ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

Backout the ORCA commit. :(

Просмотр исходного кода 
There is a linking issue on Mac OSX that needs to be addressed before this is able to come back into the trunk.

This commit was SVN r26676.
Этот коммит содержится в:
Josh Hursey 2012-06-27 01:28:28 +00:00
родитель 338cd252d3
Коммит 28681deffa
209 изменённых файлов: 1618 добавлений и 6497 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

2
VERSION
Просмотреть файл

@ -1,7 +1,6 @@
# Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved. # Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved.
# Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved. # Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved.
# Copyright (c) 2011 NVIDIA Corporation. All rights reserved. # Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
# This is the VERSION file for Open MPI, describing the precise # This is the VERSION file for Open MPI, describing the precise
# version of Open MPI in this distribution. The various components of # version of Open MPI in this distribution. The various components of
@ -95,7 +94,6 @@ libmpi_mpifh_so_version=0:0:0
libmpi_usempi_tkr_so_version=0:0:0 libmpi_usempi_tkr_so_version=0:0:0
libmpi_usempi_ignore_tkr_so_version=0:0:0 libmpi_usempi_ignore_tkr_so_version=0:0:0
libopen_rte_so_version=0:0:0 libopen_rte_so_version=0:0:0
libopen_rca_so_version=0:0:0
libopen_pal_so_version=0:0:0 libopen_pal_so_version=0:0:0
libmpi_java_so_version=0:0:0 libmpi_java_so_version=0:0:0

29
autogen.pl
Просмотреть файл

@ -2,7 +2,7 @@
# #
# Copyright (c) 2009-2012 Cisco Systems, Inc. All rights reserved. # Copyright (c) 2009-2012 Cisco Systems, Inc. All rights reserved.
# Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved. # Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. #
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -41,7 +41,6 @@ my @subdirs;
# Command line parameters # Command line parameters
my $no_ompi_arg = 0; my $no_ompi_arg = 0;
my $no_orte_arg = 0; my $no_orte_arg = 0;
my $no_orca_arg = 0;
my $quiet_arg = 0; my $quiet_arg = 0;
my $debug_arg = 0; my $debug_arg = 0;
my $help_arg = 0; my $help_arg = 0;
@ -926,7 +925,6 @@ sub patch_autotools_output {
my $ok = Getopt::Long::GetOptions("no-ompi" => \$no_ompi_arg, my $ok = Getopt::Long::GetOptions("no-ompi" => \$no_ompi_arg,
"no-orte" => \$no_orte_arg, "no-orte" => \$no_orte_arg,
"no-orca" => \$no_orca_arg,
"quiet|q" => \$quiet_arg, "quiet|q" => \$quiet_arg,
"debug|d" => \$debug_arg, "debug|d" => \$debug_arg,
"help|h" => \$help_arg, "help|h" => \$help_arg,
@ -941,7 +939,6 @@ if (!$ok || $help_arg) {
print "Options: print "Options:
--no-ompi | -no-ompi Do not build the Open MPI layer --no-ompi | -no-ompi Do not build the Open MPI layer
--no-orte | -no-orte Do not build the ORTE layer --no-orte | -no-orte Do not build the ORTE layer
--no-orca | -no-orca Do not build the ORCA and Open MPI layers
--quiet | -q Do not display normal verbose output --quiet | -q Do not display normal verbose output
--debug | -d Output lots of debug information --debug | -d Output lots of debug information
--help | -h This help list --help | -h This help list
@ -970,26 +967,12 @@ if (! -e "orte") {
$no_orte_arg = 1; $no_orte_arg = 1;
debug "No orte subdirectory found - will not build ORTE\n"; debug "No orte subdirectory found - will not build ORTE\n";
} }
if (! -e "orca") {
$no_orca_arg = 1;
$no_ompi_arg = 1;
debug "No orca subdirectory found - will not build Pinapple and MPI layers\n";
}
# --no-orca implies --no-ompi
if ($no_orca_arg) {
$no_ompi_arg = 1;
}
if ($no_ompi_arg) { if ($no_ompi_arg) {
$project_name_long = "Open MPI Runtime Collaborative Abstraction";
$project_name_short = "open-rca";
}
if ($no_ompi_arg && $no_orca_arg) {
$project_name_long = "Open MPI Run Time Environment"; $project_name_long = "Open MPI Run Time Environment";
$project_name_short = "open-rte"; $project_name_short = "open-rte";
} }
if ($no_ompi_arg && $no_orca_arg && $no_orte_arg ) { if ($no_orte_arg) {
$project_name_long = "Open Portability Access Layer"; $project_name_long = "Open Portability Access Layer";
$project_name_short = "open-pal"; $project_name_short = "open-pal";
} }
@ -1143,11 +1126,9 @@ if (! (-f "VERSION" && -f "configure.ac" && -f $topdir_file)) {
my $projects; my $projects;
push(@{$projects}, { name => "opal", dir => "opal", need_base => 1 }); push(@{$projects}, { name => "opal", dir => "opal", need_base => 1 });
push(@{$projects}, { name => "orte", dir => "orte", need_base => 1 }) push(@{$projects}, { name => "orte", dir => "orte", need_base => 1 })
if (!$no_orte_arg); if (!$no_ompi_arg || !$no_orte_arg);
push(@{$projects}, { name => "orca", dir => "orca", need_base => 1 })
if (!$no_orca_arg);
push(@{$projects}, { name => "ompi", dir => "ompi", need_base => 1 }) push(@{$projects}, { name => "ompi", dir => "ompi", need_base => 1 })
if (!$no_ompi_arg && !$no_orca_arg); if (!$no_ompi_arg);
# Save the list of projects in the m4 file # Save the list of projects in the m4 file
my $str; my $str;

6
config/distscript.csh
Просмотреть файл

@ -11,7 +11,6 @@
# Copyright (c) 2004-2005 The Regents of the University of California. # Copyright (c) 2004-2005 The Regents of the University of California.
# All rights reserved. # All rights reserved.
# Copyright (c) 2009 Cisco Systems, Inc. All rights reserved. # Copyright (c) 2009 Cisco Systems, Inc. All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -92,9 +91,6 @@ set dirs=opal
if (-d orte) then if (-d orte) then
set dirs="$dirs orte" set dirs="$dirs orte"
endif endif
if (-d orca) then
set dirs="$dirs orca"
endif
if (-d ompi) then if (-d ompi) then
set dirs="$dirs ompi" set dirs="$dirs ompi"
endif endif
@ -173,7 +169,7 @@ cd ..
echo "*** Now in: `pwd`" echo "*** Now in: `pwd`"
echo "*** Replacing config.sub/config.guess with latest from ftp.gnu.org..." echo "*** Replacing config.sub/config.guess with latest from ftp.gnu.org..."
foreach file (config.guess config.sub) foreach file (config.guess config.sub)
foreach dir (opal orte orca ompi) foreach dir (opal orte ompi)
if (-d $dir) then if (-d $dir) then
find $dir -name $file \ find $dir -name $file \
-exec chmod +w {} \; \ -exec chmod +w {} \; \

24
configure.ac
Просмотреть файл

@ -14,7 +14,7 @@
# Copyright (c) 2006-2008 Sun Microsystems, Inc. All rights reserved. # Copyright (c) 2006-2008 Sun Microsystems, Inc. All rights reserved.
# Copyright (c) 2006-2011 Los Alamos National Security, LLC. All rights # Copyright (c) 2006-2011 Los Alamos National Security, LLC. All rights
# reserved. # reserved.
# Copyright (c) 2009-2012 Oak Ridge National Labs. All rights reserved. # Copyright (c) 2009 Oak Ridge National Labs. All rights reserved.
# Copyright (c) 2011 NVIDIA Corporation. All rights reserved. # Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
# $COPYRIGHT$ # $COPYRIGHT$
# #
@ -98,7 +98,6 @@ AC_SUBST([CONFIGURE_DEPENDENCIES], ['$(top_srcdir)/VERSION'])
# Set up project specific AM_CONDITIONALs # Set up project specific AM_CONDITIONALs
AM_CONDITIONAL([PROJECT_OMPI], m4_ifdef([project_ompi], [true], [false])) AM_CONDITIONAL([PROJECT_OMPI], m4_ifdef([project_ompi], [true], [false]))
AM_CONDITIONAL([PROJECT_ORTE], m4_ifdef([project_orte], [true], [false])) AM_CONDITIONAL([PROJECT_ORTE], m4_ifdef([project_orte], [true], [false]))
AM_CONDITIONAL([PROJECT_ORCA], m4_ifdef([project_orca], [true], [false]))
ompi_show_subtitle "Checking versions" ompi_show_subtitle "Checking versions"
@ -113,11 +112,6 @@ m4_ifdef([project_orte],
[$srcdir/VERSION], [$srcdir/VERSION],
[orte/include/orte/version.h])]) [orte/include/orte/version.h])])
m4_ifdef([project_orca],
[OPAL_SAVE_VERSION([ORCA], [Open MPI Runtime Collaborative Abstraction],
[$srcdir/VERSION],
[orca/include/orca/version.h])])
OPAL_SAVE_VERSION([OPAL], [Open Portable Access Layer], [$srcdir/VERSION], OPAL_SAVE_VERSION([OPAL], [Open Portable Access Layer], [$srcdir/VERSION],
[opal/include/opal/version.h]) [opal/include/opal/version.h])
@ -142,8 +136,6 @@ m4_ifdef([project_ompi],
AC_SUBST(libmca_common_portals_so_version)]) AC_SUBST(libmca_common_portals_so_version)])
m4_ifdef([project_orte], m4_ifdef([project_orte],
[AC_SUBST(libopen_rte_so_version)]) [AC_SUBST(libopen_rte_so_version)])
m4_ifdef([project_orca],
[AC_SUBST(libopen_rca_so_version)])
AC_SUBST(libmca_opal_common_hwloc_so_version) AC_SUBST(libmca_opal_common_hwloc_so_version)
AC_SUBST(libopen_pal_so_version) AC_SUBST(libopen_pal_so_version)
@ -171,8 +163,6 @@ AC_DEFINE_UNQUOTED([OMPI_ENABLE_PROGRESS_THREADS], [$OMPI_ENABLE_PROGRESS_THREAD
# List header files to generate # List header files to generate
AM_CONFIG_HEADER([opal/include/opal_config.h]) AM_CONFIG_HEADER([opal/include/opal_config.h])
m4_ifdef([project_orca],
[AM_CONFIG_HEADER([orca/include/orca_config.h])])
m4_ifdef([project_orte], m4_ifdef([project_orte],
[AM_CONFIG_HEADER([orte/include/orte_config.h])]) [AM_CONFIG_HEADER([orte/include/orte_config.h])])
m4_ifdef([project_ompi], m4_ifdef([project_ompi],
@ -255,7 +245,6 @@ AC_SUBST(top_ompi_builddir)
OPAL_CONFIGURE_OPTIONS OPAL_CONFIGURE_OPTIONS
m4_ifdef([project_orte], [ORTE_CONFIGURE_OPTIONS]) m4_ifdef([project_orte], [ORTE_CONFIGURE_OPTIONS])
m4_ifdef([project_orca], [ORCA_CONFIGURE_OPTIONS])
m4_ifdef([project_ompi], [OMPI_CONFIGURE_OPTIONS]) m4_ifdef([project_ompi], [OMPI_CONFIGURE_OPTIONS])
if test "$enable_binaries" = "no" -a "$enable_dist" = "yes"; then if test "$enable_binaries" = "no" -a "$enable_dist" = "yes"; then
@ -577,7 +566,7 @@ m4_ifdef([project_orte], [ORTE_SETUP_JAVA])
# Java MPI Binding request # Java MPI Binding request
################################## ##################################
# Only needed for OMPI # Only needed for OMPI
m4_ifdef([project_ompi], [m4_ifdef([project_orte], [OMPI_SETUP_JAVA_BINDINGS], [OMPI_SETUP_NO_JAVA_BINDINGS])]) m4_ifdef([project_ompi], [OMPI_SETUP_JAVA_BINDINGS])
################################## ##################################
# Hadoop support # Hadoop support
@ -1134,9 +1123,9 @@ if test "$OMPI_TOP_BUILDDIR" != "$OMPI_TOP_SRCDIR"; then
# rather than have successive assignments to these shell # rather than have successive assignments to these shell
# variables, lest the $(foo) names try to get evaluated here. # variables, lest the $(foo) names try to get evaluated here.
# Yuck! # Yuck!
CPPFLAGS='-I$(top_srcdir) -I$(top_builddir) -I$(top_srcdir)/opal/include m4_ifdef([project_orca], [-I$(top_srcdir)/orca/include]) m4_ifdef([project_orte], [-I$(top_srcdir)/orte/include]) m4_ifdef([project_ompi], [-I$(top_srcdir)/ompi/include])'" $CPPFLAGS" CPPFLAGS='-I$(top_srcdir) -I$(top_builddir) -I$(top_srcdir)/opal/include m4_ifdef([project_orte], [-I$(top_srcdir)/orte/include]) m4_ifdef([project_ompi], [-I$(top_srcdir)/ompi/include])'" $CPPFLAGS"
# C++ is only relevant if we're building OMPI # C++ is only relevant if we're building OMPI
m4_ifdef([project_ompi], [CXXCPPFLAGS='-I$(top_srcdir) -I$(top_builddir) -I$(top_srcdir)/opal/include -I$(top_srcdir)/orca/include -I$(top_srcdir)/orte/include -I$(top_srcdir)/ompi/include'" $CXXCPPFLAGS"]) m4_ifdef([project_ompi], [CXXCPPFLAGS='-I$(top_srcdir) -I$(top_builddir) -I$(top_srcdir)/opal/include -I$(top_srcdir)/orte/include -I$(top_srcdir)/ompi/include'" $CXXCPPFLAGS"])
else else
CPPFLAGS='-I$(top_srcdir)'" $CPPFLAGS" CPPFLAGS='-I$(top_srcdir)'" $CPPFLAGS"
# C++ is only relevant if we're building OMPI # C++ is only relevant if we're building OMPI
@ -1147,7 +1136,7 @@ fi
# versions without optimization for debugger modules). # versions without optimization for debugger modules).
m4_ifdef([project_orte], [ORTE_SETUP_DEBUGGER_FLAGS], m4_ifdef([project_orte], [ORTE_SETUP_DEBUGGER_FLAGS],
[m4_ifdef([project_ompi], [m4_ifdef([project_orte], [ORTE_SETUP_DEBUGGER_FLAGS])] )] ) [m4_ifdef([project_ompi], [ORTE_SETUP_DEBUGGER_FLAGS])])
# #
# Delayed the substitution of CFLAGS and CXXFLAGS until now because # Delayed the substitution of CFLAGS and CXXFLAGS until now because
@ -1180,7 +1169,6 @@ AC_PATH_PROG(PERL, perl, perl)
OPAL_SETUP_WRAPPER_FINAL OPAL_SETUP_WRAPPER_FINAL
m4_ifdef([project_orte], [ORTE_SETUP_WRAPPER_FINAL]) m4_ifdef([project_orte], [ORTE_SETUP_WRAPPER_FINAL])
m4_ifdef([project_orca], [ORCA_SETUP_WRAPPER_FINAL])
m4_ifdef([project_ompi], [OMPI_SETUP_WRAPPER_FINAL]) m4_ifdef([project_ompi], [OMPI_SETUP_WRAPPER_FINAL])
# Recreate some defines prefixed with OMPI_ so that there are no bare # Recreate some defines prefixed with OMPI_ so that there are no bare
@ -1227,7 +1215,6 @@ fi
m4_ifdef([opal_CONFIG_LOCAL], [opal_CONFIG_LOCAL]) m4_ifdef([opal_CONFIG_LOCAL], [opal_CONFIG_LOCAL])
m4_ifdef([orte_CONFIG_LOCAL], [orte_CONFIG_LOCAL]) m4_ifdef([orte_CONFIG_LOCAL], [orte_CONFIG_LOCAL])
m4_ifdef([orca_CONFIG_LOCAL], [orca_CONFIG_LOCAL])
m4_ifdef([ompi_CONFIG_LOCAL], [ompi_CONFIG_LOCAL]) m4_ifdef([ompi_CONFIG_LOCAL], [ompi_CONFIG_LOCAL])
############################################################################ ############################################################################
@ -1255,7 +1242,6 @@ AC_CONFIG_FILES([
OPAL_CONFIG_FILES OPAL_CONFIG_FILES
m4_ifdef([project_orte], [ORTE_CONFIG_FILES]) m4_ifdef([project_orte], [ORTE_CONFIG_FILES])
m4_ifdef([project_orca], [ORCA_CONFIG_FILES])
m4_ifdef([project_ompi], [OMPI_CONFIG_FILES]) m4_ifdef([project_ompi], [OMPI_CONFIG_FILES])
AC_OUTPUT AC_OUTPUT

3
ompi/Makefile.am
Просмотреть файл

@ -12,7 +12,6 @@
# Copyright (c) 2008-2012 Cisco Systems, Inc. All rights reserved. # Copyright (c) 2008-2012 Cisco Systems, Inc. All rights reserved.
# Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved. # Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved.
# Copyright (c) 2010-2011 Sandia National Laboratories. All rights reserved. # Copyright (c) 2010-2011 Sandia National Laboratories. All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -139,7 +138,7 @@ libmpi_la_LIBADD = \
$(MCA_ompi_FRAMEWORK_LIBS) \ $(MCA_ompi_FRAMEWORK_LIBS) \
$(OMPI_MPIEXT_C_LIBS) \ $(OMPI_MPIEXT_C_LIBS) \
$(OMPI_LIBMPI_EXTRA_LIBS) \ $(OMPI_LIBMPI_EXTRA_LIBS) \
$(top_ompi_builddir)/orca/libopen-rca-lt.la $(top_ompi_builddir)/orte/libopen-rte-lt.la
libmpi_la_DEPENDENCIES = $(libmpi_la_LIBADD) libmpi_la_DEPENDENCIES = $(libmpi_la_LIBADD)
libmpi_la_LDFLAGS = \ libmpi_la_LDFLAGS = \
-version-info $(libmpi_so_version) \ -version-info $(libmpi_so_version) \

6
ompi/attribute/attribute_predefined.c
Просмотреть файл

@ -11,7 +11,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2006 University of Houston. All rights reserved. * Copyright (c) 2006 University of Houston. All rights reserved.
* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -93,8 +92,7 @@
#include "ompi/errhandler/errcode.h" #include "ompi/errhandler/errcode.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
#include "ompi/mca/pml/pml.h" #include "ompi/mca/pml/pml.h"
#include "orte/util/proc_info.h"
#include "orca/include/rte_orca.h"
/* /*
* Private functions * Private functions
@ -176,7 +174,7 @@ int ompi_attr_create_predefined(void)
return ret; return ret;
} }
ret = set_f(MPI_APPNUM, orca_process_info_get_app_num() ); ret = set_f(MPI_APPNUM, orte_process_info.app_num);
return ret; return ret;
} }

13
ompi/communicator/comm.c
Просмотреть файл

@ -28,8 +28,8 @@
#include "opal/mca/hwloc/base/base.h" #include "opal/mca/hwloc/base/base.h"
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "orte/util/name_fns.h"
#include "orca/include/rte_orca.h" #include "orte/mca/rml/rml_types.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
#include "opal/threads/mutex.h" #include "opal/threads/mutex.h"
@ -1252,7 +1252,7 @@ ompi_proc_t **ompi_comm_get_rprocs ( ompi_communicator_t *local_comm,
int rc; int rc;
int local_rank, local_size; int local_rank, local_size;
ompi_proc_t **rprocs=NULL; ompi_proc_t **rprocs=NULL;
orca_std_cntr_t size_len; orte_std_cntr_t size_len;
int int_len, rlen; int int_len, rlen;
opal_buffer_t *sbuf=NULL, *rbuf=NULL; opal_buffer_t *sbuf=NULL, *rbuf=NULL;
void *sendbuf; void *sendbuf;
@ -1356,7 +1356,7 @@ ompi_proc_t **ompi_comm_get_rprocs ( ompi_communicator_t *local_comm,
goto err_exit; goto err_exit;
} }
if (OMPI_SUCCESS != (rc = opal_dss.load(rbuf, recvbuf, rlen))) { if (ORTE_SUCCESS != (rc = opal_dss.load(rbuf, recvbuf, rlen))) {
goto err_exit; goto err_exit;
} }
@ -1426,6 +1426,7 @@ int ompi_comm_determine_first ( ompi_communicator_t *intercomm, int high )
int scount=0; int scount=0;
int rc; int rc;
ompi_proc_t *ourproc, *theirproc; ompi_proc_t *ourproc, *theirproc;
orte_ns_cmp_bitmask_t mask;
rank = ompi_comm_rank (intercomm); rank = ompi_comm_rank (intercomm);
rsize= ompi_comm_remote_size (intercomm); rsize= ompi_comm_remote_size (intercomm);
@ -1467,8 +1468,8 @@ int ompi_comm_determine_first ( ompi_communicator_t *intercomm, int high )
ourproc = ompi_group_peer_lookup(intercomm->c_local_group,0); ourproc = ompi_group_peer_lookup(intercomm->c_local_group,0);
theirproc = ompi_group_peer_lookup(intercomm->c_remote_group,0); theirproc = ompi_group_peer_lookup(intercomm->c_remote_group,0);
rc = orca_process_name_compare((ORCA_NAME_CMP_JOBID | ORCA_NAME_CMP_VPID), mask = ORTE_NS_CMP_JOBID | ORTE_NS_CMP_VPID;
&(ourproc->proc_name), &(theirproc->proc_name)); rc = orte_util_compare_name_fields(mask, &(ourproc->proc_name), &(theirproc->proc_name));
if ( 0 > rc ) { if ( 0 > rc ) {
flag = true; flag = true;
} }

19
ompi/communicator/comm_cid.c
Просмотреть файл

@ -27,6 +27,7 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "orte/types.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
#include "ompi/op/op.h" #include "ompi/op/op.h"
@ -40,7 +41,7 @@
#include "ompi/runtime/mpiruntime.h" #include "ompi/runtime/mpiruntime.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
#include "orca/include/rte_orca.h" #include "orte/mca/rml/rml.h"
BEGIN_C_DECLS BEGIN_C_DECLS
@ -779,11 +780,11 @@ static int ompi_comm_allreduce_intra_oob (int *inbuf, int *outbuf,
int i; int i;
int rc; int rc;
int local_leader, local_rank; int local_leader, local_rank;
orca_process_name_t *remote_leader=NULL; orte_process_name_t *remote_leader=NULL;
orca_std_cntr_t size_count; orte_std_cntr_t size_count;
local_leader = (*((int*)lleader)); local_leader = (*((int*)lleader));
remote_leader = (orca_process_name_t*)rleader; remote_leader = (orte_process_name_t*)rleader;
size_count = count; size_count = count;
if ( &ompi_mpi_op_sum.op != op && &ompi_mpi_op_prod.op != op && if ( &ompi_mpi_op_sum.op != op && &ompi_mpi_op_prod.op != op &&
@ -813,23 +814,23 @@ static int ompi_comm_allreduce_intra_oob (int *inbuf, int *outbuf,
sbuf = OBJ_NEW(opal_buffer_t); sbuf = OBJ_NEW(opal_buffer_t);
rbuf = OBJ_NEW(opal_buffer_t); rbuf = OBJ_NEW(opal_buffer_t);
if (OPAL_SUCCESS != (rc = opal_dss.pack(sbuf, tmpbuf, (orca_std_cntr_t)count, OPAL_INT))) { if (OPAL_SUCCESS != (rc = opal_dss.pack(sbuf, tmpbuf, (orte_std_cntr_t)count, OPAL_INT))) {
goto exit; goto exit;
} }
if ( send_first ) { if ( send_first ) {
if (0 > (rc = orca_oob_send_buffer(remote_leader, sbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) { if (0 > (rc = orte_rml.send_buffer(remote_leader, sbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) {
goto exit; goto exit;
} }
if (0 > (rc = orca_oob_recv_buffer(remote_leader, rbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) { if (0 > (rc = orte_rml.recv_buffer(remote_leader, rbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) {
goto exit; goto exit;
} }
} }
else { else {
if (0 > (rc = orca_oob_recv_buffer(remote_leader, rbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) { if (0 > (rc = orte_rml.recv_buffer(remote_leader, rbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) {
goto exit; goto exit;
} }
if (0 > (rc = orca_oob_send_buffer(remote_leader, sbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) { if (0 > (rc = orte_rml.send_buffer(remote_leader, sbuf, OMPI_RML_TAG_COMM_CID_INTRA, 0))) {
goto exit; goto exit;
} }
} }

27
ompi/config/ompi_configure_options.m4
Просмотреть файл

@ -15,7 +15,7 @@ dnl Copyright (c) 2007 Sun Microsystems, Inc. All rights reserved.
dnl Copyright (c) 2009 IBM Corporation. All rights reserved. dnl Copyright (c) 2009 IBM Corporation. All rights reserved.
dnl Copyright (c) 2009 Los Alamos National Security, LLC. All rights dnl Copyright (c) 2009 Los Alamos National Security, LLC. All rights
dnl reserved. dnl reserved.
dnl Copyright (c) 2009-2012 Oak Ridge National Labs. All rights reserved. dnl Copyright (c) 2009 Oak Ridge National Labs. All rights reserved.
dnl dnl
dnl $COPYRIGHT$ dnl $COPYRIGHT$
dnl dnl
@ -252,30 +252,5 @@ fi
AC_MSG_RESULT([$OMPI_FORTRAN_MAX_ARRAY_RANK]) AC_MSG_RESULT([$OMPI_FORTRAN_MAX_ARRAY_RANK])
AC_SUBST(OMPI_FORTRAN_MAX_ARRAY_RANK) AC_SUBST(OMPI_FORTRAN_MAX_ARRAY_RANK)
#
# A global check for ORTE to amke it easier to support the tools
# See note in orca_configure_options.m4
#
AC_MSG_CHECKING([if want ORTE supported OMPI tools])
if test "$ORCA_WITH_ORTE_SUPPORT" = "0"; then
list_of_frameworks="pubsub-pmi,dpm-orte,pubsub-orte"
if test -z $enable_mca_no_build ; then
enable_mca_no_build="$list_of_frameworks"
else
enable_mca_no_build="$enable_mca_no_build,$list_of_frameworks"
fi
OMPI_WITH_ORTE_SUPPORTED_TOOLS=0
AC_MSG_RESULT([no])
else
OMPI_WITH_ORTE_SUPPORTED_TOOLS=1
AC_MSG_RESULT([yes])
fi
AC_DEFINE_UNQUOTED([OMPI_WITH_ORTE_SUPPORTED_TOOLS],
[$OMPI_WITH_ORTE_SUPPORTED_TOOLS],
[Whether we want ORTE supported OMPI tools])
AM_CONDITIONAL(OMPI_WITH_ORTE_SUPPORTED_TOOLS, test "$OMPI_WITH_ORTE_SUPPORTED_TOOLS" = "1")
])dnl ])dnl

19
ompi/config/ompi_setup_java.m4
Просмотреть файл

@ -14,7 +14,6 @@ dnl Copyright (c) 2006-2012 Los Alamos National Security, LLC. All rights
dnl reserved. dnl reserved.
dnl Copyright (c) 2007-2012 Oracle and/or its affiliates. All rights reserved. dnl Copyright (c) 2007-2012 Oracle and/or its affiliates. All rights reserved.
dnl Copyright (c) 2008-2012 Cisco Systems, Inc. All rights reserved. dnl Copyright (c) 2008-2012 Cisco Systems, Inc. All rights reserved.
dnl Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
dnl $COPYRIGHT$ dnl $COPYRIGHT$
dnl dnl
dnl Additional copyrights may follow dnl Additional copyrights may follow
@ -27,29 +26,11 @@ AC_DEFUN([OMPI_SETUP_JAVA_BINDINGS_BANNER],[
ompi_show_subtitle "Java MPI bindings" ompi_show_subtitle "Java MPI bindings"
]) ])
#
# OMPI_SETUP_NO_JAVA_BINDINGS()
# -----------------------------
# Force configure to not build bindings.
# For the case: ./autogen.pl --no-orte
#
AC_DEFUN([OMPI_SETUP_NO_JAVA_BINDINGS],[
AC_MSG_CHECKING([if want Java bindings])
AC_MSG_RESULT([no (Needs ORTE support)])
WANT_MPI_JAVA_SUPPORT=0
AC_DEFINE_UNQUOTED([OMPI_WANT_JAVA_BINDINGS], [$WANT_MPI_JAVA_SUPPORT],
[do we want java mpi bindings])
AM_CONDITIONAL(OMPI_WANT_JAVA_BINDINGS, test "$WANT_MPI_JAVA_SUPPORT" = "1")
])
# OMPI_SETUP_JAVA_BINDINGS() # OMPI_SETUP_JAVA_BINDINGS()
# ---------------- # ----------------
# Do everything required to setup the Java MPI bindings. Safe to AC_REQUIRE # Do everything required to setup the Java MPI bindings. Safe to AC_REQUIRE
# this macro. # this macro.
AC_DEFUN([OMPI_SETUP_JAVA_BINDINGS],[ AC_DEFUN([OMPI_SETUP_JAVA_BINDINGS],[
# must have Java setup # must have Java setup
AC_REQUIRE([ORTE_SETUP_JAVA]) AC_REQUIRE([ORTE_SETUP_JAVA])

25
ompi/debuggers/ompi_debuggers.c
Просмотреть файл

@ -11,7 +11,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2007-2011 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -73,7 +72,9 @@
#include "ompi/datatype/ompi_datatype.h" #include "ompi/datatype/ompi_datatype.h"
#include "ompi/include/mpi.h" #include "ompi/include/mpi.h"
#include "orca/include/rte_orca.h" #include "orte/mca/errmgr/errmgr.h"
#include "orte/mca/rml/rml.h"
#include "orte/runtime/orte_globals.h"
#if defined(OMPI_MSGQ_DLL) #if defined(OMPI_MSGQ_DLL)
/* This variable is old/deprecated -- the mpimsgq_dll_locations[] /* This variable is old/deprecated -- the mpimsgq_dll_locations[]
@ -169,17 +170,17 @@ void ompi_wait_for_debugger(void)
{ {
int i, debugger; int i, debugger;
char *a, *b, **dirs, **tmp1 = NULL, **tmp2 = NULL; char *a, *b, **dirs, **tmp1 = NULL, **tmp2 = NULL;
#if ORCA_WITH_FULL_ORTE_SUPPORT #if !ORTE_DISABLE_FULL_SUPPORT
opal_buffer_t buf; opal_buffer_t buf;
int rc; int rc;
#endif #endif
/* See lengthy comment in orte/tools/orterun/debuggers.c about /* See lengthy comment in orte/tools/orterun/debuggers.c about
orte_in_parallel_debugger */ orte_in_parallel_debugger */
#if !ORCA_WITH_FULL_ORTE_SUPPORT #if ORTE_DISABLE_FULL_SUPPORT
debugger = 0; debugger = 0;
#else #else
debugger = orca_info_in_parallel_debugger(); debugger = orte_in_parallel_debugger;
#endif #endif
/* Add in environment variables for other launchers, such as yod, /* Add in environment variables for other launchers, such as yod,
@ -224,8 +225,8 @@ void ompi_wait_for_debugger(void)
mpimsgq_dll_locations = tmp1; mpimsgq_dll_locations = tmp1;
mpidbg_dll_locations = tmp2; mpidbg_dll_locations = tmp2;
#if ORCA_WITH_FULL_ORTE_SUPPORT #if !ORTE_DISABLE_FULL_SUPPORT
if (orca_info_standalone_operation()) { if (orte_standalone_operation) {
#endif #endif
/* spin until debugger attaches and releases us */ /* spin until debugger attaches and releases us */
while (MPIR_debug_gate == 0) { while (MPIR_debug_gate == 0) {
@ -237,7 +238,7 @@ void ompi_wait_for_debugger(void)
sleep(1); /* seconds */ sleep(1); /* seconds */
#endif #endif
} }
#if ORCA_WITH_FULL_ORTE_SUPPORT #if !ORTE_DISABLE_FULL_SUPPORT
} else { } else {
/* only the rank=0 proc waits for either a message from the /* only the rank=0 proc waits for either a message from the
@ -245,21 +246,21 @@ void ompi_wait_for_debugger(void)
* spin in * the grpcomm barrier in ompi_mpi_init until rank=0 * spin in * the grpcomm barrier in ompi_mpi_init until rank=0
* joins them. * joins them.
*/ */
if (0 != orca_process_info_get_vpid(ORCA_PROC_MY_NAME) ) { if (0 != ORTE_PROC_MY_NAME->vpid) {
return; return;
} }
/* VPID 0 waits for a message from the HNP */ /* VPID 0 waits for a message from the HNP */
OBJ_CONSTRUCT(&buf, opal_buffer_t); OBJ_CONSTRUCT(&buf, opal_buffer_t);
rc = orca_oob_recv_buffer(ORCA_NAME_WILDCARD, &buf, rc = orte_rml.recv_buffer(ORTE_NAME_WILDCARD, &buf,
ORCA_OOB_TAG_DEBUGGER_RELEASE, 0); ORTE_RML_TAG_DEBUGGER_RELEASE, 0);
OBJ_DESTRUCT(&buf); /* don't care about contents of message */ OBJ_DESTRUCT(&buf); /* don't care about contents of message */
if (rc < 0) { if (rc < 0) {
/* if it failed for some reason, then we are in trouble - /* if it failed for some reason, then we are in trouble -
* for now, just report the problem and give up waiting * for now, just report the problem and give up waiting
*/ */
opal_output(0, "Debugger_attach[rank=%ld]: could not wait for debugger!", opal_output(0, "Debugger_attach[rank=%ld]: could not wait for debugger!",
(long)orca_process_info_get_vpid(ORCA_PROC_MY_NAME) ); (long)ORTE_PROC_MY_NAME->vpid);
} }
} }
#endif #endif

3
ompi/errhandler/errhandler.h
Просмотреть файл

@ -12,7 +12,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2008-2012 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2008-2009 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2008-2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -36,7 +35,7 @@
#include "ompi/errhandler/errhandler_predefined.h" #include "ompi/errhandler/errhandler_predefined.h"
#include "ompi/errhandler/errcode-internal.h" #include "ompi/errhandler/errcode-internal.h"
#include "orca/include/rte_orca.h" #include "orte/types.h"
BEGIN_C_DECLS BEGIN_C_DECLS

28
ompi/errhandler/errhandler_predefined.c
Просмотреть файл

@ -12,7 +12,7 @@
* Copyright (c) 2006 University of Houston. All rights reserved. * Copyright (c) 2006 University of Houston. All rights reserved.
* Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2009 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Oak Ridge National Labs. All rights reserved. * Copyright (c) 2010-2011 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -30,8 +30,9 @@
#include <sys/param.h> #include <sys/param.h>
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "orte/util/name_fns.h"
#include "ompi/errhandler/errhandler_predefined.h" #include "ompi/errhandler/errhandler_predefined.h"
#include "ompi/errhandler/errcode.h" #include "ompi/errhandler/errcode.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
@ -162,7 +163,7 @@ static void out(char *str, char *arg)
} }
/* /*
* Use orca_show_help() to aggregate the error messages (i.e., show it * Use orte_show_help() to aggregate the error messages (i.e., show it
* once rather than N times). * once rather than N times).
* *
* Note that this function will only be invoked for errors during the * Note that this function will only be invoked for errors during the
@ -181,9 +182,8 @@ static void backend_fatal_aggregate(char *type,
arg = va_arg(arglist, char*); arg = va_arg(arglist, char*);
va_end(arglist); va_end(arglist);
asprintf(&prefix, "[%s:%d]", asprintf(&prefix, "[%s:%d]", orte_process_info.nodename,
orca_process_info_get_nodename(), (int) orte_process_info.pid);
(int) orca_process_info_get_pid() );
if (NULL != error_code) { if (NULL != error_code) {
err_msg = ompi_mpi_errnum_get_string(*error_code); err_msg = ompi_mpi_errnum_get_string(*error_code);
@ -195,22 +195,18 @@ static void backend_fatal_aggregate(char *type,
} }
if (NULL != name && ompi_mpi_initialized && !ompi_mpi_finalized) { if (NULL != name && ompi_mpi_initialized && !ompi_mpi_finalized) {
orca_show_help("help-mpi-errors.txt", orte_show_help("help-mpi-errors.txt",
"mpi_errors_are_fatal", false, "mpi_errors_are_fatal", false,
prefix, (NULL == arg) ? "" : "in", prefix, (NULL == arg) ? "" : "in",
(NULL == arg) ? "" : arg, (NULL == arg) ? "" : arg,
prefix, prefix, ORTE_PROC_MY_NAME->jobid, ORTE_PROC_MY_NAME->vpid,
orca_process_info_get_jobid(ORCA_PROC_MY_NAME),
orca_process_info_get_vpid(ORCA_PROC_MY_NAME),
prefix, type, name, prefix, err_msg, prefix, type, prefix); prefix, type, name, prefix, err_msg, prefix, type, prefix);
} else if (NULL == name) { } else if (NULL == name) {
orca_show_help("help-mpi-errors.txt", orte_show_help("help-mpi-errors.txt",
"mpi_errors_are_fatal unknown handle", false, "mpi_errors_are_fatal unknown handle", false,
prefix, (NULL == arg) ? "" : "in", prefix, (NULL == arg) ? "" : "in",
(NULL == arg) ? "" : arg, (NULL == arg) ? "" : arg,
prefix, prefix, ORTE_PROC_MY_NAME->jobid, ORTE_PROC_MY_NAME->vpid,
orca_process_info_get_jobid(ORCA_PROC_MY_NAME),
orca_process_info_get_vpid(ORCA_PROC_MY_NAME),
prefix, type, prefix, err_msg, prefix, type, prefix); prefix, type, prefix, err_msg, prefix, type, prefix);
} }
@ -329,7 +325,7 @@ static void backend_fatal(char *type, struct ompi_communicator_t *comm,
meaning that there is a better chance that the error message meaning that there is a better chance that the error message
will actually get printed). Note that we can only do will actually get printed). Note that we can only do
aggregation after MPI_INIT and before MPI_FINALIZE. */ aggregation after MPI_INIT and before MPI_FINALIZE. */
if (orca_show_help_want_aggregate() && orca_show_help_is_available()) { if (orte_help_want_aggregate && orte_show_help_is_available()) {
backend_fatal_aggregate(type, comm, name, error_code, arglist); backend_fatal_aggregate(type, comm, name, error_code, arglist);
} else { } else {
backend_fatal_no_aggregate(type, comm, name, error_code, arglist); backend_fatal_no_aggregate(type, comm, name, error_code, arglist);

5
ompi/include/ompi/constants.h
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -20,9 +19,9 @@
#ifndef OMPI_CONSTANTS_H #ifndef OMPI_CONSTANTS_H
#define OMPI_CONSTANTS_H #define OMPI_CONSTANTS_H
#include "orca/constants.h" #include "orte/constants.h"
#define OMPI_ERR_BASE ORCA_ERR_MAX #define OMPI_ERR_BASE ORTE_ERR_MAX
/* error codes */ /* error codes */
enum { enum {

13
ompi/mca/bml/r2/bml_r2.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2007-2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2007-2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -29,16 +28,14 @@
#include "opal/class/opal_bitmap.h" #include "opal/class/opal_bitmap.h"
#include "opal/util/argv.h" #include "opal/util/argv.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h"
#include "ompi/mca/bml/bml.h" #include "ompi/mca/bml/bml.h"
#include "ompi/mca/bml/base/base.h" #include "ompi/mca/bml/base/base.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
#include "ompi/mca/btl/base/base.h" #include "ompi/mca/btl/base/base.h"
#include "ompi/mca/bml/base/bml_base_btl.h" #include "ompi/mca/bml/base/bml_base_btl.h"
#include "bml_r2.h" #include "bml_r2.h"
#include "orte/util/name_fns.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
extern mca_bml_base_component_t mca_bml_r2_component; extern mca_bml_base_component_t mca_bml_r2_component;
@ -408,13 +405,13 @@ static int mca_bml_r2_add_procs( size_t nprocs,
if (mca_bml_r2.show_unreach_errors && if (mca_bml_r2.show_unreach_errors &&
OMPI_ERR_UNREACH == ret) { OMPI_ERR_UNREACH == ret) {
orca_show_help("help-mca-bml-r2.txt", orte_show_help("help-mca-bml-r2.txt",
"unreachable proc", "unreachable proc",
true, true,
ORCA_NAME_PRINT(&(ompi_proc_local_proc->proc_name)), ORTE_NAME_PRINT(&(ompi_proc_local_proc->proc_name)),
(ompi_proc_local_proc->proc_hostname ? (ompi_proc_local_proc->proc_hostname ?
ompi_proc_local_proc->proc_hostname : "unknown!"), ompi_proc_local_proc->proc_hostname : "unknown!"),
ORCA_NAME_PRINT(&(unreach_proc->proc_name)), ORTE_NAME_PRINT(&(unreach_proc->proc_name)),
(unreach_proc->proc_hostname ? (unreach_proc->proc_hostname ?
unreach_proc->proc_hostname : "unknown!"), unreach_proc->proc_hostname : "unknown!"),
btl_names); btl_names);

31
ompi/mca/bml/r2/bml_r2_ft.c
Просмотреть файл

@ -12,7 +12,6 @@
* Copyright (c) 2007-2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2007-2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -28,7 +27,8 @@
#include "opal/runtime/opal_progress.h" #include "opal/runtime/opal_progress.h"
#include "orca/include/rte_orca.h" #include "orte/mca/grpcomm/grpcomm.h"
#include "orte/util/proc_info.h"
#include "ompi/runtime/ompi_cr.h" #include "ompi/runtime/ompi_cr.h"
#include "ompi/mca/bml/base/base.h" #include "ompi/mca/bml/base/base.h"
@ -42,7 +42,7 @@
int mca_bml_r2_ft_event(int state) int mca_bml_r2_ft_event(int state)
{ {
#if ORCA_WITH_FULL_ORTE_SUPPORT #if !ORTE_DISABLE_FULL_SUPPORT
static bool first_continue_pass = false; static bool first_continue_pass = false;
ompi_proc_t** procs = NULL; ompi_proc_t** procs = NULL;
size_t num_procs; size_t num_procs;
@ -51,6 +51,7 @@ int mca_bml_r2_ft_event(int state)
int loc_state; int loc_state;
int param_type = -1; int param_type = -1;
char *param_list = NULL; char *param_list = NULL;
orte_grpcomm_collective_t coll;
if(OPAL_CRS_CHECKPOINT == state) { if(OPAL_CRS_CHECKPOINT == state) {
/* Do nothing for now */ /* Do nothing for now */
@ -59,7 +60,7 @@ int mca_bml_r2_ft_event(int state)
first_continue_pass = !first_continue_pass; first_continue_pass = !first_continue_pass;
/* Since nothing in Checkpoint, we are fine here (unless required by BTL) */ /* Since nothing in Checkpoint, we are fine here (unless required by BTL) */
if( orca_info_cr_continue_like_restart() && !first_continue_pass) { if( orte_cr_continue_like_restart && !first_continue_pass) {
procs = ompi_proc_all(&num_procs); procs = ompi_proc_all(&num_procs);
if(NULL == procs) { if(NULL == procs) {
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
@ -141,7 +142,7 @@ int mca_bml_r2_ft_event(int state)
} }
else if(OPAL_CRS_CONTINUE == state) { else if(OPAL_CRS_CONTINUE == state) {
/* Matches OPAL_CRS_RESTART_PRE */ /* Matches OPAL_CRS_RESTART_PRE */
if( orca_info_cr_continue_like_restart() && first_continue_pass) { if( orte_cr_continue_like_restart && first_continue_pass) {
if( OMPI_SUCCESS != (ret = mca_bml_r2_finalize()) ) { if( OMPI_SUCCESS != (ret = mca_bml_r2_finalize()) ) {
opal_output(0, "bml:r2: ft_event(Restart): Failed to finalize BML framework\n"); opal_output(0, "bml:r2: ft_event(Restart): Failed to finalize BML framework\n");
return ret; return ret;
@ -152,15 +153,20 @@ int mca_bml_r2_ft_event(int state)
} }
} }
/* Matches OPAL_CRS_RESTART */ /* Matches OPAL_CRS_RESTART */
else if( orca_info_cr_continue_like_restart() && !first_continue_pass ) { else if( orte_cr_continue_like_restart && !first_continue_pass ) {
/* /*
* Barrier to make all processes have been successfully restarted before * Barrier to make all processes have been successfully restarted before
* we try to remove some restart only files. * we try to remove some restart only files.
*/ */
if (ORCA_SUCCESS != (ret = orca_coll_barrier(ORCA_COLL_TYPE_BARRIER_INIT) ) ) { OBJ_CONSTRUCT(&coll, orte_grpcomm_collective_t);
opal_output(0, "bml:r2: ft_event(Restart): Failed in orca_coll_barrier (%d)", ret); coll.id = orte_process_info.peer_init_barrier;
if (OMPI_SUCCESS != (ret = orte_grpcomm.barrier(&coll))) {
opal_output(0, "bml:r2: ft_event(Restart): Failed in orte_grpcomm.barrier (%d)", ret);
return ret; return ret;
} }
while (coll.active) {
opal_progress();
}
/* /*
* Re-open the BTL framework to get the full list of components. * Re-open the BTL framework to get the full list of components.
@ -230,10 +236,15 @@ int mca_bml_r2_ft_event(int state)
* Barrier to make all processes have been successfully restarted before * Barrier to make all processes have been successfully restarted before
* we try to remove some restart only files. * we try to remove some restart only files.
*/ */
if (ORCA_SUCCESS != (ret = orca_coll_barrier(ORCA_COLL_TYPE_BARRIER_INIT) ) ) { OBJ_CONSTRUCT(&coll, orte_grpcomm_collective_t);
opal_output(0, "bml:r2: ft_event(Restart): Failed in orca_coll_barrier (%d)", ret); coll.id = orte_process_info.peer_init_barrier;
if (OMPI_SUCCESS != (ret = orte_grpcomm.barrier(&coll))) {
opal_output(0, "bml:r2: ft_event(Restart): Failed in orte_grpcomm.barrier (%d)", ret);
return ret; return ret;
} }
while (coll.active) {
opal_progress();
}
/* /*
* Re-open the BTL framework to get the full list of components. * Re-open the BTL framework to get the full list of components.

