2005-07-27 17:42:37 +04:00
# -*- shell-script -*-
#
2005-11-05 22:57:48 +03:00
# Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
# University Research and Technology
# Corporation. All rights reserved.
# Copyright (c) 2004-2005 The University of Tennessee and The University
# of Tennessee Research Foundation. All rights
# reserved.
2011-07-04 18:00:41 +04:00
# Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
2005-07-27 17:42:37 +04:00
# University of Stuttgart. All rights reserved.
# Copyright (c) 2004-2005 The Regents of the University of California.
# All rights reserved.
2013-12-07 03:40:34 +04:00
# Copyright (c) 2007-2013 Cisco Systems, Inc. All rights reserved.
2011-08-05 00:26:08 +04:00
# Copyright (c) 2008-2011 Mellanox Technologies. All rights reserved.
2011-02-16 08:37:22 +03:00
# Copyright (c) 2011 Oracle and/or its affiliates. All rights reserved.
2013-07-18 02:12:29 +04:00
# Copyright (c) 2013 NVIDIA Corporation. All rights reserved.
2005-07-27 17:42:37 +04:00
# $COPYRIGHT$
2011-07-04 18:00:41 +04:00
#
2005-07-27 17:42:37 +04:00
# Additional copyrights may follow
2011-07-04 18:00:41 +04:00
#
2005-07-27 17:42:37 +04:00
# $HEADER$
#
2008-01-15 02:22:03 +03:00
# MCA_btl_openib_POST_CONFIG([should_build])
# ------------------------------------------
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
AC_DEFUN([MCA_opal_btl_openib_POST_CONFIG], [
2008-07-08 10:32:54 +04:00
AM_CONDITIONAL([MCA_btl_openib_have_xrc], [test $1 -eq 1 -a "x$btl_openib_have_xrc" = "x1"])
2008-05-02 15:52:33 +04:00
AM_CONDITIONAL([MCA_btl_openib_have_rdmacm], [test $1 -eq 1 -a "x$btl_openib_have_rdmacm" = "x1"])
2011-08-05 00:26:08 +04:00
AM_CONDITIONAL([MCA_btl_openib_have_dynamic_sl], [test $1 -eq 1 -a "x$btl_openib_have_opensm_devel" = "x1"])
2011-09-20 01:24:58 +04:00
AM_CONDITIONAL([MCA_btl_openib_have_udcm], [test $1 -eq 1 -a "x$btl_openib_have_udcm" = "x1"])
2008-01-15 02:22:03 +03:00
])
2005-07-27 17:42:37 +04:00
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
# MCA_btl_openib_CONFIG([action-if-can-copalle],
# [action-if-cant-copalle])
2005-07-27 17:42:37 +04:00
# ------------------------------------------------
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
AC_DEFUN([MCA_opal_btl_openib_CONFIG],[
AC_CONFIG_FILES([opal/mca/btl/openib/Makefile])
2010-09-18 03:04:06 +04:00
2013-12-07 03:40:34 +04:00
OPAL_VAR_SCOPE_PUSH([cpcs have_threads LDFLAGS_save LIBS_save])
2008-05-02 15:52:33 +04:00
cpcs="oob"
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
OPAL_CHECK_OPENFABRICS([btl_openib],
2013-12-14 00:27:33 +04:00
[btl_openib_happy="yes"
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
OPAL_CHECK_OPENFABRICS_CM([btl_openib])],
2005-07-27 17:42:37 +04:00
[btl_openib_happy="no"])
AS_IF([test "$btl_openib_happy" = "yes"],
2013-01-29 04:00:43 +04:00
[# With the new openib flags, look for ibv_fork_init
2007-04-21 04:15:05 +04:00
LDFLAGS_save="$LDFLAGS"
LIBS_save="$LIBS"
LDFLAGS="$LDFLAGS $btl_openib_LDFLAGS"
LIBS="$LIBS $btl_openib_LIBS"
AC_CHECK_FUNCS([ibv_fork_init])
LDFLAGS="$LDFLAGS_save"
LIBS="$LIBS_save"
2005-07-27 17:42:37 +04:00
$1],
[$2])
2011-09-20 01:24:58 +04:00
AC_MSG_CHECKING([for thread support (needed for rdmacm/udcm)])
2008-10-06 04:46:02 +04:00
have_threads=`echo $THREAD_TYPE | awk '{ print [$]1 }'`
if test "x$have_threads" = "x"; then
have_threads=none
fi
AC_MSG_RESULT([$have_threads])
2005-07-27 17:42:37 +04:00
2008-05-02 15:52:33 +04:00
AS_IF([test "$btl_openib_happy" = "yes"],
2008-07-08 10:32:54 +04:00
[if test "x$btl_openib_have_xrc" = "x1"; then
2008-06-04 12:50:30 +04:00
cpcs="$cpcs xoob"
2008-05-02 15:52:33 +04:00
fi
2008-10-06 04:46:02 +04:00
if test "x$btl_openib_have_rdmacm" = "x1" -a \
"$have_threads" != "none"; then
2008-07-08 10:32:54 +04:00
cpcs="$cpcs rdmacm"
2014-02-27 15:29:47 +04:00
if test "$enable_openib_rdmacm_ibaddr" = "yes"; then
2014-03-24 17:36:04 +04:00
AC_MSG_CHECKING([IB addressing])
AC_EGREP_CPP(
yes,
[
#include <infiniband/ib.h>
#ifdef AF_IB
yes
#endif
],
[
AC_CHECK_HEADERS(
[rdma/rsocket.h],
[
AC_MSG_RESULT([yes])
AC_DEFINE(BTL_OPENIB_RDMACM_IB_ADDR, 1, rdmacm IB_AF addressing support)
],
[
AC_MSG_RESULT([no])
AC_DEFINE(BTL_OPENIB_RDMACM_IB_ADDR, 0, rdmacm without IB_AF addressing support)
AC_MSG_WARN([There is no IB_AF addressing support by lib rdmacm.])
