHER that incorrectly cached buffer
-  contents at init time
+# Add compatibility with CUDA 13
 
-# Fix memory allocation bug in ROMIO GPFS driver
+# Fix missing const in nondestructive request test and status query
+  functions
 
-# Fix missing error names in ch4/ofi netmod
+# Fix HCOLL support
 
-# Fix potential hang in multi-VCI active message RMA
+# Fix crash with GPU-aware build when running on systems with no GPUs
 
-# Fix bug in ch3 large count support with derived datatypes
+# Fix HIP device query
 
-# Fix manpage generation to provide aliases for large-count versions
+# Fix singleton init with Hydra
 
-# Fix potential crash in Hydra with long PMI command messages
+# Fix thread safety for Level Zero memcpy functions
 
-# Fix bug in exit status capture in Hydra when there are multiple
-  processes with non-zero exit
+# Fix potential crash when release gather collective initialization fails
 
-# Fix implementation of C/Fortran status conversion functions
+# Fix ch3 connect/accept protocol handling when a discard event arrives
+  after a connection is already established
 
-# Fix implementation of MPI_Type_create_f90_xxx functions
+# Fix inlining for posix eager modules in ch4/shm
+
+# Fix weak attribute usage in MPI ABI build
+
+# Fix potential use-after-free bug in Hydra during spawn operations
+
+# Fix bug in persistent bcast algorithm
+
+# Fix bug in fabric coordinate retrieval with PMIx
+
+# Fix integer overflow and signed/unsigned bugs in ROMIO
+
+# Fix Quobtye ROMIO driver build error
+
+# Fix broken string conversion in mpi_f08 module
+
+# Fix compilation issue with mpi_f08 and NAG compiler
+
+# Fixes for various test program bugs
 
 ===============================================================================
-                               Changes in 4.2.2
+                               Changes in 4.3.1
 ===============================================================================
+# Fix initialization in GPU-aware builds when no devices are present
+
+# Fix internal pmix.h header conflict when building with an external
+  PMIx library.
+
+# Fix build issue with Slurm by removing dependency on libslurm and
+  always using internal logic for parsing the Slurm hostfile.
 
-# Update embedded libfabric to v1.20.1 and fix compilation with GCC 14.
+# Fix potential stale GPU IPC handle usage resulting in data corruption
+  or crashes
 
-# Fix dataloop support for MPIX_Type_iov routines
+# Update XPMEM thresholds to avoid excessive buffer mapping overhead
 
-# Fix crash in Hydra when system has many local ip addresses
+# Fix potential hang in ROMIO when setting info hints on certain files
 
-# Fix RMA fallback check in ch4:ofi netmod
+# Improved detection of incompatible PMI[x] client/server configuration
 
-# Fix MPI_UNDEFINED handling in mpi_f08 module
+# Fix use of PMIX_PREFIX attribute for certain versions of OpenPMIx
 
-# Fix Slurm environment variable inheritance in Hydra
+# Fix Intel GPU output with MPIR_CVAR_DEBUG_SUMMARY
 
-# Fix multiple issues with large count support in ROMIO
+# Fix F08 binding compilation with nvfortran
 
-# Fix potential hang in init using PMIx client and nonstandard keys
+# Fix line continuation Hydra's --configfile option
 
-# Fix crash if PMIx client cannot get appnum from server during init
+# Fix valgrind uninitialized read warnings in ch3
 
-# Fix other build errors and warnings
+# Fix missing mpixxx_opts.conf file with help text for mpicc and friends
+
+# Fixes for several compiler warnings
 
 ===============================================================================
-                               Changes in 4.2.1
+                               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.
 
-# Disable flat_namespace to build libmpifort on macOS by default
+# 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.
 
-# Prefix warning messages with "MPICH"
+# Added benchmark tests test/mpi/bench/p2p_{latency,bw}. 
 
-# Add --disable-doc configure option
+# Added CMA support in CH4 IPC.
 
-# Fix support for building MPICH Fortran support with Xcode 15 on macOS
+# Added IPC read algorithm for intranode Allgather and Allgatherv.
 
-# Fix bug in MPI_WIN_SHARED_QUERY when window is not shared
+# Added CVAR MPIR_CVAR_CH4_SHM_POSIX_TOPO_ENABLE to enable non-temporal memcpy
+  for inter-numa shm communication.
 
-# Fix bug in ch4/ofi gpu pipelining
+# Added CVAR MPIR_CVAR_DEBUG_PROGRESS_TIMEOUT for debugging MPI deadlock issues.  
 
-# Fixes for Intel GPU support
+# 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.
 
-# Fix memory leak in ch4/shm collectives
+# 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.
 
-# Fix bug in MPI_COMM_SPLIT with intercommunicators and non-zero root
+# Autogen default to use -yaksa-depth=2.
 
-# Fix bug in DAOS ROMIO driver
+# Default MPIR_CVAR_CH4_ROOTS_ONLY_PMI to on.
 
-# Fix bug in cycling error code array
+# Added ch4 netmod API am_tag_send and am_tag_recv.
 
-# Return an error if there is failure to create a datatype in mpi_f08
-  module for noncontiguous user data
+# Added MPIR_CVAR_CH4_OFI_EAGER_THRESHOLD to force RNDV send mode.
 
-# Return an error when shared memory segment creation fails
+# 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
 
 ===============================================================================
-                               Changes in 4.2.0
+                               Changes in 4.2
 ===============================================================================
 # Complete support MPI 4.1 specification
 
@@ -155,8 +226,6 @@
   decided at the October 2023 meeting that this was a mistake and will
   be removed in an erratum.
 
-# Fix potential crash in GPU memory hooks
-
 
 ===============================================================================
                                Changes in 4.1