15
ompi/mca/btl/base/btl_base_error.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2006-2007 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2006-2007 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -26,7 +25,11 @@
#include "base.h" #include "base.h"
#include "btl_base_error.h" #include "btl_base_error.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/proc_info.h"
#include "orte/types.h"
#include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
int mca_btl_base_verbose = -1; int mca_btl_base_verbose = -1;
@ -60,11 +63,11 @@ void mca_btl_base_error_no_nics(const char* transport,
char *procid; char *procid;
if (mca_btl_base_warn_component_unused) { if (mca_btl_base_warn_component_unused) {
/* print out no-nic warning if user told us to */ /* print out no-nic warning if user told us to */
asprintf(&procid, "%s", ORCA_NAME_PRINT(ORCA_PROC_MY_NAME)); asprintf(&procid, "%s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
orca_show_help("help-mpi-btl-base.txt", "btl:no-nics", orte_show_help("help-mpi-btl-base.txt", "btl:no-nics",
true, procid, transport, orca_process_info_get_nodename(), true, procid, transport, orte_process_info.nodename,
nic_name); nic_name);
free(procid); free(procid);
} }
} }

21
ompi/mca/btl/base/btl_base_error.h
Просмотреть файл

@ -11,7 +11,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2007 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -27,7 +26,9 @@
#include <errno.h> #include <errno.h>
#include <stdio.h> #include <stdio.h>
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
OMPI_DECLSPEC extern int mca_btl_base_verbose; OMPI_DECLSPEC extern int mca_btl_base_verbose;
@ -37,8 +38,8 @@ OMPI_DECLSPEC extern int mca_btl_base_out(const char*, ...) __opal_attribute_for
#define BTL_OUTPUT(args) \ #define BTL_OUTPUT(args) \
do { \ do { \
mca_btl_base_out("[%s]%s[%s:%d:%s] ", \ mca_btl_base_out("[%s]%s[%s:%d:%s] ", \
orca_process_info_get_nodename(), \ orte_process_info.nodename, \
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), \ ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), \
__FILE__, __LINE__, __func__); \ __FILE__, __LINE__, __func__); \
mca_btl_base_out args; \ mca_btl_base_out args; \
mca_btl_base_out("\n"); \ mca_btl_base_out("\n"); \
@ -48,8 +49,8 @@ do { \
#define BTL_ERROR(args) \ #define BTL_ERROR(args) \
do { \ do { \
mca_btl_base_err("[%s]%s[%s:%d:%s] ", \ mca_btl_base_err("[%s]%s[%s:%d:%s] ", \
orca_process_info_get_nodename(), \ orte_process_info.nodename, \
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), \ ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), \
__FILE__, __LINE__, __func__); \ __FILE__, __LINE__, __func__); \
mca_btl_base_err args; \ mca_btl_base_err args; \
mca_btl_base_err("\n"); \ mca_btl_base_err("\n"); \
@ -58,9 +59,9 @@ do { \
#define BTL_PEER_ERROR(proc, args) \ #define BTL_PEER_ERROR(proc, args) \
do { \ do { \
mca_btl_base_err("%s[%s:%d:%s] from %s ", \ mca_btl_base_err("%s[%s:%d:%s] from %s ", \
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), \ ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), \
__FILE__, __LINE__, __func__, \ __FILE__, __LINE__, __func__, \
orca_process_info_get_nodename()); \ orte_process_info.nodename); \
if(proc && proc->proc_hostname) { \ if(proc && proc->proc_hostname) { \
mca_btl_base_err("to: %s ", proc->proc_hostname); \ mca_btl_base_err("to: %s ", proc->proc_hostname); \
} \ } \
@ -74,8 +75,8 @@ do { \
do { \ do { \
if(mca_btl_base_verbose > 0) { \ if(mca_btl_base_verbose > 0) { \
mca_btl_base_err("[%s]%s[%s:%d:%s] ", \ mca_btl_base_err("[%s]%s[%s:%d:%s] ", \
orca_process_info_get_nodename(), \ orte_process_info.nodename, \
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), \ ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), \
__FILE__, __LINE__, __func__); \ __FILE__, __LINE__, __func__); \
mca_btl_base_err args; \ mca_btl_base_err args; \
mca_btl_base_err("\n"); \ mca_btl_base_err("\n"); \

9
ompi/mca/btl/base/btl_base_select.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2007 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -22,13 +21,13 @@
#include "opal/util/argv.h" #include "opal/util/argv.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "opal/mca/base/mca_base_component_repository.h" #include "opal/mca/base/mca_base_component_repository.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
#include "ompi/mca/btl/base/base.h" #include "ompi/mca/btl/base/base.h"
#include "orte/mca/errmgr/errmgr.h"
#include "opal/runtime/opal.h" #include "opal/runtime/opal.h"
OBJ_CLASS_INSTANCE( mca_btl_base_selected_module_t, OBJ_CLASS_INSTANCE( mca_btl_base_selected_module_t,
@ -153,9 +152,9 @@ int mca_btl_base_select(bool enable_progress_threads,
/* Finished querying all components. Check for the bozo case. */ /* Finished querying all components. Check for the bozo case. */
if (0 == opal_list_get_size(&mca_btl_base_modules_initialized)) { if (0 == opal_list_get_size(&mca_btl_base_modules_initialized)) {
orca_show_help("help-mca-base.txt", "find-available:none-found", true, orte_show_help("help-mca-base.txt", "find-available:none-found", true,
"btl"); "btl");
orca_error_mgr_abort(1, NULL); orte_errmgr.abort(1, NULL);
} }
return OMPI_SUCCESS; return OMPI_SUCCESS;
} }

7
ompi/mca/btl/elan/btl_elan.c
Просмотреть файл

@ -2,7 +2,6 @@
* Copyright (c) 2004-2008 The University of Tennessee and The University * Copyright (c) 2004-2008 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights * of Tennessee Research Foundation. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -26,9 +25,7 @@
#include "elan/elan.h" #include "elan/elan.h"
#include "opal/util/os_path.h" #include "opal/util/os_path.h"
#include "opal/util/opal_environ.h" #include "opal/util/opal_environ.h"
#include "orte/util/proc_info.h"
#include "orca/include/rte_orca.h"
/** /**
* *
@ -71,7 +68,7 @@ static int mca_btl_elan_add_procs( struct mca_btl_base_module_t* btl,
FILE* file; FILE* file;
ELAN_BASE* base; ELAN_BASE* base;
filename = opal_os_path( false, orca_process_info_get_process_session_dir(), "ELAN_ID", NULL ); filename = opal_os_path( false, orte_process_info.proc_session_dir, "ELAN_ID", NULL );
file = fopen( filename, "w" ); file = fopen( filename, "w" );
fprintf( file, "%s %d\n", ompi_proc_local_proc->proc_hostname, elan_btl->elan_position ); fprintf( file, "%s %d\n", ompi_proc_local_proc->proc_hostname, elan_btl->elan_position );

3
ompi/mca/btl/mx/btl_mx.c
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -644,7 +643,7 @@ int mca_btl_mx_ft_event(int state) {
* kernel: blcr: thaw_threads returned error, aborting. -1 * kernel: blcr: thaw_threads returned error, aborting. -1
* JJH: It may be possible to, instead of restarting the entire driver, just reconnect endpoints * JJH: It may be possible to, instead of restarting the entire driver, just reconnect endpoints
*/ */
orca_info_cr_continue_like_restart() = true; orte_cr_continue_like_restart = true;
for( i = 0; i < mca_btl_mx_component.mx_num_btls; i++ ) { for( i = 0; i < mca_btl_mx_component.mx_num_btls; i++ ) {
mx_btl = mca_btl_mx_component.mx_btls[i]; mx_btl = mca_btl_mx_component.mx_btls[i];

8
ompi/mca/btl/mx/btl_mx_proc.c
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -19,8 +18,7 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "orca/include/rte_orca.h" #include "orte/util/name_fns.h"
#include "ompi/runtime/ompi_module_exchange.h" #include "ompi/runtime/ompi_module_exchange.h"
#include "btl_mx.h" #include "btl_mx.h"
@ -123,7 +121,7 @@ mca_btl_mx_proc_t* mca_btl_mx_proc_create(ompi_proc_t* ompi_proc)
ompi_proc, (void*)&mx_peers, &size ); ompi_proc, (void*)&mx_peers, &size );
if( OMPI_SUCCESS != rc ) { if( OMPI_SUCCESS != rc ) {
opal_output( 0, "mca_pml_base_modex_recv failed for peer %s", opal_output( 0, "mca_pml_base_modex_recv failed for peer %s",
ORCA_NAME_PRINT(&ompi_proc->proc_name) ); ORTE_NAME_PRINT(&ompi_proc->proc_name) );
return NULL; return NULL;
} }
@ -132,7 +130,7 @@ mca_btl_mx_proc_t* mca_btl_mx_proc_create(ompi_proc_t* ompi_proc)
} }
if( (size % sizeof(mca_btl_mx_addr_t)) != 0 ) { if( (size % sizeof(mca_btl_mx_addr_t)) != 0 ) {
opal_output( 0, "invalid mx address for peer %s", opal_output( 0, "invalid mx address for peer %s",
ORCA_NAME_PRINT(&ompi_proc->proc_name) ); ORTE_NAME_PRINT(&ompi_proc->proc_name) );
return NULL; return NULL;
} }
/* Let's see if we have a way to connect to the remote proc using MX. /* Let's see if we have a way to connect to the remote proc using MX.

17
ompi/mca/btl/ofud/btl_ofud_component.c
Просмотреть файл

@ -14,7 +14,6 @@
* Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -33,18 +32,18 @@
#include "opal_stdint.h" #include "opal_stdint.h"
#include "ompi/constants.h" #include "ompi/constants.h"
#include "opal/prefetch.h" #include "opal/prefetch.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
#include "opal/mca/timer/base/base.h" #include "opal/mca/timer/base/base.h"
#include "opal/util/argv.h" #include "opal/util/argv.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orte/mca/errmgr/errmgr.h"
#include "ompi/mca/btl/base/base.h" #include "ompi/mca/btl/base/base.h"
#include "ompi/runtime/ompi_module_exchange.h" #include "ompi/runtime/ompi_module_exchange.h"
#include "ompi/runtime/mpiruntime.h" #include "ompi/runtime/mpiruntime.h"
#include "orte/runtime/orte_globals.h"
#include "btl_ofud.h" #include "btl_ofud.h"
#include "btl_ofud_frag.h" #include "btl_ofud_frag.h"
#include "btl_ofud_endpoint.h" #include "btl_ofud_endpoint.h"
@ -440,7 +439,7 @@ mca_btl_base_module_t** mca_btl_ud_component_init(int* num_btl_modules,
*num_btl_modules = 0; *num_btl_modules = 0;
num_devs = 0; num_devs = 0;
seedv[0] = orca_process_info_get_vpid(ORCA_PROC_MY_NAME); seedv[0] = ORTE_PROC_MY_NAME->vpid;
seedv[1] = opal_timer_base_get_cycles(); seedv[1] = opal_timer_base_get_cycles();
seedv[2] = opal_timer_base_get_cycles(); seedv[2] = opal_timer_base_get_cycles();
seed48(seedv); seed48(seedv);
@ -452,7 +451,7 @@ mca_btl_base_module_t** mca_btl_ud_component_init(int* num_btl_modules,
mca_btl_ofud_component.if_list = NULL; mca_btl_ofud_component.if_list = NULL;
if (NULL != mca_btl_ofud_component.if_include && if (NULL != mca_btl_ofud_component.if_include &&
NULL != mca_btl_ofud_component.if_exclude) { NULL != mca_btl_ofud_component.if_exclude) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"specified include and exclude", true, "specified include and exclude", true,
mca_btl_ofud_component.if_include, mca_btl_ofud_component.if_include,
mca_btl_ofud_component.if_exclude, NULL); mca_btl_ofud_component.if_exclude, NULL);
@ -554,14 +553,14 @@ mca_btl_base_module_t** mca_btl_ud_component_init(int* num_btl_modules,
mca_btl_ofud_component.ud_btls = (mca_btl_ud_module_t*) mca_btl_ofud_component.ud_btls = (mca_btl_ud_module_t*)
malloc(sizeof(mca_btl_ud_module_t) * mca_btl_ofud_component.num_btls); malloc(sizeof(mca_btl_ud_module_t) * mca_btl_ofud_component.num_btls);
if(NULL == mca_btl_ofud_component.ud_btls) { if(NULL == mca_btl_ofud_component.ud_btls) {
ORCA_ERROR_LOG(ORCA_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return NULL; return NULL;
} }
btls = (struct mca_btl_base_module_t**) btls = (struct mca_btl_base_module_t**)
malloc(mca_btl_ofud_component.num_btls * sizeof(mca_btl_ud_module_t*)); malloc(mca_btl_ofud_component.num_btls * sizeof(mca_btl_ud_module_t*));
if(NULL == btls) { if(NULL == btls) {
ORCA_ERROR_LOG(ORCA_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return NULL; return NULL;
} }

5
ompi/mca/btl/ofud/btl_ofud_proc.c
Просмотреть файл

@ -11,7 +11,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2006 Sandia National Laboratories. All rights * Copyright (c) 2006 Sandia National Laboratories. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -126,14 +125,14 @@ mca_btl_ud_proc_t* mca_btl_ud_proc_create(ompi_proc_t* ompi_proc)
if(OMPI_SUCCESS != rc) { if(OMPI_SUCCESS != rc) {
opal_output(0, opal_output(0,
"[%s:%d] ompi_modex_recv failed for peer %s", "[%s:%d] ompi_modex_recv failed for peer %s",
__FILE__,__LINE__,ORCA_NAME_PRINT(&ompi_proc->proc_name)); __FILE__,__LINE__,ORTE_NAME_PRINT(&ompi_proc->proc_name));
OBJ_RELEASE(module_proc); OBJ_RELEASE(module_proc);
return NULL; return NULL;
} }
if((size % sizeof(mca_btl_ud_addr_t)) != 0) { if((size % sizeof(mca_btl_ud_addr_t)) != 0) {
opal_output(0, "[%s:%d] invalid module address for peer %s", opal_output(0, "[%s:%d] invalid module address for peer %s",
__FILE__,__LINE__,ORCA_NAME_PRINT(&ompi_proc->proc_name)); __FILE__,__LINE__,ORTE_NAME_PRINT(&ompi_proc->proc_name));
OBJ_RELEASE(module_proc); OBJ_RELEASE(module_proc);
return NULL; return NULL;
} }

37
ompi/mca/btl/openib/btl_openib.c
Просмотреть файл

@ -17,7 +17,6 @@
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2008-2012 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2008-2012 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2009 IBM Corporation. All rights reserved. * Copyright (c) 2009 IBM Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -30,8 +29,8 @@
#ifdef HAVE_INTTYPES_H #ifdef HAVE_INTTYPES_H
#include <inttypes.h> #include <inttypes.h>
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "opal/class/opal_bitmap.h" #include "opal/class/opal_bitmap.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/util/arch.h" #include "opal/util/arch.h"
@ -55,7 +54,7 @@
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
#include "ompi/mca/mpool/mpool.h" #include "ompi/mca/mpool/mpool.h"
#include "ompi/mca/mpool/grdma/mpool_grdma.h" #include "ompi/mca/mpool/grdma/mpool_grdma.h"
#include "orte/util/proc_info.h"
#include <errno.h> #include <errno.h>
#include <string.h> #include <string.h>
#include <math.h> #include <math.h>
@ -143,14 +142,14 @@ void mca_btl_openib_show_init_error(const char *file, int line,
} }
#endif #endif
orca_show_help("help-mpi-btl-openib.txt", "init-fail-no-mem", orte_show_help("help-mpi-btl-openib.txt", "init-fail-no-mem",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
file, line, func, dev, str_limit); file, line, func, dev, str_limit);
if (NULL != str_limit) free(str_limit); if (NULL != str_limit) free(str_limit);
} else { } else {
orca_show_help("help-mpi-btl-openib.txt", "init-fail-create-q", orte_show_help("help-mpi-btl-openib.txt", "init-fail-create-q",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
file, line, func, strerror(errno), errno, dev); file, line, func, strerror(errno), errno, dev);
} }
} }
@ -475,9 +474,9 @@ static int mca_btl_openib_tune_endpoint(mca_btl_openib_module_t* openib_btl,
ompi_btl_openib_ini_values_t values; ompi_btl_openib_ini_values_t values;
if(mca_btl_openib_get_transport_type(openib_btl) != endpoint->rem_info.rem_transport_type) { if(mca_btl_openib_get_transport_type(openib_btl) != endpoint->rem_info.rem_transport_type) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"conflicting transport types", true, "conflicting transport types", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
(openib_btl->device->ib_dev_attr).vendor_id, (openib_btl->device->ib_dev_attr).vendor_id,
(openib_btl->device->ib_dev_attr).vendor_part_id, (openib_btl->device->ib_dev_attr).vendor_part_id,
@ -496,9 +495,9 @@ static int mca_btl_openib_tune_endpoint(mca_btl_openib_module_t* openib_btl,
if (OMPI_SUCCESS != ret && if (OMPI_SUCCESS != ret &&
OMPI_ERR_NOT_FOUND != ret) { OMPI_ERR_NOT_FOUND != ret) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"error in device init", true, "error in device init", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev)); ibv_get_device_name(openib_btl->device->ib_dev));
return ret; return ret;
} }
@ -537,9 +536,9 @@ static int mca_btl_openib_tune_endpoint(mca_btl_openib_module_t* openib_btl,
if(0 != strcmp(mca_btl_openib_component.receive_queues, if(0 != strcmp(mca_btl_openib_component.receive_queues,
recv_qps)) { recv_qps)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"unsupported queues configuration", true, "unsupported queues configuration", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
(openib_btl->device->ib_dev_attr).vendor_id, (openib_btl->device->ib_dev_attr).vendor_id,
(openib_btl->device->ib_dev_attr).vendor_part_id, (openib_btl->device->ib_dev_attr).vendor_part_id,
@ -559,9 +558,9 @@ static int mca_btl_openib_tune_endpoint(mca_btl_openib_module_t* openib_btl,
if(NULL != values.receive_queues) { if(NULL != values.receive_queues) {
if(0 != strcmp(mca_btl_openib_component.receive_queues, if(0 != strcmp(mca_btl_openib_component.receive_queues,
values.receive_queues)) { values.receive_queues)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"unsupported queues configuration", true, "unsupported queues configuration", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
(openib_btl->device->ib_dev_attr).vendor_id, (openib_btl->device->ib_dev_attr).vendor_id,
(openib_btl->device->ib_dev_attr).vendor_part_id, (openib_btl->device->ib_dev_attr).vendor_part_id,
@ -631,8 +630,8 @@ int mca_btl_openib_add_procs(
/* OOB, XOOB, and RDMACM do not support SELF comunication, so /* OOB, XOOB, and RDMACM do not support SELF comunication, so
* mark the prco as unreachable by openib btl */ * mark the prco as unreachable by openib btl */
if (OPAL_EQUAL == orca_process_name_compare if (OPAL_EQUAL == orte_util_compare_name_fields
(ORCA_NAME_CMP_ALL, ORCA_PROC_MY_NAME, &ompi_proc->proc_name)) { (ORTE_NS_CMP_ALL, ORTE_PROC_MY_NAME, &ompi_proc->proc_name)) {
continue; continue;
} }
#if defined(HAVE_STRUCT_IBV_DEVICE_TRANSPORT_TYPE) #if defined(HAVE_STRUCT_IBV_DEVICE_TRANSPORT_TYPE)
@ -1758,7 +1757,7 @@ int mca_btl_openib_ft_event(int state) {
if(OPAL_CRS_CHECKPOINT == state) { if(OPAL_CRS_CHECKPOINT == state) {
/* Continue must reconstruct the routes (including modex), since we /* Continue must reconstruct the routes (including modex), since we
* have to tear down the devices completely. */ * have to tear down the devices completely. */
orca_info_cr_continue_like_restart() = true; orte_cr_continue_like_restart = true;
/* /*
* To keep the node from crashing we need to call ibv_close_device * To keep the node from crashing we need to call ibv_close_device

15
ompi/mca/btl/openib/btl_openib_async.c
Просмотреть файл

@ -3,7 +3,6 @@
* Copyright (c) 2007-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -21,7 +20,7 @@
#include <unistd.h> #include <unistd.h>
#include <errno.h> #include <errno.h>
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "ompi/mca/btl/base/base.h" #include "ompi/mca/btl/base/base.h"
#include "btl_openib.h" #include "btl_openib.h"
@ -349,15 +348,15 @@ static int btl_openib_async_deviceh(struct mca_btl_openib_async_poll *devices_po
case IBV_EVENT_QP_ACCESS_ERR: case IBV_EVENT_QP_ACCESS_ERR:
case IBV_EVENT_PATH_MIG_ERR: case IBV_EVENT_PATH_MIG_ERR:
case IBV_EVENT_SRQ_ERR: case IBV_EVENT_SRQ_ERR:
orca_show_help("help-mpi-btl-openib.txt", "of error event", orte_show_help("help-mpi-btl-openib.txt", "of error event",
true,orca_process_info_get_nodename(), orca_process_info_get_pid(), true,orte_process_info.nodename, orte_process_info.pid,
event_type, event_type,
openib_event_to_str((enum ibv_event_type)event_type), openib_event_to_str((enum ibv_event_type)event_type),
xrc_event ? "true" : "false"); xrc_event ? "true" : "false");
break; break;
case IBV_EVENT_PORT_ERR: case IBV_EVENT_PORT_ERR:
orca_show_help("help-mpi-btl-openib.txt", "of error event", orte_show_help("help-mpi-btl-openib.txt", "of error event",
true,orca_process_info_get_nodename(), orca_process_info_get_pid(), true,orte_process_info.nodename, orte_process_info.pid,
event_type, event_type,
openib_event_to_str((enum ibv_event_type)event_type), openib_event_to_str((enum ibv_event_type)event_type),
xrc_event ? "true" : "false"); xrc_event ? "true" : "false");
@ -386,8 +385,8 @@ static int btl_openib_async_deviceh(struct mca_btl_openib_async_poll *devices_po
break; break;
default: default:
orca_show_help("help-mpi-btl-openib.txt", "of unknown event", orte_show_help("help-mpi-btl-openib.txt", "of unknown event",
true,orca_process_info_get_nodename(), orca_process_info_get_pid(), true,orte_process_info.nodename, orte_process_info.pid,
event_type, xrc_event ? "true" : "false"); event_type, xrc_event ? "true" : "false");
} }
ibv_ack_async_event(&event); ibv_ack_async_event(&event);

143
ompi/mca/btl/openib/btl_openib_component.c
Просмотреть файл

@ -17,7 +17,6 @@
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2009-2012 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2009-2012 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -72,7 +71,9 @@ const char *ibv_get_sysfs_path(void);
#include "opal/mca/installdirs/installdirs.h" #include "opal/mca/installdirs/installdirs.h"
#include "opal_stdint.h" #include "opal_stdint.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/proc_info.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/constants.h" #include "ompi/constants.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
@ -603,8 +604,8 @@ static int openib_reg_mr(void *reg_data, void *base, size_t size,
openib_reg->mr = ibv_reg_mr(device->ib_pd, base, size, access_flag); openib_reg->mr = ibv_reg_mr(device->ib_pd, base, size, access_flag);
if (NULL == openib_reg->mr) { if (NULL == openib_reg->mr) {
orca_show_help("help-mpi-btl-openib.txt", "mem-reg-fail", orte_show_help("help-mpi-btl-openib.txt", "mem-reg-fail",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
ibv_get_device_name(device->ib_dev), ibv_get_device_name(device->ib_dev),
__func__, strerror(errno), errno); __func__, strerror(errno), errno);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
@ -696,8 +697,8 @@ static int init_one_port(opal_list_t *btl_list, mca_btl_openib_device_t *device,
size. */ size. */
if (mca_btl_openib_component.gid_index > if (mca_btl_openib_component.gid_index >
ib_port_attr->gid_tbl_len) { ib_port_attr->gid_tbl_len) {
orca_show_help("help-mpi-btl-openib.txt", "gid index too large", orte_show_help("help-mpi-btl-openib.txt", "gid index too large",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
ibv_get_device_name(device->ib_dev), port_num, ibv_get_device_name(device->ib_dev), port_num,
mca_btl_openib_component.gid_index, mca_btl_openib_component.gid_index,
ib_port_attr->gid_tbl_len); ib_port_attr->gid_tbl_len);
@ -754,8 +755,8 @@ static int init_one_port(opal_list_t *btl_list, mca_btl_openib_device_t *device,
if(mca_btl_openib_component.ib_num_btls > 0 && if(mca_btl_openib_component.ib_num_btls > 0 &&
IB_DEFAULT_GID_PREFIX == subnet_id && IB_DEFAULT_GID_PREFIX == subnet_id &&
mca_btl_openib_component.warn_default_gid_prefix) { mca_btl_openib_component.warn_default_gid_prefix) {
orca_show_help("help-mpi-btl-openib.txt", "default subnet prefix", orte_show_help("help-mpi-btl-openib.txt", "default subnet prefix",
true, orca_process_info_get_nodename()); true, orte_process_info.nodename);
} }
lmc = (1 << ib_port_attr->lmc); lmc = (1 << ib_port_attr->lmc);
@ -778,7 +779,7 @@ static int init_one_port(opal_list_t *btl_list, mca_btl_openib_device_t *device,
} else if (0 == lmc % (mca_btl_openib_component.apm_lmc + 1)) { } else if (0 == lmc % (mca_btl_openib_component.apm_lmc + 1)) {
lmc_step = mca_btl_openib_component.apm_lmc + 1; lmc_step = mca_btl_openib_component.apm_lmc + 1;
} else { } else {
orca_show_help("help-mpi-btl-openib.txt", "apm with wrong lmc",true, orte_show_help("help-mpi-btl-openib.txt", "apm with wrong lmc",true,
mca_btl_openib_component.apm_lmc, lmc); mca_btl_openib_component.apm_lmc, lmc);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -786,7 +787,7 @@ static int init_one_port(opal_list_t *btl_list, mca_btl_openib_device_t *device,
if (mca_btl_openib_component.apm_lmc) { if (mca_btl_openib_component.apm_lmc) {
/* Disable apm and report warning */ /* Disable apm and report warning */
mca_btl_openib_component.apm_lmc = 0; mca_btl_openib_component.apm_lmc = 0;
orca_show_help("help-mpi-btl-openib.txt", "apm without lmc",true); orte_show_help("help-mpi-btl-openib.txt", "apm without lmc",true);
} }
} }
#endif #endif
@ -1126,11 +1127,11 @@ static int prepare_device_for_use(mca_btl_openib_device_t *device)
*/ */
if (!(device->ib_dev_attr.device_cap_flags & IBV_DEVICE_XRC) && if (!(device->ib_dev_attr.device_cap_flags & IBV_DEVICE_XRC) &&
MCA_BTL_XRC_ENABLED) { MCA_BTL_XRC_ENABLED) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"XRC on device without XRC support", true, "XRC on device without XRC support", true,
mca_btl_openib_component.num_xrc_qps, mca_btl_openib_component.num_xrc_qps,
ibv_get_device_name(device->ib_dev), ibv_get_device_name(device->ib_dev),
orca_process_info_get_nodename()); orte_process_info.nodename);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -1435,9 +1436,9 @@ static int setup_qps(void)
queues = opal_argv_split(mca_btl_openib_component.receive_queues, ':'); queues = opal_argv_split(mca_btl_openib_component.receive_queues, ':');
if (0 == opal_argv_count(queues)) { if (0 == opal_argv_count(queues)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"no qps in receive_queues", true, "no qps in receive_queues", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_openib_component.receive_queues); mca_btl_openib_component.receive_queues);
ret = OMPI_ERROR; ret = OMPI_ERROR;
goto error; goto error;
@ -1455,16 +1456,16 @@ static int setup_qps(void)
#if HAVE_XRC #if HAVE_XRC
num_xrc_qps++; num_xrc_qps++;
#else #else
orca_show_help("help-mpi-btl-openib.txt", "No XRC support", true, orte_show_help("help-mpi-btl-openib.txt", "No XRC support", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_openib_component.receive_queues); mca_btl_openib_component.receive_queues);
ret = OMPI_ERR_NOT_AVAILABLE; ret = OMPI_ERR_NOT_AVAILABLE;
goto error; goto error;
#endif #endif
} else { } else {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid qp type in receive_queues", true, "invalid qp type in receive_queues", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_openib_component.receive_queues, mca_btl_openib_component.receive_queues,
queues[qp]); queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
@ -1475,8 +1476,8 @@ static int setup_qps(void)
/* Current XRC implementation can't used with other QP types - PP /* Current XRC implementation can't used with other QP types - PP
and SRQ */ and SRQ */
if (num_xrc_qps > 0 && (num_pp_qps > 0 || num_srq_qps > 0)) { if (num_xrc_qps > 0 && (num_pp_qps > 0 || num_srq_qps > 0)) {
orca_show_help("help-mpi-btl-openib.txt", "XRC with PP or SRQ", true, orte_show_help("help-mpi-btl-openib.txt", "XRC with PP or SRQ", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_openib_component.receive_queues); mca_btl_openib_component.receive_queues);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
@ -1484,8 +1485,8 @@ static int setup_qps(void)
/* Current XRC implementation can't used with btls_per_lid > 1 */ /* Current XRC implementation can't used with btls_per_lid > 1 */
if (num_xrc_qps > 0 && mca_btl_openib_component.btls_per_lid > 1) { if (num_xrc_qps > 0 && mca_btl_openib_component.btls_per_lid > 1) {
orca_show_help("help-mpi-btl-openib.txt", "XRC with BTLs per LID", orte_show_help("help-mpi-btl-openib.txt", "XRC with BTLs per LID",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
mca_btl_openib_component.receive_queues, num_xrc_qps); mca_btl_openib_component.receive_queues, num_xrc_qps);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
@ -1510,9 +1511,9 @@ static int setup_qps(void)
if ('P' == params[0][0]) { if ('P' == params[0][0]) {
int32_t rd_win, rd_rsv; int32_t rd_win, rd_rsv;
if (count < 3 || count > 6) { if (count < 3 || count > 6) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid pp qp specification", true, "invalid pp qp specification", true,
orca_process_info_get_nodename(), queues[qp]); orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1535,15 +1536,15 @@ static int setup_qps(void)
mca_btl_openib_component.qp_infos[qp].u.pp_qp.rd_win = rd_win; mca_btl_openib_component.qp_infos[qp].u.pp_qp.rd_win = rd_win;
mca_btl_openib_component.qp_infos[qp].u.pp_qp.rd_rsv = rd_rsv; mca_btl_openib_component.qp_infos[qp].u.pp_qp.rd_rsv = rd_rsv;
if ((rd_num - rd_low) > rd_win) { if ((rd_num - rd_low) > rd_win) {
orca_show_help("help-mpi-btl-openib.txt", "non optimal rd_win", orte_show_help("help-mpi-btl-openib.txt", "non optimal rd_win",
true, rd_win, rd_num - rd_low); true, rd_win, rd_num - rd_low);
} }
} else { } else {
int32_t sd_max, rd_init, srq_limit; int32_t sd_max, rd_init, srq_limit;
if (count < 3 || count > 7) { if (count < 3 || count > 7) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid srq specification", true, "invalid srq specification", true,
orca_process_info_get_nodename(), queues[qp]); orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1578,15 +1579,15 @@ static int setup_qps(void)
} }
if (rd_num < rd_init) { if (rd_num < rd_init) {
orca_show_help("help-mpi-btl-openib.txt", "rd_num must be >= rd_init", orte_show_help("help-mpi-btl-openib.txt", "rd_num must be >= rd_init",
true, orca_process_info_get_nodename(), queues[qp]); true, orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
if (rd_num < srq_limit) { if (rd_num < srq_limit) {
orca_show_help("help-mpi-btl-openib.txt", "srq_limit must be > rd_num", orte_show_help("help-mpi-btl-openib.txt", "srq_limit must be > rd_num",
true, orca_process_info_get_nodename(), queues[qp]); true, orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1597,8 +1598,8 @@ static int setup_qps(void)
} }
if (rd_num <= rd_low) { if (rd_num <= rd_low) {
orca_show_help("help-mpi-btl-openib.txt", "rd_num must be > rd_low", orte_show_help("help-mpi-btl-openib.txt", "rd_num must be > rd_low",
true, orca_process_info_get_nodename(), queues[qp]); true, orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1617,23 +1618,23 @@ static int setup_qps(void)
mca_btl_openib_module.super.btl_eager_limit : mca_btl_openib_module.super.btl_eager_limit :
mca_btl_openib_module.super.btl_max_send_size; mca_btl_openib_module.super.btl_max_send_size;
if (max_qp_size < max_size_needed) { if (max_qp_size < max_size_needed) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"biggest qp size is too small", true, "biggest qp size is too small", true,
orca_process_info_get_nodename(), max_qp_size, orte_process_info.nodename, max_qp_size,
max_size_needed); max_size_needed);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} else if (max_qp_size > max_size_needed) { } else if (max_qp_size > max_size_needed) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"biggest qp size is too big", true, "biggest qp size is too big", true,
orca_process_info_get_nodename(), max_qp_size, orte_process_info.nodename, max_qp_size,
max_size_needed); max_size_needed);
} }
if (mca_btl_openib_component.ib_free_list_max > 0 && if (mca_btl_openib_component.ib_free_list_max > 0 &&
min_freelist_size > mca_btl_openib_component.ib_free_list_max) { min_freelist_size > mca_btl_openib_component.ib_free_list_max) {
orca_show_help("help-mpi-btl-openib.txt", "freelist too small", true, orte_show_help("help-mpi-btl-openib.txt", "freelist too small", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_openib_component.ib_free_list_max, mca_btl_openib_component.ib_free_list_max,
min_freelist_size); min_freelist_size);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
@ -1717,9 +1718,9 @@ static int init_one_device(opal_list_t *btl_list, struct ibv_device* ib_dev)
warning that we're using default values (unless overridden warning that we're using default values (unless overridden
that we don't want to see these warnings) */ that we don't want to see these warnings) */
if (mca_btl_openib_component.warn_no_device_params_found) { if (mca_btl_openib_component.warn_no_device_params_found) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"no device params found", true, "no device params found", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(device->ib_dev), ibv_get_device_name(device->ib_dev),
device->ib_dev_attr.vendor_id, device->ib_dev_attr.vendor_id,
device->ib_dev_attr.vendor_part_id); device->ib_dev_attr.vendor_part_id);
@ -1825,8 +1826,8 @@ static int init_one_device(opal_list_t *btl_list, struct ibv_device* ib_dev)
cq = ibv_create_cq(device->ib_dev_context, 1, NULL, NULL, 0); cq = ibv_create_cq(device->ib_dev_context, 1, NULL, NULL, 0);
#endif #endif
if (NULL == cq) { if (NULL == cq) {
orca_show_help("help-mpi-btl-openib.txt", "init-fail-create-q", orte_show_help("help-mpi-btl-openib.txt", "init-fail-create-q",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
__FILE__, __LINE__, "ibv_create_cq", __FILE__, __LINE__, "ibv_create_cq",
strerror(errno), errno, strerror(errno), errno,
ibv_get_device_name(device->ib_dev)); ibv_get_device_name(device->ib_dev));
@ -1878,7 +1879,7 @@ static int init_one_device(opal_list_t *btl_list, struct ibv_device* ib_dev)
/* Eager RDMA is not currently supported with progress threads */ /* Eager RDMA is not currently supported with progress threads */
if (device->use_eager_rdma && OMPI_ENABLE_PROGRESS_THREADS) { if (device->use_eager_rdma && OMPI_ENABLE_PROGRESS_THREADS) {
device->use_eager_rdma = 0; device->use_eager_rdma = 0;
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"eager RDMA and progress threads", true); "eager RDMA and progress threads", true);
} }
@ -1964,7 +1965,7 @@ static int init_one_device(opal_list_t *btl_list, struct ibv_device* ib_dev)
if (device->btls > 0) { if (device->btls > 0) {
/* if apm was enabled it should be > 1 */ /* if apm was enabled it should be > 1 */
if (1 == mca_btl_openib_component.apm_ports) { if (1 == mca_btl_openib_component.apm_ports) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"apm not enough ports", true); "apm not enough ports", true);
mca_btl_openib_component.apm_ports = 0; mca_btl_openib_component.apm_ports = 0;
} }
@ -2223,10 +2224,10 @@ static int init_one_device(opal_list_t *btl_list, struct ibv_device* ib_dev)
if (NULL != values.receive_queues) { if (NULL != values.receive_queues) {
if (0 != strcmp(values.receive_queues, if (0 != strcmp(values.receive_queues,
mca_btl_openib_component.receive_queues)) { mca_btl_openib_component.receive_queues)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"locally conflicting receive_queues", true, "locally conflicting receive_queues", true,
opal_install_dirs.pkgdatadir, opal_install_dirs.pkgdatadir,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(receive_queues_device->ib_dev), ibv_get_device_name(receive_queues_device->ib_dev),
receive_queues_device->ib_dev_attr.vendor_id, receive_queues_device->ib_dev_attr.vendor_id,
receive_queues_device->ib_dev_attr.vendor_part_id, receive_queues_device->ib_dev_attr.vendor_part_id,
@ -2247,10 +2248,10 @@ static int init_one_device(opal_list_t *btl_list, struct ibv_device* ib_dev)
device's INI file, we must error. */ device's INI file, we must error. */
else if (BTL_OPENIB_RQ_SOURCE_DEVICE_INI == else if (BTL_OPENIB_RQ_SOURCE_DEVICE_INI ==
mca_btl_openib_component.receive_queues_source) { mca_btl_openib_component.receive_queues_source) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"locally conflicting receive_queues", true, "locally conflicting receive_queues", true,
opal_install_dirs.pkgdatadir, opal_install_dirs.pkgdatadir,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(receive_queues_device->ib_dev), ibv_get_device_name(receive_queues_device->ib_dev),
receive_queues_device->ib_dev_attr.vendor_id, receive_queues_device->ib_dev_attr.vendor_id,
receive_queues_device->ib_dev_attr.vendor_part_id, receive_queues_device->ib_dev_attr.vendor_part_id,
@ -2286,9 +2287,9 @@ error:
} }
if (OMPI_SUCCESS != ret) { if (OMPI_SUCCESS != ret) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"error in device init", true, "error in device init", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(device->ib_dev)); ibv_get_device_name(device->ib_dev));
} }
@ -2547,7 +2548,7 @@ sort_devs_by_distance(struct ibv_device **ib_devs, int count)
for (i = 0; i < count; i++) { for (i = 0; i < count; i++) {
devs[i].ib_dev = ib_devs[i]; devs[i].ib_dev = ib_devs[i];
if (OPAL_HAVE_HWLOC && orca_process_info_is_bound()) { if (OPAL_HAVE_HWLOC && orte_proc_is_bound) {
/* If this process is bound to one or more PUs, we can get /* If this process is bound to one or more PUs, we can get
an accurate distance. */ an accurate distance. */
devs[i].distance = get_ib_dev_distance(ib_devs[i]); devs[i].distance = get_ib_dev_distance(ib_devs[i]);
@ -2613,7 +2614,7 @@ btl_openib_component_init(int *num_btl_modules,
} }
#ifndef __WINDOWS__ #ifndef __WINDOWS__
seedv[0] = orca_process_info_get_vpid(ORCA_PROC_MY_NAME); seedv[0] = ORTE_PROC_MY_NAME->vpid;
seedv[1] = opal_timer_base_get_cycles(); seedv[1] = opal_timer_base_get_cycles();
seedv[2] = opal_timer_base_get_cycles(); seedv[2] = opal_timer_base_get_cycles();
seed48(seedv); seed48(seedv);
@ -2645,9 +2646,9 @@ btl_openib_component_init(int *num_btl_modules,
#if !OPAL_HAVE_THREADS #if !OPAL_HAVE_THREADS
if ((OPAL_MEMORY_FREE_SUPPORT | OPAL_MEMORY_MUNMAP_SUPPORT) == if ((OPAL_MEMORY_FREE_SUPPORT | OPAL_MEMORY_MUNMAP_SUPPORT) ==
((OPAL_MEMORY_FREE_SUPPORT | OPAL_MEMORY_MUNMAP_SUPPORT) & value)) { ((OPAL_MEMORY_FREE_SUPPORT | OPAL_MEMORY_MUNMAP_SUPPORT) & value)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"ptmalloc2 with no threads", true, "ptmalloc2 with no threads", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
goto no_btls; goto no_btls;
} }
#endif #endif
@ -2771,9 +2772,9 @@ btl_openib_component_init(int *num_btl_modules,
couldn't provide it. So print an error and deactivate couldn't provide it. So print an error and deactivate
this BTL. */ this BTL. */
if (mca_btl_openib_component.want_fork_support > 0) { if (mca_btl_openib_component.want_fork_support > 0) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"ibv_fork_init fail", true, "ibv_fork_init fail", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
goto no_btls; goto no_btls;
} }
} }
@ -2795,7 +2796,7 @@ btl_openib_component_init(int *num_btl_modules,
list_count++; list_count++;
if (list_count > 1) { if (list_count > 1) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"specified include and exclude", true, "specified include and exclude", true,
NULL == mca_btl_openib_component.if_include ? NULL == mca_btl_openib_component.if_include ?
"<not specified>" : mca_btl_openib_component.if_include, "<not specified>" : mca_btl_openib_component.if_include,
@ -2863,7 +2864,7 @@ btl_openib_component_init(int *num_btl_modules,
continue; continue;
} }
#else #else
orca_show_help("help-mpi-btl-openib.txt", "no iwarp support", orte_show_help("help-mpi-btl-openib.txt", "no iwarp support",
true); true);
#endif #endif
break; break;
@ -2881,8 +2882,8 @@ btl_openib_component_init(int *num_btl_modules,
} }
free(dev_sorted); free(dev_sorted);
if (!found) { if (!found) {
orca_show_help("help-mpi-btl-openib.txt", "no devices right type", orte_show_help("help-mpi-btl-openib.txt", "no devices right type",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
((BTL_OPENIB_DT_IB == mca_btl_openib_component.device_type) ? ((BTL_OPENIB_DT_IB == mca_btl_openib_component.device_type) ?
"InfiniBand" : "InfiniBand" :
(BTL_OPENIB_DT_IWARP == mca_btl_openib_component.device_type) ? (BTL_OPENIB_DT_IWARP == mca_btl_openib_component.device_type) ?
@ -2898,16 +2899,16 @@ btl_openib_component_init(int *num_btl_modules,
if (0 != opal_argv_count(mca_btl_openib_component.if_list) && if (0 != opal_argv_count(mca_btl_openib_component.if_list) &&
mca_btl_openib_component.warn_nonexistent_if) { mca_btl_openib_component.warn_nonexistent_if) {
char *str = opal_argv_join(mca_btl_openib_component.if_list, ','); char *str = opal_argv_join(mca_btl_openib_component.if_list, ',');
orca_show_help("help-mpi-btl-openib.txt", "nonexistent port", orte_show_help("help-mpi-btl-openib.txt", "nonexistent port",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
((NULL != mca_btl_openib_component.if_include) ? ((NULL != mca_btl_openib_component.if_include) ?
"in" : "ex"), str); "in" : "ex"), str);
free(str); free(str);
} }
if(0 == mca_btl_openib_component.ib_num_btls) { if(0 == mca_btl_openib_component.ib_num_btls) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"no active ports found", true, orca_process_info_get_nodename()); "no active ports found", true, orte_process_info.nodename);
goto no_btls; goto no_btls;
} }
@ -3018,9 +3019,9 @@ btl_openib_component_init(int *num_btl_modules,
/* Do finial init on device */ /* Do finial init on device */
ret = prepare_device_for_use(device); ret = prepare_device_for_use(device);
if (OMPI_SUCCESS != ret) { if (OMPI_SUCCESS != ret) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"error in device init", true, "error in device init", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(device->ib_dev)); ibv_get_device_name(device->ib_dev));
goto no_btls; goto no_btls;
} }
@ -3577,16 +3578,16 @@ error:
ibv_get_device_name(endpoint->qps[qp].qp->lcl_qp->context->device); ibv_get_device_name(endpoint->qps[qp].qp->lcl_qp->context->device);
if (IBV_WC_RNR_RETRY_EXC_ERR == wc->status) { if (IBV_WC_RNR_RETRY_EXC_ERR == wc->status) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
BTL_OPENIB_QP_TYPE_PP(qp) ? BTL_OPENIB_QP_TYPE_PP(qp) ?
"pp rnr retry exceeded" : "pp rnr retry exceeded" :
"srq rnr retry exceeded", true, "srq rnr retry exceeded", true,
orca_process_info_get_nodename(), device_name, orte_process_info.nodename, device_name,
peer_hostname); peer_hostname);
} else if (IBV_WC_RETRY_EXC_ERR == wc->status) { } else if (IBV_WC_RETRY_EXC_ERR == wc->status) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"pp retry exceeded", true, "pp retry exceeded", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
device_name, peer_hostname); device_name, peer_hostname);
} }
} }

8
ompi/mca/btl/openib/btl_openib_endpoint.c
Просмотреть файл

@ -17,7 +17,7 @@
* Copyright (c) 2006-2009 Mellanox Technologies, Inc. All rights reserved. * Copyright (c) 2006-2009 Mellanox Technologies, Inc. All rights reserved.
* Copyright (c) 2010-2011 IBM Corporation. All rights reserved. * Copyright (c) 2010-2011 IBM Corporation. All rights reserved.
* Copyright (c) 2010-2011 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2010-2011 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -37,7 +37,7 @@
#include "opal_stdint.h" #include "opal_stdint.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "ompi/types.h" #include "ompi/types.h"
#include "ompi/class/ompi_free_list.h" #include "ompi/class/ompi_free_list.h"
@ -1042,9 +1042,9 @@ void *mca_btl_openib_endpoint_invoke_error(void *context)
/* If we didn't find a BTL, then just bail :-( */ /* If we didn't find a BTL, then just bail :-( */
if (NULL == btl || NULL == btl->error_cb) { if (NULL == btl || NULL == btl->error_cb) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"cannot raise btl error", true, "cannot raise btl error", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
__FILE__, __LINE__); __FILE__, __LINE__);
exit(1); exit(1);
} }