]
)],
[
AC_MSG_RESULT([no])
AC_DEFINE(BTL_OPENIB_RDMACM_IB_ADDR, 0, rdmacm without IB_AF addressing support)
AC_MSG_WARN([There is no IB_AF addressing support by lib rdmacm.])
])
2014-03-04 12:54:17 +04:00
else
2014-03-24 17:36:04 +04:00
AC_DEFINE(BTL_OPENIB_RDMACM_IB_ADDR, 0, rdmacm without IB_AF addressing support)
2014-02-27 15:29:47 +04:00
fi
2008-05-02 15:52:33 +04:00
fi
2011-09-20 01:24:58 +04:00
if test "x$btl_openib_have_udcm" = "x1" -a \
"$have_threads" != "none"; then
cpcs="$cpcs udcm"
fi
2008-05-02 15:52:33 +04:00
AC_MSG_CHECKING([which openib btl cpcs will be built])
AC_MSG_RESULT([$cpcs])])
2011-01-19 23:58:22 +03:00
# Enable openib device failover. It is disabled by default.
2012-05-09 00:42:09 +04:00
AC_MSG_CHECKING([whether openib failover is enabled])
2011-01-19 23:58:22 +03:00
AC_ARG_ENABLE([btl-openib-failover],
[AC_HELP_STRING([--enable-btl-openib-failover],
[enable openib BTL failover (default: disabled)])])
if test "$enable_btl_openib_failover" = "yes"; then
AC_MSG_RESULT([yes])
btl_openib_failover_enabled=1
else
AC_MSG_RESULT([no])
btl_openib_failover_enabled=0
fi
AC_DEFINE_UNQUOTED([BTL_OPENIB_FAILOVER_ENABLED], [$btl_openib_failover_enabled],
[enable openib BTL failover])
AM_CONDITIONAL([MCA_btl_openib_enable_failover], [test "x$btl_openib_failover_enabled" = "x1"])
2014-11-20 09:22:43 +03:00
2012-05-09 00:42:09 +04:00
# Check for __malloc_hook availability
AC_ARG_ENABLE(btl-openib-malloc-alignment,
AC_HELP_STRING([--enable-btl-openib-malloc-alignment], [Enable support for allocated memory alignment. Default: enabled if supported, disabled otherwise.]))
btl_openib_malloc_hooks_enabled=0
AS_IF([test "$enable_btl_openib_malloc_alignment" != "no"],
[AC_CHECK_HEADER([malloc.h],
[AC_CHECK_FUNC([__malloc_hook],
[AC_CHECK_FUNC([__realloc_hook],
[AC_CHECK_FUNC([__free_hook],
[btl_openib_malloc_hooks_enabled=1])])])])])
AS_IF([test "$enable_btl_openib_malloc_alignment" = "yes" -a "$btl_openib_malloc_hooks_enabled" = "0"],
[AC_MSG_ERROR([openib malloc alignment is requested but __malloc_hook is not available])])
AC_MSG_CHECKING([whether the openib BTL will use malloc hooks])
AS_IF([test "$btl_openib_malloc_hooks_enabled" = "0"],
[AC_MSG_RESULT([no])],
[AC_MSG_RESULT([yes])])
AC_DEFINE_UNQUOTED(BTL_OPENIB_MALLOC_HOOKS_ENABLED, [$btl_openib_malloc_hooks_enabled],
[Whether the openib BTL malloc hooks are enabled])
2013-07-18 02:12:29 +04:00
# make sure that CUDA-aware checks have been done
AC_REQUIRE([OPAL_CHECK_CUDA])
2012-05-09 00:42:09 +04:00
2005-07-27 17:42:37 +04:00
# substitute in the things needed to build openib
AC_SUBST([btl_openib_CFLAGS])
AC_SUBST([btl_openib_CPPFLAGS])
AC_SUBST([btl_openib_LDFLAGS])
AC_SUBST([btl_openib_LIBS])
2008-05-02 15:52:33 +04:00
2010-11-13 02:22:11 +03:00
OPAL_VAR_SCOPE_POP
2007-04-21 04:15:05 +04:00
])dnl