9
ompi/mca/btl/openib/btl_openib_failover.c
Просмотреть файл

@ -3,7 +3,6 @@
* Copyright (c) 2011 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -373,14 +372,14 @@ void btl_openib_handle_failover_control_messages(mca_btl_openib_control_header_t
opal_output_verbose(20, mca_btl_openib_component.verbose_failover, opal_output_verbose(20, mca_btl_openib_component.verbose_failover,
"IB: rank=%d, control message (remote=%d), " "IB: rank=%d, control message (remote=%d), "
"moved local head by one (new=%d)", "moved local head by one (new=%d)",
orca_process_info_get_vpid(ORCA_PROC_MY_NAME), ORTE_PROC_MY_NAME->vpid,
newep->endpoint_proc->proc_ompi->proc_name.vpid, newep->endpoint_proc->proc_ompi->proc_name.vpid,
newep->eager_rdma_local.head); newep->eager_rdma_local.head);
} else { } else {
opal_output_verbose(20, mca_btl_openib_component.verbose_failover, opal_output_verbose(20, mca_btl_openib_component.verbose_failover,
"IB: rank=%d, control message (remote=%d), " "IB: rank=%d, control message (remote=%d), "
"did not move local head by one (still=%d)", "did not move local head by one (still=%d)",
orca_process_info_get_vpid(ORCA_PROC_MY_NAME), ORTE_PROC_MY_NAME->vpid,
newep->endpoint_proc->proc_ompi->proc_name.vpid, newep->endpoint_proc->proc_ompi->proc_name.vpid,
newep->eager_rdma_local.head); newep->eager_rdma_local.head);
} }
@ -685,7 +684,7 @@ static void mca_btl_openib_endpoint_notify(mca_btl_base_endpoint_t* endpoint, ui
bc_hdr->control.type = type; bc_hdr->control.type = type;
bc_hdr->lid = endpoint->endpoint_btl->port_info.lid; bc_hdr->lid = endpoint->endpoint_btl->port_info.lid;
bc_hdr->subnet_id = endpoint->endpoint_btl->port_info.subnet_id; bc_hdr->subnet_id = endpoint->endpoint_btl->port_info.subnet_id;
bc_hdr->vpid = orca_process_info_get_vpid(ORCA_PROC_MY_NAME); bc_hdr->vpid = ORTE_PROC_MY_NAME->vpid;
bc_hdr->index = index; bc_hdr->index = index;
if(newep->nbo) { if(newep->nbo) {
@ -740,7 +739,7 @@ void mca_btl_openib_dump_all_local_rdma_frags(mca_btl_openib_device_t *device) {
mca_btl_openib_endpoint_t* endpoint; mca_btl_openib_endpoint_t* endpoint;
c = device->eager_rdma_buffers_count; c = device->eager_rdma_buffers_count;
opal_output(0, "rank=%d, device=%s", orca_process_info_get_vpid(ORCA_PROC_MY_NAME), device->ib_dev->name); opal_output(0, "rank=%d, device=%s", ORTE_PROC_MY_NAME->vpid, device->ib_dev->name);
for(i = 0; i < c; i++) { for(i = 0; i < c; i++) {
endpoint = device->eager_rdma_buffers[i]; endpoint = device->eager_rdma_buffers[i];

9
ompi/mca/btl/openib/btl_openib_ini.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2008 Mellanox Technologies. All rights reserved. * Copyright (c) 2008 Mellanox Technologies. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -31,7 +30,7 @@
#include <unistd.h> #include <unistd.h>
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "btl_openib.h" #include "btl_openib.h"
@ -248,7 +247,7 @@ static int parse_file(char *filename)
ini_filename = filename; ini_filename = filename;
btl_openib_ini_yyin = fopen(filename, "r"); btl_openib_ini_yyin = fopen(filename, "r");
if (NULL == btl_openib_ini_yyin) { if (NULL == btl_openib_ini_yyin) {
orca_show_help("help-mpi-btl-openib.txt", "ini file:file not found", orte_show_help("help-mpi-btl-openib.txt", "ini file:file not found",
true, filename); true, filename);
ret = OMPI_ERR_NOT_FOUND; ret = OMPI_ERR_NOT_FOUND;
goto cleanup; goto cleanup;
@ -425,7 +424,7 @@ static int parse_line(parsed_section_values_t *sv)
/* Have no idea what this parameter is. Not an error -- just /* Have no idea what this parameter is. Not an error -- just
ignore it */ ignore it */
if (!showed_unknown_field_warning) { if (!showed_unknown_field_warning) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"ini file:unknown field", true, "ini file:unknown field", true,
ini_filename, btl_openib_ini_yynewlines, ini_filename, btl_openib_ini_yynewlines,
key_buffer); key_buffer);
@ -694,7 +693,7 @@ static inline void show_help(const char *topic)
if (0 == strcmp("\n", btl_openib_ini_yytext)) { if (0 == strcmp("\n", btl_openib_ini_yytext)) {
btl_openib_ini_yytext = "<end of line>"; btl_openib_ini_yytext = "<end of line>";
} }
orca_show_help("help-mpi-btl-openib.txt", topic, true, orte_show_help("help-mpi-btl-openib.txt", topic, true,
ini_filename, btl_openib_ini_yynewlines, ini_filename, btl_openib_ini_yynewlines,
btl_openib_ini_yytext); btl_openib_ini_yytext);
btl_openib_ini_yytext = save; btl_openib_ini_yytext = save;

20
ompi/mca/btl/openib/btl_openib_ip.c
Просмотреть файл

@ -2,8 +2,6 @@
* Copyright (c) 2008 Chelsio, Inc. All rights reserved. * Copyright (c) 2008 Chelsio, Inc. All rights reserved.
* Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
* *
@ -24,7 +22,7 @@
#include "opal/util/argv.h" #include "opal/util/argv.h"
#include "opal/util/if.h" #include "opal/util/if.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "connect/connect.h" #include "connect/connect.h"
#endif #endif
@ -199,9 +197,9 @@ static int ipaddr_specified(struct sockaddr_in *ipaddr, uint32_t netmask)
if (NULL == temp || NULL == temp[0] || NULL == temp[1] || if (NULL == temp || NULL == temp[0] || NULL == temp[1] ||
NULL != temp[2]) { NULL != temp[2]) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid ipaddr_inexclude", true, "include", "invalid ipaddr_inexclude", true, "include",
orca_process_info_get_nodename(), list[i], orte_process_info.nodename, list[i],
"Invalid specification (missing \"/\")"); "Invalid specification (missing \"/\")");
if (NULL != temp) { if (NULL != temp) {
opal_argv_free(temp); opal_argv_free(temp);
@ -210,9 +208,9 @@ static int ipaddr_specified(struct sockaddr_in *ipaddr, uint32_t netmask)
} }
if (1 != inet_pton(ipaddr->sin_family, temp[0], &ipae)) { if (1 != inet_pton(ipaddr->sin_family, temp[0], &ipae)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid ipaddr_inexclude", true, "include", "invalid ipaddr_inexclude", true, "include",
orca_process_info_get_nodename(), list[i], orte_process_info.nodename, list[i],
"Invalid specification (inet_pton() failed)"); "Invalid specification (inet_pton() failed)");
opal_argv_free(temp); opal_argv_free(temp);
continue; continue;
@ -241,9 +239,9 @@ static int ipaddr_specified(struct sockaddr_in *ipaddr, uint32_t netmask)
if (NULL == temp || NULL == temp[0] || NULL == temp[1] || if (NULL == temp || NULL == temp[0] || NULL == temp[1] ||
NULL != temp[2]) { NULL != temp[2]) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid ipaddr_inexclude", true, "exclude", "invalid ipaddr_inexclude", true, "exclude",
orca_process_info_get_nodename(), list[i], orte_process_info.nodename, list[i],
"Invalid specification (missing \"/\")"); "Invalid specification (missing \"/\")");
if (NULL != temp) { if (NULL != temp) {
opal_argv_free(temp); opal_argv_free(temp);
@ -252,9 +250,9 @@ static int ipaddr_specified(struct sockaddr_in *ipaddr, uint32_t netmask)
} }
if (1 != inet_pton(ipaddr->sin_family, temp[0], &ipae)) { if (1 != inet_pton(ipaddr->sin_family, temp[0], &ipae)) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"invalid ipaddr_inexclude", true, "exclude", "invalid ipaddr_inexclude", true, "exclude",
orca_process_info_get_nodename(), list[i], orte_process_info.nodename, list[i],
"Invalid specification (inet_pton() failed)"); "Invalid specification (inet_pton() failed)");
opal_argv_free(temp); opal_argv_free(temp);
continue; continue;

41
ompi/mca/btl/openib/btl_openib_mca.c
Просмотреть файл

@ -15,7 +15,6 @@
* reserved. * reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -31,7 +30,7 @@
#include "opal/mca/installdirs/installdirs.h" #include "opal/mca/installdirs/installdirs.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "btl_openib.h" #include "btl_openib.h"
#include "btl_openib_mca.h" #include "btl_openib_mca.h"
#include "btl_openib_ini.h" #include "btl_openib_ini.h"
@ -180,9 +179,9 @@ int btl_openib_register_mca_params(void)
mca_btl_openib_component.want_fork_support = ival; mca_btl_openib_component.want_fork_support = ival;
#else #else
if (0 != ival) { if (0 != ival) {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"ibv_fork requested but not supported", true, "ibv_fork requested but not supported", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
return OMPI_ERROR; return OMPI_ERROR;
} }
#endif #endif
@ -211,9 +210,9 @@ int btl_openib_register_mca_params(void)
} else if (0 == strcasecmp(str, "all")) { } else if (0 == strcasecmp(str, "all")) {
mca_btl_openib_component.device_type = BTL_OPENIB_DT_ALL; mca_btl_openib_component.device_type = BTL_OPENIB_DT_ALL;
} else { } else {
orca_show_help("help-mpi-btl-openib.txt", orte_show_help("help-mpi-btl-openib.txt",
"ibv_fork requested but not supported", true, "ibv_fork requested but not supported", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
return OMPI_ERROR; return OMPI_ERROR;
} }
free(str); free(str);
@ -298,7 +297,7 @@ int btl_openib_register_mca_params(void)
CHECK(reg_int("mtu", "ib_mtu", msg, IBV_MTU_1024, &ival, 0)); CHECK(reg_int("mtu", "ib_mtu", msg, IBV_MTU_1024, &ival, 0));
free(msg); free(msg);
if (ival < IBV_MTU_1024 || ival > IBV_MTU_4096) { if (ival < IBV_MTU_1024 || ival > IBV_MTU_4096) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "invalid value for btl_openib_ib_mtu", true, "invalid value for btl_openib_ib_mtu",
"btl_openib_ib_mtu reset to 1024"); "btl_openib_ib_mtu reset to 1024");
mca_btl_openib_component.ib_mtu = IBV_MTU_1024; mca_btl_openib_component.ib_mtu = IBV_MTU_1024;
@ -311,12 +310,12 @@ int btl_openib_register_mca_params(void)
"(must be >= 0 and <= 31)", "(must be >= 0 and <= 31)",
25, &ival, 0)); 25, &ival, 0));
if (ival > 31) { if (ival > 31) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_min_rnr_timer > 31", true, "btl_openib_ib_min_rnr_timer > 31",
"btl_openib_ib_min_rnr_timer reset to 31"); "btl_openib_ib_min_rnr_timer reset to 31");
ival = 31; ival = 31;
} else if (ival < 0){ } else if (ival < 0){
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_min_rnr_timer < 0", true, "btl_openib_ib_min_rnr_timer < 0",
"btl_openib_ib_min_rnr_timer reset to 0"); "btl_openib_ib_min_rnr_timer reset to 0");
ival = 0; ival = 0;
@ -328,12 +327,12 @@ int btl_openib_register_mca_params(void)
"(must be >= 0 and <= 31)", "(must be >= 0 and <= 31)",
20, &ival, 0)); 20, &ival, 0));
if (ival > 31) { if (ival > 31) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_timeout > 31", true, "btl_openib_ib_timeout > 31",
"btl_openib_ib_timeout reset to 31"); "btl_openib_ib_timeout reset to 31");
ival = 31; ival = 31;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_timeout < 0", true, "btl_openib_ib_timeout < 0",
"btl_openib_ib_timeout reset to 0"); "btl_openib_ib_timeout reset to 0");
ival = 0; ival = 0;
@ -345,12 +344,12 @@ int btl_openib_register_mca_params(void)
"(must be >= 0 and <= 7)", "(must be >= 0 and <= 7)",
7, &ival, 0)); 7, &ival, 0));
if (ival > 7) { if (ival > 7) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_retry_count > 7", true, "btl_openib_ib_retry_count > 7",
"btl_openib_ib_retry_count reset to 7"); "btl_openib_ib_retry_count reset to 7");
ival = 7; ival = 7;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_retry_count < 0", true, "btl_openib_ib_retry_count < 0",
"btl_openib_ib_retry_count reset to 0"); "btl_openib_ib_retry_count reset to 0");
ival = 0; ival = 0;
@ -365,12 +364,12 @@ int btl_openib_register_mca_params(void)
"(must be >= 0 and <= 7; 7 = \"infinite\")", "(must be >= 0 and <= 7; 7 = \"infinite\")",
7, &ival, 0)); 7, &ival, 0));
if (ival > 7) { if (ival > 7) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_rnr_retry > 7", true, "btl_openib_ib_rnr_retry > 7",
"btl_openib_ib_rnr_retry reset to 7"); "btl_openib_ib_rnr_retry reset to 7");
ival = 7; ival = 7;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_rnr_retry < 0", true, "btl_openib_ib_rnr_retry < 0",
"btl_openib_ib_rnr_retry reset to 0"); "btl_openib_ib_rnr_retry reset to 0");
ival = 0; ival = 0;
@ -387,12 +386,12 @@ int btl_openib_register_mca_params(void)
"(must be >= 0 and <= 15)", "(must be >= 0 and <= 15)",
0, &ival, 0)); 0, &ival, 0));
if (ival > 15) { if (ival > 15) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_service_level > 15", true, "btl_openib_ib_service_level > 15",
"btl_openib_ib_service_level reset to 15"); "btl_openib_ib_service_level reset to 15");
ival = 15; ival = 15;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_ib_service_level < 0", true, "btl_openib_ib_service_level < 0",
"btl_openib_ib_service_level reset to 0"); "btl_openib_ib_service_level reset to 0");
ival = 0; ival = 0;
@ -495,8 +494,8 @@ int btl_openib_register_mca_params(void)
"(must be > 0 and power of two)", "(must be > 0 and power of two)",
64, &ival, REGINT_GE_ZERO)); 64, &ival, REGINT_GE_ZERO));
if(ival <= 1 || (ival & (ival - 1))) { if(ival <= 1 || (ival & (ival - 1))) {
orca_show_help("help-mpi-btl-openib.txt", "wrong buffer alignment", orte_show_help("help-mpi-btl-openib.txt", "wrong buffer alignment",
true, ival, orca_process_info_get_nodename(), 64); true, ival, orte_process_info.nodename, 64);
mca_btl_openib_component.buffer_alignment = 64; mca_btl_openib_component.buffer_alignment = 64;
} else { } else {
mca_btl_openib_component.buffer_alignment = (uint32_t) ival; mca_btl_openib_component.buffer_alignment = (uint32_t) ival;
@ -619,7 +618,7 @@ int btl_openib_register_mca_params(void)
if (mca_btl_openib_component.use_memalign != 32 if (mca_btl_openib_component.use_memalign != 32
&& mca_btl_openib_component.use_memalign != 64 && mca_btl_openib_component.use_memalign != 64
&& mca_btl_openib_component.use_memalign != 0){ && mca_btl_openib_component.use_memalign != 0){
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "Wrong btl_openib_memalign parameter value. Allowed values: 64, 32, 0.", true, "Wrong btl_openib_memalign parameter value. Allowed values: 64, 32, 0.",
"btl_openib_memalign is reset to 32"); "btl_openib_memalign is reset to 32");
mca_btl_openib_component.use_memalign = 32; mca_btl_openib_component.use_memalign = 32;
@ -632,7 +631,7 @@ int btl_openib_register_mca_params(void)
&ival, &ival,
REGINT_GE_ZERO); REGINT_GE_ZERO);
if (ival < 0){ if (ival < 0){
orca_show_help("help-mpi-btl-openib.txt", "invalid mca param value", orte_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
true, "btl_openib_memalign_threshold must be positive", true, "btl_openib_memalign_threshold must be positive",
"btl_openib_memalign_threshold is reset to btl_openib_eager_limit"); "btl_openib_memalign_threshold is reset to btl_openib_eager_limit");
ival = mca_btl_openib_component.eager_limit; ival = mca_btl_openib_component.eager_limit;

3
ompi/mca/btl/openib/btl_openib_proc.c
Просмотреть файл

@ -11,7 +11,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -155,7 +154,7 @@ mca_btl_openib_proc_t* mca_btl_openib_proc_create(ompi_proc_t* ompi_proc)
if (OMPI_SUCCESS != rc) { if (OMPI_SUCCESS != rc) {
BTL_ERROR(("[%s:%d] ompi_modex_recv failed for peer %s", BTL_ERROR(("[%s:%d] ompi_modex_recv failed for peer %s",
__FILE__, __LINE__, __FILE__, __LINE__,
ORCA_NAME_PRINT(&ompi_proc->proc_name))); ORTE_NAME_PRINT(&ompi_proc->proc_name)));
OBJ_RELEASE(module_proc); OBJ_RELEASE(module_proc);
return NULL; return NULL;
} }

7
ompi/mca/btl/openib/btl_openib_xrc.c
Просмотреть файл

@ -1,7 +1,6 @@
/* /*
* Copyright (c) 2007-2008 Mellanox Technologies. All rights reserved. * Copyright (c) 2007-2008 Mellanox Technologies. All rights reserved.
* Copyright (c) 2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -45,7 +44,7 @@ int mca_btl_openib_open_xrc_domain(struct mca_btl_openib_device_t *device)
dev_name = ibv_get_device_name(device->ib_dev); dev_name = ibv_get_device_name(device->ib_dev);
len = asprintf(&xrc_file_name, len = asprintf(&xrc_file_name,
"%s"OPAL_PATH_SEP"openib_xrc_domain_%s", "%s"OPAL_PATH_SEP"openib_xrc_domain_%s",
orca_process_info_get_job_session_dir(), dev_name); orte_process_info.job_session_dir, dev_name);
if (0 > len) { if (0 > len) {
BTL_ERROR(("Failed to allocate memomry for XRC file name: %s\n", BTL_ERROR(("Failed to allocate memomry for XRC file name: %s\n",
strerror(errno))); strerror(errno)));
@ -112,7 +111,7 @@ static void ib_address_destructor(ib_address_t *ib_addr)
OBJ_DESTRUCT(&ib_addr->pending_ep); OBJ_DESTRUCT(&ib_addr->pending_ep);
} }
static int ib_address_init(ib_address_t *ib_addr, uint16_t lid, uint64_t s_id, orca_jobid_t ep_jobid) static int ib_address_init(ib_address_t *ib_addr, uint16_t lid, uint64_t s_id, orte_jobid_t ep_jobid)
{ {
ib_addr->key = malloc(SIZE_OF3(s_id, lid, ep_jobid)); ib_addr->key = malloc(SIZE_OF3(s_id, lid, ep_jobid));
if (NULL == ib_addr->key) { if (NULL == ib_addr->key) {
@ -137,7 +136,7 @@ static int ib_address_init(ib_address_t *ib_addr, uint16_t lid, uint64_t s_id, o
* Before call to this function you need to protect with * Before call to this function you need to protect with
*/ */
int mca_btl_openib_ib_address_add_new (uint16_t lid, uint64_t s_id, int mca_btl_openib_ib_address_add_new (uint16_t lid, uint64_t s_id,
orca_jobid_t ep_jobid, mca_btl_openib_endpoint_t *ep) orte_jobid_t ep_jobid, mca_btl_openib_endpoint_t *ep)
{ {
void *tmp; void *tmp;
int ret = OMPI_SUCCESS; int ret = OMPI_SUCCESS;

3
ompi/mca/btl/openib/btl_openib_xrc.h
Просмотреть файл

@ -1,6 +1,5 @@
/* /*
* Copyright (c) 2007-2008 Mellanox Technologies. All rights reserved. * Copyright (c) 2007-2008 Mellanox Technologies. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -45,6 +44,6 @@ typedef struct ib_address_t ib_address_t;
int mca_btl_openib_open_xrc_domain(struct mca_btl_openib_device_t *device); int mca_btl_openib_open_xrc_domain(struct mca_btl_openib_device_t *device);
int mca_btl_openib_close_xrc_domain(struct mca_btl_openib_device_t *device); int mca_btl_openib_close_xrc_domain(struct mca_btl_openib_device_t *device);
int mca_btl_openib_ib_address_add_new (uint16_t lid, uint64_t s_id, int mca_btl_openib_ib_address_add_new (uint16_t lid, uint64_t s_id,
orca_jobid_t ep_jobid, mca_btl_openib_endpoint_t *ep); orte_jobid_t ep_jobid, mca_btl_openib_endpoint_t *ep);
#endif #endif

16
ompi/mca/btl/openib/connect/btl_openib_connect_base.c
Просмотреть файл

@ -3,7 +3,7 @@
* Copyright (c) 2007 Mellanox Technologies, Inc. All rights reserved. * Copyright (c) 2007 Mellanox Technologies, Inc. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -27,7 +27,7 @@
#include "connect/btl_openib_connect_udcm.h" #include "connect/btl_openib_connect_udcm.h"
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/util/argv.h" #include "opal/util/argv.h"
#include "opal/util/output.h" #include "opal/util/output.h"
@ -121,9 +121,9 @@ int ompi_btl_openib_connect_base_register(void)
} }
} }
if (NULL == all[i]) { if (NULL == all[i]) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"cpc name not found", true, "cpc name not found", true,
"include", orca_process_info_get_nodename(), "include", orte_process_info.nodename,
"include", cpc_include, temp[j], "include", cpc_include, temp[j],
all_cpc_names); all_cpc_names);
opal_argv_free(temp); opal_argv_free(temp);
@ -147,9 +147,9 @@ int ompi_btl_openib_connect_base_register(void)
} }
} }
if (NULL == all[i]) { if (NULL == all[i]) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"cpc name not found", true, "cpc name not found", true,
"exclude", orca_process_info_get_nodename(), "exclude", orte_process_info.nodename,
"exclude", cpc_exclude, temp[j], "exclude", cpc_exclude, temp[j],
all_cpc_names); all_cpc_names);
opal_argv_free(temp); opal_argv_free(temp);
@ -292,9 +292,9 @@ int ompi_btl_openib_connect_base_select_for_local_port(mca_btl_openib_module_t *
/* If we got an empty array, then no CPCs were eligible. Doh! */ /* If we got an empty array, then no CPCs were eligible. Doh! */
if (0 == cpc_index) { if (0 == cpc_index) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"no cpcs for port", true, "no cpcs for port", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(btl->device->ib_dev), ibv_get_device_name(btl->device->ib_dev),
btl->port_num, msg); btl->port_num, msg);
free(cpcs); free(cpcs);

120
ompi/mca/btl/openib/connect/btl_openib_connect_oob.c
Просмотреть файл

@ -15,7 +15,7 @@
* Copyright (c) 2008-2011 Mellanox Technologies. All rights reserved. * Copyright (c) 2008-2011 Mellanox Technologies. All rights reserved.
* Copyright (c) 2009-2011 IBM Corporation. All rights reserved. * Copyright (c) 2009-2011 IBM Corporation. All rights reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -27,11 +27,14 @@
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "opal_stdint.h" #include "opal_stdint.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/util/error.h" #include "opal/util/error.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orte/mca/rml/rml.h"
#include "orte/mca/rml/rml_types.h"
#include "orte/mca/errmgr/errmgr.h"
#include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
#include "btl_openib.h" #include "btl_openib.h"
@ -74,11 +77,11 @@ static int qp_create_one(mca_btl_base_endpoint_t* endpoint, int qp,
static int send_connect_data(mca_btl_base_endpoint_t* endpoint, static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
uint8_t message_type); uint8_t message_type);
static void rml_send_cb(int status, orca_process_name_t* endpoint, static void rml_send_cb(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
static void rml_recv_cb(int status, orca_process_name_t* process_name, static void rml_recv_cb(int status, orte_process_name_t* process_name,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
/* /*
@ -146,12 +149,12 @@ static int oob_component_query(mca_btl_openib_module_t *btl,
ensure to only post it *once*, because another btl may have ensure to only post it *once*, because another btl may have
come in before this and already posted it. */ come in before this and already posted it. */
if (!rml_recv_posted) { if (!rml_recv_posted) {
rc = orca_oob_recv_buffer_nb(ORCA_NAME_WILDCARD, rc = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
OMPI_RML_TAG_OPENIB, OMPI_RML_TAG_OPENIB,
ORCA_OOB_PERSISTENT, ORTE_RML_PERSISTENT,
rml_recv_cb, rml_recv_cb,
NULL); NULL);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
opal_output_verbose(5, mca_btl_base_output, opal_output_verbose(5, mca_btl_base_output,
"openib BTL: oob CPC system error %d (%s)", "openib BTL: oob CPC system error %d (%s)",
rc, opal_strerror(rc)); rc, opal_strerror(rc));
@ -162,7 +165,7 @@ static int oob_component_query(mca_btl_openib_module_t *btl,
*cpc = (ompi_btl_openib_connect_base_module_t *) malloc(sizeof(ompi_btl_openib_connect_base_module_t)); *cpc = (ompi_btl_openib_connect_base_module_t *) malloc(sizeof(ompi_btl_openib_connect_base_module_t));
if (NULL == *cpc) { if (NULL == *cpc) {
orca_oob_recv_cancel(ORCA_NAME_WILDCARD, OMPI_RML_TAG_OPENIB); orte_rml.recv_cancel(ORTE_NAME_WILDCARD, OMPI_RML_TAG_OPENIB);
rml_recv_posted = false; rml_recv_posted = false;
opal_output_verbose(5, mca_btl_base_output, opal_output_verbose(5, mca_btl_base_output,
"openib BTL: oob CPC system error (malloc failed)"); "openib BTL: oob CPC system error (malloc failed)");
@ -218,7 +221,7 @@ static int oob_module_start_connect(ompi_btl_openib_connect_base_module_t *cpc,
static int oob_component_finalize(void) static int oob_component_finalize(void)
{ {
if (rml_recv_posted) { if (rml_recv_posted) {
orca_oob_recv_cancel(ORCA_NAME_WILDCARD, OMPI_RML_TAG_OPENIB); orte_rml.recv_cancel(ORTE_NAME_WILDCARD, OMPI_RML_TAG_OPENIB);
rml_recv_posted = false; rml_recv_posted = false;
} }
#if (ENABLE_DYNAMIC_SL) #if (ENABLE_DYNAMIC_SL)
@ -483,9 +486,9 @@ static int qp_create_one(mca_btl_base_endpoint_t* endpoint, int qp,
my_qp = ibv_create_qp(openib_btl->device->ib_pd, &init_attr); my_qp = ibv_create_qp(openib_btl->device->ib_pd, &init_attr);
if (NULL == my_qp) { if (NULL == my_qp) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"ibv_create_qp failed", true, "ibv_create_qp failed", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
"Reliable connected (RC)"); "Reliable connected (RC)");
return OMPI_ERROR; return OMPI_ERROR;
@ -494,8 +497,8 @@ static int qp_create_one(mca_btl_base_endpoint_t* endpoint, int qp,
if (init_attr.cap.max_inline_data < req_inline) { if (init_attr.cap.max_inline_data < req_inline) {
endpoint->qps[qp].ib_inline_max = init_attr.cap.max_inline_data; endpoint->qps[qp].ib_inline_max = init_attr.cap.max_inline_data;
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"inline truncated", true, orca_process_info_get_nodename(), "inline truncated", true, orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
openib_btl->port_num, openib_btl->port_num,
req_inline, init_attr.cap.max_inline_data); req_inline, init_attr.cap.max_inline_data);
@ -536,7 +539,7 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
int rc; int rc;
if (NULL == buffer) { if (NULL == buffer) {
ORCA_ERROR_LOG(ORCA_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -544,14 +547,14 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT8)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT8));
rc = opal_dss.pack(buffer, &message_type, 1, OPAL_UINT8); rc = opal_dss.pack(buffer, &message_type, 1, OPAL_UINT8);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT64)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT64));
rc = opal_dss.pack(buffer, &endpoint->subnet_id, 1, OPAL_UINT64); rc = opal_dss.pack(buffer, &endpoint->subnet_id, 1, OPAL_UINT64);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -562,13 +565,13 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
&endpoint->rem_info.rem_qps[0].rem_qp_num, 1, &endpoint->rem_info.rem_qps[0].rem_qp_num, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16));
rc = opal_dss.pack(buffer, &endpoint->rem_info.rem_lid, 1, OPAL_UINT16); rc = opal_dss.pack(buffer, &endpoint->rem_info.rem_lid, 1, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -581,14 +584,14 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_qp->qp_num, rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_qp->qp_num,
1, OPAL_UINT32); 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_psn, 1, rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_psn, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -596,30 +599,30 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16));
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->lid, 1, OPAL_UINT16); rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->lid, 1, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->device->mtu, 1, rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->device->mtu, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->index, 1, OPAL_UINT32); rc = opal_dss.pack(buffer, &endpoint->index, 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
/* send to remote endpoint */ /* send to remote endpoint */
rc = orca_oob_send_buffer_nb(&endpoint->endpoint_proc->proc_ompi->proc_name, rc = orte_rml.send_buffer_nb(&endpoint->endpoint_proc->proc_ompi->proc_name,
buffer, OMPI_RML_TAG_OPENIB, 0, buffer, OMPI_RML_TAG_OPENIB, 0,
rml_send_cb, NULL); rml_send_cb, NULL);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("Sent QP Info, LID = %d, SUBNET = %" PRIx64 "\n", BTL_VERBOSE(("Sent QP Info, LID = %d, SUBNET = %" PRIx64 "\n",
@ -634,8 +637,8 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
* Callback when we have finished RML sending the connect data to a * Callback when we have finished RML sending the connect data to a
* remote peer * remote peer
*/ */
static void rml_send_cb(int status, orca_process_name_t* endpoint, static void rml_send_cb(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
OBJ_RELEASE(buffer); OBJ_RELEASE(buffer);
@ -647,8 +650,8 @@ static void rml_send_cb(int status, orca_process_name_t* endpoint,
* and if this endpoint is trying to connect, reply with our QP info, * and if this endpoint is trying to connect, reply with our QP info,
* otherwise try to modify QP's and establish reliable connection * otherwise try to modify QP's and establish reliable connection
*/ */
static void rml_recv_cb(int status, orca_process_name_t* process_name, static void rml_recv_cb(int status, orte_process_name_t* process_name,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
mca_btl_openib_proc_t *ib_proc; mca_btl_openib_proc_t *ib_proc;
@ -671,16 +674,16 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
our door */ our door */
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT8)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT8));
rc = opal_dss.unpack(buffer, &message_type, &cnt, OPAL_UINT8); rc = opal_dss.unpack(buffer, &message_type, &cnt, OPAL_UINT8);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT64)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT64));
rc = opal_dss.unpack(buffer, &rem_info.rem_subnet_id, &cnt, OPAL_UINT64); rc = opal_dss.unpack(buffer, &rem_info.rem_subnet_id, &cnt, OPAL_UINT64);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
@ -688,15 +691,15 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
if (ENDPOINT_CONNECT_REQUEST != message_type) { if (ENDPOINT_CONNECT_REQUEST != message_type) {
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &lcl_qp, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &lcl_qp, &cnt, OPAL_UINT32);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16));
rc = opal_dss.unpack(buffer, &lcl_lid, &cnt, OPAL_UINT16); rc = opal_dss.unpack(buffer, &lcl_lid, &cnt, OPAL_UINT16);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
@ -713,16 +716,16 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_qp_num, &cnt, rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_qp_num, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_psn, &cnt, rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_psn, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
@ -730,22 +733,22 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16));
rc = opal_dss.unpack(buffer, &rem_info.rem_lid, &cnt, OPAL_UINT16); rc = opal_dss.unpack(buffer, &rem_info.rem_lid, &cnt, OPAL_UINT16);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &rem_info.rem_mtu, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &rem_info.rem_mtu, &cnt, OPAL_UINT32);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &rem_info.rem_index, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &rem_info.rem_index, &cnt, OPAL_UINT32);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_openib_endpoint_invoke_error(NULL); mca_btl_openib_endpoint_invoke_error(NULL);
return; return;
} }
@ -755,9 +758,8 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
rem_info.rem_lid, rem_info.rem_lid,
rem_info.rem_subnet_id)); rem_info.rem_subnet_id));
master = orca_process_name_compare(ORCA_NAME_CMP_ALL, master = orte_util_compare_name_fields(ORTE_NS_CMP_ALL, ORTE_PROC_MY_NAME,
ORCA_PROC_MY_NAME, process_name) > 0 ? true : false;
process_name) > 0 ? true : false;
/* Need to protect the ib_procs list */ /* Need to protect the ib_procs list */
OPAL_THREAD_LOCK(&mca_btl_openib_component.ib_lock); OPAL_THREAD_LOCK(&mca_btl_openib_component.ib_lock);
@ -769,7 +771,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
ib_proc = (mca_btl_openib_proc_t*)opal_list_get_next(ib_proc)) { ib_proc = (mca_btl_openib_proc_t*)opal_list_get_next(ib_proc)) {
bool found = false; bool found = false;
if (OPAL_EQUAL != orca_process_name_compare(ORCA_NAME_CMP_ALL, if (OPAL_EQUAL != orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&ib_proc->proc_ompi->proc_name, process_name)) { &ib_proc->proc_ompi->proc_name, process_name)) {
continue; continue;
} }

30
ompi/mca/btl/openib/connect/btl_openib_connect_rdmacm.c
Просмотреть файл

@ -6,7 +6,7 @@
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -48,7 +48,7 @@
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/util/error.h" #include "opal/util/error.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "btl_openib_fd.h" #include "btl_openib_fd.h"
#include "btl_openib_proc.h" #include "btl_openib_proc.h"
@ -247,7 +247,7 @@ static void rdmacm_component_register(void)
if (value >= 0 && value < 65536) { if (value >= 0 && value < 65536) {
rdmacm_port = (uint16_t) value; rdmacm_port = (uint16_t) value;
} else { } else {
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"illegal tcp port", true, value); "illegal tcp port", true, value);
} }
@ -258,7 +258,7 @@ static void rdmacm_component_register(void)
if (value > 0) { if (value > 0) {
rdmacm_resolve_timeout = value; rdmacm_resolve_timeout = value;
} else { } else {
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"illegal timeout", true, value); "illegal timeout", true, value);
} }
@ -269,7 +269,7 @@ static void rdmacm_component_register(void)
if (value > 0) { if (value > 0) {
rdmacm_resolve_max_retry_count = value; rdmacm_resolve_max_retry_count = value;
} else { } else {
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"illegal retry count", true, value); "illegal retry count", true, value);
} }
@ -453,9 +453,9 @@ static int rdmacm_setup_qp(rdmacm_contents_t *contents,
endpoint->qps[qpnum].credit_frag = NULL; endpoint->qps[qpnum].credit_frag = NULL;
if (attr.cap.max_inline_data < req_inline) { if (attr.cap.max_inline_data < req_inline) {
endpoint->qps[qpnum].ib_inline_max = attr.cap.max_inline_data; endpoint->qps[qpnum].ib_inline_max = attr.cap.max_inline_data;
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"inline truncated", true, "inline truncated", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(contents->openib_btl->device->ib_dev), ibv_get_device_name(contents->openib_btl->device->ib_dev),
contents->openib_btl->port_num, contents->openib_btl->port_num,
req_inline, attr.cap.max_inline_data); req_inline, attr.cap.max_inline_data);
@ -753,16 +753,16 @@ static void *show_help_cant_find_endpoint(void *context)
if (NULL != c) { if (NULL != c) {
msg = stringify(c->peer_ip_addr); msg = stringify(c->peer_ip_addr);
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"could not find matching endpoint", true, "could not find matching endpoint", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
c->device_name, c->device_name,
c->peer_tcp_port); c->peer_tcp_port);
free(msg); free(msg);
} else { } else {
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"could not find matching endpoint", true, "could not find matching endpoint", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
"<unknown>", "<unknown>", -1); "<unknown>", "<unknown>", -1);
} }
free(context); free(context);
@ -1463,9 +1463,9 @@ static void *show_help_rdmacm_event_error(void *c)
id_context_t *context = (id_context_t*) event->id->context; id_context_t *context = (id_context_t*) event->id->context;
if (RDMA_CM_EVENT_DEVICE_REMOVAL == event->event) { if (RDMA_CM_EVENT_DEVICE_REMOVAL == event->event) {
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"rdma cm device removal", true, "rdma cm device removal", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(event->id->verbs->device)); ibv_get_device_name(event->id->verbs->device));
} else { } else {
const char *device = "Unknown"; const char *device = "Unknown";
@ -1474,9 +1474,9 @@ static void *show_help_rdmacm_event_error(void *c)
NULL != event->id->verbs->device) { NULL != event->id->verbs->device) {
device = ibv_get_device_name(event->id->verbs->device); device = ibv_get_device_name(event->id->verbs->device);
} }
orca_show_help("help-mpi-btl-openib-cpc-rdmacm.txt", orte_show_help("help-mpi-btl-openib-cpc-rdmacm.txt",
"rdma cm event error", true, "rdma cm event error", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
device, device,
rdma_event_str(event->event), rdma_event_str(event->event),
context->endpoint->endpoint_proc->proc_ompi->proc_hostname); context->endpoint->endpoint_proc->proc_ompi->proc_hostname);

8
ompi/mca/btl/openib/connect/btl_openib_connect_sl.c
Просмотреть файл

@ -1,6 +1,6 @@
/* /*
* Copyright (c) 2011 Mellanox Technologies. All rights reserved. * Copyright (c) 2011 Mellanox Technologies. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -9,7 +9,7 @@
*/ */
#include "btl_openib.h" #include "btl_openib.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "connect/btl_openib_connect_sl.h" #include "connect/btl_openib_connect_sl.h"
#include <infiniband/iba/ib_types.h> #include <infiniband/iba/ib_types.h>
@ -108,8 +108,8 @@ static int init_ud_qp(struct ibv_context *context_arg,
cache->cq = ibv_create_cq(cache->context, 4, NULL, NULL, 0); cache->cq = ibv_create_cq(cache->context, 4, NULL, NULL, 0);
if (NULL == cache->cq) { if (NULL == cache->cq) {
BTL_ERROR(("error creating cq, errno says %s", strerror(errno))); BTL_ERROR(("error creating cq, errno says %s", strerror(errno)));
orca_show_help("help-mpi-btl-openib.txt", "init-fail-create-q", orte_show_help("help-mpi-btl-openib.txt", "init-fail-create-q",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
__FILE__, __LINE__, "ibv_create_cq", __FILE__, __LINE__, "ibv_create_cq",
strerror(errno), errno, strerror(errno), errno,
ibv_get_device_name(context_arg->device)); ibv_get_device_name(context_arg->device));

20
ompi/mca/btl/openib/connect/btl_openib_connect_udcm.c
Просмотреть файл

@ -5,7 +5,7 @@
* Copyright (c) 2009 IBM Corporation. All rights reserved. * Copyright (c) 2009 IBM Corporation. All rights reserved.
* Copyright (c) 2011 Los Alamos National Security, LLC. All * Copyright (c) 2011 Los Alamos National Security, LLC. All
* rights reserved. * rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -70,7 +70,8 @@
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/util/error.h" #include "opal/util/error.h"
#include "opal_stdint.h" #include "opal_stdint.h"
#include "orca/include/rte_orca.h" #include "orte/mca/ess/ess.h"
#include "orte/util/show_help.h"
#include "btl_openib_endpoint.h" #include "btl_openib_endpoint.h"
#include "btl_openib_proc.h" #include "btl_openib_proc.h"
@ -1051,9 +1052,8 @@ static int udcm_create_sync_qp (mca_btl_base_endpoint_t *lcl_ep)
udep->sync_qp = ibv_create_qp(m->btl->device->ib_pd, &init_attr); udep->sync_qp = ibv_create_qp(m->btl->device->ib_pd, &init_attr);
if (NULL == udep->sync_qp) { if (NULL == udep->sync_qp) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"ibv_create_qp failed", true, "ibv_create_qp failed", true, orte_process_info.nodename,
orca_process_info_get_nodename(),
ibv_get_device_name(m->btl->device->ib_dev), ibv_get_device_name(m->btl->device->ib_dev),
"Reliable connected (RC)"); "Reliable connected (RC)");
@ -1166,9 +1166,8 @@ static int udcm_qp_create_one(udcm_module_t *m, mca_btl_base_endpoint_t* lcl_ep,
lcl_ep->qps[qp].qp->lcl_qp = ibv_create_qp(m->btl->device->ib_pd, lcl_ep->qps[qp].qp->lcl_qp = ibv_create_qp(m->btl->device->ib_pd,
&init_attr); &init_attr);
if (NULL == lcl_ep->qps[qp].qp->lcl_qp) { if (NULL == lcl_ep->qps[qp].qp->lcl_qp) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"ibv_create_qp failed", true, "ibv_create_qp failed", true, orte_process_info.nodename,
orca_process_info_get_nodename(),
ibv_get_device_name(m->btl->device->ib_dev), ibv_get_device_name(m->btl->device->ib_dev),
"Reliable connected (RC)"); "Reliable connected (RC)");
@ -1177,9 +1176,8 @@ static int udcm_qp_create_one(udcm_module_t *m, mca_btl_base_endpoint_t* lcl_ep,
if (init_attr.cap.max_inline_data < req_inline) { if (init_attr.cap.max_inline_data < req_inline) {
lcl_ep->qps[qp].ib_inline_max = init_attr.cap.max_inline_data; lcl_ep->qps[qp].ib_inline_max = init_attr.cap.max_inline_data;
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"inline truncated", true, "inline truncated", true, orte_process_info.nodename,
orca_process_info_get_nodename(),
ibv_get_device_name(m->btl->device->ib_dev), ibv_get_device_name(m->btl->device->ib_dev),
m->btl->port_num, req_inline, m->btl->port_num, req_inline,
init_attr.cap.max_inline_data); init_attr.cap.max_inline_data);

95
ompi/mca/btl/openib/connect/btl_openib_connect_xoob.c
Просмотреть файл

@ -7,7 +7,7 @@
* reserved. * reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -21,9 +21,11 @@
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "opal/util/error.h" #include "opal/util/error.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h" #include "orte/util/name_fns.h"
#include "orte/mca/rml/rml.h"
#include "orte/mca/rml/rml_types.h"
#include "orte/mca/errmgr/errmgr.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
#include "btl_openib.h" #include "btl_openib.h"
@ -32,6 +34,7 @@
#include "btl_openib_xrc.h" #include "btl_openib_xrc.h"
#include "btl_openib_async.h" #include "btl_openib_async.h"
#include "connect/connect.h" #include "connect/connect.h"
#include "orte/util/show_help.h"
#if (ENABLE_DYNAMIC_SL) #if (ENABLE_DYNAMIC_SL)
#include "connect/btl_openib_connect_sl.h" #include "connect/btl_openib_connect_sl.h"
#endif #endif
@ -91,9 +94,9 @@ static int xoob_priority = 60;
* Callback when we have finished RML sending the connect data to a * Callback when we have finished RML sending the connect data to a
* remote peer * remote peer
*/ */
static void xoob_rml_send_cb(int status, orca_process_name_t* endpoint, static void xoob_rml_send_cb(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
OBJ_RELEASE(buffer); OBJ_RELEASE(buffer);
} }
@ -108,7 +111,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT8)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT8));
rc = opal_dss.unpack(buffer, message_type, &cnt, OPAL_UINT8); rc = opal_dss.unpack(buffer, message_type, &cnt, OPAL_UINT8);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack Message type = %d\n", *message_type)); BTL_VERBOSE(("Recv unpack Message type = %d\n", *message_type));
@ -116,7 +119,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT64)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT64));
rc = opal_dss.unpack(buffer, &info->rem_subnet_id, &cnt, OPAL_UINT64); rc = opal_dss.unpack(buffer, &info->rem_subnet_id, &cnt, OPAL_UINT64);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack sid = %" PRIx64 "\n", info->rem_subnet_id)); BTL_VERBOSE(("Recv unpack sid = %" PRIx64 "\n", info->rem_subnet_id));
@ -124,7 +127,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16));
rc = opal_dss.unpack(buffer, &info->rem_lid, &cnt, OPAL_UINT16); rc = opal_dss.unpack(buffer, &info->rem_lid, &cnt, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack lid = %d", info->rem_lid)); BTL_VERBOSE(("Recv unpack lid = %d", info->rem_lid));
@ -138,7 +141,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
rc = opal_dss.unpack(buffer, &info->rem_qps->rem_qp_num, &cnt, rc = opal_dss.unpack(buffer, &info->rem_qps->rem_qp_num, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack remote qp = %x", info->rem_qps->rem_qp_num)); BTL_VERBOSE(("Recv unpack remote qp = %x", info->rem_qps->rem_qp_num));
@ -147,7 +150,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
rc = opal_dss.unpack(buffer, &info->rem_qps->rem_psn, &cnt, rc = opal_dss.unpack(buffer, &info->rem_qps->rem_psn, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack remote psn = %d", info->rem_qps->rem_psn)); BTL_VERBOSE(("Recv unpack remote psn = %d", info->rem_qps->rem_psn));
@ -155,7 +158,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &info->rem_mtu, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &info->rem_mtu, &cnt, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack remote mtu = %d", info->rem_mtu)); BTL_VERBOSE(("Recv unpack remote mtu = %d", info->rem_mtu));
@ -167,7 +170,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16));
rc = opal_dss.unpack(buffer, lid, &cnt, OPAL_UINT16); rc = opal_dss.unpack(buffer, lid, &cnt, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack requested lid = %d", *lid)); BTL_VERBOSE(("Recv unpack requested lid = %d", *lid));
@ -180,7 +183,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
rc = opal_dss.unpack(buffer, &info->rem_qps->rem_qp_num, &cnt, rc = opal_dss.unpack(buffer, &info->rem_qps->rem_qp_num, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("Recv unpack requested qp = %x", info->rem_qps->rem_qp_num)); BTL_VERBOSE(("Recv unpack requested qp = %x", info->rem_qps->rem_qp_num));
@ -191,7 +194,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &info->rem_index, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &info->rem_index, &cnt, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack remote index = %d", info->rem_index)); BTL_VERBOSE(("Recv unpack remote index = %d", info->rem_index));
@ -200,7 +203,7 @@ static int xoob_receive_connect_data(mca_btl_openib_rem_info_t *info, uint16_t *
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &info->rem_srqs[srq].rem_srq_num, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &info->rem_srqs[srq].rem_srq_num, &cnt, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return OMPI_ERROR; return OMPI_ERROR;
} }
BTL_VERBOSE(("Recv unpack remote index srq num[%d]= %d", srq, info->rem_srqs[srq].rem_srq_num)); BTL_VERBOSE(("Recv unpack remote index srq num[%d]= %d", srq, info->rem_srqs[srq].rem_srq_num));
@ -219,7 +222,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
int rc, srq; int rc, srq;
if (NULL == buffer) { if (NULL == buffer) {
ORCA_ERROR_LOG(ORCA_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -233,7 +236,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT8)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT8));
rc = opal_dss.pack(buffer, &message_type, 1, OPAL_UINT8); rc = opal_dss.pack(buffer, &message_type, 1, OPAL_UINT8);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -241,7 +244,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT64)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT64));
rc = opal_dss.pack(buffer, &endpoint->subnet_id, 1, OPAL_UINT64); rc = opal_dss.pack(buffer, &endpoint->subnet_id, 1, OPAL_UINT64);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -249,7 +252,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16));
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->lid, 1, OPAL_UINT16); rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->lid, 1, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -276,14 +279,14 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &qp_num, 1, OPAL_UINT32); rc = opal_dss.pack(buffer, &qp_num, 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("Send pack lpsn = %d", psn)); BTL_VERBOSE(("Send pack lpsn = %d", psn));
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &psn, 1, OPAL_UINT32); rc = opal_dss.pack(buffer, &psn, 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -292,7 +295,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->device->mtu, 1, rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->device->mtu, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -309,7 +312,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16));
rc = opal_dss.pack(buffer, &endpoint->ib_addr->lid, 1, OPAL_UINT16); rc = opal_dss.pack(buffer, &endpoint->ib_addr->lid, 1, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -322,7 +325,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
rc = opal_dss.pack(buffer, &endpoint->ib_addr->remote_xrc_rcv_qp_num, rc = opal_dss.pack(buffer, &endpoint->ib_addr->remote_xrc_rcv_qp_num,
1, OPAL_UINT32); 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -338,7 +341,7 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->index, 1, OPAL_UINT32); rc = opal_dss.pack(buffer, &endpoint->index, 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
/* on response we add all SRQ numbers */ /* on response we add all SRQ numbers */
@ -348,18 +351,18 @@ static int xoob_send_connect_data(mca_btl_base_endpoint_t* endpoint,
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->qps[srq].u.srq_qp.srq->xrc_srq_num, rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->qps[srq].u.srq_qp.srq->xrc_srq_num,
1, OPAL_UINT32); 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
} }
/* send to remote endpoint */ /* send to remote endpoint */
rc = orca_oob_send_buffer_nb(&endpoint->endpoint_proc->proc_ompi->proc_name, rc = orte_rml.send_buffer_nb(&endpoint->endpoint_proc->proc_ompi->proc_name,
buffer, OMPI_RML_TAG_XOPENIB, 0, buffer, OMPI_RML_TAG_XOPENIB, 0,
xoob_rml_send_cb, NULL); xoob_rml_send_cb, NULL);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -411,9 +414,9 @@ static int xoob_send_qp_create (mca_btl_base_endpoint_t* endpoint)
qp_init_attr.xrc_domain = openib_btl->device->xrc_domain; qp_init_attr.xrc_domain = openib_btl->device->xrc_domain;
*qp = ibv_create_qp(openib_btl->device->ib_pd, &qp_init_attr); *qp = ibv_create_qp(openib_btl->device->ib_pd, &qp_init_attr);
if (NULL == *qp) { if (NULL == *qp) {
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"ibv_create_qp failed", true, "ibv_create_qp failed", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
"Reliable connected (XRC)"); "Reliable connected (XRC)");
return OMPI_ERROR; return OMPI_ERROR;
@ -421,8 +424,8 @@ static int xoob_send_qp_create (mca_btl_base_endpoint_t* endpoint)
if (qp_init_attr.cap.max_inline_data < req_inline) { if (qp_init_attr.cap.max_inline_data < req_inline) {
endpoint->qps[0].ib_inline_max = qp_init_attr.cap.max_inline_data; endpoint->qps[0].ib_inline_max = qp_init_attr.cap.max_inline_data;
orca_show_help("help-mpi-btl-openib-cpc-base.txt", orte_show_help("help-mpi-btl-openib-cpc-base.txt",
"inline truncated", orca_process_info_get_nodename, "inline truncated", orte_process_info.nodename,
ibv_get_device_name(openib_btl->device->ib_dev), ibv_get_device_name(openib_btl->device->ib_dev),
openib_btl->port_num, openib_btl->port_num,
req_inline, qp_init_attr.cap.max_inline_data); req_inline, qp_init_attr.cap.max_inline_data);
@ -686,7 +689,7 @@ static int xoob_reply_first_connect(mca_btl_openib_endpoint_t *endpoint,
} }
/* Find endpoint for specific subnet/lid/message */ /* Find endpoint for specific subnet/lid/message */
static mca_btl_openib_endpoint_t* xoob_find_endpoint(orca_process_name_t* process_name, static mca_btl_openib_endpoint_t* xoob_find_endpoint(orte_process_name_t* process_name,
uint64_t subnet_id, uint16_t lid, uint8_t message_type) uint64_t subnet_id, uint16_t lid, uint8_t message_type)
{ {
size_t i; size_t i;
@ -708,7 +711,7 @@ static mca_btl_openib_endpoint_t* xoob_find_endpoint(orca_process_name_t* proces
ib_proc != (mca_btl_openib_proc_t*) ib_proc != (mca_btl_openib_proc_t*)
opal_list_get_end(&mca_btl_openib_component.ib_procs); opal_list_get_end(&mca_btl_openib_component.ib_procs);
ib_proc = (mca_btl_openib_proc_t*)opal_list_get_next(ib_proc)) { ib_proc = (mca_btl_openib_proc_t*)opal_list_get_next(ib_proc)) {
if (OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, if (OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&ib_proc->proc_ompi->proc_name, process_name)) { &ib_proc->proc_ompi->proc_name, process_name)) {
found = true; found = true;
break; break;
@ -819,8 +822,8 @@ static void free_rem_info(mca_btl_openib_rem_info_t *rem_info)
* and if this endpoint is trying to connect, reply with our QP info, * and if this endpoint is trying to connect, reply with our QP info,
* otherwise try to modify QP's and establish reliable connection * otherwise try to modify QP's and establish reliable connection
*/ */
static void xoob_rml_recv_cb(int status, orca_process_name_t* process_name, static void xoob_rml_recv_cb(int status, orte_process_name_t* process_name,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
int rc; int rc;
@ -1025,12 +1028,12 @@ static int xoob_component_query(mca_btl_openib_module_t *openib_btl,
ensure to only post it *once*, because another btl may have ensure to only post it *once*, because another btl may have
come in before this and already posted it. */ come in before this and already posted it. */
if (!rml_recv_posted) { if (!rml_recv_posted) {
rc = orca_oob_recv_buffer_nb(ORCA_NAME_WILDCARD, rc = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
OMPI_RML_TAG_XOPENIB, OMPI_RML_TAG_XOPENIB,
ORCA_RML_PERSISTENT, ORTE_RML_PERSISTENT,
xoob_rml_recv_cb, xoob_rml_recv_cb,
NULL); NULL);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
opal_output_verbose(5, mca_btl_base_output, opal_output_verbose(5, mca_btl_base_output,
"openib BTL: xoob CPC system error %d (%s)", "openib BTL: xoob CPC system error %d (%s)",
rc, opal_strerror(rc)); rc, opal_strerror(rc));
@ -1138,7 +1141,7 @@ static int xoob_module_start_connect(ompi_btl_openib_connect_base_module_t *cpc,
static int xoob_component_finalize(void) static int xoob_component_finalize(void)
{ {
if (rml_recv_posted) { if (rml_recv_posted) {
orca_oob_recv_cancel(ORCA_NAME_WILDCARD, OMPI_RML_TAG_XOPENIB); orte_rml.recv_cancel(ORTE_NAME_WILDCARD, OMPI_RML_TAG_XOPENIB);
rml_recv_posted = false; rml_recv_posted = false;
} }
#if (ENABLE_DYNAMIC_SL) #if (ENABLE_DYNAMIC_SL)

5
ompi/mca/btl/sctp/btl_sctp_component.c
Просмотреть файл

@ -11,7 +11,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -943,7 +942,7 @@ void mca_btl_sctp_component_accept(void)
*/ */
static void mca_btl_sctp_component_recv_handler(int sd, short flags, void* user) static void mca_btl_sctp_component_recv_handler(int sd, short flags, void* user)
{ {
orca_process_name_t guid; orte_process_name_t guid;
struct sockaddr_in addr; struct sockaddr_in addr;
int retval; int retval;
mca_btl_sctp_proc_t* btl_proc; mca_btl_sctp_proc_t* btl_proc;
@ -967,7 +966,7 @@ static void mca_btl_sctp_component_recv_handler(int sd, short flags, void* user)
} }
SCTP_BTL_ERROR(("mca_btl_sctp_component_recv_handler() sd=%d, got %d byte guid.\n", sd, retval)); SCTP_BTL_ERROR(("mca_btl_sctp_component_recv_handler() sd=%d, got %d byte guid.\n", sd, retval));
ORCA_PROCESS_NAME_NTOH(guid); ORTE_PROCESS_NAME_NTOH(guid);
/* lookup the corresponding process */ /* lookup the corresponding process */
btl_proc = mca_btl_sctp_proc_lookup(&guid); btl_proc = mca_btl_sctp_proc_lookup(&guid);

24
ompi/mca/btl/sctp/btl_sctp_endpoint.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2009 Sun Microsystems, Inc All rights reserved. * Copyright (c) 2009 Sun Microsystems, Inc All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -60,8 +59,7 @@
#include "opal/mca/event/event.h" #include "opal/mca/event/event.h"
#include "orca/include/rte_orca.h" #include "orte/util/name_fns.h"
#include "btl_sctp.h" #include "btl_sctp.h"
#include "btl_sctp_endpoint.h" #include "btl_sctp_endpoint.h"
#include "btl_sctp_proc.h" #include "btl_sctp_proc.h"
@ -370,7 +368,7 @@ int mca_btl_sctp_endpoint_send(mca_btl_base_endpoint_t* btl_endpoint, mca_btl_sc
int rc = OMPI_SUCCESS; int rc = OMPI_SUCCESS;
/* What if there are multiple procs on this endpoint? Possible? */ /* What if there are multiple procs on this endpoint? Possible? */
orca_vpid_t vpid = btl_endpoint->endpoint_proc->proc_ompi->proc_name.vpid; orte_vpid_t vpid = btl_endpoint->endpoint_proc->proc_ompi->proc_name.vpid;
OPAL_THREAD_LOCK(&btl_endpoint->endpoint_send_lock); OPAL_THREAD_LOCK(&btl_endpoint->endpoint_send_lock);
if((mca_btl_sctp_proc_check_vpid(vpid, sender_proc_table)) == INVALID_ENTRY) { if((mca_btl_sctp_proc_check_vpid(vpid, sender_proc_table)) == INVALID_ENTRY) {
@ -547,9 +545,9 @@ static int mca_btl_sctp_endpoint_send_connect_ack(mca_btl_base_endpoint_t* btl_e
{ {
/* send process identifier to remote endpoint */ /* send process identifier to remote endpoint */
mca_btl_sctp_proc_t* btl_proc = mca_btl_sctp_proc_local(); mca_btl_sctp_proc_t* btl_proc = mca_btl_sctp_proc_local();
orca_process_name_t guid = btl_proc->proc_ompi->proc_name; orte_process_name_t guid = btl_proc->proc_ompi->proc_name;
ORCA_PROCESS_NAME_HTON(guid); ORTE_PROCESS_NAME_HTON(guid);
if(mca_btl_sctp_endpoint_send_blocking(btl_endpoint, &guid, sizeof(guid)) != if(mca_btl_sctp_endpoint_send_blocking(btl_endpoint, &guid, sizeof(guid)) !=
sizeof(guid)) { sizeof(guid)) {
return OMPI_ERR_UNREACH; return OMPI_ERR_UNREACH;
@ -580,7 +578,7 @@ bool mca_btl_sctp_endpoint_accept(mca_btl_base_endpoint_t* btl_endpoint, struct
btl_addr->addr_inet.s_addr == addr->sin_addr.s_addr) btl_addr->addr_inet.s_addr == addr->sin_addr.s_addr)
{ {
mca_btl_sctp_proc_t *endpoint_proc = btl_endpoint->endpoint_proc; mca_btl_sctp_proc_t *endpoint_proc = btl_endpoint->endpoint_proc;
cmpval = orca_process_name_compare(ORCA_NAME_CMP_ALL, cmpval = orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&endpoint_proc->proc_ompi->proc_name, &endpoint_proc->proc_ompi->proc_name,
&this_proc->proc_ompi->proc_name); &this_proc->proc_ompi->proc_name);
if((btl_endpoint->endpoint_sd < 0) || if((btl_endpoint->endpoint_sd < 0) ||
@ -830,19 +828,19 @@ static int mca_btl_sctp_endpoint_recv_blocking(mca_btl_base_endpoint_t* btl_endp
static int mca_btl_sctp_endpoint_recv_connect_ack(mca_btl_base_endpoint_t* btl_endpoint) static int mca_btl_sctp_endpoint_recv_connect_ack(mca_btl_base_endpoint_t* btl_endpoint)
{ {
orca_process_name_t guid; orte_process_name_t guid;
mca_btl_sctp_proc_t* btl_proc = btl_endpoint->endpoint_proc; mca_btl_sctp_proc_t* btl_proc = btl_endpoint->endpoint_proc;
if((mca_btl_sctp_endpoint_recv_blocking(btl_endpoint, &guid, if((mca_btl_sctp_endpoint_recv_blocking(btl_endpoint, &guid,
sizeof(orca_process_name_t))) != sizeof(orca_process_name_t)) { sizeof(orte_process_name_t))) != sizeof(orte_process_name_t)) {
return OMPI_ERR_UNREACH; return OMPI_ERR_UNREACH;
} }
ORCA_PROCESS_NAME_NTOH(guid); ORTE_PROCESS_NAME_NTOH(guid);
/* compare this to the expected values */ /* compare this to the expected values */
if(memcmp(&btl_proc->proc_ompi->proc_name, &guid, sizeof(orca_process_name_t)) != 0) { if(memcmp(&btl_proc->proc_ompi->proc_name, &guid, sizeof(orte_process_name_t)) != 0) {
BTL_ERROR(("received unexpected process identifier %s", BTL_ERROR(("received unexpected process identifier %s",
ORCA_NAME_PRINT(&guid))); ORTE_NAME_PRINT(&guid)));
mca_btl_sctp_endpoint_close(btl_endpoint); mca_btl_sctp_endpoint_close(btl_endpoint);
return OMPI_ERR_UNREACH; return OMPI_ERR_UNREACH;
} }
@ -1200,7 +1198,7 @@ static void mca_btl_sctp_endpoint_send_handler(int sd, short flags, void* user)
/* 1 to many */ /* 1 to many */
mca_btl_sctp_endpoint_t* btl_endpoint = (mca_btl_sctp_endpoint_t *)user; mca_btl_sctp_endpoint_t* btl_endpoint = (mca_btl_sctp_endpoint_t *)user;
our_sctp_endpoint *current_our_endpoint = NULL; our_sctp_endpoint *current_our_endpoint = NULL;
orca_vpid_t vpid; orte_vpid_t vpid;
send_handler_1_to_many_different_endpoint: send_handler_1_to_many_different_endpoint:
vpid = btl_endpoint->endpoint_proc->proc_ompi->proc_name.vpid; vpid = btl_endpoint->endpoint_proc->proc_ompi->proc_name.vpid;
OPAL_THREAD_LOCK(&btl_endpoint->endpoint_send_lock); OPAL_THREAD_LOCK(&btl_endpoint->endpoint_send_lock);

19
ompi/mca/btl/sctp/btl_sctp_proc.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2010 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2010 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -68,7 +67,7 @@ void mca_btl_sctp_proc_destruct(mca_btl_sctp_proc_t* stcp_proc)
/* remove from list of all proc instances */ /* remove from list of all proc instances */
OPAL_THREAD_LOCK(&mca_btl_sctp_component.sctp_lock); OPAL_THREAD_LOCK(&mca_btl_sctp_component.sctp_lock);
opal_hash_table_remove_value_uint64(&mca_btl_sctp_component.sctp_procs, opal_hash_table_remove_value_uint64(&mca_btl_sctp_component.sctp_procs,
orca_process_info_hash_name(&stcp_proc->proc_ompi->proc_name)); orte_util_hash_name(&stcp_proc->proc_ompi->proc_name));
OPAL_THREAD_UNLOCK(&mca_btl_sctp_component.sctp_lock); OPAL_THREAD_UNLOCK(&mca_btl_sctp_component.sctp_lock);
/* release resources */ /* release resources */
@ -114,7 +113,7 @@ mca_btl_sctp_proc_t* mca_btl_sctp_proc_create(ompi_proc_t* ompi_proc)
int rc; int rc;
size_t size; size_t size;
mca_btl_sctp_proc_t* btl_proc; mca_btl_sctp_proc_t* btl_proc;
uint64_t hash = orca_process_info_hash_name(&ompi_proc->proc_name); uint64_t hash = orte_util_hash_name(&ompi_proc->proc_name);
OPAL_THREAD_LOCK(&mca_btl_sctp_component.sctp_lock); OPAL_THREAD_LOCK(&mca_btl_sctp_component.sctp_lock);
rc = opal_hash_table_get_value_uint64(&mca_btl_sctp_component.sctp_procs, rc = opal_hash_table_get_value_uint64(&mca_btl_sctp_component.sctp_procs,
@ -335,12 +334,12 @@ int mca_btl_sctp_proc_remove(mca_btl_sctp_proc_t* btl_proc, mca_btl_base_endpoin
* Look for an existing SCTP process instance based on the globally unique * Look for an existing SCTP process instance based on the globally unique
* process identifier. * process identifier.
*/ */
mca_btl_sctp_proc_t* mca_btl_sctp_proc_lookup(const orca_process_name_t *name) mca_btl_sctp_proc_t* mca_btl_sctp_proc_lookup(const orte_process_name_t *name)
{ {
mca_btl_sctp_proc_t* proc = NULL; mca_btl_sctp_proc_t* proc = NULL;
OPAL_THREAD_LOCK(&mca_btl_sctp_component.sctp_lock); OPAL_THREAD_LOCK(&mca_btl_sctp_component.sctp_lock);
opal_hash_table_get_value_uint64(&mca_btl_sctp_component.sctp_procs, opal_hash_table_get_value_uint64(&mca_btl_sctp_component.sctp_procs,
orca_process_info_hash_name(name), (void**)&proc); orte_util_hash_name(name), (void**)&proc);
OPAL_THREAD_UNLOCK(&mca_btl_sctp_component.sctp_lock); OPAL_THREAD_UNLOCK(&mca_btl_sctp_component.sctp_lock);
return proc; return proc;
} }
@ -374,11 +373,11 @@ bool mca_btl_sctp_proc_accept(mca_btl_sctp_proc_t* btl_proc, struct sockaddr_in*
* *
* TODO - change this to use a hash for constant time performance * TODO - change this to use a hash for constant time performance
*/ */
static int mca_btl_sctp_proc_check(int is_vpid, sctp_assoc_t id, orca_vpid_t vpid, struct mca_btl_sctp_proc_table_node *table) { static int mca_btl_sctp_proc_check(int is_vpid, sctp_assoc_t id, orte_vpid_t vpid, struct mca_btl_sctp_proc_table_node *table) {
#if MCA_BTL_SCTP_DONT_USE_HASH #if MCA_BTL_SCTP_DONT_USE_HASH
int i; int i;
for(i = 0; i < MCA_BTL_SCTP_PROC_TABLE_SIZE; i++) { for(i = 0; i < MCA_BTL_SCTP_PROC_TABLE_SIZE; i++) {
/* sender_proc_table uses orca_vpid_t. /* sender_proc_table uses orte_vpid_t.
* recvr_proc_table uses sctp_assoc_id. * recvr_proc_table uses sctp_assoc_id.
* Calls using this function use one or the other. * Calls using this function use one or the other.
*/ */
@ -404,7 +403,7 @@ static int mca_btl_sctp_proc_check(int is_vpid, sctp_assoc_t id, orca_vpid_t vpi
#endif #endif
} }
int mca_btl_sctp_proc_check_vpid(orca_vpid_t vpid, struct mca_btl_sctp_proc_table_node *table) { int mca_btl_sctp_proc_check_vpid(orte_vpid_t vpid, struct mca_btl_sctp_proc_table_node *table) {
return mca_btl_sctp_proc_check(1, 0, vpid, table); return mca_btl_sctp_proc_check(1, 0, vpid, table);
} }
@ -422,7 +421,7 @@ int mca_btl_sctp_proc_check_assoc_id(sctp_assoc_t id, struct mca_btl_sctp_proc_t
* TODO change this to a hash table that can expand to eliminate * TODO change this to a hash table that can expand to eliminate
* MCA_BTL_SCTP_PROC_TABLE_SIZE limitation * MCA_BTL_SCTP_PROC_TABLE_SIZE limitation
*/ */
static void mca_btl_sctp_proc_add(sctp_assoc_t id, orca_vpid_t vpid, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table) { static void mca_btl_sctp_proc_add(sctp_assoc_t id, orte_vpid_t vpid, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table) {
#if MCA_BTL_SCTP_DONT_USE_HASH #if MCA_BTL_SCTP_DONT_USE_HASH
int i; int i;
for(i = 0; i < MCA_BTL_SCTP_PROC_TABLE_SIZE; i++) { for(i = 0; i < MCA_BTL_SCTP_PROC_TABLE_SIZE; i++) {
@ -441,7 +440,7 @@ static void mca_btl_sctp_proc_add(sctp_assoc_t id, orca_vpid_t vpid, struct mca_
#endif #endif
} }
void mca_btl_sctp_proc_add_vpid(orca_vpid_t vpid, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table) { void mca_btl_sctp_proc_add_vpid(orte_vpid_t vpid, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table) {
mca_btl_sctp_proc_add(0, vpid, proc, table); mca_btl_sctp_proc_add(0, vpid, proc, table);
} }

9
ompi/mca/btl/sctp/btl_sctp_proc.h
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -61,7 +60,7 @@ typedef struct mca_btl_sctp_proc_t mca_btl_sctp_proc_t;
OBJ_CLASS_DECLARATION(mca_btl_sctp_proc_t); OBJ_CLASS_DECLARATION(mca_btl_sctp_proc_t);
mca_btl_sctp_proc_t* mca_btl_sctp_proc_create(ompi_proc_t* ompi_proc); mca_btl_sctp_proc_t* mca_btl_sctp_proc_create(ompi_proc_t* ompi_proc);
mca_btl_sctp_proc_t* mca_btl_sctp_proc_lookup(const orca_process_name_t* name); mca_btl_sctp_proc_t* mca_btl_sctp_proc_lookup(const orte_process_name_t* name);
int mca_btl_sctp_proc_insert(mca_btl_sctp_proc_t*, mca_btl_base_endpoint_t*); int mca_btl_sctp_proc_insert(mca_btl_sctp_proc_t*, mca_btl_base_endpoint_t*);
int mca_btl_sctp_proc_remove(mca_btl_sctp_proc_t*, mca_btl_base_endpoint_t*); int mca_btl_sctp_proc_remove(mca_btl_sctp_proc_t*, mca_btl_base_endpoint_t*);
bool mca_btl_sctp_proc_accept(mca_btl_sctp_proc_t*, struct sockaddr_in*, int); bool mca_btl_sctp_proc_accept(mca_btl_sctp_proc_t*, struct sockaddr_in*, int);
@ -91,7 +90,7 @@ enum {
struct mca_btl_sctp_proc_table_node { struct mca_btl_sctp_proc_table_node {
int valid; int valid;
sctp_assoc_t sctp_assoc_id; sctp_assoc_t sctp_assoc_id;
orca_vpid_t vpid; orte_vpid_t vpid;
struct mca_btl_sctp_proc_t *proc; struct mca_btl_sctp_proc_t *proc;
}; };
typedef struct mca_btl_sctp_proc_table_node mca_btl_sctp_proc_table_node; typedef struct mca_btl_sctp_proc_table_node mca_btl_sctp_proc_table_node;
@ -99,9 +98,9 @@ typedef struct mca_btl_sctp_proc_table_node mca_btl_sctp_proc_table_node;
extern struct mca_btl_sctp_proc_table_node *recvr_proc_table; extern struct mca_btl_sctp_proc_table_node *recvr_proc_table;
extern struct mca_btl_sctp_proc_table_node *sender_proc_table; extern struct mca_btl_sctp_proc_table_node *sender_proc_table;
int mca_btl_sctp_proc_check_vpid(orca_vpid_t vpid, struct mca_btl_sctp_proc_table_node *table); int mca_btl_sctp_proc_check_vpid(orte_vpid_t vpid, struct mca_btl_sctp_proc_table_node *table);
int mca_btl_sctp_proc_check_assoc_id(sctp_assoc_t id, struct mca_btl_sctp_proc_table_node *table); int mca_btl_sctp_proc_check_assoc_id(sctp_assoc_t id, struct mca_btl_sctp_proc_table_node *table);
void mca_btl_sctp_proc_add_vpid(orca_vpid_t vpid, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table); void mca_btl_sctp_proc_add_vpid(orte_vpid_t vpid, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table);
void mca_btl_sctp_proc_add_assoc_id(sctp_assoc_t id, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table); void mca_btl_sctp_proc_add_assoc_id(sctp_assoc_t id, struct mca_btl_sctp_proc_t *proc, struct mca_btl_sctp_proc_table_node *table);
mca_btl_sctp_proc_t *mca_btl_sctp_proc_get(sctp_assoc_t id, struct mca_btl_sctp_proc_table_node *table); mca_btl_sctp_proc_t *mca_btl_sctp_proc_get(sctp_assoc_t id, struct mca_btl_sctp_proc_table_node *table);

5
ompi/mca/btl/sctp/btl_sctp_recv_handler.c
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -98,7 +97,7 @@ void mca_btl_sctp_recv_handler(int sd, short flags, void *user) {
/* allocated this elsewhere only once per BTL to avoid repeatedly calling malloc */ /* allocated this elsewhere only once per BTL to avoid repeatedly calling malloc */
char *buf = sctp_recv_buf; char *buf = sctp_recv_buf;
orca_process_name_t guid; orte_process_name_t guid;
struct sockaddr_in their_addr; struct sockaddr_in their_addr;
int retval; int retval;
mca_btl_sctp_proc_t *btl_proc; mca_btl_sctp_proc_t *btl_proc;
@ -215,7 +214,7 @@ data_still_pending_on_endpoint:
/* Setup guid. */ /* Setup guid. */
memcpy(&guid, buf, retval); memcpy(&guid, buf, retval);
ORCA_PROCESS_NAME_NTOH(guid); ORTE_PROCESS_NAME_NTOH(guid);
/* lookup the corresponding process */ /* lookup the corresponding process */
btl_proc = mca_btl_sctp_proc_lookup(&guid); btl_proc = mca_btl_sctp_proc_lookup(&guid);

12
ompi/mca/btl/sm/btl_sm.c
Просмотреть файл

@ -14,7 +14,6 @@
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2010-2012 IBM Corporation. All rights reserved. * Copyright (c) 2010-2012 IBM Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -43,8 +42,7 @@
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/util/printf.h" #include "opal/util/printf.h"
#include "opal/mca/hwloc/base/base.h" #include "opal/mca/hwloc/base/base.h"
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "opal/datatype/opal_convertor.h" #include "opal/datatype/opal_convertor.h"
#include "ompi/class/ompi_free_list.h" #include "ompi/class/ompi_free_list.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
@ -254,8 +252,8 @@ static int sm_btl_first_time_init(mca_btl_sm_t *sm_btl, int n)
/* set file name */ /* set file name */
if (asprintf(&sm_ctl_file, "%s"OPAL_PATH_SEP"shared_mem_btl_module.%s", if (asprintf(&sm_ctl_file, "%s"OPAL_PATH_SEP"shared_mem_btl_module.%s",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
orca_process_info_get_nodename() ) < 0) { orte_process_info.nodename) < 0) {
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -389,7 +387,7 @@ create_sm_endpoint(int local_proc, struct ompi_proc_t *proc)
OBJ_CONSTRUCT(&ep->endpoint_lock, opal_mutex_t); OBJ_CONSTRUCT(&ep->endpoint_lock, opal_mutex_t);
#if OMPI_ENABLE_PROGRESS_THREADS == 1 #if OMPI_ENABLE_PROGRESS_THREADS == 1
sprintf(path, "%s"OPAL_PATH_SEP"sm_fifo.%lu", sprintf(path, "%s"OPAL_PATH_SEP"sm_fifo.%lu",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
(unsigned long)proc->proc_name.vpid); (unsigned long)proc->proc_name.vpid);
ep->fifo_fd = open(path, O_WRONLY); ep->fifo_fd = open(path, O_WRONLY);
if(ep->fifo_fd < 0) { if(ep->fifo_fd < 0) {
@ -1163,7 +1161,7 @@ int mca_btl_sm_ft_event(int state) {
} }
} }
else if(OPAL_CRS_CONTINUE == state) { else if(OPAL_CRS_CONTINUE == state) {
if( orca_info_cr_continue_like_restart() ) { if( orte_cr_continue_like_restart ) {
if( NULL != mca_btl_sm_component.sm_seg ) { if( NULL != mca_btl_sm_component.sm_seg ) {
/* Add shared memory file */ /* Add shared memory file */
opal_crs_base_cleanup_append(mca_btl_sm_component.sm_seg->shmem_ds.seg_name, false); opal_crs_base_cleanup_append(mca_btl_sm_component.sm_seg->shmem_ds.seg_name, false);

41
ompi/mca/btl/sm/btl_sm_component.c
Просмотреть файл

@ -15,7 +15,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2011 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2010-2012 IBM Corporation. All rights reserved. * Copyright (c) 2010-2012 IBM Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -47,7 +46,9 @@
#include "opal/mca/event/event.h" #include "opal/mca/event/event.h"
#include "opal/util/bit_ops.h" #include "opal/util/bit_ops.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
@ -151,9 +152,9 @@ static int sm_register(void)
mca_btl_sm_component.use_knem = i; mca_btl_sm_component.use_knem = i;
} else { } else {
if (i > 0) { if (i > 0) {
orca_show_help("help-mpi-btl-sm.txt", orte_show_help("help-mpi-btl-sm.txt",
"knem requested but not supported", true, "knem requested but not supported", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
return OMPI_ERROR; return OMPI_ERROR;
} }
mca_btl_sm_component.use_knem = 0; mca_btl_sm_component.use_knem = 0;
@ -366,7 +367,7 @@ static mca_btl_base_module_t** mca_btl_sm_component_init(
*num_btls = 0; *num_btls = 0;
/* if no session directory was created, then we cannot be used */ /* if no session directory was created, then we cannot be used */
if (!orca_process_info_create_session_dirs()) { if (!orte_create_session_dirs) {
return NULL; return NULL;
} }
@ -377,8 +378,8 @@ static mca_btl_base_module_t** mca_btl_sm_component_init(
#if OMPI_ENABLE_PROGRESS_THREADS == 1 #if OMPI_ENABLE_PROGRESS_THREADS == 1
/* create a named pipe to receive events */ /* create a named pipe to receive events */
sprintf( mca_btl_sm_component.sm_fifo_path, sprintf( mca_btl_sm_component.sm_fifo_path,
"%s"OPAL_PATH_SEP"sm_fifo.%lu", orca_process_info_get_job_session_dir(), "%s"OPAL_PATH_SEP"sm_fifo.%lu", orte_process_info.job_session_dir,
(unsigned long)orca_process_info_get_vpid(ORCA_PROC_MY_NAME) ); (unsigned long)ORTE_PROC_MY_NAME->vpid );
if(mkfifo(mca_btl_sm_component.sm_fifo_path, 0660) < 0) { if(mkfifo(mca_btl_sm_component.sm_fifo_path, 0660) < 0) {
opal_output(0, "mca_btl_sm_component_init: mkfifo failed with errno=%d\n",errno); opal_output(0, "mca_btl_sm_component_init: mkfifo failed with errno=%d\n",errno);
return NULL; return NULL;
@ -441,11 +442,11 @@ static mca_btl_base_module_t** mca_btl_sm_component_init(
if (0 != stat("/dev/knem", &sbuf)) { if (0 != stat("/dev/knem", &sbuf)) {
sbuf.st_mode = 0; sbuf.st_mode = 0;
} }
orca_show_help("help-mpi-btl-sm.txt", "knem permission denied", orte_show_help("help-mpi-btl-sm.txt", "knem permission denied",
true, orca_process_info_get_nodename(), sbuf.st_mode); true, orte_process_info.nodename, sbuf.st_mode);
} else { } else {
orca_show_help("help-mpi-btl-sm.txt", "knem fail open", orte_show_help("help-mpi-btl-sm.txt", "knem fail open",
true, orca_process_info_get_nodename(), errno, true, orte_process_info.nodename, errno,
strerror(errno)); strerror(errno));
} }
goto no_knem; goto no_knem;
@ -456,14 +457,14 @@ static mca_btl_base_module_t** mca_btl_sm_component_init(
rc = ioctl(mca_btl_sm.knem_fd, KNEM_CMD_GET_INFO, rc = ioctl(mca_btl_sm.knem_fd, KNEM_CMD_GET_INFO,
&mca_btl_sm_component.knem_info); &mca_btl_sm_component.knem_info);
if (rc < 0) { if (rc < 0) {
orca_show_help("help-mpi-btl-sm.txt", "knem get ABI fail", orte_show_help("help-mpi-btl-sm.txt", "knem get ABI fail",
true, orca_process_info_get_nodename(), errno, true, orte_process_info.nodename, errno,
strerror(errno)); strerror(errno));
goto no_knem; goto no_knem;
} }
if (KNEM_ABI_VERSION != mca_btl_sm_component.knem_info.abi) { if (KNEM_ABI_VERSION != mca_btl_sm_component.knem_info.abi) {
orca_show_help("help-mpi-btl-sm.txt", "knem ABI mismatch", orte_show_help("help-mpi-btl-sm.txt", "knem ABI mismatch",
true, orca_process_info_get_nodename(), KNEM_ABI_VERSION, true, orte_process_info.nodename, KNEM_ABI_VERSION,
mca_btl_sm_component.knem_info.abi); mca_btl_sm_component.knem_info.abi);
goto no_knem; goto no_knem;
} }
@ -484,8 +485,8 @@ static mca_btl_base_module_t** mca_btl_sm_component_init(
MAP_SHARED, mca_btl_sm.knem_fd, MAP_SHARED, mca_btl_sm.knem_fd,
KNEM_STATUS_ARRAY_FILE_OFFSET); KNEM_STATUS_ARRAY_FILE_OFFSET);
if (MAP_FAILED == mca_btl_sm.knem_status_array) { if (MAP_FAILED == mca_btl_sm.knem_status_array) {
orca_show_help("help-mpi-btl-sm.txt", "knem mmap fail", orte_show_help("help-mpi-btl-sm.txt", "knem mmap fail",
true, orca_process_info_get_nodename(), errno, true, orte_process_info.nodename, errno,
strerror(errno)); strerror(errno));
goto no_knem; goto no_knem;
} }
@ -496,8 +497,8 @@ static mca_btl_base_module_t** mca_btl_sm_component_init(
malloc(sizeof(mca_btl_sm_frag_t *) * malloc(sizeof(mca_btl_sm_frag_t *) *
mca_btl_sm_component.knem_max_simultaneous); mca_btl_sm_component.knem_max_simultaneous);
if (NULL == mca_btl_sm.knem_frag_array) { if (NULL == mca_btl_sm.knem_frag_array) {
orca_show_help("help-mpi-btl-sm.txt", "knem init fail", orte_show_help("help-mpi-btl-sm.txt", "knem init fail",
true, orca_process_info_get_nodename(), "malloc", true, orte_process_info.nodename, "malloc",
errno, strerror(errno)); errno, strerror(errno));
goto no_knem; goto no_knem;
} }

13
ompi/mca/btl/smcuda/btl_smcuda.c
Просмотреть файл

@ -14,7 +14,6 @@
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 NVIDIA Corporation. All rights reserved. * Copyright (c) 2012 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -39,9 +38,7 @@
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/util/printf.h" #include "opal/util/printf.h"
#include "opal/mca/hwloc/base/base.h" #include "opal/mca/hwloc/base/base.h"
#include "orte/util/proc_info.h"
#include "orca/include/rte_orca.h"
#include "opal/datatype/opal_convertor.h" #include "opal/datatype/opal_convertor.h"
#include "ompi/class/ompi_free_list.h" #include "ompi/class/ompi_free_list.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
@ -268,8 +265,8 @@ static int smcuda_btl_first_time_init(mca_btl_smcuda_t *smcuda_btl, int n)
/* set file name */ /* set file name */
if (asprintf(&sm_ctl_file, "%s"OPAL_PATH_SEP"shared_mem_btl_module.%s", if (asprintf(&sm_ctl_file, "%s"OPAL_PATH_SEP"shared_mem_btl_module.%s",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
orca_process_info_get_nodename()) < 0) { orte_process_info.nodename) < 0) {
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -403,7 +400,7 @@ create_sm_endpoint(int local_proc, struct ompi_proc_t *proc)
OBJ_CONSTRUCT(&ep->endpoint_lock, opal_mutex_t); OBJ_CONSTRUCT(&ep->endpoint_lock, opal_mutex_t);
#if OMPI_ENABLE_PROGRESS_THREADS == 1 #if OMPI_ENABLE_PROGRESS_THREADS == 1
sprintf(path, "%s"OPAL_PATH_SEP"sm_fifo.%lu", sprintf(path, "%s"OPAL_PATH_SEP"sm_fifo.%lu",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
(unsigned long)proc->proc_name.vpid); (unsigned long)proc->proc_name.vpid);
ep->fifo_fd = open(path, O_WRONLY); ep->fifo_fd = open(path, O_WRONLY);
if(ep->fifo_fd < 0) { if(ep->fifo_fd < 0) {
@ -1078,7 +1075,7 @@ int mca_btl_smcuda_ft_event(int state) {
} }
} }
else if(OPAL_CRS_CONTINUE == state) { else if(OPAL_CRS_CONTINUE == state) {
if( orca_info_cr_continue_like_restart() ) { if( orte_cr_continue_like_restart ) {
if( NULL != mca_btl_smcuda_component.sm_seg ) { if( NULL != mca_btl_smcuda_component.sm_seg ) {
/* Add shared memory file */ /* Add shared memory file */
opal_crs_base_cleanup_append(mca_btl_smcuda_component.sm_seg->shmem_ds.seg_name, false); opal_crs_base_cleanup_append(mca_btl_smcuda_component.sm_seg->shmem_ds.seg_name, false);

12
ompi/mca/btl/smcuda/btl_smcuda_component.c
Просмотреть файл

@ -14,7 +14,6 @@
* Copyright (c) 2010-2011 Los Alamos National Security, LLC. * Copyright (c) 2010-2011 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2011-2012 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011-2012 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -46,8 +45,9 @@
#include "opal/mca/event/event.h" #include "opal/mca/event/event.h"
#include "opal/util/bit_ops.h" #include "opal/util/bit_ops.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orte/util/proc_info.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
@ -297,7 +297,7 @@ static mca_btl_base_module_t** mca_btl_smcuda_component_init(
*num_btls = 0; *num_btls = 0;
/* if no session directory was created, then we cannot be used */ /* if no session directory was created, then we cannot be used */
if (!orca_process_info_create_session_dirs()) { if (!orte_create_session_dirs) {
return NULL; return NULL;
} }
@ -308,8 +308,8 @@ static mca_btl_base_module_t** mca_btl_smcuda_component_init(
#if OMPI_ENABLE_PROGRESS_THREADS == 1 #if OMPI_ENABLE_PROGRESS_THREADS == 1
/* create a named pipe to receive events */ /* create a named pipe to receive events */
sprintf( mca_btl_smcuda_component.sm_fifo_path, sprintf( mca_btl_smcuda_component.sm_fifo_path,
"%s"OPAL_PATH_SEP"sm_fifo.%lu", orca_process_info_get_job_session_dir(), "%s"OPAL_PATH_SEP"sm_fifo.%lu", orte_process_info.job_session_dir,
(unsigned long)orca_process_info_get_vpid(ORCA_PROC_MY_NAME) ); (unsigned long)ORTE_PROC_MY_NAME->vpid );
if(mkfifo(mca_btl_smcuda_component.sm_fifo_path, 0660) < 0) { if(mkfifo(mca_btl_smcuda_component.sm_fifo_path, 0660) < 0) {
opal_output(0, "mca_btl_smcuda_component_init: mkfifo failed with errno=%d\n",errno); opal_output(0, "mca_btl_smcuda_component_init: mkfifo failed with errno=%d\n",errno);
return NULL; return NULL;

38
ompi/mca/btl/tcp/btl_tcp_component.c
Просмотреть файл

@ -11,7 +11,7 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2007-2011 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2009-2012 Oak Ridge National Laboratory * Copyright (c) 2009 Oak Ridge National Laboratory
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* $COPYRIGHT$ * $COPYRIGHT$
@ -57,7 +57,9 @@
#include "opal/util/net.h" #include "opal/util/net.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orca/include/rte_orca.h" #include "orte/types.h"
#include "orte/util/show_help.h"
#include "orte/mca/ess/ess.h"
#include "ompi/constants.h" #include "ompi/constants.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
@ -213,8 +215,8 @@ static int mca_btl_tcp_component_register(void)
mca_btl_tcp_param_register_int( "port_min_v4", mca_btl_tcp_param_register_int( "port_min_v4",
"The minimum port where the TCP BTL will try to bind (default 1024)", 1024 ); "The minimum port where the TCP BTL will try to bind (default 1024)", 1024 );
if( mca_btl_tcp_component.tcp_port_min > USHRT_MAX ) { if( mca_btl_tcp_component.tcp_port_min > USHRT_MAX ) {
orca_show_help("help-mpi-btl-tcp.txt", "invalid minimum port", orte_show_help("help-mpi-btl-tcp.txt", "invalid minimum port",
true, "v4", orca_process_info_get_nodename(), true, "v4", orte_process_info.nodename,
mca_btl_tcp_component.tcp_port_min ); mca_btl_tcp_component.tcp_port_min );
mca_btl_tcp_component.tcp_port_min = 1024; mca_btl_tcp_component.tcp_port_min = 1024;
} }
@ -232,8 +234,8 @@ static int mca_btl_tcp_component_register(void)
mca_btl_tcp_param_register_int( "port_min_v6", mca_btl_tcp_param_register_int( "port_min_v6",
"The minimum port where the TCP BTL will try to bind (default 1024)", 1024 ); "The minimum port where the TCP BTL will try to bind (default 1024)", 1024 );
if( mca_btl_tcp_component.tcp6_port_min > USHRT_MAX ) { if( mca_btl_tcp_component.tcp6_port_min > USHRT_MAX ) {
orca_show_help("help-mpi-btl-tcp.txt", "invalid minimum port", orte_show_help("help-mpi-btl-tcp.txt", "invalid minimum port",
true, "v6", orca_process_info_get_nodename(), true, "v6", orte_process_info.nodename,
mca_btl_tcp_component.tcp6_port_min ); mca_btl_tcp_component.tcp6_port_min );
mca_btl_tcp_component.tcp6_port_min = 1024; mca_btl_tcp_component.tcp6_port_min = 1024;
} }
@ -279,10 +281,10 @@ static int mca_btl_tcp_component_register(void)
if (NULL != argv && '\0' != *(argv[0])) { if (NULL != argv && '\0' != *(argv[0])) {
int if_index, rc, count; int if_index, rc, count;
orca_node_rank_t node_rank; orte_node_rank_t node_rank;
char name[256]; char name[256];
node_rank = orca_node_info_get_rank(ORCA_PROC_MY_NAME); node_rank = orte_ess.get_node_rank(ORTE_PROC_MY_NAME);
/* Now that we've got that local rank, take the /* Now that we've got that local rank, take the
corresponding entry from the tcp_if_seq list (wrapping corresponding entry from the tcp_if_seq list (wrapping
@ -304,10 +306,10 @@ static int mca_btl_tcp_component_register(void)
} }
} }
if (if_index < 0) { if (if_index < 0) {
orca_show_help("help-mpi-btl-tcp.txt", orte_show_help("help-mpi-btl-tcp.txt",
"invalid if_inexclude", "invalid if_inexclude",
true, "if_seq", true, "if_seq",
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_tcp_component.tcp_if_seq, mca_btl_tcp_component.tcp_if_seq,
"Interface does not exist"); "Interface does not exist");
return OMPI_ERR_BAD_PARAM; return OMPI_ERR_BAD_PARAM;
@ -518,8 +520,8 @@ static char **split_and_resolve(char **orig_str, char *name)
tmp = strdup(argv[i]); tmp = strdup(argv[i]);
str = strchr(argv[i], '/'); str = strchr(argv[i], '/');
if (NULL == str) { if (NULL == str) {
orca_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude", orte_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude",
true, name, orca_process_info_get_nodename(), true, name, orte_process_info.nodename,
tmp, "Invalid specification (missing \"/\")"); tmp, "Invalid specification (missing \"/\")");
free(argv[i]); free(argv[i]);
free(tmp); free(tmp);
@ -535,8 +537,8 @@ static char **split_and_resolve(char **orig_str, char *name)
free(argv[i]); free(argv[i]);
if (1 != ret) { if (1 != ret) {
orca_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude", orte_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude",
true, name, orca_process_info_get_nodename(), tmp, true, name, orte_process_info.nodename, tmp,
"Invalid specification (inet_pton() failed)"); "Invalid specification (inet_pton() failed)");
free(tmp); free(tmp);
continue; continue;
@ -562,8 +564,8 @@ static char **split_and_resolve(char **orig_str, char *name)
/* If we didn't find a match, keep trying */ /* If we didn't find a match, keep trying */
if (if_index < 0) { if (if_index < 0) {
orca_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude", orte_show_help("help-mpi-btl-tcp.txt", "invalid if_inexclude",
true, name, orca_process_info_get_nodename(), tmp, true, name, orte_process_info.nodename, tmp,
"Did not find interface matching this subnet"); "Did not find interface matching this subnet");
free(tmp); free(tmp);
continue; continue;
@ -1126,7 +1128,7 @@ static void mca_btl_tcp_component_accept_handler( int incoming_sd,
*/ */
static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user) static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user)
{ {
orca_process_name_t guid; orte_process_name_t guid;
struct sockaddr_storage addr; struct sockaddr_storage addr;
int retval; int retval;
mca_btl_tcp_proc_t* btl_proc; mca_btl_tcp_proc_t* btl_proc;
@ -1141,7 +1143,7 @@ static void mca_btl_tcp_component_recv_handler(int sd, short flags, void* user)
CLOSE_THE_SOCKET(sd); CLOSE_THE_SOCKET(sd);
return; return;
} }
ORCA_PROCESS_NAME_NTOH(guid); ORTE_PROCESS_NAME_NTOH(guid);
/* now set socket up to be non-blocking */ /* now set socket up to be non-blocking */
if((flags = fcntl(sd, F_GETFL, 0)) < 0) { if((flags = fcntl(sd, F_GETFL, 0)) < 0) {

19
ompi/mca/btl/tcp/btl_tcp_endpoint.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2007-2008 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007-2008 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -316,9 +315,9 @@ static int mca_btl_tcp_endpoint_send_connect_ack(mca_btl_base_endpoint_t* btl_en
{ {
/* send process identifier to remote endpoint */ /* send process identifier to remote endpoint */
mca_btl_tcp_proc_t* btl_proc = mca_btl_tcp_proc_local(); mca_btl_tcp_proc_t* btl_proc = mca_btl_tcp_proc_local();
orca_process_name_t guid = btl_proc->proc_ompi->proc_name; orte_process_name_t guid = btl_proc->proc_ompi->proc_name;
ORCA_PROCESS_NAME_HTON(guid); ORTE_PROCESS_NAME_HTON(guid);
if(mca_btl_tcp_endpoint_send_blocking(btl_endpoint, &guid, sizeof(guid)) != if(mca_btl_tcp_endpoint_send_blocking(btl_endpoint, &guid, sizeof(guid)) !=
sizeof(guid)) { sizeof(guid)) {
return OMPI_ERR_UNREACH; return OMPI_ERR_UNREACH;
@ -351,7 +350,7 @@ bool mca_btl_tcp_endpoint_accept(mca_btl_base_endpoint_t* btl_endpoint,
return false; return false;
} }
cmpval = orca_process_name_compare(ORCA_NAME_CMP_ALL, cmpval = orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&endpoint_proc->proc_ompi->proc_name, &endpoint_proc->proc_ompi->proc_name,
&this_proc->proc_ompi->proc_name); &this_proc->proc_ompi->proc_name);
if((btl_endpoint->endpoint_sd < 0) || if((btl_endpoint->endpoint_sd < 0) ||
@ -472,19 +471,19 @@ static int mca_btl_tcp_endpoint_recv_blocking(mca_btl_base_endpoint_t* btl_endpo
*/ */
static int mca_btl_tcp_endpoint_recv_connect_ack(mca_btl_base_endpoint_t* btl_endpoint) static int mca_btl_tcp_endpoint_recv_connect_ack(mca_btl_base_endpoint_t* btl_endpoint)
{ {
orca_process_name_t guid; orte_process_name_t guid;
mca_btl_tcp_proc_t* btl_proc = btl_endpoint->endpoint_proc; mca_btl_tcp_proc_t* btl_proc = btl_endpoint->endpoint_proc;
if((mca_btl_tcp_endpoint_recv_blocking(btl_endpoint, &guid, sizeof(orca_process_name_t))) != sizeof(orca_process_name_t)) { if((mca_btl_tcp_endpoint_recv_blocking(btl_endpoint, &guid, sizeof(orte_process_name_t))) != sizeof(orte_process_name_t)) {
return OMPI_ERR_UNREACH; return OMPI_ERR_UNREACH;
} }
ORCA_PROCESS_NAME_NTOH(guid); ORTE_PROCESS_NAME_NTOH(guid);
/* compare this to the expected values */ /* compare this to the expected values */
if (OPAL_EQUAL != orca_process_name_compare(ORCA_NAME_CMP_ALL, if (OPAL_EQUAL != orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&btl_proc->proc_ompi->proc_name, &btl_proc->proc_ompi->proc_name,
&guid)) { &guid)) {
BTL_ERROR(("received unexpected process identifier %s", BTL_ERROR(("received unexpected process identifier %s",
ORCA_NAME_PRINT(&guid))); ORTE_NAME_PRINT(&guid)));
mca_btl_tcp_endpoint_close(btl_endpoint); mca_btl_tcp_endpoint_close(btl_endpoint);
return OMPI_ERR_UNREACH; return OMPI_ERR_UNREACH;
} }
@ -571,7 +570,7 @@ static int mca_btl_tcp_endpoint_start_connect(mca_btl_base_endpoint_t* btl_endpo
opal_output_verbose(20, mca_btl_base_output, opal_output_verbose(20, mca_btl_base_output,
"btl: tcp: attempting to connect() to %s address %s on port %d", "btl: tcp: attempting to connect() to %s address %s on port %d",
ORCA_NAME_PRINT(&btl_endpoint->endpoint_proc->proc_ompi->proc_name), ORTE_NAME_PRINT(&btl_endpoint->endpoint_proc->proc_ompi->proc_name),
opal_net_get_hostname((struct sockaddr*) &endpoint_addr), opal_net_get_hostname((struct sockaddr*) &endpoint_addr),
ntohs(btl_endpoint->endpoint_addr->addr_port)); ntohs(btl_endpoint->endpoint_addr->addr_port));

9
ompi/mca/btl/tcp/btl_tcp_proc.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2008-2010 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2008-2010 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -77,7 +76,7 @@ void mca_btl_tcp_proc_destruct(mca_btl_tcp_proc_t* tcp_proc)
/* remove from list of all proc instances */ /* remove from list of all proc instances */
OPAL_THREAD_LOCK(&mca_btl_tcp_component.tcp_lock); OPAL_THREAD_LOCK(&mca_btl_tcp_component.tcp_lock);
opal_hash_table_remove_value_uint64(&mca_btl_tcp_component.tcp_procs, opal_hash_table_remove_value_uint64(&mca_btl_tcp_component.tcp_procs,
orca_process_info_hash_name(&tcp_proc->proc_ompi->proc_name)); orte_util_hash_name(&tcp_proc->proc_ompi->proc_name));
OPAL_THREAD_UNLOCK(&mca_btl_tcp_component.tcp_lock); OPAL_THREAD_UNLOCK(&mca_btl_tcp_component.tcp_lock);
/* release resources */ /* release resources */
@ -100,7 +99,7 @@ mca_btl_tcp_proc_t* mca_btl_tcp_proc_create(ompi_proc_t* ompi_proc)
int rc; int rc;
size_t size; size_t size;
mca_btl_tcp_proc_t* btl_proc; mca_btl_tcp_proc_t* btl_proc;
uint64_t hash = orca_process_info_hash_name(&ompi_proc->proc_name); uint64_t hash = orte_util_hash_name(&ompi_proc->proc_name);
OPAL_THREAD_LOCK(&mca_btl_tcp_component.tcp_lock); OPAL_THREAD_LOCK(&mca_btl_tcp_component.tcp_lock);
rc = opal_hash_table_get_value_uint64(&mca_btl_tcp_component.tcp_procs, rc = opal_hash_table_get_value_uint64(&mca_btl_tcp_component.tcp_procs,
@ -707,12 +706,12 @@ int mca_btl_tcp_proc_remove(mca_btl_tcp_proc_t* btl_proc, mca_btl_base_endpoint_
* Look for an existing TCP process instance based on the globally unique * Look for an existing TCP process instance based on the globally unique
* process identifier. * process identifier.
*/ */
mca_btl_tcp_proc_t* mca_btl_tcp_proc_lookup(const orca_process_name_t *name) mca_btl_tcp_proc_t* mca_btl_tcp_proc_lookup(const orte_process_name_t *name)
{ {
mca_btl_tcp_proc_t* proc = NULL; mca_btl_tcp_proc_t* proc = NULL;
OPAL_THREAD_LOCK(&mca_btl_tcp_component.tcp_lock); OPAL_THREAD_LOCK(&mca_btl_tcp_component.tcp_lock);
opal_hash_table_get_value_uint64(&mca_btl_tcp_component.tcp_procs, opal_hash_table_get_value_uint64(&mca_btl_tcp_component.tcp_procs,
orca_process_info_hash_name(name), (void**)&proc); orte_util_hash_name(name), (void**)&proc);
OPAL_THREAD_UNLOCK(&mca_btl_tcp_component.tcp_lock); OPAL_THREAD_UNLOCK(&mca_btl_tcp_component.tcp_lock);
return proc; return proc;
} }

5
ompi/mca/btl/tcp/btl_tcp_proc.h
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -23,7 +22,7 @@
#include "opal/class/opal_object.h" #include "opal/class/opal_object.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
#include "orca/include/rte_orca.h" #include "orte/types.h"
#include "btl_tcp.h" #include "btl_tcp.h"
#include "btl_tcp_addr.h" #include "btl_tcp_addr.h"
#include "btl_tcp_endpoint.h" #include "btl_tcp_endpoint.h"
@ -107,7 +106,7 @@ enum mca_btl_tcp_connection_quality {
mca_btl_tcp_proc_t* mca_btl_tcp_proc_create(ompi_proc_t* ompi_proc); mca_btl_tcp_proc_t* mca_btl_tcp_proc_create(ompi_proc_t* ompi_proc);
mca_btl_tcp_proc_t* mca_btl_tcp_proc_lookup(const orca_process_name_t* name); mca_btl_tcp_proc_t* mca_btl_tcp_proc_lookup(const orte_process_name_t* name);
int mca_btl_tcp_proc_insert(mca_btl_tcp_proc_t*, mca_btl_base_endpoint_t*); int mca_btl_tcp_proc_insert(mca_btl_tcp_proc_t*, mca_btl_base_endpoint_t*);
int mca_btl_tcp_proc_remove(mca_btl_tcp_proc_t*, mca_btl_base_endpoint_t*); int mca_btl_tcp_proc_remove(mca_btl_tcp_proc_t*, mca_btl_base_endpoint_t*);
bool mca_btl_tcp_proc_accept(mca_btl_tcp_proc_t*, struct sockaddr*, int); bool mca_btl_tcp_proc_accept(mca_btl_tcp_proc_t*, struct sockaddr*, int);

8
ompi/mca/btl/udapl/btl_udapl.c
Просмотреть файл

@ -13,7 +13,7 @@
* Copyright (c) 2006 Sandia National Laboratories. All rights * Copyright (c) 2006 Sandia National Laboratories. All rights
* reserved. * reserved.
* Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -812,7 +812,7 @@ static int mca_btl_udapl_assign_netmask(mca_btl_udapl_module_t* udapl_btl)
BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP, BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP,
("help-mpi-btl-udapl.txt", "interface not found", ("help-mpi-btl-udapl.txt", "interface not found",
true, orca_process_info_get_nodename(), btl_addr_string)); true, orte_process_info.nodename, btl_addr_string));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -826,7 +826,7 @@ static int mca_btl_udapl_assign_netmask(mca_btl_udapl_module_t* udapl_btl)
BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP, BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP,
("help-mpi-btl-udapl.txt", "netmask not found", ("help-mpi-btl-udapl.txt", "netmask not found",
true, orca_process_info_get_nodename(), btl_addr_string)); true, orte_process_info.nodename, btl_addr_string));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -840,7 +840,7 @@ static int mca_btl_udapl_assign_netmask(mca_btl_udapl_module_t* udapl_btl)
/* current uDAPL BTL does not support IPv6 */ /* current uDAPL BTL does not support IPv6 */
BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP, BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP,
("help-mpi-btl-udapl.txt", "IPv4 only", ("help-mpi-btl-udapl.txt", "IPv4 only",
true, orca_process_info_get_nodename())); true, orte_process_info.nodename));
return OMPI_ERROR; return OMPI_ERROR;
} }

8
ompi/mca/btl/udapl/btl_udapl.h
Просмотреть файл

@ -12,7 +12,7 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -37,9 +37,7 @@
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
#include "ompi/mca/btl/base/base.h" #include "ompi/mca/btl/base/base.h"
#include "ompi/mca/btl/base/btl_base_error.h" #include "ompi/mca/btl/base/btl_base_error.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h"
#include "ompi/mca/mpool/mpool.h" #include "ompi/mca/mpool/mpool.h"
#include "ompi/mca/btl/btl.h" #include "ompi/mca/btl/btl.h"
#include "btl_udapl_endpoint.h" #include "btl_udapl_endpoint.h"
@ -234,7 +232,7 @@ do { \
#define BTL_UDAPL_VERBOSE_HELP(verbose_level, args) \ #define BTL_UDAPL_VERBOSE_HELP(verbose_level, args) \
do { \ do { \
if (verbose_level <= mca_btl_udapl_component.udapl_verbosity) { \ if (verbose_level <= mca_btl_udapl_component.udapl_verbosity) { \
orca_show_help args; \ orte_show_help args; \
} \ } \
} while(0); } while(0);

8
ompi/mca/btl/udapl/btl_udapl_component.c
Просмотреть файл

@ -14,7 +14,7 @@
* reserved. * reserved.
* Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -39,9 +39,7 @@
#include "ompi/mca/btl/base/base.h" #include "ompi/mca/btl/base/base.h"
#include "ompi/mca/btl/base/btl_base_error.h" #include "ompi/mca/btl/base/btl_base_error.h"
#include "btl_udapl_endpoint.h" #include "btl_udapl_endpoint.h"
#include "orte/util/proc_info.h"
#include "orca/include/rte_orca.h"
#include "ompi/runtime/ompi_module_exchange.h" #include "ompi/runtime/ompi_module_exchange.h"
#include "ompi/runtime/mpiruntime.h" #include "ompi/runtime/mpiruntime.h"
@ -419,7 +417,7 @@ static int mca_btl_udapl_modify_ia_list(DAT_COUNT *num_info_entries,
char *str = opal_argv_join(mca_btl_udapl_component.if_list, ','); char *str = opal_argv_join(mca_btl_udapl_component.if_list, ',');
BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP, BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP,
("help-mpi-btl-udapl.txt", "nonexistent entry", ("help-mpi-btl-udapl.txt", "nonexistent entry",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
((NULL != mca_btl_udapl_component.if_include) ? ((NULL != mca_btl_udapl_component.if_include) ?
"in" : "ex"), str)); "in" : "ex"), str));
free(str); free(str);

48
ompi/mca/btl/udapl/btl_udapl_endpoint.c
Просмотреть файл

@ -13,7 +13,7 @@
* Copyright (c) 2006 Sandia National Laboratories. All rights * Copyright (c) 2006 Sandia National Laboratories. All rights
* reserved. * reserved.
* Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2007-2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -28,8 +28,8 @@
#include "ompi/types.h" #include "ompi/types.h"
#include "opal/align.h" #include "opal/align.h"
#include "orca/include/rte_orca.h" #include "orte/mca/rml/rml.h"
#include "orte/mca/errmgr/errmgr.h"
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "opal/class/opal_pointer_array.h" #include "opal/class/opal_pointer_array.h"
@ -46,9 +46,9 @@
static void mca_btl_udapl_endpoint_send_cb( static void mca_btl_udapl_endpoint_send_cb(
int status, int status,
orca_process_name_t* endpoint, orte_process_name_t* endpoint,
opal_buffer_t* buffer, opal_buffer_t* buffer,
orca_oob_tag_t tag, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
static int mca_btl_udapl_start_connect(mca_btl_base_endpoint_t* endpoint); static int mca_btl_udapl_start_connect(mca_btl_base_endpoint_t* endpoint);
static int mca_btl_udapl_endpoint_post_recv( static int mca_btl_udapl_endpoint_post_recv(
@ -57,9 +57,9 @@ static int mca_btl_udapl_endpoint_post_recv(
void mca_btl_udapl_endpoint_connect(mca_btl_udapl_endpoint_t* endpoint); void mca_btl_udapl_endpoint_connect(mca_btl_udapl_endpoint_t* endpoint);
void mca_btl_udapl_endpoint_recv( void mca_btl_udapl_endpoint_recv(
int status, int status,
orca_process_name_t* endpoint, orte_process_name_t* endpoint,
opal_buffer_t* buffer, opal_buffer_t* buffer,
orca_oob_tag_t tag, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
static int mca_btl_udapl_endpoint_finish_eager(mca_btl_udapl_endpoint_t*); static int mca_btl_udapl_endpoint_finish_eager(mca_btl_udapl_endpoint_t*);
static int mca_btl_udapl_endpoint_finish_max(mca_btl_udapl_endpoint_t*); static int mca_btl_udapl_endpoint_finish_max(mca_btl_udapl_endpoint_t*);
@ -322,8 +322,8 @@ int mca_btl_udapl_endpoint_send(mca_btl_base_endpoint_t* endpoint,
} }
static void mca_btl_udapl_endpoint_send_cb(int status, orca_process_name_t* endpoint, static void mca_btl_udapl_endpoint_send_cb(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, void* cbdata) opal_buffer_t* buffer, orte_rml_tag_t tag, void* cbdata)
{ {
OBJ_RELEASE(buffer); OBJ_RELEASE(buffer);
} }
@ -523,7 +523,7 @@ static int mca_btl_udapl_start_connect(mca_btl_base_endpoint_t* endpoint)
int rc; int rc;
if(NULL == buf) { if(NULL == buf) {
ORCA_ERROR_LOG(ORCA_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -532,21 +532,21 @@ static int mca_btl_udapl_start_connect(mca_btl_base_endpoint_t* endpoint)
/* Pack our address information */ /* Pack our address information */
rc = opal_dss.pack(buf, &addr->port, 1, OPAL_UINT64); rc = opal_dss.pack(buf, &addr->port, 1, OPAL_UINT64);
if(OPAL_SUCCESS != rc) { if(OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
rc = opal_dss.pack(buf, &addr->addr, sizeof(DAT_SOCK_ADDR), OPAL_UINT8); rc = opal_dss.pack(buf, &addr->addr, sizeof(DAT_SOCK_ADDR), OPAL_UINT8);
if(OPAL_SUCCESS != rc) { if(OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
/* Send the buffer */ /* Send the buffer */
rc = orca_oob_send_buffer_nb(&endpoint->endpoint_proc->proc_ompi->proc_name, buf, rc = orte_rml.send_buffer_nb(&endpoint->endpoint_proc->proc_ompi->proc_name, buf,
OMPI_RML_TAG_UDAPL, 0, mca_btl_udapl_endpoint_send_cb, NULL); OMPI_RML_TAG_UDAPL, 0, mca_btl_udapl_endpoint_send_cb, NULL);
if(0 > rc) { if(0 > rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -555,8 +555,8 @@ static int mca_btl_udapl_start_connect(mca_btl_base_endpoint_t* endpoint)
} }
void mca_btl_udapl_endpoint_recv(int status, orca_process_name_t* endpoint, void mca_btl_udapl_endpoint_recv(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, void* cbdata) opal_buffer_t* buffer, orte_rml_tag_t tag, void* cbdata)
{ {
mca_btl_udapl_addr_t addr; mca_btl_udapl_addr_t addr;
mca_btl_udapl_proc_t* proc; mca_btl_udapl_proc_t* proc;
@ -568,14 +568,14 @@ void mca_btl_udapl_endpoint_recv(int status, orca_process_name_t* endpoint,
/* Unpack data */ /* Unpack data */
rc = opal_dss.unpack(buffer, &addr.port, &cnt, OPAL_UINT64); rc = opal_dss.unpack(buffer, &addr.port, &cnt, OPAL_UINT64);
if(OPAL_SUCCESS != rc) { if(OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return; return;
} }
cnt = sizeof(mca_btl_udapl_addr_t); cnt = sizeof(mca_btl_udapl_addr_t);
rc = opal_dss.unpack(buffer, &addr.addr, &cnt, OPAL_UINT8); rc = opal_dss.unpack(buffer, &addr.addr, &cnt, OPAL_UINT8);
if(OPAL_SUCCESS != rc) { if(OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return; return;
} }
@ -587,7 +587,7 @@ void mca_btl_udapl_endpoint_recv(int status, orca_process_name_t* endpoint,
opal_list_get_end(&mca_btl_udapl_component.udapl_procs); opal_list_get_end(&mca_btl_udapl_component.udapl_procs);
proc = (mca_btl_udapl_proc_t*)opal_list_get_next(proc)) { proc = (mca_btl_udapl_proc_t*)opal_list_get_next(proc)) {
if(OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, &proc->proc_ompi->proc_name, endpoint)) { if(OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL, &proc->proc_ompi->proc_name, endpoint)) {
for(i = 0; i < proc->proc_endpoint_count; i++) { for(i = 0; i < proc->proc_endpoint_count; i++) {
ep = proc->proc_endpoints[i]; ep = proc->proc_endpoints[i];
@ -613,8 +613,8 @@ void mca_btl_udapl_endpoint_recv(int status, orca_process_name_t* endpoint,
void mca_btl_udapl_endpoint_post_oob_recv(void) void mca_btl_udapl_endpoint_post_oob_recv(void)
{ {
orca_oob_recv_buffer_nb(ORCA_NAME_WILDCARD, OMPI_RML_TAG_UDAPL, orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, OMPI_RML_TAG_UDAPL,
ORCA_OOB_PERSISTENT, mca_btl_udapl_endpoint_recv, NULL); ORTE_RML_PERSISTENT, mca_btl_udapl_endpoint_recv, NULL);
} }
@ -631,7 +631,7 @@ void mca_btl_udapl_endpoint_connect(mca_btl_udapl_endpoint_t* endpoint)
/* Nasty test to prevent deadlock and unwanted connection attempts */ /* Nasty test to prevent deadlock and unwanted connection attempts */
/* This right here is the whole point of using the ORTE/RML handshake */ /* This right here is the whole point of using the ORTE/RML handshake */
if((MCA_BTL_UDAPL_CONN_EAGER == endpoint->endpoint_state && if((MCA_BTL_UDAPL_CONN_EAGER == endpoint->endpoint_state &&
0 > orca_process_name_compare(ORCA_NAME_CMP_ALL, 0 > orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&endpoint->endpoint_proc->proc_ompi->proc_name, &endpoint->endpoint_proc->proc_ompi->proc_name,
&ompi_proc_local()->proc_name)) || &ompi_proc_local()->proc_name)) ||
(MCA_BTL_UDAPL_CLOSED != endpoint->endpoint_state && (MCA_BTL_UDAPL_CLOSED != endpoint->endpoint_state &&
@ -782,7 +782,7 @@ static int mca_btl_udapl_endpoint_finish_eager(
} }
/* Only one side does dat_ep_connect() */ /* Only one side does dat_ep_connect() */
if(0 < orca_process_name_compare(ORCA_NAME_CMP_ALL, if(0 < orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&endpoint->endpoint_proc->proc_ompi->proc_name, &endpoint->endpoint_proc->proc_ompi->proc_name,
&ompi_proc_local()->proc_name)) { &ompi_proc_local()->proc_name)) {
@ -970,7 +970,7 @@ static int mca_btl_udapl_endpoint_pd_finish_eager(
* with this. * with this.
*/ */
if((BTL_UDAPL_NUM_CONNECTION != endpoint->endpoint_connections_completed) if((BTL_UDAPL_NUM_CONNECTION != endpoint->endpoint_connections_completed)
&& (0 < orca_process_name_compare(ORCA_NAME_CMP_ALL, && (0 < orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&endpoint->endpoint_proc->proc_ompi->proc_name, &endpoint->endpoint_proc->proc_ompi->proc_name,
&ompi_proc_local()->proc_name))) { &ompi_proc_local()->proc_name))) {

9
ompi/mca/btl/udapl/btl_udapl_proc.c
Просмотреть файл

@ -12,7 +12,6 @@
* Copyright (c) 2006 Sandia National Laboratories. All rights * Copyright (c) 2006 Sandia National Laboratories. All rights
* reserved. * reserved.
* Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved. * Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -132,7 +131,7 @@ mca_btl_udapl_proc_t* mca_btl_udapl_proc_create(ompi_proc_t* ompi_proc)
if(OMPI_SUCCESS != rc) { if(OMPI_SUCCESS != rc) {
BTL_UDAPL_VERBOSE_OUTPUT(VERBOSE_CRITICAL, BTL_UDAPL_VERBOSE_OUTPUT(VERBOSE_CRITICAL,
("ompi_modex_recv failed for peer %s", ("ompi_modex_recv failed for peer %s",
ORCA_NAME_PRINT(&ompi_proc->proc_name))); ORTE_NAME_PRINT(&ompi_proc->proc_name)));
OBJ_RELEASE(udapl_proc); OBJ_RELEASE(udapl_proc);
return NULL; return NULL;
} }
@ -140,7 +139,7 @@ mca_btl_udapl_proc_t* mca_btl_udapl_proc_create(ompi_proc_t* ompi_proc)
if((size % sizeof(mca_btl_udapl_addr_t)) != 0) { if((size % sizeof(mca_btl_udapl_addr_t)) != 0) {
BTL_UDAPL_VERBOSE_OUTPUT(VERBOSE_CRITICAL, BTL_UDAPL_VERBOSE_OUTPUT(VERBOSE_CRITICAL,
("invalid udapl address for peer %s", ("invalid udapl address for peer %s",
ORCA_NAME_PRINT(&ompi_proc->proc_name))); ORTE_NAME_PRINT(&ompi_proc->proc_name)));
OBJ_RELEASE(udapl_proc); OBJ_RELEASE(udapl_proc);
return NULL; return NULL;
} }
@ -252,14 +251,14 @@ static int mca_btl_udapl_proc_address_match(
/* current uDAPL BTL only supports IPv4 */ /* current uDAPL BTL only supports IPv4 */
BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP, BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP,
("help-mpi-btl-udapl.txt", "IPv4 only", ("help-mpi-btl-udapl.txt", "IPv4 only",
true, orca_process_info_get_nodename())); true, orte_process_info.nodename));
return OMPI_ERROR; return OMPI_ERROR;
} }
if (MCA_BTL_UDAPL_INVALID_PEER_ADDR_IDX == *peer_addr_idx) { if (MCA_BTL_UDAPL_INVALID_PEER_ADDR_IDX == *peer_addr_idx) {
BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP, BTL_UDAPL_VERBOSE_HELP(VERBOSE_SHOW_HELP,
("help-mpi-btl-udapl.txt", "no network match", ("help-mpi-btl-udapl.txt", "no network match",
true, btl_addr_string, orca_process_info_get_nodename(), true, btl_addr_string, orte_process_info.nodename,
peer_proc->proc_ompi->proc_hostname)); peer_proc->proc_ompi->proc_hostname));
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }

5
ompi/mca/btl/vader/btl_vader.c
Просмотреть файл

@ -14,7 +14,6 @@
* Copyright (c) 2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -173,8 +172,8 @@ static int vader_btl_first_time_init(mca_btl_vader_t *vader_btl, int n)
/* set file name */ /* set file name */
if(asprintf(&vader_ctl_file, "%s"OPAL_PATH_SEP"vader_btl_module.%s", if(asprintf(&vader_ctl_file, "%s"OPAL_PATH_SEP"vader_btl_module.%s",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
orca_process_info_get_nodename()) < 0) orte_process_info.nodename) < 0)
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
/* Pass in a data segment alignment of 0 to get no data /* Pass in a data segment alignment of 0 to get no data

8
ompi/mca/btl/vader/btl_vader_component.c
Просмотреть файл

@ -15,7 +15,6 @@
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2011 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -26,8 +25,9 @@
#include "ompi/constants.h" #include "ompi/constants.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orte/util/proc_info.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "ompi/mca/btl/base/btl_base_error.h" #include "ompi/mca/btl/base/btl_base_error.h"
@ -229,7 +229,7 @@ static mca_btl_base_module_t **mca_btl_vader_component_init (int *num_btls,
/* if no session directory was created, then we cannot be used */ /* if no session directory was created, then we cannot be used */
/* XXX LANL FIXME -- this is not the case. we can use an anonymous segment */ /* XXX LANL FIXME -- this is not the case. we can use an anonymous segment */
if (!orca_process_info_create_session_dirs()) { if (!orte_create_session_dirs) {
return NULL; return NULL;
} }

35
ompi/mca/btl/wv/btl_wv.c
Просмотреть файл

@ -17,7 +17,6 @@
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2008-2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2008-2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2009 IBM Corporation. All rights reserved. * Copyright (c) 2009 IBM Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -28,8 +27,8 @@
#include "ompi_config.h" #include "ompi_config.h"
#include <string.h> #include <string.h>
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "opal/class/opal_bitmap.h" #include "opal/class/opal_bitmap.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/util/arch.h" #include "opal/util/arch.h"
@ -46,7 +45,7 @@
#include "opal/datatype/opal_convertor.h" #include "opal/datatype/opal_convertor.h"
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
#include "ompi/mca/mpool/grdma/mpool_grdma.h" #include "ompi/mca/mpool/grdma/mpool_grdma.h"
#include "orte/util/proc_info.h"
#include <errno.h> #include <errno.h>
#include <string.h> #include <string.h>
#include <math.h> #include <math.h>
@ -115,14 +114,14 @@ void mca_btl_wv_show_init_error(const char *file, int line,
{ {
if (ENOMEM == errno) {char *str_limit = NULL; if (ENOMEM == errno) {char *str_limit = NULL;
orca_show_help("help-mpi-btl-wv.txt", "init-fail-no-mem", orte_show_help("help-mpi-btl-wv.txt", "init-fail-no-mem",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
file, line, func, dev, str_limit); file, line, func, dev, str_limit);
if (NULL != str_limit) free(str_limit); if (NULL != str_limit) free(str_limit);
} else { } else {
orca_show_help("help-mpi-btl-wv.txt", "init-fail-create-q", orte_show_help("help-mpi-btl-wv.txt", "init-fail-create-q",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
file, line, func, strerror(errno), errno, dev); file, line, func, strerror(errno), errno, dev);
} }
} }
@ -288,9 +287,9 @@ static int mca_btl_wv_tune_endpoint(mca_btl_wv_module_t* wv_btl,
ompi_btl_wv_ini_values_t values; ompi_btl_wv_ini_values_t values;
if(mca_btl_wv_get_transport_type(wv_btl) != endpoint->rem_info.rem_transport_type) { if(mca_btl_wv_get_transport_type(wv_btl) != endpoint->rem_info.rem_transport_type) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"conflicting transport types", true, "conflicting transport types", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
wv_btl->device->ib_dev->name, wv_btl->device->ib_dev->name,
(wv_btl->device->ib_dev_attr).VendorId, (wv_btl->device->ib_dev_attr).VendorId,
(wv_btl->device->ib_dev_attr).VendorPartId, (wv_btl->device->ib_dev_attr).VendorPartId,
@ -309,9 +308,9 @@ static int mca_btl_wv_tune_endpoint(mca_btl_wv_module_t* wv_btl,
if (OMPI_SUCCESS != ret && if (OMPI_SUCCESS != ret &&
OMPI_ERR_NOT_FOUND != ret) { OMPI_ERR_NOT_FOUND != ret) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"error in device init", true, "error in device init", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
wv_btl->device->ib_dev->name); wv_btl->device->ib_dev->name);
return ret; return ret;
} }
@ -350,9 +349,9 @@ static int mca_btl_wv_tune_endpoint(mca_btl_wv_module_t* wv_btl,
if(0 != strcmp(mca_btl_wv_component.receive_queues, if(0 != strcmp(mca_btl_wv_component.receive_queues,
recv_qps)) { recv_qps)) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"unsupported queues configuration", true, "unsupported queues configuration", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
wv_btl->device->ib_dev->name, wv_btl->device->ib_dev->name,
(wv_btl->device->ib_dev_attr).VendorId, (wv_btl->device->ib_dev_attr).VendorId,
(wv_btl->device->ib_dev_attr).VendorPartId, (wv_btl->device->ib_dev_attr).VendorPartId,
@ -372,9 +371,9 @@ static int mca_btl_wv_tune_endpoint(mca_btl_wv_module_t* wv_btl,
if(NULL != values.receive_queues) { if(NULL != values.receive_queues) {
if(0 != strcmp(mca_btl_wv_component.receive_queues, if(0 != strcmp(mca_btl_wv_component.receive_queues,
values.receive_queues)) { values.receive_queues)) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"unsupported queues configuration", true, "unsupported queues configuration", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
wv_btl->device->ib_dev->name, wv_btl->device->ib_dev->name,
(wv_btl->device->ib_dev_attr).VendorId, (wv_btl->device->ib_dev_attr).VendorId,
(wv_btl->device->ib_dev_attr).VendorPartId, (wv_btl->device->ib_dev_attr).VendorPartId,
@ -431,8 +430,8 @@ int mca_btl_wv_add_procs(struct mca_btl_base_module_t* btl,
/* OOB, XOOB, RDMACM, IBCM does not support SELF comunication, so /* OOB, XOOB, RDMACM, IBCM does not support SELF comunication, so
* mark the prco as unreachable by wv btl */ * mark the prco as unreachable by wv btl */
if (OPAL_EQUAL == orca_process_name_compare if (OPAL_EQUAL == orte_util_compare_name_fields
(ORCA_NAME_CMP_ALL, ORCA_PROC_MY_NAME, &ompi_proc->proc_name)) { (ORTE_NS_CMP_ALL, ORTE_PROC_MY_NAME, &ompi_proc->proc_name)) {
continue; continue;
} }

97
ompi/mca/btl/wv/btl_wv_component.c
Просмотреть файл

@ -16,7 +16,6 @@
* reserved. * reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -46,7 +45,9 @@
#include "opal/mca/installdirs/installdirs.h" #include "opal/mca/installdirs/installdirs.h"
#include "opal_stdint.h" #include "opal_stdint.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/proc_info.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/constants.h" #include "ompi/constants.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
@ -515,8 +516,8 @@ static int init_one_port(opal_list_t *btl_list, mca_btl_wv_device_t *device,
if(mca_btl_wv_component.ib_num_btls > 0 && if(mca_btl_wv_component.ib_num_btls > 0 &&
IB_DEFAULT_GID_PREFIX == subnet_id && IB_DEFAULT_GID_PREFIX == subnet_id &&
mca_btl_wv_component.warn_default_gid_prefix) { mca_btl_wv_component.warn_default_gid_prefix) {
orca_show_help("help-mpi-btl-wv.txt", "default subnet prefix", orte_show_help("help-mpi-btl-wv.txt", "default subnet prefix",
true, orca_process_info_get_nodename()); true, orte_process_info.nodename);
} }
lmc = (1 << ib_port_attr->Lmc); lmc = (1 << ib_port_attr->Lmc);
@ -1037,9 +1038,9 @@ static int setup_qps(void)
queues = opal_argv_split(mca_btl_wv_component.receive_queues, ':'); queues = opal_argv_split(mca_btl_wv_component.receive_queues, ':');
if (0 == opal_argv_count(queues)) { if (0 == opal_argv_count(queues)) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"no qps in receive_queues", true, "no qps in receive_queues", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_wv_component.receive_queues); mca_btl_wv_component.receive_queues);
ret = OMPI_ERROR; ret = OMPI_ERROR;
goto error; goto error;
@ -1054,9 +1055,9 @@ static int setup_qps(void)
} else if (0 == strncmp("S,", queues[qp], 2)) { } else if (0 == strncmp("S,", queues[qp], 2)) {
num_srq_qps++; num_srq_qps++;
}else { }else {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"invalid qp type in receive_queues", true, "invalid qp type in receive_queues", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_wv_component.receive_queues, mca_btl_wv_component.receive_queues,
queues[qp]); queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
@ -1083,9 +1084,9 @@ static int setup_qps(void)
if ('P' == params[0][0]) { if ('P' == params[0][0]) {
int32_t rd_win, rd_rsv; int32_t rd_win, rd_rsv;
if (count < 3 || count > 6) { if (count < 3 || count > 6) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"invalid pp qp specification", true, "invalid pp qp specification", true,
orca_process_info_get_nodename(), queues[qp]); orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1108,15 +1109,15 @@ static int setup_qps(void)
mca_btl_wv_component.qp_infos[qp].u.pp_qp.rd_win = rd_win; mca_btl_wv_component.qp_infos[qp].u.pp_qp.rd_win = rd_win;
mca_btl_wv_component.qp_infos[qp].u.pp_qp.rd_rsv = rd_rsv; mca_btl_wv_component.qp_infos[qp].u.pp_qp.rd_rsv = rd_rsv;
if ((rd_num - rd_low) > rd_win) { if ((rd_num - rd_low) > rd_win) {
orca_show_help("help-mpi-btl-wv.txt", "non optimal rd_win", orte_show_help("help-mpi-btl-wv.txt", "non optimal rd_win",
true, rd_win, rd_num - rd_low); true, rd_win, rd_num - rd_low);
} }
} else { } else {
int32_t sd_max, rd_init, srq_limit; int32_t sd_max, rd_init, srq_limit;
if (count < 3 || count > 7) { if (count < 3 || count > 7) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"invalid srq specification", true, "invalid srq specification", true,
orca_process_info_get_nodename(), queues[qp]); orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1145,15 +1146,15 @@ static int setup_qps(void)
} }
if (rd_num < rd_init) { if (rd_num < rd_init) {
orca_show_help("help-mpi-btl-wv.txt", "rd_num must be >= rd_init", orte_show_help("help-mpi-btl-wv.txt", "rd_num must be >= rd_init",
true, orca_process_info_get_nodename(), queues[qp]); true, orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
if (rd_num < srq_limit) { if (rd_num < srq_limit) {
orca_show_help("help-mpi-btl-wv.txt", "srq_limit must be > rd_num", orte_show_help("help-mpi-btl-wv.txt", "srq_limit must be > rd_num",
true, orca_process_info_get_nodename(), queues[qp]); true, orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1164,8 +1165,8 @@ static int setup_qps(void)
} }
if (rd_num <= rd_low) { if (rd_num <= rd_low) {
orca_show_help("help-mpi-btl-wv.txt", "rd_num must be > rd_low", orte_show_help("help-mpi-btl-wv.txt", "rd_num must be > rd_low",
true, orca_process_info_get_nodename(), queues[qp]); true, orte_process_info.nodename, queues[qp]);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} }
@ -1184,23 +1185,23 @@ static int setup_qps(void)
mca_btl_wv_module.super.btl_eager_limit : mca_btl_wv_module.super.btl_eager_limit :
mca_btl_wv_module.super.btl_max_send_size; mca_btl_wv_module.super.btl_max_send_size;
if (max_qp_size < max_size_needed) { if (max_qp_size < max_size_needed) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"biggest qp size is too small", true, "biggest qp size is too small", true,
orca_process_info_get_nodename(), max_qp_size, orte_process_info.nodename, max_qp_size,
max_size_needed); max_size_needed);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
goto error; goto error;
} else if (max_qp_size > max_size_needed) { } else if (max_qp_size > max_size_needed) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"biggest qp size is too big", true, "biggest qp size is too big", true,
orca_process_info_get_nodename(), max_qp_size, orte_process_info.nodename, max_qp_size,
max_size_needed); max_size_needed);
} }
if (mca_btl_wv_component.ib_free_list_max > 0 && if (mca_btl_wv_component.ib_free_list_max > 0 &&
min_freelist_size > mca_btl_wv_component.ib_free_list_max) { min_freelist_size > mca_btl_wv_component.ib_free_list_max) {
orca_show_help("help-mpi-btl-wv.txt", "freelist too small", true, orte_show_help("help-mpi-btl-wv.txt", "freelist too small", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
mca_btl_wv_component.ib_free_list_max, mca_btl_wv_component.ib_free_list_max,
min_freelist_size); min_freelist_size);
ret = OMPI_ERR_BAD_PARAM; ret = OMPI_ERR_BAD_PARAM;
@ -1316,9 +1317,9 @@ static int init_one_device(opal_list_t *btl_list, struct wv_device* ib_dev)
warning that we're using default values (unless overridden warning that we're using default values (unless overridden
that we don't want to see these warnings) */ that we don't want to see these warnings) */
if (mca_btl_wv_component.warn_no_device_params_found) { if (mca_btl_wv_component.warn_no_device_params_found) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"no device params found", true, "no device params found", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
device->ib_dev->name, device->ib_dev->name,
device->ib_dev_attr.VendorId, device->ib_dev_attr.VendorId,
device->ib_dev_attr.VendorPartId); device->ib_dev_attr.VendorPartId);
@ -1491,7 +1492,7 @@ static int init_one_device(opal_list_t *btl_list, struct wv_device* ib_dev)
if (device->btls > 0) { if (device->btls > 0) {
/* if apm was enabled it should be > 1 */ /* if apm was enabled it should be > 1 */
if (1 == mca_btl_wv_component.apm_ports) { if (1 == mca_btl_wv_component.apm_ports) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"apm not enough ports", true); "apm not enough ports", true);
mca_btl_wv_component.apm_ports = 0; mca_btl_wv_component.apm_ports = 0;
} }
@ -1750,10 +1751,10 @@ static int init_one_device(opal_list_t *btl_list, struct wv_device* ib_dev)
if (NULL != values.receive_queues) { if (NULL != values.receive_queues) {
if (0 != strcmp(values.receive_queues, if (0 != strcmp(values.receive_queues,
mca_btl_wv_component.receive_queues)) { mca_btl_wv_component.receive_queues)) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"locally conflicting receive_queues", true, "locally conflicting receive_queues", true,
opal_install_dirs.pkgdatadir, opal_install_dirs.pkgdatadir,
orca_process_info_get_nodename(), orte_process_info.nodename,
receive_queues_device->ib_dev->name, receive_queues_device->ib_dev->name,
receive_queues_device->ib_dev_attr.VendorId, receive_queues_device->ib_dev_attr.VendorId,
receive_queues_device->ib_dev_attr.VendorPartId, receive_queues_device->ib_dev_attr.VendorPartId,
@ -1774,10 +1775,10 @@ static int init_one_device(opal_list_t *btl_list, struct wv_device* ib_dev)
device's INI file, we must error. */ device's INI file, we must error. */
else if (BTL_WV_RQ_SOURCE_DEVICE_INI == else if (BTL_WV_RQ_SOURCE_DEVICE_INI ==
mca_btl_wv_component.receive_queues_source) { mca_btl_wv_component.receive_queues_source) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"locally conflicting receive_queues", true, "locally conflicting receive_queues", true,
opal_install_dirs.pkgdatadir, opal_install_dirs.pkgdatadir,
orca_process_info_get_nodename(), orte_process_info.nodename,
receive_queues_device->ib_dev->name, receive_queues_device->ib_dev->name,
receive_queues_device->ib_dev_attr.VendorId, receive_queues_device->ib_dev_attr.VendorId,
receive_queues_device->ib_dev_attr.VendorPartId, receive_queues_device->ib_dev_attr.VendorPartId,
@ -1808,9 +1809,9 @@ error:
} }
if (OMPI_SUCCESS != ret) { if (OMPI_SUCCESS != ret) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"error in device init", true, "error in device init", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
device->ib_dev->name); device->ib_dev->name);
} }
device->ib_dev_context->device_if->Release(); device->ib_dev_context->device_if->Release();
@ -2092,7 +2093,7 @@ sort_devs_by_distance(struct wv_device **ib_devs, int count)
for (i = 0; i < count; i++) { for (i = 0; i < count; i++) {
devs[i].ib_dev = ib_devs[i]; devs[i].ib_dev = ib_devs[i];
if (orca_process_info_is_bound()) { if (orte_proc_is_bound) {
/* If this process is bound to one or more PUs, we can get /* If this process is bound to one or more PUs, we can get
an accurate distance. */ an accurate distance. */
devs[i].distance = get_ib_dev_distance(ib_devs[i]); devs[i].distance = get_ib_dev_distance(ib_devs[i]);
@ -2281,7 +2282,7 @@ btl_wv_component_init(int *num_btl_modules,
list_count++; list_count++;
if (list_count > 1) { if (list_count > 1) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"specified include and exclude", true, "specified include and exclude", true,
NULL == mca_btl_wv_component.if_include ? NULL == mca_btl_wv_component.if_include ?
"<not specified>" : mca_btl_wv_component.if_include, "<not specified>" : mca_btl_wv_component.if_include,
@ -2334,8 +2335,8 @@ btl_wv_component_init(int *num_btl_modules,
} }
free(dev_sorted); free(dev_sorted);
if (!found) { if (!found) {
orca_show_help("help-mpi-btl-wv.txt", "no devices right type", orte_show_help("help-mpi-btl-wv.txt", "no devices right type",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
((BTL_WV_DT_IB == mca_btl_wv_component.device_type) ? ((BTL_WV_DT_IB == mca_btl_wv_component.device_type) ?
"InfiniBand" : "InfiniBand" :
(BTL_WV_DT_IWARP == mca_btl_wv_component.device_type) ? (BTL_WV_DT_IWARP == mca_btl_wv_component.device_type) ?
@ -2351,16 +2352,16 @@ btl_wv_component_init(int *num_btl_modules,
if (0 != opal_argv_count(mca_btl_wv_component.if_list) && if (0 != opal_argv_count(mca_btl_wv_component.if_list) &&
mca_btl_wv_component.warn_nonexistent_if) { mca_btl_wv_component.warn_nonexistent_if) {
char *str = opal_argv_join(mca_btl_wv_component.if_list, ','); char *str = opal_argv_join(mca_btl_wv_component.if_list, ',');
orca_show_help("help-mpi-btl-wv.txt", "nonexistent port", orte_show_help("help-mpi-btl-wv.txt", "nonexistent port",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
((NULL != mca_btl_wv_component.if_include) ? ((NULL != mca_btl_wv_component.if_include) ?
"in" : "ex"), str); "in" : "ex"), str);
free(str); free(str);
} }
if(0 == mca_btl_wv_component.ib_num_btls) { if(0 == mca_btl_wv_component.ib_num_btls) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"no active ports found", true, orca_process_info_get_nodename()); "no active ports found", true, orte_process_info.nodename);
goto no_btls; goto no_btls;
} }
@ -2449,9 +2450,9 @@ btl_wv_component_init(int *num_btl_modules,
/* Do finial init on device */ /* Do finial init on device */
ret = prepare_device_for_use(device); ret = prepare_device_for_use(device);
if (OMPI_SUCCESS != ret) { if (OMPI_SUCCESS != ret) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"error in device init", true, "error in device init", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
device->ib_dev->name); device->ib_dev->name);
goto no_btls; goto no_btls;
} }
@ -2973,16 +2974,16 @@ error:
(endpoint->qps[qp].qp->lcl_qp->context->device->name); (endpoint->qps[qp].qp->lcl_qp->context->device->name);
if (WvWcRnrRetryError == wc->Status) { if (WvWcRnrRetryError == wc->Status) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
BTL_WV_QP_TYPE_PP(qp) ? BTL_WV_QP_TYPE_PP(qp) ?
"pp rnr retry exceeded" : "pp rnr retry exceeded" :
"srq rnr retry exceeded", true, "srq rnr retry exceeded", true,
orca_process_info_get_nodename(), device_name, orte_process_info.nodename, device_name,
peer_hostname); peer_hostname);
} else if (-2 == wc->Status) { } else if (-2 == wc->Status) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"pp retry exceeded", true, "pp retry exceeded", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
device_name, peer_hostname); device_name, peer_hostname);
} }
} }

8
ompi/mca/btl/wv/btl_wv_endpoint.c
Просмотреть файл

@ -17,7 +17,7 @@
* Copyright (c) 2006-2009 Mellanox Technologies, Inc. All rights reserved. * Copyright (c) 2006-2009 Mellanox Technologies, Inc. All rights reserved.
* Copyright (c) 2010 IBM Corporation. All rights reserved. * Copyright (c) 2010 IBM Corporation. All rights reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -34,7 +34,7 @@
#include "opal_stdint.h" #include "opal_stdint.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "ompi/types.h" #include "ompi/types.h"
#include "ompi/class/ompi_free_list.h" #include "ompi/class/ompi_free_list.h"
@ -930,9 +930,9 @@ void *mca_btl_wv_endpoint_invoke_error(void *context)
/* If we didn't find a BTL, then just bail :-( */ /* If we didn't find a BTL, then just bail :-( */
if (NULL == btl || NULL == btl->error_cb) { if (NULL == btl || NULL == btl->error_cb) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"cannot raise btl error", true, "cannot raise btl error", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
__FILE__, __LINE__); __FILE__, __LINE__);
exit(1); exit(1);
} }

10
ompi/mca/btl/wv/btl_wv_ini.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2008 Mellanox Technologies. All rights reserved. * Copyright (c) 2008 Mellanox Technologies. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -31,8 +30,7 @@
#include <unistd.h> #include <unistd.h>
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "btl_wv.h" #include "btl_wv.h"
@ -243,7 +241,7 @@ static int parse_file(char *filename)
ini_filename = filename; ini_filename = filename;
btl_wv_ini_yyin = fopen(filename, "r"); btl_wv_ini_yyin = fopen(filename, "r");
if (NULL == btl_wv_ini_yyin) { if (NULL == btl_wv_ini_yyin) {
orca_show_help("help-mpi-btl-wv.txt", "ini file:file not found", orte_show_help("help-mpi-btl-wv.txt", "ini file:file not found",
true, filename); true, filename);
ret = OMPI_ERR_NOT_FOUND; ret = OMPI_ERR_NOT_FOUND;
goto cleanup; goto cleanup;
@ -420,7 +418,7 @@ static int parse_line(parsed_section_values_t *sv)
/* Have no idea what this parameter is. Not an error -- just /* Have no idea what this parameter is. Not an error -- just
ignore it */ ignore it */
if (!showed_unknown_field_warning) { if (!showed_unknown_field_warning) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"ini file:unknown field", true, "ini file:unknown field", true,
ini_filename, btl_wv_ini_yynewlines, ini_filename, btl_wv_ini_yynewlines,
key_buffer); key_buffer);
@ -689,7 +687,7 @@ static inline void show_help(const char *topic)
if (0 == strcmp("\n", btl_wv_ini_yytext)) { if (0 == strcmp("\n", btl_wv_ini_yytext)) {
btl_wv_ini_yytext = "<end of line>"; btl_wv_ini_yytext = "<end of line>";
} }
orca_show_help("help-mpi-btl-wv.txt", topic, true, orte_show_help("help-mpi-btl-wv.txt", topic, true,
ini_filename, btl_wv_ini_yynewlines, ini_filename, btl_wv_ini_yynewlines,
btl_wv_ini_yytext); btl_wv_ini_yytext);
btl_wv_ini_yytext = save; btl_wv_ini_yytext = save;

39
ompi/mca/btl/wv/btl_wv_mca.c
Просмотреть файл

@ -15,7 +15,6 @@
* reserved. * reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2009-2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -31,9 +30,7 @@
#include "opal/mca/installdirs/installdirs.h" #include "opal/mca/installdirs/installdirs.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h"
#include "btl_wv.h" #include "btl_wv.h"
#include "btl_wv_mca.h" #include "btl_wv_mca.h"
#include "btl_wv_ini.h" #include "btl_wv_ini.h"
@ -169,9 +166,9 @@ int btl_wv_register_mca_params(void)
"(negative = try to enable fork support, but continue even if it is not available, 0 = do not enable fork support, positive = try to enable fork support and fail if it is not available)", "(negative = try to enable fork support, but continue even if it is not available, 0 = do not enable fork support, positive = try to enable fork support and fail if it is not available)",
ival2, &ival, 0)); ival2, &ival, 0));
if (0 != ival) { if (0 != ival) {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"ib_fork requested but not supported", true, "ib_fork requested but not supported", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -199,9 +196,9 @@ int btl_wv_register_mca_params(void)
} else if (0 == strcasecmp(str, "all")) { } else if (0 == strcasecmp(str, "all")) {
mca_btl_wv_component.device_type = BTL_WV_DT_ALL; mca_btl_wv_component.device_type = BTL_WV_DT_ALL;
} else { } else {
orca_show_help("help-mpi-btl-wv.txt", orte_show_help("help-mpi-btl-wv.txt",
"ib_fork requested but not supported", true, "ib_fork requested but not supported", true,
orca_process_info_get_nodename()); orte_process_info.nodename);
return OMPI_ERROR; return OMPI_ERROR;
} }
free(str); free(str);
@ -285,7 +282,7 @@ int btl_wv_register_mca_params(void)
CHECK(reg_int("mtu", "ib_mtu", msg, WV_MTU_1024, &ival, 0)); CHECK(reg_int("mtu", "ib_mtu", msg, WV_MTU_1024, &ival, 0));
free(msg); free(msg);
if (ival < WV_MTU_1024 || ival > WV_MTU_4096) { if (ival < WV_MTU_1024 || ival > WV_MTU_4096) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "invalid value for btl_wv_ib_mtu", true, "invalid value for btl_wv_ib_mtu",
"btl_wv_ib_mtu reset to 1024"); "btl_wv_ib_mtu reset to 1024");
mca_btl_wv_component.ib_mtu = WV_MTU_1024; mca_btl_wv_component.ib_mtu = WV_MTU_1024;
@ -298,12 +295,12 @@ int btl_wv_register_mca_params(void)
"(must be >= 0 and <= 31)", "(must be >= 0 and <= 31)",
25, &ival, 0)); 25, &ival, 0));
if (ival > 31) { if (ival > 31) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_min_rnr_timer > 31", true, "btl_wv_ib_min_rnr_timer > 31",
"btl_wv_ib_min_rnr_timer reset to 31"); "btl_wv_ib_min_rnr_timer reset to 31");
ival = 31; ival = 31;
} else if (ival < 0){ } else if (ival < 0){
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_min_rnr_timer < 0", true, "btl_wv_ib_min_rnr_timer < 0",
"btl_wv_ib_min_rnr_timer reset to 0"); "btl_wv_ib_min_rnr_timer reset to 0");
ival = 0; ival = 0;
@ -315,12 +312,12 @@ int btl_wv_register_mca_params(void)
"(must be >= 0 and <= 31)", "(must be >= 0 and <= 31)",
20, &ival, 0)); 20, &ival, 0));
if (ival > 31) { if (ival > 31) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_timeout > 31", true, "btl_wv_ib_timeout > 31",
"btl_wv_ib_timeout reset to 31"); "btl_wv_ib_timeout reset to 31");
ival = 31; ival = 31;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_timeout < 0", true, "btl_wv_ib_timeout < 0",
"btl_wv_ib_timeout reset to 0"); "btl_wv_ib_timeout reset to 0");
ival = 0; ival = 0;
@ -332,12 +329,12 @@ int btl_wv_register_mca_params(void)
"(must be >= 0 and <= 7)", "(must be >= 0 and <= 7)",
7, &ival, 0)); 7, &ival, 0));
if (ival > 7) { if (ival > 7) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_retry_count > 7", true, "btl_wv_ib_retry_count > 7",
"btl_wv_ib_retry_count reset to 7"); "btl_wv_ib_retry_count reset to 7");
ival = 7; ival = 7;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_retry_count < 0", true, "btl_wv_ib_retry_count < 0",
"btl_wv_ib_retry_count reset to 0"); "btl_wv_ib_retry_count reset to 0");
ival = 0; ival = 0;
@ -352,12 +349,12 @@ int btl_wv_register_mca_params(void)
"(must be >= 0 and <= 7; 7 = \"infinite\")", "(must be >= 0 and <= 7; 7 = \"infinite\")",
7, &ival, 0)); 7, &ival, 0));
if (ival > 7) { if (ival > 7) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_rnr_retry > 7", true, "btl_wv_ib_rnr_retry > 7",
"btl_wv_ib_rnr_retry reset to 7"); "btl_wv_ib_rnr_retry reset to 7");
ival = 7; ival = 7;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_rnr_retry < 0", true, "btl_wv_ib_rnr_retry < 0",
"btl_wv_ib_rnr_retry reset to 0"); "btl_wv_ib_rnr_retry reset to 0");
ival = 0; ival = 0;
@ -368,12 +365,12 @@ int btl_wv_register_mca_params(void)
"(must be >= 0 and <= 15)", "(must be >= 0 and <= 15)",
0, &ival, 0)); 0, &ival, 0));
if (ival > 15) { if (ival > 15) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_service_level > 15", true, "btl_wv_ib_service_level > 15",
"btl_wv_ib_service_level reset to 15"); "btl_wv_ib_service_level reset to 15");
ival = 15; ival = 15;
} else if (ival < 0) { } else if (ival < 0) {
orca_show_help("help-mpi-btl-wv.txt", "invalid mca param value", orte_show_help("help-mpi-btl-wv.txt", "invalid mca param value",
true, "btl_wv_ib_service_level < 0", true, "btl_wv_ib_service_level < 0",
"btl_wv_ib_service_level reset to 0"); "btl_wv_ib_service_level reset to 0");
ival = 0; ival = 0;
@ -430,8 +427,8 @@ int btl_wv_register_mca_params(void)
"(must be > 0 and power of two)", "(must be > 0 and power of two)",
64, &ival, REGINT_GE_ZERO)); 64, &ival, REGINT_GE_ZERO));
if(ival <= 1 || (ival & (ival - 1))) { if(ival <= 1 || (ival & (ival - 1))) {
orca_show_help("help-mpi-btl-wv.txt", "wrong buffer alignment", orte_show_help("help-mpi-btl-wv.txt", "wrong buffer alignment",
true, ival, orca_process_info_get_nodename(), 64); true, ival, orte_process_info.nodename, 64);
mca_btl_wv_component.buffer_alignment = 64; mca_btl_wv_component.buffer_alignment = 64;
} else { } else {
mca_btl_wv_component.buffer_alignment = (uint32_t) ival; mca_btl_wv_component.buffer_alignment = (uint32_t) ival;

3
ompi/mca/btl/wv/btl_wv_proc.h
Просмотреть файл

@ -11,7 +11,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2006-2007 Voltaire All rights reserved. * Copyright (c) 2006-2007 Voltaire All rights reserved.
* Copyright (c) 2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -67,7 +66,7 @@ struct mca_btl_wv_proc_t {
ompi_proc_t *proc_ompi; ompi_proc_t *proc_ompi;
/** globally unique identifier for the process */ /** globally unique identifier for the process */
orca_process_name_t proc_guid; orte_process_name_t proc_guid;
/** modex messages from this proc; one for each port in the peer */ /** modex messages from this proc; one for each port in the peer */
mca_btl_wv_proc_modex_t *proc_ports; mca_btl_wv_proc_modex_t *proc_ports;

17
ompi/mca/btl/wv/connect/btl_wv_connect_base.c
Просмотреть файл

@ -3,7 +3,7 @@
* Copyright (c) 2007 Mellanox Technologies, Inc. All rights reserved. * Copyright (c) 2007 Mellanox Technologies, Inc. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -18,8 +18,7 @@
#include "connect/btl_wv_connect_oob.h" #include "connect/btl_wv_connect_oob.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/util/argv.h" #include "opal/util/argv.h"
#include "opal/util/output.h" #include "opal/util/output.h"
@ -86,9 +85,9 @@ int ompi_btl_wv_connect_base_register(void)
} }
} }
if (NULL == all[i]) { if (NULL == all[i]) {
orca_show_help("help-mpi-btl-wv-cpc-base.txt", orte_show_help("help-mpi-btl-wv-cpc-base.txt",
"cpc name not found", true, "cpc name not found", true,
"include", orca_process_info_get_nodename(), "include", orte_process_info.nodename,
"include", cpc_include, temp[j], "include", cpc_include, temp[j],
all_cpc_names); all_cpc_names);
opal_argv_free(temp); opal_argv_free(temp);
@ -112,9 +111,9 @@ int ompi_btl_wv_connect_base_register(void)
} }
} }
if (NULL == all[i]) { if (NULL == all[i]) {
orca_show_help("help-mpi-btl-wv-cpc-base.txt", orte_show_help("help-mpi-btl-wv-cpc-base.txt",
"cpc name not found", true, "cpc name not found", true,
"exclude", orca_process_info_get_nodename(), "exclude", orte_process_info.nodename,
"exclude", cpc_exclude, temp[j], "exclude", cpc_exclude, temp[j],
all_cpc_names); all_cpc_names);
opal_argv_free(temp); opal_argv_free(temp);
@ -258,9 +257,9 @@ int ompi_btl_wv_connect_base_select_for_local_port(mca_btl_wv_module_t *btl)
/* If we got an empty array, then no CPCs were eligible. Doh! */ /* If we got an empty array, then no CPCs were eligible. Doh! */
if (0 == cpc_index) { if (0 == cpc_index) {
orca_show_help("help-mpi-btl-wv-cpc-base.txt", orte_show_help("help-mpi-btl-wv-cpc-base.txt",
"no cpcs for port", true, "no cpcs for port", true,
orca_process_info_get_nodename(), orte_process_info.nodename,
btl->device->ib_dev->name, btl->device->ib_dev->name,
btl->port_num, msg); btl->port_num, msg);
free(cpcs); free(cpcs);

91
ompi/mca/btl/wv/connect/btl_wv_connect_oob.c
Просмотреть файл

@ -15,7 +15,7 @@
* Copyright (c) 2008-2009 Mellanox Technologies. All rights reserved. * Copyright (c) 2008-2009 Mellanox Technologies. All rights reserved.
* Copyright (c) 2009 IBM Corporation. All rights reserved. * Copyright (c) 2009 IBM Corporation. All rights reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -25,17 +25,20 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "opal_stdint.h" #include "opal_stdint.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h"
#include "opal/util/error.h" #include "opal/util/error.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orte/mca/rml/rml.h"
#include "orte/mca/rml/rml_types.h"
#include "orte/mca/errmgr/errmgr.h"
#include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
#include "btl_wv.h" #include "btl_wv.h"
#include "btl_wv_endpoint.h" #include "btl_wv_endpoint.h"
#include "btl_wv_proc.h" #include "btl_wv_proc.h"
#include "connect/connect.h" #include "connect/connect.h"
#include "orte/util/show_help.h"
#include <rdma/winverbs.h> #include <rdma/winverbs.h>
#include <malloc.h> #include <malloc.h>
@ -167,11 +170,11 @@ static int qp_create_one(mca_btl_base_endpoint_t* endpoint, int qp,
static int send_connect_data(mca_btl_base_endpoint_t* endpoint, static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
uint8_t message_type); uint8_t message_type);
static void rml_send_cb(int status, orca_process_name_t* endpoint, static void rml_send_cb(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
static void rml_recv_cb(int status, orca_process_name_t* process_name, static void rml_recv_cb(int status, orte_process_name_t* process_name,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
static int init_ud_qp(struct wv_context *context_arg, static int init_ud_qp(struct wv_context *context_arg,
struct mca_btl_wv_sa_qp_cache *cache); struct mca_btl_wv_sa_qp_cache *cache);
@ -239,12 +242,12 @@ static int oob_component_query(mca_btl_wv_module_t *btl,
ensure to only post it *once*, because another btl may have ensure to only post it *once*, because another btl may have
come in before this and already posted it. */ come in before this and already posted it. */
if (!rml_recv_posted) { if (!rml_recv_posted) {
rc = orca_oob_recv_buffer_nb(ORCA_NAME_WILDCARD, rc = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
OMPI_RML_TAG_OPENIB, OMPI_RML_TAG_OPENIB,
ORCA_OOB_PERSISTENT, ORTE_RML_PERSISTENT,
rml_recv_cb, rml_recv_cb,
NULL); NULL);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
opal_output_verbose(5, mca_btl_base_output, opal_output_verbose(5, mca_btl_base_output,
"wv BTL: oob CPC system error %d (%s)", "wv BTL: oob CPC system error %d (%s)",
rc, opal_strerror(rc)); rc, opal_strerror(rc));
@ -255,7 +258,7 @@ static int oob_component_query(mca_btl_wv_module_t *btl,
*cpc = (ompi_btl_wv_connect_base_module_t *) malloc(sizeof(ompi_btl_wv_connect_base_module_t)); *cpc = (ompi_btl_wv_connect_base_module_t *) malloc(sizeof(ompi_btl_wv_connect_base_module_t));
if (NULL == *cpc) { if (NULL == *cpc) {
orca_oob_recv_cancel(ORCA_NAME_WILDCARD, OMPI_RML_TAG_OPENIB); orte_rml.recv_cancel(ORTE_NAME_WILDCARD, OMPI_RML_TAG_OPENIB);
rml_recv_posted = false; rml_recv_posted = false;
opal_output_verbose(5, mca_btl_base_output, opal_output_verbose(5, mca_btl_base_output,
"wv BTL: oob CPC system error (malloc failed)"); "wv BTL: oob CPC system error (malloc failed)");
@ -311,7 +314,7 @@ static int oob_module_start_connect(ompi_btl_wv_connect_base_module_t *cpc,
static int oob_component_finalize(void) static int oob_component_finalize(void)
{ {
if (rml_recv_posted) { if (rml_recv_posted) {
orca_oob_recv_cancel(ORCA_NAME_WILDCARD, OMPI_RML_TAG_OPENIB); orte_rml.recv_cancel(ORTE_NAME_WILDCARD, OMPI_RML_TAG_OPENIB);
rml_recv_posted = false; rml_recv_posted = false;
} }
@ -610,7 +613,7 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
int rc; int rc;
if (NULL == buffer) { if (NULL == buffer) {
ORCA_ERROR_LOG(ORCA_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -618,14 +621,14 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT8)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT8));
rc = opal_dss.pack(buffer, &message_type, 1, OPAL_UINT8); rc = opal_dss.pack(buffer, &message_type, 1, OPAL_UINT8);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT64)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT64));
rc = opal_dss.pack(buffer, &endpoint->subnet_id, 1, OPAL_UINT64); rc = opal_dss.pack(buffer, &endpoint->subnet_id, 1, OPAL_UINT64);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
@ -636,13 +639,13 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
&endpoint->rem_info.rem_qps[0].rem_qp_num, 1, &endpoint->rem_info.rem_qps[0].rem_qp_num, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16));
rc = opal_dss.pack(buffer, &endpoint->rem_info.rem_lid, 1, OPAL_UINT16); rc = opal_dss.pack(buffer, &endpoint->rem_info.rem_lid, 1, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -655,14 +658,14 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_qp->qp_num, rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_qp->qp_num,
1, OPAL_UINT32); 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_psn, 1, rc = opal_dss.pack(buffer, &endpoint->qps[qp].qp->lcl_psn, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
@ -670,30 +673,30 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT16));
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->lid, 1, OPAL_UINT16); rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->lid, 1, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->device->mtu, 1, rc = opal_dss.pack(buffer, &endpoint->endpoint_btl->device->mtu, 1,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32)); BTL_VERBOSE(("packing %d of %d\n", 1, OPAL_UINT32));
rc = opal_dss.pack(buffer, &endpoint->index, 1, OPAL_UINT32); rc = opal_dss.pack(buffer, &endpoint->index, 1, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
} }
/* send to remote endpoint */ /* send to remote endpoint */
rc = orca_oob_send_buffer_nb(&endpoint->endpoint_proc->proc_guid, rc = orte_rml.send_buffer_nb(&endpoint->endpoint_proc->proc_guid,
buffer, OMPI_RML_TAG_OPENIB, 0, buffer, OMPI_RML_TAG_OPENIB, 0,
rml_send_cb, NULL); rml_send_cb, NULL);
if (ORCA_SUCCESS != rc) { if (ORTE_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
return rc; return rc;
} }
BTL_VERBOSE(("Sent QP Info, LID = %d, SUBNET = %" PRIx64 "\n", BTL_VERBOSE(("Sent QP Info, LID = %d, SUBNET = %" PRIx64 "\n",
@ -708,8 +711,8 @@ static int send_connect_data(mca_btl_base_endpoint_t* endpoint,
* Callback when we have finished RML sending the connect data to a * Callback when we have finished RML sending the connect data to a
* remote peer * remote peer
*/ */
static void rml_send_cb(int status, orca_process_name_t* endpoint, static void rml_send_cb(int status, orte_process_name_t* endpoint,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
OBJ_RELEASE(buffer); OBJ_RELEASE(buffer);
@ -721,8 +724,8 @@ static void rml_send_cb(int status, orca_process_name_t* endpoint,
* and if this endpoint is trying to connect, reply with our QP info, * and if this endpoint is trying to connect, reply with our QP info,
* otherwise try to modify QP's and establish reliable connection * otherwise try to modify QP's and establish reliable connection
*/ */
static void rml_recv_cb(int status, orca_process_name_t* process_name, static void rml_recv_cb(int status, orte_process_name_t* process_name,
opal_buffer_t* buffer, orca_oob_tag_t tag, opal_buffer_t* buffer, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
mca_btl_wv_proc_t *ib_proc; mca_btl_wv_proc_t *ib_proc;
@ -741,7 +744,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT8)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT8));
rc = opal_dss.unpack(buffer, &message_type, &cnt, OPAL_UINT8); rc = opal_dss.unpack(buffer, &message_type, &cnt, OPAL_UINT8);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
@ -749,7 +752,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT64)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT64));
rc = opal_dss.unpack(buffer, &rem_info.rem_subnet_id, &cnt, OPAL_UINT64); rc = opal_dss.unpack(buffer, &rem_info.rem_subnet_id, &cnt, OPAL_UINT64);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
@ -758,14 +761,14 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &lcl_qp, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &lcl_qp, &cnt, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16));
rc = opal_dss.unpack(buffer, &lcl_lid, &cnt, OPAL_UINT16); rc = opal_dss.unpack(buffer, &lcl_lid, &cnt, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
@ -783,7 +786,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_qp_num, &cnt, rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_qp_num, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
@ -791,7 +794,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_psn, &cnt, rc = opal_dss.unpack(buffer, &rem_info.rem_qps[qp].rem_psn, &cnt,
OPAL_UINT32); OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
@ -800,21 +803,21 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT16));
rc = opal_dss.unpack(buffer, &rem_info.rem_lid, &cnt, OPAL_UINT16); rc = opal_dss.unpack(buffer, &rem_info.rem_lid, &cnt, OPAL_UINT16);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &rem_info.rem_mtu, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &rem_info.rem_mtu, &cnt, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32)); BTL_VERBOSE(("unpacking %d of %d\n", cnt, OPAL_UINT32));
rc = opal_dss.unpack(buffer, &rem_info.rem_index, &cnt, OPAL_UINT32); rc = opal_dss.unpack(buffer, &rem_info.rem_index, &cnt, OPAL_UINT32);
if (OPAL_SUCCESS != rc) { if (OPAL_SUCCESS != rc) {
ORCA_ERROR_LOG(rc); ORTE_ERROR_LOG(rc);
mca_btl_wv_endpoint_invoke_error(NULL); mca_btl_wv_endpoint_invoke_error(NULL);
return; return;
} }
@ -824,7 +827,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
rem_info.rem_lid, rem_info.rem_lid,
rem_info.rem_subnet_id)); rem_info.rem_subnet_id));
master = orca_process_name_compare(ORCA_NAME_CMP_ALL, ORCA_PROC_MY_NAME, master = orte_util_compare_name_fields(ORTE_NS_CMP_ALL, ORTE_PROC_MY_NAME,
process_name) > 0 ? true : false; process_name) > 0 ? true : false;
/* Need to protect the ib_procs list */ /* Need to protect the ib_procs list */
@ -837,7 +840,7 @@ static void rml_recv_cb(int status, orca_process_name_t* process_name,
ib_proc = (mca_btl_wv_proc_t*)opal_list_get_next(ib_proc)) { ib_proc = (mca_btl_wv_proc_t*)opal_list_get_next(ib_proc)) {
bool found = false; bool found = false;
if (orca_process_name_compare(ORCA_NAME_CMP_ALL, if (orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&ib_proc->proc_guid, process_name) != OPAL_EQUAL) { &ib_proc->proc_guid, process_name) != OPAL_EQUAL) {
continue; continue;
} }
@ -1018,8 +1021,8 @@ static int init_ud_qp(struct wv_context *context_arg,
cache->cq->cqe = (uint32_t) entries; cache->cq->cqe = (uint32_t) entries;
if (NULL == cache->cq) { if (NULL == cache->cq) {
BTL_ERROR(("error creating cq, errno says %s", strerror(errno))); BTL_ERROR(("error creating cq, errno says %s", strerror(errno)));
orca_show_help("help-mpi-btl-wv.txt", "init-fail-create-q", orte_show_help("help-mpi-btl-wv.txt", "init-fail-create-q",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
__FILE__, __LINE__, "create_cq", __FILE__, __LINE__, "create_cq",
strerror(errno), errno, strerror(errno), errno,
context_arg->device->name); context_arg->device->name);

4
ompi/mca/coll/base/coll_base_comm_select.c
Просмотреть файл

@ -30,7 +30,7 @@
#include "mpi.h" #include "mpi.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/class/opal_list.h" #include "opal/class/opal_list.h"
#include "opal/class/opal_object.h" #include "opal/class/opal_object.h"
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
@ -118,7 +118,7 @@ int mca_coll_base_comm_select(ompi_communicator_t * comm)
collective modules available, then print error and return. */ collective modules available, then print error and return. */
if (NULL == selectable) { if (NULL == selectable) {
/* There's no modules available */ /* There's no modules available */
orca_show_help("help-mca-coll-base", orte_show_help("help-mca-coll-base",
"comm-select:none-available", true); "comm-select:none-available", true);
return OMPI_ERROR; return OMPI_ERROR;
} }

5
ompi/mca/coll/base/coll_base_find_available.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -28,7 +27,7 @@
#include "ompi/constants.h" #include "ompi/constants.h"
#include "opal/class/opal_list.h" #include "opal/class/opal_list.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "opal/mca/base/mca_base_component_repository.h" #include "opal/mca/base/mca_base_component_repository.h"
@ -131,7 +130,7 @@ int mca_coll_base_find_available(bool enable_progress_threads,
mca_coll_base_components_available_valid = false; mca_coll_base_components_available_valid = false;
opal_output_verbose(10, mca_coll_base_output, opal_output_verbose(10, mca_coll_base_output,
"coll:find_available: no coll components available!"); "coll:find_available: no coll components available!");
orca_show_help("help-mca-base", "find-available:none-found", true, orte_show_help("help-mca-base", "find-available:none-found", true,
"coll"); "coll");
return OMPI_ERROR; return OMPI_ERROR;
} }

3
ompi/mca/coll/sm/coll_sm.h
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -27,7 +26,7 @@
#include "mpi.h" #include "mpi.h"
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
#include "opal/datatype/opal_convertor.h" #include "opal/datatype/opal_convertor.h"
#include "orca/include/rte_orca.h" #include "orte/types.h"
#include "ompi/mca/coll/coll.h" #include "ompi/mca/coll/coll.h"
#include "ompi/mca/common/sm/common_sm.h" #include "ompi/mca/common/sm/common_sm.h"

7
ompi/mca/coll/sm/coll_sm_component.c
Просмотреть файл

@ -12,7 +12,6 @@
* Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2011 Los Alamos National Security, LLC. * Copyright (c) 2011 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -30,7 +29,7 @@
#include "ompi/constants.h" #include "ompi/constants.h"
#include "ompi/mca/coll/coll.h" #include "ompi/mca/coll/coll.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "coll_sm.h" #include "coll_sm.h"
@ -192,13 +191,13 @@ static int sm_register(void)
cs->sm_tree_degree, cs->sm_tree_degree,
&cs->sm_tree_degree); &cs->sm_tree_degree);
if (cs->sm_tree_degree > cs->sm_control_size) { if (cs->sm_tree_degree > cs->sm_control_size) {
orca_show_help("help-mpi-coll-sm.txt", orte_show_help("help-mpi-coll-sm.txt",
"tree-degree-larger-than-control", true, "tree-degree-larger-than-control", true,
cs->sm_tree_degree, cs->sm_control_size); cs->sm_tree_degree, cs->sm_control_size);
cs->sm_tree_degree = cs->sm_control_size; cs->sm_tree_degree = cs->sm_control_size;
} }
if (cs->sm_tree_degree > 255) { if (cs->sm_tree_degree > 255) {
orca_show_help("help-mpi-coll-sm.txt", orte_show_help("help-mpi-coll-sm.txt",
"tree-degree-larger-than-255", true, "tree-degree-larger-than-255", true,
cs->sm_tree_degree); cs->sm_tree_degree);
cs->sm_tree_degree = 255; cs->sm_tree_degree = 255;

14
ompi/mca/coll/sm/coll_sm_module.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2009-2012 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2009-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -52,7 +51,8 @@
#include "opal/mca/hwloc/base/base.h" #include "opal/mca/hwloc/base/base.h"
#include "opal/util/os_path.h" #include "opal/util/os_path.h"
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "orte/util/name_fns.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
#include "ompi/group/group.h" #include "ompi/group/group.h"
@ -528,33 +528,33 @@ static int bootstrap_comm(ompi_communicator_t *comm,
int num_in_use = c->sm_comm_num_in_use_flags; int num_in_use = c->sm_comm_num_in_use_flags;
int frag_size = c->sm_fragment_size; int frag_size = c->sm_fragment_size;
int control_size = c->sm_control_size; int control_size = c->sm_control_size;
orca_process_name_t *lowest_name = NULL; orte_process_name_t *lowest_name = NULL;
size_t size; size_t size;
ompi_proc_t *proc; ompi_proc_t *proc;
/* Make the rendezvous filename for this communicators shmem data /* Make the rendezvous filename for this communicators shmem data
segment. The CID is not guaranteed to be unique among all segment. The CID is not guaranteed to be unique among all
procs on this node, so also pair it with the PID of the proc procs on this node, so also pair it with the PID of the proc
with the lowest RTE name to form a unique filename. */ with the lowest ORTE name to form a unique filename. */
proc = ompi_group_peer_lookup(comm->c_local_group, 0); proc = ompi_group_peer_lookup(comm->c_local_group, 0);
lowest_name = &(proc->proc_name); lowest_name = &(proc->proc_name);
for (i = 1; i < comm_size; ++i) { for (i = 1; i < comm_size; ++i) {
proc = ompi_group_peer_lookup(comm->c_local_group, i); proc = ompi_group_peer_lookup(comm->c_local_group, i);
if (orca_process_name_compare(ORCA_NAME_CMP_ALL, if (orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&(proc->proc_name), &(proc->proc_name),
lowest_name) < 0) { lowest_name) < 0) {
lowest_name = &(proc->proc_name); lowest_name = &(proc->proc_name);
} }
} }
asprintf(&shortpath, "coll-sm-cid-%d-name-%s.mmap", comm->c_contextid, asprintf(&shortpath, "coll-sm-cid-%d-name-%s.mmap", comm->c_contextid,
ORCA_NAME_PRINT(lowest_name)); ORTE_NAME_PRINT(lowest_name));
if (NULL == shortpath) { if (NULL == shortpath) {
opal_output_verbose(10, mca_coll_base_output, opal_output_verbose(10, mca_coll_base_output,
"coll:sm:enable:bootstrap comm (%d/%s): asprintf failed", "coll:sm:enable:bootstrap comm (%d/%s): asprintf failed",
comm->c_contextid, comm->c_name); comm->c_contextid, comm->c_name);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
fullpath = opal_os_path(false, orca_process_info_get_job_session_dir(), fullpath = opal_os_path(false, orte_process_info.job_session_dir,
shortpath, NULL); shortpath, NULL);
free(shortpath); free(shortpath);
if (NULL == fullpath) { if (NULL == fullpath) {

98
ompi/mca/common/cuda/common_cuda.c
Просмотреть файл

@ -10,7 +10,6 @@
* Copyright (c) 2004-2006 The Regents of the University of California. * Copyright (c) 2004-2006 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2011-2012 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011-2012 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -36,7 +35,8 @@
#include "opal/datatype/opal_datatype_cuda.h" #include "opal/datatype/opal_datatype_cuda.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/proc_info.h"
#include "common_cuda.h" #include "common_cuda.h"
static bool common_cuda_initialized = false; static bool common_cuda_initialized = false;
@ -175,10 +175,10 @@ static int mca_common_cuda_init(void)
/* Check for the not initialized error since we can make suggestions to /* Check for the not initialized error since we can make suggestions to
* user for this error. */ * user for this error. */
if (CUDA_ERROR_NOT_INITIALIZED == res) { if (CUDA_ERROR_NOT_INITIALIZED == res) {
orca_show_help("help-mpi-common-cuda.txt", "cuCtxGetCurrent failed not initialized", orte_show_help("help-mpi-common-cuda.txt", "cuCtxGetCurrent failed not initialized",
true); true);
} else { } else {
orca_show_help("help-mpi-common-cuda.txt", "cuCtxGetCurrent failed", orte_show_help("help-mpi-common-cuda.txt", "cuCtxGetCurrent failed",
true, res); true, res);
} }
} }
@ -186,7 +186,7 @@ static int mca_common_cuda_init(void)
mca_common_cuda_register_memory = false; mca_common_cuda_register_memory = false;
} else if ((CUDA_SUCCESS == res) && (NULL == cuContext)) { } else if ((CUDA_SUCCESS == res) && (NULL == cuContext)) {
if (mca_common_cuda_warning) { if (mca_common_cuda_warning) {
orca_show_help("help-mpi-common-cuda.txt", "cuCtxGetCurrent returned NULL", orte_show_help("help-mpi-common-cuda.txt", "cuCtxGetCurrent returned NULL",
true); true);
} }
mca_common_cuda_enabled = false; mca_common_cuda_enabled = false;
@ -211,7 +211,7 @@ static int mca_common_cuda_init(void)
cuda_event_ipc_array = (CUevent *) malloc(sizeof(CUevent) * cuda_event_max); cuda_event_ipc_array = (CUevent *) malloc(sizeof(CUevent) * cuda_event_max);
if (NULL == cuda_event_ipc_array) { if (NULL == cuda_event_ipc_array) {
orca_show_help("help-mpi-common-cuda.txt", "No memory", orte_show_help("help-mpi-common-cuda.txt", "No memory",
true, errno, strerror(errno)); true, errno, strerror(errno));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -220,7 +220,7 @@ static int mca_common_cuda_init(void)
for (i = 0; i < cuda_event_max; i++) { for (i = 0; i < cuda_event_max; i++) {
res = cuEventCreate(&cuda_event_ipc_array[i], CU_EVENT_DISABLE_TIMING); res = cuEventCreate(&cuda_event_ipc_array[i], CU_EVENT_DISABLE_TIMING);
if (CUDA_SUCCESS != res) { if (CUDA_SUCCESS != res) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed",
true, res); true, res);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -231,7 +231,7 @@ static int mca_common_cuda_init(void)
cuda_event_ipc_frag_array = (struct mca_btl_base_descriptor_t **) cuda_event_ipc_frag_array = (struct mca_btl_base_descriptor_t **)
malloc(sizeof(struct mca_btl_base_descriptor_t *) * cuda_event_max); malloc(sizeof(struct mca_btl_base_descriptor_t *) * cuda_event_max);
if (NULL == cuda_event_ipc_frag_array) { if (NULL == cuda_event_ipc_frag_array) {
orca_show_help("help-mpi-common-cuda.txt", "No memory", orte_show_help("help-mpi-common-cuda.txt", "No memory",
true, errno, strerror(errno)); true, errno, strerror(errno));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -249,7 +249,7 @@ static int mca_common_cuda_init(void)
cuda_event_dtoh_array = (CUevent *) malloc(sizeof(CUevent) * cuda_event_max); cuda_event_dtoh_array = (CUevent *) malloc(sizeof(CUevent) * cuda_event_max);
if (NULL == cuda_event_dtoh_array) { if (NULL == cuda_event_dtoh_array) {
orca_show_help("help-mpi-common-cuda.txt", "No memory", orte_show_help("help-mpi-common-cuda.txt", "No memory",
true, errno, strerror(errno)); true, errno, strerror(errno));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -258,7 +258,7 @@ static int mca_common_cuda_init(void)
for (i = 0; i < cuda_event_max; i++) { for (i = 0; i < cuda_event_max; i++) {
res = cuEventCreate(&cuda_event_dtoh_array[i], CU_EVENT_DISABLE_TIMING); res = cuEventCreate(&cuda_event_dtoh_array[i], CU_EVENT_DISABLE_TIMING);
if (CUDA_SUCCESS != res) { if (CUDA_SUCCESS != res) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed",
true, res); true, res);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -269,7 +269,7 @@ static int mca_common_cuda_init(void)
cuda_event_dtoh_frag_array = (struct mca_btl_base_descriptor_t **) cuda_event_dtoh_frag_array = (struct mca_btl_base_descriptor_t **)
malloc(sizeof(struct mca_btl_base_descriptor_t *) * cuda_event_max); malloc(sizeof(struct mca_btl_base_descriptor_t *) * cuda_event_max);
if (NULL == cuda_event_dtoh_frag_array) { if (NULL == cuda_event_dtoh_frag_array) {
orca_show_help("help-mpi-common-cuda.txt", "No memory", orte_show_help("help-mpi-common-cuda.txt", "No memory",
true, errno, strerror(errno)); true, errno, strerror(errno));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -284,7 +284,7 @@ static int mca_common_cuda_init(void)
cuda_event_htod_array = (CUevent *) malloc(sizeof(CUevent) * cuda_event_max); cuda_event_htod_array = (CUevent *) malloc(sizeof(CUevent) * cuda_event_max);
if (NULL == cuda_event_htod_array) { if (NULL == cuda_event_htod_array) {
orca_show_help("help-mpi-common-cuda.txt", "No memory", orte_show_help("help-mpi-common-cuda.txt", "No memory",
true, errno, strerror(errno)); true, errno, strerror(errno));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -293,7 +293,7 @@ static int mca_common_cuda_init(void)
for (i = 0; i < cuda_event_max; i++) { for (i = 0; i < cuda_event_max; i++) {
res = cuEventCreate(&cuda_event_htod_array[i], CU_EVENT_DISABLE_TIMING); res = cuEventCreate(&cuda_event_htod_array[i], CU_EVENT_DISABLE_TIMING);
if (CUDA_SUCCESS != res) { if (CUDA_SUCCESS != res) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed",
true, res); true, res);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -304,7 +304,7 @@ static int mca_common_cuda_init(void)
cuda_event_htod_frag_array = (struct mca_btl_base_descriptor_t **) cuda_event_htod_frag_array = (struct mca_btl_base_descriptor_t **)
malloc(sizeof(struct mca_btl_base_descriptor_t *) * cuda_event_max); malloc(sizeof(struct mca_btl_base_descriptor_t *) * cuda_event_max);
if (NULL == cuda_event_htod_frag_array) { if (NULL == cuda_event_htod_frag_array) {
orca_show_help("help-mpi-common-cuda.txt", "No memory", orte_show_help("help-mpi-common-cuda.txt", "No memory",
true, errno, strerror(errno)); true, errno, strerror(errno));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -319,9 +319,9 @@ static int mca_common_cuda_init(void)
if (res != CUDA_SUCCESS) { if (res != CUDA_SUCCESS) {
/* If registering the memory fails, print a message and continue. /* If registering the memory fails, print a message and continue.
* This is not a fatal error. */ * This is not a fatal error. */
orca_show_help("help-mpi-common-cuda.txt", "cuMemHostRegister failed", orte_show_help("help-mpi-common-cuda.txt", "cuMemHostRegister failed",
true, mem_reg->ptr, mem_reg->amount, true, mem_reg->ptr, mem_reg->amount,
orca_process_info_get_nodename(), res, mem_reg->msg); orte_process_info.nodename, res, mem_reg->msg);
} else { } else {
opal_output_verbose(20, mca_common_cuda_output, opal_output_verbose(20, mca_common_cuda_output,
"CUDA: cuMemHostRegister OK on mpool %s: " "CUDA: cuMemHostRegister OK on mpool %s: "
@ -336,7 +336,7 @@ static int mca_common_cuda_init(void)
/* Create stream for use in ipc asynchronous copies */ /* Create stream for use in ipc asynchronous copies */
res = cuStreamCreate(&ipcStream, 0); res = cuStreamCreate(&ipcStream, 0);
if (res != CUDA_SUCCESS) { if (res != CUDA_SUCCESS) {
orca_show_help("help-mpi-common-cuda.txt", "cuStreamCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuStreamCreate failed",
true, res); true, res);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -344,7 +344,7 @@ static int mca_common_cuda_init(void)
/* Create stream for use in dtoh asynchronous copies */ /* Create stream for use in dtoh asynchronous copies */
res = cuStreamCreate(&dtohStream, 0); res = cuStreamCreate(&dtohStream, 0);
if (res != CUDA_SUCCESS) { if (res != CUDA_SUCCESS) {
orca_show_help("help-mpi-common-cuda.txt", "cuStreamCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuStreamCreate failed",
true, res); true, res);
return OMPI_ERROR; return OMPI_ERROR;
@ -353,7 +353,7 @@ static int mca_common_cuda_init(void)
/* Create stream for use in htod asynchronous copies */ /* Create stream for use in htod asynchronous copies */
res = cuStreamCreate(&htodStream, 0); res = cuStreamCreate(&htodStream, 0);
if (res != CUDA_SUCCESS) { if (res != CUDA_SUCCESS) {
orca_show_help("help-mpi-common-cuda.txt", "cuStreamCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuStreamCreate failed",
true, res); true, res);
return OMPI_ERROR; return OMPI_ERROR;
@ -393,9 +393,9 @@ void mca_common_cuda_register(void *ptr, size_t amount, char *msg) {
if (res != CUDA_SUCCESS) { if (res != CUDA_SUCCESS) {
/* If registering the memory fails, print a message and continue. /* If registering the memory fails, print a message and continue.
* This is not a fatal error. */ * This is not a fatal error. */
orca_show_help("help-mpi-common-cuda.txt", "cuMemHostRegister failed", orte_show_help("help-mpi-common-cuda.txt", "cuMemHostRegister failed",
true, ptr, amount, true, ptr, amount,
orca_process_info_get_nodename(), res, msg); orte_process_info.nodename, res, msg);
} else { } else {
opal_output_verbose(20, mca_common_cuda_output, opal_output_verbose(20, mca_common_cuda_output,
"CUDA: cuMemHostRegister OK on mpool %s: " "CUDA: cuMemHostRegister OK on mpool %s: "
@ -432,9 +432,9 @@ void mca_common_cuda_unregister(void *ptr, char *msg) {
if (res != CUDA_SUCCESS) { if (res != CUDA_SUCCESS) {
/* If unregistering the memory fails, print a message and continue. /* If unregistering the memory fails, print a message and continue.
* This is not a fatal error. */ * This is not a fatal error. */
orca_show_help("help-mpi-common-cuda.txt", "cuMemHostUnregister failed", orte_show_help("help-mpi-common-cuda.txt", "cuMemHostUnregister failed",
true, ptr, true, ptr,
orca_process_info_get_nodename(), res, msg); orte_process_info.nodename, res, msg);
} else { } else {
opal_output_verbose(20, mca_common_cuda_output, opal_output_verbose(20, mca_common_cuda_output,
"CUDA: cuMemHostUnregister OK on mpool %s: " "CUDA: cuMemHostUnregister OK on mpool %s: "
@ -473,7 +473,7 @@ int cuda_getmemhandle(void *base, size_t size, mca_mpool_base_registration_t *ne
CUDA_DUMP_MEMHANDLE((100, &memHandle, "GetMemHandle-After")); CUDA_DUMP_MEMHANDLE((100, &memHandle, "GetMemHandle-After"));
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuIpcGetMemHandle failed", orte_show_help("help-mpi-common-cuda.txt", "cuIpcGetMemHandle failed",
true, result, base); true, result, base);
return OMPI_ERROR; return OMPI_ERROR;
} else { } else {
@ -486,7 +486,7 @@ int cuda_getmemhandle(void *base, size_t size, mca_mpool_base_registration_t *ne
* how the remote side saves the handles in a cache. */ * how the remote side saves the handles in a cache. */
result = cuMemGetAddressRange(&pbase, &psize, (CUdeviceptr)base); result = cuMemGetAddressRange(&pbase, &psize, (CUdeviceptr)base);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuMemGetAddressRange failed", orte_show_help("help-mpi-common-cuda.txt", "cuMemGetAddressRange failed",
true, result, base); true, result, base);
return OMPI_ERROR; return OMPI_ERROR;
} else { } else {
@ -509,7 +509,7 @@ int cuda_getmemhandle(void *base, size_t size, mca_mpool_base_registration_t *ne
* with. */ * with. */
result = cuEventRecord((CUevent)cuda_reg->event, 0); result = cuEventRecord((CUevent)cuda_reg->event, 0);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed",
true, result, base); true, result, base);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -561,7 +561,7 @@ int cuda_openmemhandle(void *base, size_t size, mca_mpool_base_registration_t *n
return OMPI_ERR_WOULD_BLOCK; return OMPI_ERR_WOULD_BLOCK;
} }
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuIpcOpenMemHandle failed", orte_show_help("help-mpi-common-cuda.txt", "cuIpcOpenMemHandle failed",
true, result, base); true, result, base);
/* Currently, this is a non-recoverable error */ /* Currently, this is a non-recoverable error */
return OMPI_ERROR; return OMPI_ERROR;
@ -585,7 +585,7 @@ int cuda_closememhandle(void *reg_data, mca_mpool_base_registration_t *reg)
result = cuIpcCloseMemHandle((CUdeviceptr)cuda_reg->base.alloc_base); result = cuIpcCloseMemHandle((CUdeviceptr)cuda_reg->base.alloc_base);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuIpcCloseMemHandle failed", orte_show_help("help-mpi-common-cuda.txt", "cuIpcCloseMemHandle failed",
true, result, cuda_reg->base.alloc_base); true, result, cuda_reg->base.alloc_base);
/* We will just continue on and hope things continue to work. */ /* We will just continue on and hope things continue to work. */
} else { } else {
@ -604,13 +604,13 @@ void mca_common_cuda_construct_event_and_handle(uint64_t **event, void **handle)
result = cuEventCreate((CUevent *)event, CU_EVENT_INTERPROCESS | CU_EVENT_DISABLE_TIMING); result = cuEventCreate((CUevent *)event, CU_EVENT_INTERPROCESS | CU_EVENT_DISABLE_TIMING);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventCreate failed",
true, result); true, result);
} }
result = cuIpcGetEventHandle((CUipcEventHandle *)handle, (CUevent)*event); result = cuIpcGetEventHandle((CUipcEventHandle *)handle, (CUevent)*event);
if (CUDA_SUCCESS != result){ if (CUDA_SUCCESS != result){
orca_show_help("help-mpi-common-cuda.txt", "cuIpcGetEventHandle failed", orte_show_help("help-mpi-common-cuda.txt", "cuIpcGetEventHandle failed",
true, result); true, result);
} }
@ -624,7 +624,7 @@ void mca_common_cuda_destruct_event(uint64_t *event)
result = cuEventDestroy((CUevent)event); result = cuEventDestroy((CUevent)event);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventDestroy failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventDestroy failed",
true, result); true, result);
} }
} }
@ -645,7 +645,7 @@ void mca_common_wait_stream_synchronize(mca_mpool_common_cuda_reg_t *rget_reg)
result = cuIpcOpenEventHandle(&event, evtHandle); result = cuIpcOpenEventHandle(&event, evtHandle);
if (CUDA_SUCCESS != result){ if (CUDA_SUCCESS != result){
orca_show_help("help-mpi-common-cuda.txt", "cuIpcOpenEventHandle failed", orte_show_help("help-mpi-common-cuda.txt", "cuIpcOpenEventHandle failed",
true, result); true, result);
} }
@ -656,21 +656,21 @@ void mca_common_wait_stream_synchronize(mca_mpool_common_cuda_reg_t *rget_reg)
*/ */
result = cuEventRecord(event, 0); result = cuEventRecord(event, 0);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed",
true, result); true, result);
} }
/* END of Workaround */ /* END of Workaround */
result = cuStreamWaitEvent(0, event, 0); result = cuStreamWaitEvent(0, event, 0);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuStreamWaitEvent failed", orte_show_help("help-mpi-common-cuda.txt", "cuStreamWaitEvent failed",
true, result); true, result);
} }
/* All done with this event. */ /* All done with this event. */
result = cuEventDestroy(event); result = cuEventDestroy(event);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventDestroy failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventDestroy failed",
true, result); true, result);
} }
} }
@ -689,7 +689,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
* return an error. The error message will tell the user to try and * return an error. The error message will tell the user to try and
* run again, but with a larger array for storing events. */ * run again, but with a larger array for storing events. */
if (cuda_event_ipc_num_used == cuda_event_max) { if (cuda_event_ipc_num_used == cuda_event_max) {
orca_show_help("help-mpi-common-cuda.txt", "Out of cuEvent handles", orte_show_help("help-mpi-common-cuda.txt", "Out of cuEvent handles",
true, cuda_event_max, cuda_event_max+100, cuda_event_max+100); true, cuda_event_max, cuda_event_max+100, cuda_event_max+100);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
@ -699,7 +699,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
if (OPAL_LIKELY(mca_common_cuda_async)) { if (OPAL_LIKELY(mca_common_cuda_async)) {
result = cuMemcpyAsync((CUdeviceptr)dst, (CUdeviceptr)src, amount, ipcStream); result = cuMemcpyAsync((CUdeviceptr)dst, (CUdeviceptr)src, amount, ipcStream);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuMemcpyAsync failed", orte_show_help("help-mpi-common-cuda.txt", "cuMemcpyAsync failed",
true, dst, src, amount, result); true, dst, src, amount, result);
return OMPI_ERROR; return OMPI_ERROR;
} else { } else {
@ -709,7 +709,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
} }
result = cuEventRecord(cuda_event_ipc_array[cuda_event_ipc_first_avail], ipcStream); result = cuEventRecord(cuda_event_ipc_array[cuda_event_ipc_first_avail], ipcStream);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed",
true, result); true, result);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -727,7 +727,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
/* Mimic the async function so they use the same memcpy call. */ /* Mimic the async function so they use the same memcpy call. */
result = cuMemcpyAsync((CUdeviceptr)dst, (CUdeviceptr)src, amount, ipcStream); result = cuMemcpyAsync((CUdeviceptr)dst, (CUdeviceptr)src, amount, ipcStream);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuMemcpyAsync failed", orte_show_help("help-mpi-common-cuda.txt", "cuMemcpyAsync failed",
true, dst, src, amount, result); true, dst, src, amount, result);
return OMPI_ERROR; return OMPI_ERROR;
} else { } else {
@ -739,7 +739,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
/* Record an event, then wait for it to complete with calls to cuEventQuery */ /* Record an event, then wait for it to complete with calls to cuEventQuery */
result = cuEventRecord(cuda_event_ipc_array[cuda_event_ipc_first_avail], ipcStream); result = cuEventRecord(cuda_event_ipc_array[cuda_event_ipc_first_avail], ipcStream);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed",
true, result); true, result);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -755,7 +755,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
result = cuEventQuery(cuda_event_ipc_array[cuda_event_ipc_first_used]); result = cuEventQuery(cuda_event_ipc_array[cuda_event_ipc_first_used]);
if ((CUDA_SUCCESS != result) && (CUDA_ERROR_NOT_READY != result)) { if ((CUDA_SUCCESS != result) && (CUDA_ERROR_NOT_READY != result)) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed",
true, result); true, result);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -767,7 +767,7 @@ int mca_common_cuda_memcpy(void *dst, void *src, size_t amount, char *msg,
} }
result = cuEventQuery(cuda_event_ipc_array[cuda_event_ipc_first_used]); result = cuEventQuery(cuda_event_ipc_array[cuda_event_ipc_first_used]);
if ((CUDA_SUCCESS != result) && (CUDA_ERROR_NOT_READY != result)) { if ((CUDA_SUCCESS != result) && (CUDA_ERROR_NOT_READY != result)) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed",
true, result); true, result);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -796,14 +796,14 @@ int mca_common_cuda_record_dtoh_event(char *msg, struct mca_btl_base_descriptor_
* return an error. The error message will tell the user to try and * return an error. The error message will tell the user to try and
* run again, but with a larger array for storing events. */ * run again, but with a larger array for storing events. */
if (cuda_event_dtoh_num_used == cuda_event_max) { if (cuda_event_dtoh_num_used == cuda_event_max) {
orca_show_help("help-mpi-common-cuda.txt", "Out of cuEvent handles", orte_show_help("help-mpi-common-cuda.txt", "Out of cuEvent handles",
true, cuda_event_max, cuda_event_max+100, cuda_event_max+100); true, cuda_event_max, cuda_event_max+100, cuda_event_max+100);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
result = cuEventRecord(cuda_event_dtoh_array[cuda_event_dtoh_first_avail], dtohStream); result = cuEventRecord(cuda_event_dtoh_array[cuda_event_dtoh_first_avail], dtohStream);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed",
true, result); true, result);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -831,14 +831,14 @@ int mca_common_cuda_record_htod_event(char *msg, struct mca_btl_base_descriptor_
* return an error. The error message will tell the user to try and * return an error. The error message will tell the user to try and
* run again, but with a larger array for storing events. */ * run again, but with a larger array for storing events. */
if (cuda_event_htod_num_used == cuda_event_max) { if (cuda_event_htod_num_used == cuda_event_max) {
orca_show_help("help-mpi-common-cuda.txt", "Out of cuEvent handles", orte_show_help("help-mpi-common-cuda.txt", "Out of cuEvent handles",
true, cuda_event_max, cuda_event_max+100, cuda_event_max+100); true, cuda_event_max, cuda_event_max+100, cuda_event_max+100);
return OMPI_ERR_OUT_OF_RESOURCE; return OMPI_ERR_OUT_OF_RESOURCE;
} }
result = cuEventRecord(cuda_event_htod_array[cuda_event_htod_first_avail], htodStream); result = cuEventRecord(cuda_event_htod_array[cuda_event_htod_first_avail], htodStream);
if (CUDA_SUCCESS != result) { if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventRecord failed",
true, result); true, result);
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -890,7 +890,7 @@ int progress_one_cuda_ipc_event(struct mca_btl_base_descriptor_t **frag) {
*frag = NULL; *frag = NULL;
return 0; return 0;
} else if (CUDA_SUCCESS != result) { } else if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed",
true, result); true, result);
*frag = NULL; *frag = NULL;
return OMPI_ERROR; return OMPI_ERROR;
@ -932,7 +932,7 @@ int progress_one_cuda_dtoh_event(struct mca_btl_base_descriptor_t **frag) {
*frag = NULL; *frag = NULL;
return 0; return 0;
} else if (CUDA_SUCCESS != result) { } else if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed",
true, result); true, result);
*frag = NULL; *frag = NULL;
return OMPI_ERROR; return OMPI_ERROR;
@ -974,7 +974,7 @@ int progress_one_cuda_htod_event(struct mca_btl_base_descriptor_t **frag) {
*frag = NULL; *frag = NULL;
return 0; return 0;
} else if (CUDA_SUCCESS != result) { } else if (CUDA_SUCCESS != result) {
orca_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed", orte_show_help("help-mpi-common-cuda.txt", "cuEventQuery failed",
true, result); true, result);
*frag = NULL; *frag = NULL;
return OMPI_ERROR; return OMPI_ERROR;

25
ompi/mca/common/sm/common_sm.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -47,7 +46,11 @@
#include "opal/runtime/opal_cr.h" #include "opal/runtime/opal_cr.h"
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "orte/util/name_fns.h"
#include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "orte/mca/errmgr/errmgr.h"
#include "ompi/constants.h" #include "ompi/constants.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
@ -88,7 +91,7 @@ attach_and_init(size_t size_ctl_structure,
/* set up the map object */ /* set up the map object */
if (NULL == (map = OBJ_NEW(mca_common_sm_module_t))) { if (NULL == (map = OBJ_NEW(mca_common_sm_module_t))) {
ORCA_ERROR_LOG(OMPI_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(OMPI_ERR_OUT_OF_RESOURCE);
return NULL; return NULL;
} }
@ -111,8 +114,8 @@ attach_and_init(size_t size_ctl_structure,
addr = OPAL_ALIGN_PTR(addr, data_seg_alignment, unsigned char *); addr = OPAL_ALIGN_PTR(addr, data_seg_alignment, unsigned char *);
/* is addr past end of the shared memory segment? */ /* is addr past end of the shared memory segment? */
if ((unsigned char *)seg + shmem_ds.seg_size < addr) { if ((unsigned char *)seg + shmem_ds.seg_size < addr) {
orca_show_help("help-mpi-common-sm.txt", "mmap too small", 1, orte_show_help("help-mpi-common-sm.txt", "mmap too small", 1,
orca_process_info_get_nodename(), orte_process_info.nodename,
(unsigned long)shmem_ds.seg_size, (unsigned long)shmem_ds.seg_size,
(unsigned long)size_ctl_structure, (unsigned long)size_ctl_structure,
(unsigned long)data_seg_alignment); (unsigned long)data_seg_alignment);
@ -173,8 +176,8 @@ mca_common_sm_init(ompi_proc_t **procs,
/* if we have a new lowest, swap it with position 0 /* if we have a new lowest, swap it with position 0
* so that procs[0] is always the lowest named proc * so that procs[0] is always the lowest named proc
*/ */
if (OPAL_VALUE2_GREATER == orca_process_name_compare( if (OPAL_VALUE2_GREATER == orte_util_compare_name_fields(
ORCA_NAME_CMP_ALL, ORTE_NS_CMP_ALL,
&(procs[p]->proc_name), &(procs[p]->proc_name),
&(procs[0]->proc_name))) { &(procs[0]->proc_name))) {
temp_proc = procs[0]; temp_proc = procs[0];
@ -195,9 +198,9 @@ mca_common_sm_init(ompi_proc_t **procs,
} }
/* determine whether or not i am the lowest local process */ /* determine whether or not i am the lowest local process */
lowest_local_proc = (0 == orca_process_name_compare( lowest_local_proc = (0 == orte_util_compare_name_fields(
ORCA_NAME_CMP_ALL, ORTE_NS_CMP_ALL,
ORCA_PROC_MY_NAME, ORTE_PROC_MY_NAME,
&(procs[0]->proc_name))); &(procs[0]->proc_name)));
/* figure out if i am the lowest rank in the group. /* figure out if i am the lowest rank in the group.
@ -286,7 +289,7 @@ mca_common_sm_init_group(ompi_group_t *group,
} }
else if (NULL == (procs = (ompi_proc_t **) else if (NULL == (procs = (ompi_proc_t **)
malloc(sizeof(ompi_proc_t *) * group_size))) { malloc(sizeof(ompi_proc_t *) * group_size))) {
ORCA_ERROR_LOG(OMPI_ERR_OUT_OF_RESOURCE); ORTE_ERROR_LOG(OMPI_ERR_OUT_OF_RESOURCE);
goto out; goto out;
} }
/* make sure that all the procs in the group are local */ /* make sure that all the procs in the group are local */

28
ompi/mca/common/sm/common_sm_rml.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Los Alamos National Security, LLC. * Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -26,7 +25,12 @@
#include "opal/types.h" #include "opal/types.h"
#include "opal/dss/dss.h" #include "opal/dss/dss.h"
#include "orca/include/rte_orca.h" #include "orte/mca/rml/rml.h"
#include "orte/util/proc_info.h"
#include "orte/util/name_fns.h"
#include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "orte/mca/errmgr/errmgr.h"
#include "ompi/constants.h" #include "ompi/constants.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
@ -69,7 +73,7 @@ mca_common_sm_rml_info_bcast(opal_shmem_ds_t *out_ds_buf,
* note in common_sm.c for more details. */ * note in common_sm.c for more details. */
tmprc = opal_dss.pack(buffer, &msg_id_str, 1, OPAL_STRING); tmprc = opal_dss.pack(buffer, &msg_id_str, 1, OPAL_STRING);
if (OPAL_SUCCESS != tmprc) { if (OPAL_SUCCESS != tmprc) {
ORCA_ERROR_LOG(ORCA_ERR_PACK_FAILURE); ORTE_ERROR_LOG(ORTE_ERR_PACK_FAILURE);
rc = OMPI_ERR_PACK_FAILURE; rc = OMPI_ERR_PACK_FAILURE;
goto out; goto out;
} }
@ -77,7 +81,7 @@ mca_common_sm_rml_info_bcast(opal_shmem_ds_t *out_ds_buf,
(int32_t)sizeof(opal_shmem_ds_t), (int32_t)sizeof(opal_shmem_ds_t),
OPAL_BYTE); OPAL_BYTE);
if (OPAL_SUCCESS != tmprc) { if (OPAL_SUCCESS != tmprc) {
ORCA_ERROR_LOG(ORCA_ERR_PACK_FAILURE); ORTE_ERROR_LOG(ORTE_ERR_PACK_FAILURE);
rc = OMPI_ERR_PACK_FAILURE; rc = OMPI_ERR_PACK_FAILURE;
goto out; goto out;
} }
@ -85,10 +89,10 @@ mca_common_sm_rml_info_bcast(opal_shmem_ds_t *out_ds_buf,
/* first num_local_procs items should be local procs */ /* first num_local_procs items should be local procs */
for (p = 1; p < num_local_procs; ++p) { for (p = 1; p < num_local_procs; ++p) {
/* a potential future optimization: use non-blocking routines */ /* a potential future optimization: use non-blocking routines */
tmprc = orca_oob_send_buffer(&(procs[p]->proc_name), buffer, tag, tmprc = orte_rml.send_buffer(&(procs[p]->proc_name), buffer, tag,
0); 0);
if (0 > tmprc) { if (0 > tmprc) {
ORCA_ERROR_LOG(tmprc); ORTE_ERROR_LOG(tmprc);
opal_progress_event_users_decrement(); opal_progress_event_users_decrement();
rc = OMPI_ERROR; rc = OMPI_ERROR;
goto out; goto out;
@ -102,10 +106,10 @@ mca_common_sm_rml_info_bcast(opal_shmem_ds_t *out_ds_buf,
/* bump up the libevent polling frequency while we're in this RML recv, /* bump up the libevent polling frequency while we're in this RML recv,
* just to ensure we're checking libevent frequently. */ * just to ensure we're checking libevent frequently. */
opal_progress_event_users_increment(); opal_progress_event_users_increment();
tmprc = orca_oob_recv_buffer(&(procs[0]->proc_name), buffer, tag, 0); tmprc = orte_rml.recv_buffer(&(procs[0]->proc_name), buffer, tag, 0);
opal_progress_event_users_decrement(); opal_progress_event_users_decrement();
if (0 > tmprc) { if (0 > tmprc) {
ORCA_ERROR_LOG(tmprc); ORTE_ERROR_LOG(tmprc);
rc = OMPI_ERROR; rc = OMPI_ERROR;
goto out; goto out;
} }
@ -114,14 +118,14 @@ mca_common_sm_rml_info_bcast(opal_shmem_ds_t *out_ds_buf,
tmprc = opal_dss.unpack(buffer, &msg_id_str_to_tx, &num_vals, tmprc = opal_dss.unpack(buffer, &msg_id_str_to_tx, &num_vals,
OPAL_STRING); OPAL_STRING);
if (0 > tmprc) { if (0 > tmprc) {
ORCA_ERROR_LOG(ORCA_ERR_UNPACK_FAILURE); ORTE_ERROR_LOG(ORTE_ERR_UNPACK_FAILURE);
rc = OMPI_ERROR; rc = OMPI_ERROR;
goto out; goto out;
} }
num_vals = (int32_t)sizeof(opal_shmem_ds_t); num_vals = (int32_t)sizeof(opal_shmem_ds_t);
tmprc = opal_dss.unpack(buffer, out_ds_buf, &num_vals, OPAL_BYTE); tmprc = opal_dss.unpack(buffer, out_ds_buf, &num_vals, OPAL_BYTE);
if (0 > tmprc) { if (0 > tmprc) {
ORCA_ERROR_LOG(ORCA_ERR_UNPACK_FAILURE); ORTE_ERROR_LOG(ORTE_ERR_UNPACK_FAILURE);
rc = OMPI_ERROR; rc = OMPI_ERROR;
goto out; goto out;
} }
@ -130,8 +134,8 @@ mca_common_sm_rml_info_bcast(opal_shmem_ds_t *out_ds_buf,
* outside of our current scope of assumptions. see "RML Messaging and * outside of our current scope of assumptions. see "RML Messaging and
* Our Assumptions" note in common_sm.c */ * Our Assumptions" note in common_sm.c */
if (0 != strcmp(msg_id_str_to_tx, msg_id_str)) { if (0 != strcmp(msg_id_str_to_tx, msg_id_str)) {
orca_show_help("help-mpi-common-sm.txt", "unexpected message id", orte_show_help("help-mpi-common-sm.txt", "unexpected message id",
true, orca_process_info_get_nodename(), true, orte_process_info.nodename,
msg_id_str, msg_id_str_to_tx); msg_id_str, msg_id_str_to_tx);
rc = OMPI_ERROR; rc = OMPI_ERROR;
/* here for extra debug info only */ /* here for extra debug info only */

346
ompi/mca/crcp/bkmrk/crcp_bkmrk_pml.c
Просмотреть файл

@ -5,7 +5,6 @@
* of Tennessee Research Foundation. All rights * of Tennessee Research Foundation. All rights
* reserved. * reserved.
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -32,7 +31,11 @@
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orca/include/rte_orca.h" #include "orte/runtime/orte_globals.h"
#include "orte/util/name_fns.h"
#include "orte/mca/grpcomm/grpcomm.h"
#include "orte/mca/rml/rml.h"
#include "orte/mca/ess/ess.h"
#include "ompi/request/request.h" #include "ompi/request/request.h"
#include "ompi/mca/dpm/dpm.h" #include "ompi/mca/dpm/dpm.h"
@ -299,9 +302,9 @@ static int drain_message_copy_remove_persistent(ompi_crcp_bkmrk_pml_drain_messag
ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref); ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref);
/* /*
* Peer List: Find the peer reference matching the Open RTE/ORCA process name * Peer List: Find the peer reference matching the ORTE process name
*/ */
static ompi_crcp_bkmrk_pml_peer_ref_t* find_peer(orca_process_name_t proc); static ompi_crcp_bkmrk_pml_peer_ref_t* find_peer(orte_process_name_t proc);
/* /*
* Peer List: Find the peer reference matching the index into the communicator * Peer List: Find the peer reference matching the index into the communicator
@ -343,9 +346,9 @@ static int recv_bookmarks(int peer_idx);
* Callback to receive the bookmarks from a peer * Callback to receive the bookmarks from a peer
*/ */
static void recv_bookmarks_cbfunc(int status, static void recv_bookmarks_cbfunc(int status,
orca_process_name_t* sender, orte_process_name_t* sender,
opal_buffer_t *buffer, opal_buffer_t *buffer,
orca_oob_tag_t tag, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
static int total_recv_bookmarks = 0; static int total_recv_bookmarks = 0;
@ -416,9 +419,9 @@ static int ft_event_post_drain_acks(void);
* Callback to service drain message acks. * Callback to service drain message acks.
*/ */
static void drain_message_ack_cbfunc(int status, static void drain_message_ack_cbfunc(int status,
orca_process_name_t* sender, orte_process_name_t* sender,
opal_buffer_t *buffer, opal_buffer_t *buffer,
orca_oob_tag_t tag, orte_rml_tag_t tag,
void* cbdata); void* cbdata);
/* /*
@ -697,8 +700,8 @@ OBJ_CLASS_INSTANCE(ompi_crcp_bkmrk_pml_peer_ref_t,
ompi_crcp_bkmrk_pml_peer_ref_destruct); ompi_crcp_bkmrk_pml_peer_ref_destruct);
void ompi_crcp_bkmrk_pml_peer_ref_construct(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref) { void ompi_crcp_bkmrk_pml_peer_ref_construct(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref) {
peer_ref->proc_name.jobid = ORCA_JOBID_INVALID; peer_ref->proc_name.jobid = ORTE_JOBID_INVALID;
peer_ref->proc_name.vpid = ORCA_VPID_INVALID; peer_ref->proc_name.vpid = ORTE_VPID_INVALID;
OBJ_CONSTRUCT(&peer_ref->send_list, opal_list_t); OBJ_CONSTRUCT(&peer_ref->send_list, opal_list_t);
OBJ_CONSTRUCT(&peer_ref->isend_list, opal_list_t); OBJ_CONSTRUCT(&peer_ref->isend_list, opal_list_t);
@ -724,8 +727,8 @@ void ompi_crcp_bkmrk_pml_peer_ref_construct(ompi_crcp_bkmrk_pml_peer_ref_t *peer
void ompi_crcp_bkmrk_pml_peer_ref_destruct( ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref) { void ompi_crcp_bkmrk_pml_peer_ref_destruct( ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref) {
opal_list_item_t* item = NULL; opal_list_item_t* item = NULL;
peer_ref->proc_name.jobid = ORCA_JOBID_INVALID; peer_ref->proc_name.jobid = ORTE_JOBID_INVALID;
peer_ref->proc_name.vpid = ORCA_VPID_INVALID; peer_ref->proc_name.vpid = ORTE_VPID_INVALID;
while( NULL != (item = opal_list_remove_first(&peer_ref->send_list)) ) { while( NULL != (item = opal_list_remove_first(&peer_ref->send_list)) ) {
HOKE_TRAFFIC_MSG_REF_RETURN(item); HOKE_TRAFFIC_MSG_REF_RETURN(item);
@ -833,8 +836,8 @@ void ompi_crcp_bkmrk_pml_traffic_message_ref_construct(ompi_crcp_bkmrk_pml_traff
OBJ_CONSTRUCT(&msg_ref->msg_contents, opal_list_t); OBJ_CONSTRUCT(&msg_ref->msg_contents, opal_list_t);
msg_ref->proc_name.jobid = ORCA_JOBID_INVALID; msg_ref->proc_name.jobid = ORTE_JOBID_INVALID;
msg_ref->proc_name.vpid = ORCA_VPID_INVALID; msg_ref->proc_name.vpid = ORTE_VPID_INVALID;
msg_ref->matched = INVALID_INT; msg_ref->matched = INVALID_INT;
msg_ref->done = INVALID_INT; msg_ref->done = INVALID_INT;
@ -860,8 +863,8 @@ void ompi_crcp_bkmrk_pml_traffic_message_ref_destruct( ompi_crcp_bkmrk_pml_traff
} }
OBJ_DESTRUCT(&(msg_ref->msg_contents)); OBJ_DESTRUCT(&(msg_ref->msg_contents));
msg_ref->proc_name.jobid = ORCA_JOBID_INVALID; msg_ref->proc_name.jobid = ORTE_JOBID_INVALID;
msg_ref->proc_name.vpid = ORCA_VPID_INVALID; msg_ref->proc_name.vpid = ORTE_VPID_INVALID;
msg_ref->matched = INVALID_INT; msg_ref->matched = INVALID_INT;
msg_ref->done = INVALID_INT; msg_ref->done = INVALID_INT;
@ -893,8 +896,8 @@ void ompi_crcp_bkmrk_pml_drain_message_ref_construct(ompi_crcp_bkmrk_pml_drain_m
OBJ_CONSTRUCT(&msg_ref->msg_contents, opal_list_t); OBJ_CONSTRUCT(&msg_ref->msg_contents, opal_list_t);
msg_ref->proc_name.jobid = ORCA_JOBID_INVALID; msg_ref->proc_name.jobid = ORTE_JOBID_INVALID;
msg_ref->proc_name.vpid = ORCA_VPID_INVALID; msg_ref->proc_name.vpid = ORTE_VPID_INVALID;
msg_ref->done = INVALID_INT; msg_ref->done = INVALID_INT;
msg_ref->active = INVALID_INT; msg_ref->active = INVALID_INT;
@ -924,8 +927,8 @@ void ompi_crcp_bkmrk_pml_drain_message_ref_destruct( ompi_crcp_bkmrk_pml_drain_m
} }
OBJ_DESTRUCT(&(msg_ref->msg_contents)); OBJ_DESTRUCT(&(msg_ref->msg_contents));
msg_ref->proc_name.jobid = ORCA_JOBID_INVALID; msg_ref->proc_name.jobid = ORTE_JOBID_INVALID;
msg_ref->proc_name.vpid = ORCA_VPID_INVALID; msg_ref->proc_name.vpid = ORTE_VPID_INVALID;
msg_ref->done = INVALID_INT; msg_ref->done = INVALID_INT;
msg_ref->active = INVALID_INT; msg_ref->active = INVALID_INT;
@ -943,15 +946,15 @@ OBJ_CLASS_INSTANCE(ompi_crcp_bkmrk_pml_drain_message_ack_ref_t,
void ompi_crcp_bkmrk_pml_drain_message_ack_ref_construct(ompi_crcp_bkmrk_pml_drain_message_ack_ref_t *msg_ack_ref) { void ompi_crcp_bkmrk_pml_drain_message_ack_ref_construct(ompi_crcp_bkmrk_pml_drain_message_ack_ref_t *msg_ack_ref) {
msg_ack_ref->complete = false; msg_ack_ref->complete = false;
msg_ack_ref->peer.jobid = ORCA_JOBID_INVALID; msg_ack_ref->peer.jobid = ORTE_JOBID_INVALID;
msg_ack_ref->peer.vpid = ORCA_VPID_INVALID; msg_ack_ref->peer.vpid = ORTE_VPID_INVALID;
} }
void ompi_crcp_bkmrk_pml_drain_message_ack_ref_destruct( ompi_crcp_bkmrk_pml_drain_message_ack_ref_t *msg_ack_ref) { void ompi_crcp_bkmrk_pml_drain_message_ack_ref_destruct( ompi_crcp_bkmrk_pml_drain_message_ack_ref_t *msg_ack_ref) {
msg_ack_ref->complete = false; msg_ack_ref->complete = false;
msg_ack_ref->peer.jobid = ORCA_JOBID_INVALID; msg_ack_ref->peer.jobid = ORTE_JOBID_INVALID;
msg_ack_ref->peer.vpid = ORCA_VPID_INVALID; msg_ack_ref->peer.vpid = ORTE_VPID_INVALID;
} }
@ -1066,7 +1069,7 @@ do { \
#define UNPACK_BUFFER(buffer, var, count, type, error_msg) \ #define UNPACK_BUFFER(buffer, var, count, type, error_msg) \
{ \ { \
orca_std_cntr_t n = count; \ orte_std_cntr_t n = count; \
if (OPAL_SUCCESS != (ret = opal_dss.unpack(buffer, &(var), &n, type)) ) { \ if (OPAL_SUCCESS != (ret = opal_dss.unpack(buffer, &(var), &n, type)) ) { \
opal_output(mca_crcp_bkmrk_component.super.output_handle, \ opal_output(mca_crcp_bkmrk_component.super.output_handle, \
"%s (Return %d)", error_msg, ret); \ "%s (Return %d)", error_msg, ret); \
@ -1484,7 +1487,7 @@ ompi_crcp_base_pml_state_t* ompi_crcp_bkmrk_pml_del_procs(
if(NULL == item) { if(NULL == item) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: del_procs: Unable to find peer %s\n", "crcp:bkmrk: del_procs: Unable to find peer %s\n",
ORCA_NAME_PRINT(&(procs[i]->proc_name))); ORTE_NAME_PRINT(&(procs[i]->proc_name)));
exit_status = OMPI_ERROR; exit_status = OMPI_ERROR;
goto DONE; goto DONE;
} }
@ -3033,7 +3036,7 @@ ompi_crcp_base_pml_state_t* ompi_crcp_bkmrk_pml_ft_event(
if( opal_cr_timing_barrier_enabled ) { if( opal_cr_timing_barrier_enabled ) {
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_CRCPBR0); OPAL_CR_SET_TIMER(OPAL_CR_TIMER_CRCPBR0);
orca_coll_barrier(ORCA_COLL_TYPE_BARRIER_CR); orte_grpcomm.barrier();
} }
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_CRCP0); OPAL_CR_SET_TIMER(OPAL_CR_TIMER_CRCP0);
@ -3048,7 +3051,7 @@ ompi_crcp_base_pml_state_t* ompi_crcp_bkmrk_pml_ft_event(
if( OMPI_SUCCESS != (ret = ompi_crcp_bkmrk_pml_quiesce_start(QUIESCE_TAG_CKPT)) ) { if( OMPI_SUCCESS != (ret = ompi_crcp_bkmrk_pml_quiesce_start(QUIESCE_TAG_CKPT)) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s ft_event: Checkpoint Coordination Failed %d", "crcp:bkmrk: %s ft_event: Checkpoint Coordination Failed %d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ret); ret);
exit_status = ret; exit_status = ret;
goto DONE; goto DONE;
@ -3078,7 +3081,7 @@ ompi_crcp_base_pml_state_t* ompi_crcp_bkmrk_pml_ft_event(
first_continue_pass = !first_continue_pass; first_continue_pass = !first_continue_pass;
/* Only finalize the Protocol after the PML has been rebuilt */ /* Only finalize the Protocol after the PML has been rebuilt */
if( orca_info_cr_continue_like_restart() && first_continue_pass ) { if( orte_cr_continue_like_restart && first_continue_pass ) {
goto DONE; goto DONE;
} }
@ -3101,7 +3104,7 @@ ompi_crcp_base_pml_state_t* ompi_crcp_bkmrk_pml_ft_event(
if( opal_cr_timing_barrier_enabled ) { if( opal_cr_timing_barrier_enabled ) {
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_COREBR1); OPAL_CR_SET_TIMER(OPAL_CR_TIMER_COREBR1);
orca_coll_barrier(ORCA_COLL_TYPE_BARRIER_CR); orte_grpcomm.barrier();
} }
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_CORE2); OPAL_CR_SET_TIMER(OPAL_CR_TIMER_CORE2);
} }
@ -3124,7 +3127,7 @@ ompi_crcp_base_pml_state_t* ompi_crcp_bkmrk_pml_ft_event(
cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
/* JJH - Assuming only one global jobid at the moment */ /* JJH - Assuming only one global jobid at the moment */
cur_peer_ref->proc_name.jobid = orca_process_info_get_jobid(ORCA_PROC_MY_NAME); cur_peer_ref->proc_name.jobid = ORTE_PROC_MY_NAME->jobid;
} }
/* /*
@ -3227,7 +3230,7 @@ static int traffic_message_append(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
} else { } else {
CREATE_NEW_MSG((*msg_ref), msg_type, CREATE_NEW_MSG((*msg_ref), msg_type,
count, ddt_size, tag, dest, comm, count, ddt_size, tag, dest, comm,
ORCA_JOBID_INVALID, ORCA_VPID_INVALID); ORTE_JOBID_INVALID, ORTE_VPID_INVALID);
} }
if( msg_type == COORD_MSG_TYPE_P_SEND || if( msg_type == COORD_MSG_TYPE_P_SEND ||
@ -3392,7 +3395,7 @@ static int traffic_message_find_mark_persistent(ompi_crcp_bkmrk_pml_traffic_mess
else if( loc_breq->req_sequence == breq->req_sequence ) { else if( loc_breq->req_sequence == breq->req_sequence ) {
OPAL_OUTPUT_VERBOSE((25, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((25, mca_crcp_bkmrk_component.super.output_handle,
"%s %8s Request [%d] (%s) %d : %d", "%s %8s Request [%d] (%s) %d : %d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(set_is_active ? "Start" : (NULL != c_ref ? "Drain" : "Complete")), (set_is_active ? "Start" : (NULL != c_ref ? "Drain" : "Complete")),
(int)msg_ref->msg_id, (int)msg_ref->msg_id,
(content_ref->active ? "T" : "F"), (content_ref->active ? "T" : "F"),
@ -3493,8 +3496,8 @@ static int traffic_message_create_drain_message(bool post_drain,
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
" --> Create Drain Msg: %s %4d = min(%4d / %4d)", " --> Create Drain Msg: %s %4d = min(%4d / %4d)",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
(post_drain ? "Posting" : "Not Posting"), (post_drain ? "Posting" : "Not Posting"),
m_total, (*posted_msg_ref)->active, max_post )); m_total, (*posted_msg_ref)->active, max_post ));
@ -3521,8 +3524,8 @@ static int traffic_message_create_drain_message(bool post_drain,
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
" \t--> Find Content: %s (%4d of %4d)", " \t--> Find Content: %s (%4d of %4d)",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
(post_drain ? "Posting" : "Not Posting"), (post_drain ? "Posting" : "Not Posting"),
m_iter, m_total)); m_iter, m_total));
@ -3599,8 +3602,8 @@ static int traffic_message_create_drain_message(bool post_drain,
OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Added %d messages to the drained list (size = %d)", "Added %d messages to the drained list (size = %d)",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
(*num_posted), (*num_posted),
(int)opal_list_get_size(&(peer_ref->drained_list)) )); (int)opal_list_get_size(&(peer_ref->drained_list)) ));
@ -3917,7 +3920,7 @@ static int drain_message_find_any(size_t count, int tag, int peer,
continue; continue;
} }
if( OPAL_EQUAL != orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL != orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&(cur_peer_ref->proc_name), &(cur_peer_ref->proc_name),
&(comm->c_local_group->grp_proc_pointers[peer]->proc_name)) ) { &(comm->c_local_group->grp_proc_pointers[peer]->proc_name)) ) {
continue; continue;
@ -4136,10 +4139,10 @@ static int drain_message_copy_remove(ompi_crcp_bkmrk_pml_drain_message_ref_t *dr
/************************************************ /************************************************
* Peer List Utility Functions * Peer List Utility Functions
************************************************/ ************************************************/
static ompi_crcp_bkmrk_pml_peer_ref_t * find_peer(orca_process_name_t proc) static ompi_crcp_bkmrk_pml_peer_ref_t * find_peer(orte_process_name_t proc)
{ {
opal_list_item_t* item = NULL; opal_list_item_t* item = NULL;
orca_name_cmp_bitmask_t mask; orte_ns_cmp_bitmask_t mask;
for(item = opal_list_get_first(&ompi_crcp_bkmrk_pml_peer_refs); for(item = opal_list_get_first(&ompi_crcp_bkmrk_pml_peer_refs);
item != opal_list_get_end(&ompi_crcp_bkmrk_pml_peer_refs); item != opal_list_get_end(&ompi_crcp_bkmrk_pml_peer_refs);
@ -4147,9 +4150,9 @@ static ompi_crcp_bkmrk_pml_peer_ref_t * find_peer(orca_process_name_t proc)
ompi_crcp_bkmrk_pml_peer_ref_t *cur_peer_ref; ompi_crcp_bkmrk_pml_peer_ref_t *cur_peer_ref;
cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
mask = ORCA_NAME_CMP_JOBID | ORCA_NAME_CMP_VPID; mask = ORTE_NS_CMP_JOBID | ORTE_NS_CMP_VPID;
if( OPAL_EQUAL == orca_process_name_compare(mask, if( OPAL_EQUAL == orte_util_compare_name_fields(mask,
&(cur_peer_ref->proc_name), &(cur_peer_ref->proc_name),
&proc) ) { &proc) ) {
return cur_peer_ref; return cur_peer_ref;
@ -4251,7 +4254,7 @@ static int ft_event_coordinate_peers(void)
if( stall_for_completion ) { if( stall_for_completion ) {
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s **** STALLING %s in PID %d ***", "crcp:bkmrk: %s **** STALLING %s in PID %d ***",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(current_msg_type == COORD_MSG_TYPE_B_SEND ? "Send" : "Recv"), (current_msg_type == COORD_MSG_TYPE_B_SEND ? "Send" : "Recv"),
getpid() )); getpid() ));
step_to_return_to = 1; step_to_return_to = 1;
@ -4278,7 +4281,7 @@ static int ft_event_coordinate_peers(void)
OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s Coordination Finished...\n", "crcp:bkmrk: %s Coordination Finished...\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME))); ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* /*
* Now that all our peer channels are marked as drained * Now that all our peer channels are marked as drained
@ -4310,8 +4313,8 @@ static int ft_event_finalize_exchange(void)
ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref; ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref;
peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
if( OPAL_EQUAL != orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL != orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
(ORCA_PROC_MY_NAME), (ORTE_PROC_MY_NAME),
&(peer_ref->proc_name)) ) { &(peer_ref->proc_name)) ) {
TRAFFIC_MSG_DUMP_PEER(10, (peer_ref, "finalize_exchange", false)); TRAFFIC_MSG_DUMP_PEER(10, (peer_ref, "finalize_exchange", false));
} }
@ -4418,11 +4421,10 @@ static int ft_event_finalize_exchange(void)
static int ft_event_exchange_bookmarks(void) static int ft_event_exchange_bookmarks(void)
{ {
int peer_idx = 0; int peer_idx = 0;
int my_idx; int my_idx = ORTE_PROC_MY_NAME->vpid;
int iter = 0; int iter = 0;
int num_peers = 0; int num_peers = 0;
my_idx = orca_process_info_get_vpid(ORCA_PROC_MY_NAME);
num_peers = opal_list_get_size(&ompi_crcp_bkmrk_pml_peer_refs); num_peers = opal_list_get_size(&ompi_crcp_bkmrk_pml_peer_refs);
for( peer_idx = (num_peers - my_idx - 1), iter = 0; for( peer_idx = (num_peers - my_idx - 1), iter = 0;
@ -4462,15 +4464,15 @@ static int ft_event_check_bookmarks(void)
int p_n_from_p_m = 0; int p_n_from_p_m = 0;
if( 10 <= mca_crcp_bkmrk_component.super.verbose ) { if( 10 <= mca_crcp_bkmrk_component.super.verbose ) {
sleep(orca_process_info_get_vpid(ORCA_PROC_MY_NAME)); sleep(ORTE_PROC_MY_NAME->vpid);
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"---------------------------------------------")); "---------------------------------------------"));
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"Process %s Match Table", "Process %s Match Table",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME))); ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"%s %5s | %7s | %7s | %7s | %7s |", "%s %5s | %7s | %7s | %7s | %7s |",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
"Vpid", "T_Send", "M_Recv", "M_Send", "T_Recv")); "Vpid", "T_Send", "M_Recv", "M_Send", "T_Recv"));
for(item = opal_list_get_first(&ompi_crcp_bkmrk_pml_peer_refs); for(item = opal_list_get_first(&ompi_crcp_bkmrk_pml_peer_refs);
@ -4488,7 +4490,7 @@ static int ft_event_check_bookmarks(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"%s %5d | %7d | %7d | %7d | %7d |", "%s %5d | %7d | %7d | %7d | %7d |",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
peer_ref->proc_name.vpid, peer_ref->proc_name.vpid,
t_send, m_recv, m_send, t_recv)); t_send, m_recv, m_send, t_recv));
} }
@ -4507,8 +4509,8 @@ static int ft_event_check_bookmarks(void)
ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref; ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref;
peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
if( OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
(ORCA_PROC_MY_NAME), (ORTE_PROC_MY_NAME),
&(peer_ref->proc_name)) ) { &(peer_ref->proc_name)) ) {
continue; continue;
} }
@ -4516,7 +4518,7 @@ static int ft_event_check_bookmarks(void)
TRAFFIC_MSG_DUMP_PEER(15, (peer_ref, "-- Bookmark Details --", false)); TRAFFIC_MSG_DUMP_PEER(15, (peer_ref, "-- Bookmark Details --", false));
/* Lowest Rank sends first */ /* Lowest Rank sends first */
if( orca_process_info_get_vpid(ORCA_PROC_MY_NAME) < peer_ref->proc_name.vpid ) { if( ORTE_PROC_MY_NAME->vpid < peer_ref->proc_name.vpid ) {
/******************** /********************
* Check P_n --> P_m * Check P_n --> P_m
* Has the peer received all the messages that I have put on the wire? * Has the peer received all the messages that I have put on the wire?
@ -4530,8 +4532,8 @@ static int ft_event_check_bookmarks(void)
"crcp:bkmrk: %s --> %s " "crcp:bkmrk: %s --> %s "
"Total Sent (%4d) = Matched Recv. (%4d) => Diff (%4d). " "Total Sent (%4d) = Matched Recv. (%4d) => Diff (%4d). "
" WARNING: Peer received more than was sent. :(\n", " WARNING: Peer received more than was sent. :(\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4544,8 +4546,8 @@ static int ft_event_check_bookmarks(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s --> %s " "crcp:bkmrk: %s --> %s "
"Total Sent (%4d) = Matched Recv. (%4d). Peer needs %4d.\n", "Total Sent (%4d) = Matched Recv. (%4d). Peer needs %4d.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4558,7 +4560,7 @@ static int ft_event_check_bookmarks(void)
if( OMPI_SUCCESS != (ret = send_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) { if( OMPI_SUCCESS != (ret = send_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: check_bookmarks: Unable to send message details to peer %s: Return %d\n", "crcp:bkmrk: check_bookmarks: Unable to send message details to peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_ref->proc_name), ORTE_NAME_PRINT(&peer_ref->proc_name),
ret); ret);
return ret; return ret;
} }
@ -4577,8 +4579,8 @@ static int ft_event_check_bookmarks(void)
"crcp:bkmrk: %s --> %s " "crcp:bkmrk: %s --> %s "
"Matched Sent (%4d) = Total Recv. (%4d) => Diff (%4d). " "Matched Sent (%4d) = Total Recv. (%4d) => Diff (%4d). "
" WARNING: I received more than the peer sent. :(\n", " WARNING: I received more than the peer sent. :(\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4591,8 +4593,8 @@ static int ft_event_check_bookmarks(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Matched Sent (%4d) = Total Recv. (%4d). I need %4d.\n", "Matched Sent (%4d) = Total Recv. (%4d). I need %4d.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4604,7 +4606,7 @@ static int ft_event_check_bookmarks(void)
if( OMPI_SUCCESS != (ret = recv_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) { if( OMPI_SUCCESS != (ret = recv_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: check_bookmarks: Unable to recv message details from peer %s: Return %d\n", "crcp:bkmrk: check_bookmarks: Unable to recv message details from peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_ref->proc_name), ORTE_NAME_PRINT(&peer_ref->proc_name),
ret); ret);
return ret; return ret;
} }
@ -4625,8 +4627,8 @@ static int ft_event_check_bookmarks(void)
"crcp:bkmrk: %s --> %s " "crcp:bkmrk: %s --> %s "
"Matched Sent (%4d) = Total Recv. (%4d) => Diff (%4d). " "Matched Sent (%4d) = Total Recv. (%4d) => Diff (%4d). "
" WARNING: I received more than the peer sent. :(\n", " WARNING: I received more than the peer sent. :(\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4639,8 +4641,8 @@ static int ft_event_check_bookmarks(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Matched Sent (%4d) = Total Recv. (%4d). I need %4d.\n", "Matched Sent (%4d) = Total Recv. (%4d). I need %4d.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4652,7 +4654,7 @@ static int ft_event_check_bookmarks(void)
if( OMPI_SUCCESS != (ret = recv_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) { if( OMPI_SUCCESS != (ret = recv_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: check_bookmarks: Unable to recv message details from peer %s: Return %d\n", "crcp:bkmrk: check_bookmarks: Unable to recv message details from peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_ref->proc_name), ORTE_NAME_PRINT(&peer_ref->proc_name),
ret); ret);
return ret; return ret;
} }
@ -4671,8 +4673,8 @@ static int ft_event_check_bookmarks(void)
"crcp:bkmrk: %s --> %s " "crcp:bkmrk: %s --> %s "
"Total Sent (%4d) = Matched Recv. (%4d) => Diff (%4d). " "Total Sent (%4d) = Matched Recv. (%4d) => Diff (%4d). "
" WARNING: Peer received more than was sent. :(\n", " WARNING: Peer received more than was sent. :(\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4685,8 +4687,8 @@ static int ft_event_check_bookmarks(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s --> %s " "crcp:bkmrk: %s --> %s "
"Total Sent (%4d) = Matched Recv. (%4d). Peer needs %4d.\n", "Total Sent (%4d) = Matched Recv. (%4d). Peer needs %4d.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
p_n_to_p_m, p_n_to_p_m,
p_n_from_p_m, p_n_from_p_m,
(p_n_to_p_m - p_n_from_p_m) (p_n_to_p_m - p_n_from_p_m)
@ -4699,7 +4701,7 @@ static int ft_event_check_bookmarks(void)
if( OMPI_SUCCESS != (ret = send_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) { if( OMPI_SUCCESS != (ret = send_msg_details(peer_ref, p_n_to_p_m, p_n_from_p_m) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: check_bookmarks: Unable to send message details to peer %s: Return %d\n", "crcp:bkmrk: check_bookmarks: Unable to send message details to peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_ref->proc_name), ORTE_NAME_PRINT(&peer_ref->proc_name),
ret); ret);
return ret; return ret;
} }
@ -4724,7 +4726,7 @@ static int ft_event_post_drain_acks(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s Wait on %d Drain ACK Messages.\n", "crcp:bkmrk: %s Wait on %d Drain ACK Messages.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(int)req_size)); (int)req_size));
/* /*
@ -4737,15 +4739,15 @@ static int ft_event_post_drain_acks(void)
drain_msg_ack = (ompi_crcp_bkmrk_pml_drain_message_ack_ref_t*)item; drain_msg_ack = (ompi_crcp_bkmrk_pml_drain_message_ack_ref_t*)item;
/* Post the receive */ /* Post the receive */
if( ORCA_SUCCESS != (ret = orca_oob_recv_buffer_nb( &drain_msg_ack->peer, if( ORTE_SUCCESS != (ret = orte_rml.recv_buffer_nb( &drain_msg_ack->peer,
OMPI_CRCP_COORD_BOOKMARK_TAG, OMPI_CRCP_COORD_BOOKMARK_TAG,
0, 0,
drain_message_ack_cbfunc, drain_message_ack_cbfunc,
NULL) ) ) { NULL) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s: Failed to post a RML receive to the peer\n", "crcp:bkmrk: %s <-- %s: Failed to post a RML receive to the peer\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(drain_msg_ack->peer))); ORTE_NAME_PRINT(&(drain_msg_ack->peer)));
return ret; return ret;
} }
} }
@ -4754,9 +4756,9 @@ static int ft_event_post_drain_acks(void)
} }
static void drain_message_ack_cbfunc(int status, static void drain_message_ack_cbfunc(int status,
orca_process_name_t* sender, orte_process_name_t* sender,
opal_buffer_t *buffer, opal_buffer_t *buffer,
orca_oob_tag_t tag, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
int ret, exit_status = OMPI_SUCCESS; int ret, exit_status = OMPI_SUCCESS;
@ -4780,15 +4782,15 @@ static void drain_message_ack_cbfunc(int status,
/* If this ACK has not completed yet */ /* If this ACK has not completed yet */
if(!drain_msg_ack->complete) { if(!drain_msg_ack->complete) {
/* If it is the correct peer */ /* If it is the correct peer */
if( OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&(drain_msg_ack->peer), &(drain_msg_ack->peer),
sender) ) { sender) ) {
/* We found it! */ /* We found it! */
drain_msg_ack->complete = true; drain_msg_ack->complete = true;
OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s --> %s Received ACK of FLUSH from peer\n", "crcp:bkmrk: %s --> %s Received ACK of FLUSH from peer\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(sender) )); ORTE_NAME_PRINT(sender) ));
return; return;
} }
} }
@ -4796,8 +4798,8 @@ static void drain_message_ack_cbfunc(int status,
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s --> %s ERROR: Unable to match ACK to peer\n", "crcp:bkmrk: %s --> %s ERROR: Unable to match ACK to peer\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(sender) ); ORTE_NAME_PRINT(sender) );
cleanup: cleanup:
return; return;
@ -4818,8 +4820,8 @@ static int ft_event_post_drained(void)
item = opal_list_get_next(item) ) { item = opal_list_get_next(item) ) {
cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
if( OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
(ORCA_PROC_MY_NAME), (ORTE_PROC_MY_NAME),
&(cur_peer_ref->proc_name)) ) { &(cur_peer_ref->proc_name)) ) {
continue; continue;
} }
@ -4883,8 +4885,8 @@ static int ft_event_post_drained(void)
cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
peer_total = 0; peer_total = 0;
if( OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
(ORCA_PROC_MY_NAME), (ORTE_PROC_MY_NAME),
&(cur_peer_ref->proc_name)) ) { &(cur_peer_ref->proc_name)) ) {
continue; continue;
} }
@ -4930,8 +4932,8 @@ static int ft_event_post_drained(void)
if( peer_total > 0 || stall_for_completion ) { if( peer_total > 0 || stall_for_completion ) {
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s Will be draining %4d messages from this peer. Total %4d %s\n", "crcp:bkmrk: %s <-- %s Will be draining %4d messages from this peer. Total %4d %s\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(cur_peer_ref->proc_name)), ORTE_NAME_PRINT(&(cur_peer_ref->proc_name)),
peer_total, peer_total,
quiesce_request_count, quiesce_request_count,
(stall_for_completion ? "(And Stalling)" : "") )); (stall_for_completion ? "(And Stalling)" : "") ));
@ -4961,8 +4963,8 @@ static int ft_event_post_drain_message(ompi_crcp_bkmrk_pml_drain_message_ref_t
if( content_ref->already_posted ) { if( content_ref->already_posted ) {
OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s Found a message that we do not need to post.\n", "crcp:bkmrk: %s <-- %s Found a message that we do not need to post.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(drain_msg_ref->proc_name)) )); ORTE_NAME_PRINT(&(drain_msg_ref->proc_name)) ));
return OMPI_SUCCESS; return OMPI_SUCCESS;
} }
@ -4975,8 +4977,8 @@ static int ft_event_post_drain_message(ompi_crcp_bkmrk_pml_drain_message_ref_t
*/ */
OPAL_OUTPUT_VERBOSE((20, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((20, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s Posting a message to be drained from rank %d.\n", "crcp:bkmrk: %s <-- %s Posting a message to be drained from rank %d.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(drain_msg_ref->proc_name)), ORTE_NAME_PRINT(&(drain_msg_ref->proc_name)),
drain_msg_ref->rank)); drain_msg_ref->rank));
if( OMPI_SUCCESS != (ret = wrapped_pml_module->pml_irecv(content_ref->buffer, if( OMPI_SUCCESS != (ret = wrapped_pml_module->pml_irecv(content_ref->buffer,
(drain_msg_ref->count * drain_msg_ref->ddt_size), (drain_msg_ref->count * drain_msg_ref->ddt_size),
@ -4987,8 +4989,8 @@ static int ft_event_post_drain_message(ompi_crcp_bkmrk_pml_drain_message_ref_t
&(content_ref->request) ) ) ) { &(content_ref->request) ) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s Failed to post the Draining PML iRecv\n", "crcp:bkmrk: %s <-- %s Failed to post the Draining PML iRecv\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(drain_msg_ref->proc_name)) ); ORTE_NAME_PRINT(&(drain_msg_ref->proc_name)) );
return ret; return ret;
} }
@ -5006,7 +5008,7 @@ static int ft_event_wait_quiesce(void)
if( OMPI_SUCCESS != (ret = wait_quiesce_drained() ) ) { if( OMPI_SUCCESS != (ret = wait_quiesce_drained() ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: wait_quiesce: %s Failed to quiesce drained messages\n", "crcp:bkmrk: wait_quiesce: %s Failed to quiesce drained messages\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME) ); ORTE_NAME_PRINT(ORTE_PROC_MY_NAME) );
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
} }
@ -5017,7 +5019,7 @@ static int ft_event_wait_quiesce(void)
if( OMPI_SUCCESS != (ret = wait_quiesce_drain_ack() ) ) { if( OMPI_SUCCESS != (ret = wait_quiesce_drain_ack() ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: wait_quiesce: %s Failed to recv all drain ACKs\n", "crcp:bkmrk: wait_quiesce: %s Failed to recv all drain ACKs\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME) ); ORTE_NAME_PRINT(ORTE_PROC_MY_NAME) );
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
} }
@ -5039,7 +5041,7 @@ static int wait_quiesce_drained(void)
OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s Waiting on %d messages to drain\n", "crcp:bkmrk: %s Waiting on %d messages to drain\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(int)quiesce_request_count)); (int)quiesce_request_count));
/* /*
@ -5068,8 +5070,8 @@ static int wait_quiesce_drained(void)
item = opal_list_get_next(item) ) { item = opal_list_get_next(item) ) {
cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item; cur_peer_ref = (ompi_crcp_bkmrk_pml_peer_ref_t*)item;
if( OPAL_EQUAL == orca_process_name_compare(ORCA_NAME_CMP_ALL, if( OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
(ORCA_PROC_MY_NAME), (ORTE_PROC_MY_NAME),
&(cur_peer_ref->proc_name)) ) { &(cur_peer_ref->proc_name)) ) {
continue; continue;
} }
@ -5083,8 +5085,8 @@ static int wait_quiesce_drained(void)
OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((5, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s --> %s Send ACKs to Peer\n", "crcp:bkmrk: %s --> %s Send ACKs to Peer\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(cur_peer_ref->proc_name)) )); ORTE_NAME_PRINT(&(cur_peer_ref->proc_name)) ));
/* Send All Clear to Peer */ /* Send All Clear to Peer */
if (NULL == (buffer = OBJ_NEW(opal_buffer_t))) { if (NULL == (buffer = OBJ_NEW(opal_buffer_t))) {
@ -5095,7 +5097,7 @@ static int wait_quiesce_drained(void)
PACK_BUFFER(buffer, response, 1, OPAL_SIZE, ""); PACK_BUFFER(buffer, response, 1, OPAL_SIZE, "");
/* JJH - Performance Optimization? - Why not post all isends, then wait? */ /* JJH - Performance Optimization? - Why not post all isends, then wait? */
if ( 0 > ( ret = orca_oob_send_buffer(&(cur_peer_ref->proc_name), buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) { if ( 0 > ( ret = orte_rml.send_buffer(&(cur_peer_ref->proc_name), buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) {
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
} }
@ -5184,7 +5186,7 @@ static int coord_request_wait_all( size_t count,
OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s Request Wait: Done with idx %d of %d\n", "crcp:bkmrk: %s Request Wait: Done with idx %d of %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(int)i, (int)count)); (int)i, (int)count));
} }
@ -5223,7 +5225,7 @@ static int wait_quiesce_drain_ack(void)
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s Waiting on %d Drain ACK messages\n", "crcp:bkmrk: %s Waiting on %d Drain ACK messages\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
num_outstanding)); num_outstanding));
while(0 < num_outstanding) { while(0 < num_outstanding) {
@ -5256,7 +5258,7 @@ static int wait_quiesce_drain_ack(void)
static int send_bookmarks(int peer_idx) static int send_bookmarks(int peer_idx)
{ {
ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref; ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref;
orca_process_name_t peer_name; orte_process_name_t peer_name;
opal_buffer_t *buffer = NULL; opal_buffer_t *buffer = NULL;
int exit_status = OMPI_SUCCESS; int exit_status = OMPI_SUCCESS;
int ret; int ret;
@ -5265,7 +5267,7 @@ static int send_bookmarks(int peer_idx)
/* /*
* Find the peer structure for this peer * Find the peer structure for this peer
*/ */
peer_name.jobid = orca_process_info_get_jobid(ORCA_PROC_MY_NAME); peer_name.jobid = ORTE_PROC_MY_NAME->jobid;
peer_name.vpid = peer_idx; peer_name.vpid = peer_idx;
if( NULL == (peer_ref = find_peer(peer_name))) { if( NULL == (peer_ref = find_peer(peer_name))) {
@ -5278,8 +5280,8 @@ static int send_bookmarks(int peer_idx)
OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s --> %s Sending bookmark (S[%6d] R[%6d])\n", "crcp:bkmrk: %s --> %s Sending bookmark (S[%6d] R[%6d])\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&peer_name), ORTE_NAME_PRINT(&peer_name),
peer_ref->total_msgs_sent, peer_ref->total_msgs_sent,
peer_ref->total_msgs_recvd)); peer_ref->total_msgs_recvd));
@ -5296,10 +5298,10 @@ static int send_bookmarks(int peer_idx)
PACK_BUFFER(buffer, (peer_ref->total_msgs_recvd), 1, OPAL_UINT32, PACK_BUFFER(buffer, (peer_ref->total_msgs_recvd), 1, OPAL_UINT32,
"crcp:bkmrk: send_bookmarks: Unable to pack total_msgs_recvd"); "crcp:bkmrk: send_bookmarks: Unable to pack total_msgs_recvd");
if ( 0 > ( ret = orca_oob_send_buffer(&peer_name, buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) { if ( 0 > ( ret = orte_rml.send_buffer(&peer_name, buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: send_bookmarks: Failed to send bookmark to peer %s: Return %d\n", "crcp:bkmrk: send_bookmarks: Failed to send bookmark to peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_name), ORTE_NAME_PRINT(&peer_name),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5319,23 +5321,23 @@ static int send_bookmarks(int peer_idx)
/* Paired with send_bookmarks */ /* Paired with send_bookmarks */
static int recv_bookmarks(int peer_idx) static int recv_bookmarks(int peer_idx)
{ {
orca_process_name_t peer_name; orte_process_name_t peer_name;
int exit_status = OMPI_SUCCESS; int exit_status = OMPI_SUCCESS;
int ret; int ret;
START_TIMER(CRCP_TIMER_CKPT_EX_PEER_R); START_TIMER(CRCP_TIMER_CKPT_EX_PEER_R);
peer_name.jobid = orca_process_info_get_jobid(ORCA_PROC_MY_NAME); peer_name.jobid = ORTE_PROC_MY_NAME->jobid;
peer_name.vpid = peer_idx; peer_name.vpid = peer_idx;
if ( 0 > (ret = orca_oob_recv_buffer_nb(&peer_name, if ( 0 > (ret = orte_rml.recv_buffer_nb(&peer_name,
OMPI_CRCP_COORD_BOOKMARK_TAG, OMPI_CRCP_COORD_BOOKMARK_TAG,
0, 0,
recv_bookmarks_cbfunc, recv_bookmarks_cbfunc,
NULL) ) ) { NULL) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: recv_bookmarks: Failed to post receive bookmark from peer %s: Return %d\n", "crcp:bkmrk: recv_bookmarks: Failed to post receive bookmark from peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_name), ORTE_NAME_PRINT(&peer_name),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5352,15 +5354,15 @@ static int recv_bookmarks(int peer_idx)
} }
static void recv_bookmarks_cbfunc(int status, static void recv_bookmarks_cbfunc(int status,
orca_process_name_t* sender, orte_process_name_t* sender,
opal_buffer_t *buffer, opal_buffer_t *buffer,
orca_oob_tag_t tag, orte_rml_tag_t tag,
void* cbdata) void* cbdata)
{ {
ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref; ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref;
int exit_status = OMPI_SUCCESS; int exit_status = OMPI_SUCCESS;
int ret, tmp_int; int ret, tmp_int;
orca_vpid_t peer_idx; orte_vpid_t peer_idx;
peer_idx = sender->vpid; peer_idx = sender->vpid;
@ -5385,8 +5387,8 @@ static void recv_bookmarks_cbfunc(int status,
OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((15, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s Received bookmark (S[%6d] R[%6d]) vs. (S[%6d] R[%6d])\n", "crcp:bkmrk: %s <-- %s Received bookmark (S[%6d] R[%6d]) vs. (S[%6d] R[%6d])\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(sender), ORTE_NAME_PRINT(sender),
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
peer_ref->matched_msgs_recvd, peer_ref->matched_msgs_recvd,
peer_ref->total_msgs_sent, peer_ref->total_msgs_sent,
@ -5450,8 +5452,8 @@ static int send_msg_details(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
if(OMPI_SUCCESS != (ret = do_send_msg_detail(peer_ref, msg_ref, &num_matches, &p_total_found, &finished)) ) { if(OMPI_SUCCESS != (ret = do_send_msg_detail(peer_ref, msg_ref, &num_matches, &p_total_found, &finished)) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: send_msg_details: %s --> %s Failed to send message details to peer. Return %d\n", "crcp:bkmrk: send_msg_details: %s --> %s Failed to send message details to peer. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
} }
@ -5513,8 +5515,8 @@ static int send_msg_details(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
opal_list_append(&drained_msg_ack_list, &(d_msg_ack->super)); opal_list_append(&drained_msg_ack_list, &(d_msg_ack->super));
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-> %s Message Inflight! Will wait on ACK from this peer.\n", "crcp:bkmrk: %s <-> %s Message Inflight! Will wait on ACK from this peer.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)))); ORTE_NAME_PRINT(&(peer_ref->proc_name))));
END_TIMER(CRCP_TIMER_CKPT_CHECK_PEER_S); END_TIMER(CRCP_TIMER_CKPT_CHECK_PEER_S);
DISPLAY_INDV_TIMER(CRCP_TIMER_CKPT_CHECK_PEER_S, peer_ref->proc_name.vpid, total_details_sent); DISPLAY_INDV_TIMER(CRCP_TIMER_CKPT_CHECK_PEER_S, peer_ref->proc_name.vpid, total_details_sent);
@ -5587,11 +5589,11 @@ static int do_send_msg_detail(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
/* /*
* Do the send... * Do the send...
*/ */
if ( 0 > ( ret = orca_oob_send_buffer(&peer_ref->proc_name, buffer, if ( 0 > ( ret = orte_rml.send_buffer(&peer_ref->proc_name, buffer,
OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) { OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: do_send_msg_detail: Unable to send message details to peer %s: Return %d\n", "crcp:bkmrk: do_send_msg_detail: Unable to send message details to peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_ref->proc_name), ORTE_NAME_PRINT(&peer_ref->proc_name),
ret); ret);
exit_status = OMPI_ERROR; exit_status = OMPI_ERROR;
@ -5614,12 +5616,12 @@ static int do_send_msg_detail(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
/* /*
* Recv the ACK msg * Recv the ACK msg
*/ */
if ( 0 > (ret = orca_oob_recv_buffer(&peer_ref->proc_name, buffer, if ( 0 > (ret = orte_rml.recv_buffer(&peer_ref->proc_name, buffer,
OMPI_CRCP_COORD_BOOKMARK_TAG, 0) ) ) { OMPI_CRCP_COORD_BOOKMARK_TAG, 0) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: do_send_msg_detail: %s --> %s Failed to receive ACK buffer from peer. Return %d\n", "crcp:bkmrk: do_send_msg_detail: %s --> %s Failed to receive ACK buffer from peer. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5705,8 +5707,8 @@ static int recv_msg_details(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: recv_msg_details: %s <-- %s " "crcp:bkmrk: recv_msg_details: %s <-- %s "
"Failed to receive message detail from peer. Return %d\n", "Failed to receive message detail from peer. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5726,8 +5728,8 @@ static int recv_msg_details(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: recv_msg_details: %s <-- %s " "crcp:bkmrk: recv_msg_details: %s <-- %s "
"Failed to check message detail from peer. Return %d\n", "Failed to check message detail from peer. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5738,8 +5740,8 @@ static int recv_msg_details(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s Recv Detail: Stage --: [%3d / %3d] [%3d, %3d, %s]", "crcp:bkmrk: %s <-- %s Recv Detail: Stage --: [%3d / %3d] [%3d, %3d, %s]",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
need, found, need, found,
num_resolved, total_details_recv, num_resolved, total_details_recv,
( need <= found ? "T" : "F") )); ( need <= found ? "T" : "F") ));
@ -5756,8 +5758,8 @@ static int recv_msg_details(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
if(OMPI_SUCCESS != (ret = do_recv_msg_detail_resp(peer_ref, response, num_resolved, found))) { if(OMPI_SUCCESS != (ret = do_recv_msg_detail_resp(peer_ref, response, num_resolved, found))) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: recv_msg_details: %s <-- %s Failed to respond to peer. Return %d\n", "crcp:bkmrk: recv_msg_details: %s <-- %s Failed to respond to peer. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5788,11 +5790,11 @@ static int do_recv_msg_detail(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
/* /*
* Recv the msg * Recv the msg
*/ */
if ( 0 > (ret = orca_oob_recv_buffer(&peer_ref->proc_name, buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0) ) ) { if ( 0 > (ret = orte_rml.recv_buffer(&peer_ref->proc_name, buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0) ) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: do_recv_msg_detail: %s <-- %s Failed to receive buffer from peer. Return %d\n", "crcp:bkmrk: do_recv_msg_detail: %s <-- %s Failed to receive buffer from peer. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5851,8 +5853,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Stage 0: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d", "Stage 0: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
peer_ref->total_msgs_recvd, peer_ref->total_msgs_recvd,
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
p_num_sent, p_num_sent,
@ -5874,8 +5876,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: recv_msg_detail_check: %s -- %s " "crcp:bkmrk: recv_msg_detail_check: %s -- %s "
"Failed to determine if we have received this message. Return %d\n", "Failed to determine if we have received this message. Return %d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
ret); ret);
exit_status = ret; exit_status = ret;
goto cleanup; goto cleanup;
@ -5924,8 +5926,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Stage 1: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d", "Stage 1: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
peer_ref->total_msgs_recvd, peer_ref->total_msgs_recvd,
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
p_num_sent, p_num_sent,
@ -6009,8 +6011,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Stage 2: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d, active %4d", "Stage 2: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d, active %4d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
peer_ref->total_msgs_recvd, peer_ref->total_msgs_recvd,
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
p_num_sent, p_num_sent,
@ -6058,8 +6060,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Recv Check: Found a message that is 'active'! Prepare to STALL.\n", "Recv Check: Found a message that is 'active'! Prepare to STALL.\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)) )); ORTE_NAME_PRINT(&(peer_ref->proc_name)) ));
stall_for_completion = true; stall_for_completion = true;
} }
else { else {
@ -6067,8 +6069,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Recv Check: Found a message that is 'active', but is not the current recv! " "Recv Check: Found a message that is 'active', but is not the current recv! "
"No stall required [%3d, %3d, %3d, %3d].\n", "No stall required [%3d, %3d, %3d, %3d].\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
(int)current_msg_id, (int)current_msg_id,
(int)current_msg_type, (int)current_msg_type,
(int)posted_recv_msg_ref->msg_id, (int)posted_recv_msg_ref->msg_id,
@ -6126,8 +6128,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Stage 3: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d, active %4d", "Stage 3: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d, active %4d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
peer_ref->total_msgs_recvd, peer_ref->total_msgs_recvd,
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
p_num_sent, p_num_sent,
@ -6163,8 +6165,8 @@ static int do_recv_msg_detail_check_drain(ompi_crcp_bkmrk_pml_peer_ref_t *peer_r
OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle, OPAL_OUTPUT_VERBOSE((10, mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: %s <-- %s " "crcp:bkmrk: %s <-- %s "
"Stage 4: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d, active %4d", "Stage 4: Ck.Drain: [TR %3d/MS %3d] sent %4d, unres %4d, res %4d, active %4d",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
peer_ref->total_msgs_recvd, peer_ref->total_msgs_recvd,
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
p_num_sent, p_num_sent,
@ -6197,10 +6199,10 @@ static int do_recv_msg_detail_resp(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref,
PACK_BUFFER(buffer, total_found, 1, OPAL_UINT32, PACK_BUFFER(buffer, total_found, 1, OPAL_UINT32,
"crcp:bkmrk: recv_msg_details: Unable to ask peer for more messages"); "crcp:bkmrk: recv_msg_details: Unable to ask peer for more messages");
if ( 0 > ( ret = orca_oob_send_buffer(&peer_ref->proc_name, buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) { if ( 0 > ( ret = orte_rml.send_buffer(&peer_ref->proc_name, buffer, OMPI_CRCP_COORD_BOOKMARK_TAG, 0)) ) {
opal_output(mca_crcp_bkmrk_component.super.output_handle, opal_output(mca_crcp_bkmrk_component.super.output_handle,
"crcp:bkmrk: recv_msg_detail_resp: Unable to send message detail response to peer %s: Return %d\n", "crcp:bkmrk: recv_msg_detail_resp: Unable to send message detail response to peer %s: Return %d\n",
ORCA_NAME_PRINT(&peer_ref->proc_name), ORTE_NAME_PRINT(&peer_ref->proc_name),
ret); ret);
exit_status = OMPI_ERROR; exit_status = OMPI_ERROR;
goto cleanup; goto cleanup;
@ -6259,12 +6261,12 @@ static void display_all_timers(int state) {
double barrier_start, barrier_stop; double barrier_start, barrier_stop;
int i; int i;
if( 0 != orca_process_info_get_vpid(ORCA_PROC_MY_NAME) ) { if( 0 != ORTE_PROC_MY_NAME->vpid ) {
if( 2 > timing_enabled ) { if( 2 > timing_enabled ) {
return; return;
} }
else if( 2 == timing_enabled ) { else if( 2 == timing_enabled ) {
orca_coll_barrier(ORCA_COLL_TYPE_BARRIER_CR); orte_grpcomm.barrier();
return; return;
} }
} }
@ -6285,7 +6287,7 @@ static void display_all_timers(int state) {
if( timing_enabled >= 2) { if( timing_enabled >= 2) {
barrier_start = get_time(); barrier_start = get_time();
orca_coll_barrier(ORCA_COLL_TYPE_BARRIER_CR); orte_grpcomm.barrier();
barrier_stop = get_time(); barrier_stop = get_time();
opal_output(0, opal_output(0,
"crcp:bkmrk: timing(%20s): %20s = %10.2f s\n", "crcp:bkmrk: timing(%20s): %20s = %10.2f s\n",
@ -6305,7 +6307,7 @@ static void display_indv_timer_core(int idx, int proc, int msgs, bool direct) {
double diff = timer_end[idx] - timer_start[idx]; double diff = timer_end[idx] - timer_start[idx];
char * str = NULL; char * str = NULL;
if( 0 != orca_process_info_get_vpid(ORCA_PROC_MY_NAME) && timing_enabled < 3 ) { if( 0 != ORTE_PROC_MY_NAME->vpid && timing_enabled < 3 ) {
return; return;
} }
@ -6507,16 +6509,16 @@ static void traffic_message_dump_msg_list(opal_list_t *msg_list, bool is_drain)
static void traffic_message_dump_peer(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref, char * msg, bool root_only) static void traffic_message_dump_peer(ompi_crcp_bkmrk_pml_peer_ref_t *peer_ref, char * msg, bool root_only)
{ {
if( root_only && orca_process_info_get_vpid(ORCA_PROC_MY_NAME) != 0 ) { if( root_only && orte_process_info.my_name.vpid != 0 ) {
return; return;
} else { } else {
sleep(orca_process_info_get_vpid(ORCA_PROC_MY_NAME) * 2); sleep(orte_process_info.my_name.vpid * 2);
} }
opal_output(0, "------------- %s ---------------------------------", msg); opal_output(0, "------------- %s ---------------------------------", msg);
opal_output(0, "%s <-> %s Totals Sent [ %3d / %3d ] Recv [ %3d / %3d ]", opal_output(0, "%s <-> %s Totals Sent [ %3d / %3d ] Recv [ %3d / %3d ]",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORCA_NAME_PRINT(&(peer_ref->proc_name)), ORTE_NAME_PRINT(&(peer_ref->proc_name)),
peer_ref->total_msgs_sent, peer_ref->total_msgs_sent,
peer_ref->matched_msgs_sent, peer_ref->matched_msgs_sent,
peer_ref->total_msgs_recvd, peer_ref->total_msgs_recvd,

10
ompi/mca/crcp/bkmrk/crcp_bkmrk_pml.h
Просмотреть файл

@ -7,7 +7,7 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2011-2012 Oak Ridge National Labs. All rights reserved. * Copyright (c) 2011 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -234,7 +234,7 @@ BEGIN_C_DECLS
opal_list_t msg_contents; opal_list_t msg_contents;
/** Peer which we received from */ /** Peer which we received from */
orca_process_name_t proc_name; orte_process_name_t proc_name;
/** /**
* Count of the number of completed PML messages that match this reference. * Count of the number of completed PML messages that match this reference.
@ -271,7 +271,7 @@ BEGIN_C_DECLS
bool complete; bool complete;
/** Peer which we received from */ /** Peer which we received from */
orca_process_name_t peer; orte_process_name_t peer;
}; };
typedef struct ompi_crcp_bkmrk_pml_drain_message_ack_ref_t ompi_crcp_bkmrk_pml_drain_message_ack_ref_t; typedef struct ompi_crcp_bkmrk_pml_drain_message_ack_ref_t ompi_crcp_bkmrk_pml_drain_message_ack_ref_t;
@ -312,7 +312,7 @@ BEGIN_C_DECLS
opal_list_t msg_contents; opal_list_t msg_contents;
/** Peer which we received from */ /** Peer which we received from */
orca_process_name_t proc_name; orte_process_name_t proc_name;
/* Sample movement of values (mirrored for send): /* Sample movement of values (mirrored for send):
* Recv() iRecv() irecv_init() start() req_complete() * Recv() iRecv() irecv_init() start() req_complete()
@ -379,7 +379,7 @@ BEGIN_C_DECLS
opal_list_item_t super; opal_list_item_t super;
/** Name of peer */ /** Name of peer */
orca_process_name_t proc_name; orte_process_name_t proc_name;
/** List of messages sent to this peer */ /** List of messages sent to this peer */
opal_list_t send_list; /**< pml_send */ opal_list_t send_list; /**< pml_send */

2
ompi/mca/crcp/ompi_crcp.7in
Просмотреть файл

@ -75,7 +75,7 @@ error.
.SS none CRCP Component .SS none CRCP Component
.PP .PP
The \fInone\fP component simply selects no CRCP component. All of the CRCP The \fInone\fP component simply selects no CRCP component. All of the CRCP
function calls return immediately with OMPI_SUCCESS. function calls return immediately with ORTE_SUCCESS.
. .
.PP .PP
This component is the last component to be selected by default. This means that if This component is the last component to be selected by default. This means that if

3
ompi/mca/dpm/base/Makefile.am
Просмотреть файл

@ -9,7 +9,6 @@
# University of Stuttgart. All rights reserved. # University of Stuttgart. All rights reserved.
# Copyright (c) 2004-2005 The Regents of the University of California. # Copyright (c) 2004-2005 The Regents of the University of California.
# All rights reserved. # All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -28,7 +27,7 @@ libmca_dpm_la_SOURCES += \
base/dpm_base_select.c \ base/dpm_base_select.c \
base/dpm_base_null_fns.c base/dpm_base_null_fns.c
if ORCA_WITH_FULL_ORTE_SUPPORT if !ORTE_DISABLE_FULL_SUPPORT
libmca_dpm_la_SOURCES += \ libmca_dpm_la_SOURCES += \
base/dpm_base_common_fns.c base/dpm_base_common_fns.c
endif endif

9
ompi/mca/dpm/base/base.h
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -68,7 +67,7 @@ OMPI_DECLSPEC int ompi_dpm_base_select(void);
*/ */
OMPI_DECLSPEC int ompi_dpm_base_close(void); OMPI_DECLSPEC int ompi_dpm_base_close(void);
#if ORCA_WITH_FULL_ORTE_SUPPORT #if !ORTE_DISABLE_FULL_SUPPORT
/* Internal support functions */ /* Internal support functions */
OMPI_DECLSPEC char* ompi_dpm_base_dyn_init (void); OMPI_DECLSPEC char* ompi_dpm_base_dyn_init (void);
@ -92,10 +91,10 @@ int ompi_dpm_base_null_spawn(int count, char **array_of_commands,
int ompi_dpm_base_null_dyn_init(void); int ompi_dpm_base_null_dyn_init(void);
int ompi_dpm_base_null_dyn_finalize (void); int ompi_dpm_base_null_dyn_finalize (void);
void ompi_dpm_base_null_mark_dyncomm (ompi_communicator_t *comm); void ompi_dpm_base_null_mark_dyncomm (ompi_communicator_t *comm);
int ompi_dpm_base_null_open_port(char *port_name, orca_oob_tag_t given_tag); int ompi_dpm_base_null_open_port(char *port_name, orte_rml_tag_t given_tag);
int ompi_dpm_base_null_parse_port(char *port_name, int ompi_dpm_base_null_parse_port(char *port_name,
char **hnp_uri, char **rml_uri, orca_oob_tag_t *tag); char **hnp_uri, char **rml_uri, orte_rml_tag_t *tag);
int ompi_dpm_base_null_route_to_port(char *rml_uri, orca_process_name_t *rproc); int ompi_dpm_base_null_route_to_port(char *rml_uri, orte_process_name_t *rproc);
int ompi_dpm_base_null_close_port(char *port_name); int ompi_dpm_base_null_close_port(char *port_name);
/* useful globals */ /* useful globals */

4
ompi/mca/dpm/base/dpm_base_common_fns.c
Просмотреть файл

@ -14,7 +14,7 @@
* Copyright (c) 2006-2007 Los Alamos National Security, LLC. All rights * Copyright (c) 2006-2007 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -237,7 +237,7 @@ void ompi_dpm_base_mark_dyncomm (ompi_communicator_t *comm)
int i; int i;
int size, rsize; int size, rsize;
bool found=false; bool found=false;
orca_jobid_t thisjobid; orte_jobid_t thisjobid;
ompi_group_t *grp=NULL; ompi_group_t *grp=NULL;
ompi_proc_t *proc = NULL; ompi_proc_t *proc = NULL;

8
ompi/mca/dpm/base/dpm_base_null_fns.c
Просмотреть файл

@ -14,7 +14,7 @@
* Copyright (c) 2006-2007 Los Alamos National Security, LLC. All rights * Copyright (c) 2006-2007 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -66,18 +66,18 @@ void ompi_dpm_base_null_mark_dyncomm (ompi_communicator_t *comm)
return; return;
} }
int ompi_dpm_base_null_open_port(char *port_name, orca_oob_tag_t given_tag) int ompi_dpm_base_null_open_port(char *port_name, orte_rml_tag_t given_tag)
{ {
return OMPI_ERR_NOT_SUPPORTED; return OMPI_ERR_NOT_SUPPORTED;
} }
int ompi_dpm_base_null_parse_port(char *port_name, int ompi_dpm_base_null_parse_port(char *port_name,
char **hnp_uri, char **rml_uri, orca_oob_tag_t *tag) char **hnp_uri, char **rml_uri, orte_rml_tag_t *tag)
{ {
return OMPI_ERR_NOT_SUPPORTED; return OMPI_ERR_NOT_SUPPORTED;
} }
int ompi_dpm_base_null_route_to_port(char *rml_uri, orca_process_name_t *rproc) int ompi_dpm_base_null_route_to_port(char *rml_uri, orte_process_name_t *rproc)
{ {
return OMPI_ERR_NOT_SUPPORTED; return OMPI_ERR_NOT_SUPPORTED;
} }

13
ompi/mca/dpm/dpm.h
Просмотреть файл

@ -9,7 +9,6 @@
* University of Stuttgart. All rights reserved. * University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -31,7 +30,7 @@
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "orca/include/rte_orca.h" #include "orte/mca/rml/rml_types.h"
#include "ompi/info/info.h" #include "ompi/info/info.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
@ -39,9 +38,9 @@ BEGIN_C_DECLS
/* OMPI port definitions */ /* OMPI port definitions */
/* carry over the INVALID def */ /* carry over the INVALID def */
#define OMPI_RML_TAG_INVALID ORCA_OOB_TAG_INVALID #define OMPI_RML_TAG_INVALID ORTE_RML_TAG_INVALID
/* define a starting point to avoid conflicts */ /* define a starting point to avoid conflicts */
#define OMPI_RML_TAG_BASE ORCA_OOB_TAG_MAX #define OMPI_RML_TAG_BASE ORTE_RML_TAG_MAX
#define OMPI_RML_TAG_UDAPL OMPI_RML_TAG_BASE+1 #define OMPI_RML_TAG_UDAPL OMPI_RML_TAG_BASE+1
#define OMPI_RML_TAG_OPENIB OMPI_RML_TAG_BASE+2 #define OMPI_RML_TAG_OPENIB OMPI_RML_TAG_BASE+2
@ -117,20 +116,20 @@ typedef void (*ompi_dpm_base_module_mark_dyncomm_fn_t)(ompi_communicator_t *comm
* specified tag is valid, then it will be used to form the port. Otherwise, * specified tag is valid, then it will be used to form the port. Otherwise,
* a dynamically assigned tag that is unique to this request will be provided * a dynamically assigned tag that is unique to this request will be provided
*/ */
typedef int (*ompi_dpm_base_module_open_port_fn_t)(char *port_name, orca_oob_tag_t tag); typedef int (*ompi_dpm_base_module_open_port_fn_t)(char *port_name, orte_rml_tag_t tag);
/* /*
* Converts an opaque port string to a RML process nane and tag. * Converts an opaque port string to a RML process nane and tag.
*/ */
typedef int (*ompi_dpm_base_module_parse_port_name_t)(char *port_name, typedef int (*ompi_dpm_base_module_parse_port_name_t)(char *port_name,
char **hnp_uri, char **rml_uri, char **hnp_uri, char **rml_uri,
orca_oob_tag_t *tag); orte_rml_tag_t *tag);
/* /*
* Update the routed component to make sure that the RML can send messages to * Update the routed component to make sure that the RML can send messages to
* the remote port * the remote port
*/ */
typedef int (*ompi_dpm_base_module_route_to_port_t)(char *rml_uri, orca_process_name_t *rproc); typedef int (*ompi_dpm_base_module_route_to_port_t)(char *rml_uri, orte_process_name_t *rproc);
/* /*

4
ompi/mca/dpm/orte/Makefile.am
Просмотреть файл

@ -4,7 +4,6 @@
# Copyright (c) 2009 High Performance Computing Center Stuttgart, # Copyright (c) 2009 High Performance Computing Center Stuttgart,
# University of Stuttgart. All rights reserved. # University of Stuttgart. All rights reserved.
# Copyright (c) 2010 Cisco Systems, Inc. All rights reserved. # Copyright (c) 2010 Cisco Systems, Inc. All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -15,9 +14,6 @@
EXTRA_DIST = .windows EXTRA_DIST = .windows
CFLAGS = $(dpm_orte_CFLAGS)
AM_CPPFLAGS = $(dpm_orte_CPPFLAGS)
dist_pkgdata_DATA = help-ompi-dpm-orte.txt dist_pkgdata_DATA = help-ompi-dpm-orte.txt
# Make the output library in this directory, and name it either # Make the output library in this directory, and name it either

14
ompi/mca/dpm/orte/configure.m4
Просмотреть файл

@ -2,7 +2,7 @@
# #
# Copyright (c) 2011 Los Alamos National Security, LLC. # Copyright (c) 2011 Los Alamos National Security, LLC.
# All rights reserved. # All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. #
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -15,15 +15,7 @@
AC_DEFUN([MCA_ompi_dpm_orte_CONFIG],[ AC_DEFUN([MCA_ompi_dpm_orte_CONFIG],[
AC_CONFIG_FILES([ompi/mca/dpm/orte/Makefile]) AC_CONFIG_FILES([ompi/mca/dpm/orte/Makefile])
ORCA_CHECK_ORTE([dpm_orte], AS_IF([test "$orte_without_full_support" = 0],
[dpm_orte_happy="yes"], [$1],
[dpm_orte_happy="no"])
AS_IF([test "$dpm_orte_happy" = "yes" -a "$orca_without_orte_full_support" = 0],
[AC_SUBST([dpm_orte_CFLAGS])
AC_SUBST([dpm_orte_CPPFLAGS])
AC_SUBST([dpm_orte_LDFLAGS])
AC_SUBST([dpm_orte_LIBS])
$1],
[$2]) [$2])
])dnl ])dnl

5
ompi/mca/io/romio/src/Makefile.extra
Просмотреть файл

@ -10,7 +10,6 @@
# University of Stuttgart. All rights reserved. # University of Stuttgart. All rights reserved.
# Copyright (c) 2004-2005 The Regents of the University of California. # Copyright (c) 2004-2005 The Regents of the University of California.
# All rights reserved. # All rights reserved.
# Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
# $COPYRIGHT$ # $COPYRIGHT$
# #
# Additional copyrights may follow # Additional copyrights may follow
@ -28,10 +27,6 @@ AM_CPPFLAGS = \
-I$(top_ompi_builddir)/orte/include \ -I$(top_ompi_builddir)/orte/include \
-I$(top_ompi_srcdir)/orte \ -I$(top_ompi_srcdir)/orte \
-I$(top_ompi_srcdir)/orte/include \ -I$(top_ompi_srcdir)/orte/include \
-I$(top_ompi_builddir)/orca \
-I$(top_ompi_builddir)/orca/include \
-I$(top_ompi_srcdir)/orca \
-I$(top_ompi_srcdir)/orca/include \
-I$(top_ompi_builddir)/ompi \ -I$(top_ompi_builddir)/ompi \
-I$(top_ompi_builddir)/ompi/include \ -I$(top_ompi_builddir)/ompi/include \
-I$(top_ompi_srcdir)/ompi \ -I$(top_ompi_srcdir)/ompi \

14
ompi/mca/mpool/base/mpool_base_lookup.c
Просмотреть файл

@ -12,7 +12,6 @@
* All rights reserved. * All rights reserved.
* Copyright (c) 2006-2007 Mellanox Technologies. All rights reserved. * Copyright (c) 2006-2007 Mellanox Technologies. All rights reserved.
* Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -31,9 +30,10 @@
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "orte/util/show_help.h"
#include "orca/include/rte_orca.h" #include "orte/util/name_fns.h"
#include "orte/util/proc_info.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/mca/mpool/mpool.h" #include "ompi/mca/mpool/mpool.h"
#include "ompi/runtime/params.h" #include "ompi/runtime/params.h"
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
@ -115,9 +115,9 @@ mca_mpool_base_module_t* mca_mpool_base_module_create(
opal_mem_hooks_support_level())) { opal_mem_hooks_support_level())) {
opal_mem_hooks_register_release(mca_mpool_base_mem_cb, NULL); opal_mem_hooks_register_release(mca_mpool_base_mem_cb, NULL);
} else { } else {
orca_show_help("help-mpool-base.txt", "leave pinned failed", orte_show_help("help-mpool-base.txt", "leave pinned failed",
true, name, ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), true, name, ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orca_process_info_get_nodename() ); orte_process_info.nodename);
return NULL; return NULL;
} }

20
ompi/mca/mpool/base/mpool_base_mem_cb.c
Просмотреть файл

@ -11,7 +11,6 @@
* Copyright (c) 2004-2005 The Regents of the University of California. * Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved. * All rights reserved.
* Copyright (c) 2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -28,7 +27,8 @@
#endif #endif
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/proc_info.h"
#include "ompi/mca/mpool/base/mpool_base_mem_cb.h" #include "ompi/mca/mpool/base/mpool_base_mem_cb.h"
#include "ompi/mca/mpool/base/base.h" #include "ompi/mca/mpool/base/base.h"
@ -72,17 +72,17 @@ void mca_mpool_base_mem_cb(void* base, size_t size, void* cbdata,
if (from_alloc) { if (from_alloc) {
int len; int len;
len = snprintf(msg, sizeof(msg), "[%s:%d] Attempt to free memory that is still in use by an ongoing MPI communication (buffer %p, size %lu). MPI job will now abort.\n", len = snprintf(msg, sizeof(msg), "[%s:%d] Attempt to free memory that is still in use by an ongoing MPI communication (buffer %p, size %lu). MPI job will now abort.\n",
orca_process_info_get_nodename(), orte_process_info.nodename,
getpid(), getpid(),
base, (unsigned long) size); base, (unsigned long) size);
msg[sizeof(msg) - 1] = '\0'; msg[sizeof(msg) - 1] = '\0';
write(2, msg, len); write(2, msg, len);
} else { } else {
orca_show_help("help-mpool-base.txt", orte_show_help("help-mpool-base.txt",
"cannot deregister in-use memory", true, "cannot deregister in-use memory", true,
current->mpool_component->mpool_version.mca_component_name, current->mpool_component->mpool_version.mca_component_name,
orca_process_info_get_nodename(), orte_process_info.nodename,
base, (unsigned long) size); base, (unsigned long) size);
} }
/* We're in a callback from somewhere; we can't do /* We're in a callback from somewhere; we can't do

26
ompi/mca/mpool/base/mpool_base_tree.c
Просмотреть файл

@ -12,7 +12,6 @@
* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2007 Voltaire. All rights reserved. * Copyright (c) 2007 Voltaire. All rights reserved.
* Copyright (c) 2010 IBM Corporation. All rights reserved. * Copyright (c) 2010 IBM Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -27,7 +26,10 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "opal/mca/mca.h" #include "opal/mca/mca.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/name_fns.h"
#include "orte/util/proc_info.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/runtime/params.h" #include "ompi/runtime/params.h"
#include "ompi/class/ompi_free_list.h" #include "ompi/class/ompi_free_list.h"
#include "ompi/class/ompi_rb_tree.h" #include "ompi/class/ompi_rb_tree.h"
@ -176,18 +178,18 @@ void mca_mpool_base_tree_print(void)
if (num_leaks <= ompi_debug_show_mpi_alloc_mem_leaks || if (num_leaks <= ompi_debug_show_mpi_alloc_mem_leaks ||
ompi_debug_show_mpi_alloc_mem_leaks < 0) { ompi_debug_show_mpi_alloc_mem_leaks < 0) {
orca_show_help("help-mpool-base.txt", "all mem leaks", orte_show_help("help-mpool-base.txt", "all mem leaks",
true, ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), true, ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orca_process_info_get_nodename(), orte_process_info.nodename,
orca_process_info_get_pid(), leak_msg); orte_process_info.pid, leak_msg);
} else { } else {
int i = num_leaks - ompi_debug_show_mpi_alloc_mem_leaks; int i = num_leaks - ompi_debug_show_mpi_alloc_mem_leaks;
orca_show_help("help-mpool-base.txt", "some mem leaks", orte_show_help("help-mpool-base.txt", "some mem leaks",
true, ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), true, ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orca_process_info_get_nodename(), orte_process_info.nodename,
orca_process_info_get_pid(), i, orte_process_info.pid, leak_msg, i,
(i > 1) ? "s were" : " was", (i > 1) ? "s were" : " was",
(i > 1) ? "are" : "is"); (i > 1) ? "are" : "is");
} }
free(leak_msg); free(leak_msg);
leak_msg = NULL; leak_msg = NULL;

4
ompi/mca/mpool/grdma/mpool_grdma_component.c
Просмотреть файл

@ -14,7 +14,7 @@
* Copyright (c) 2007-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2007-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -26,7 +26,7 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "ompi/runtime/params.h" #include "ompi/runtime/params.h"
#include "mpool_grdma.h" #include "mpool_grdma.h"
#ifdef HAVE_UNISTD_H #ifdef HAVE_UNISTD_H

7
ompi/mca/mpool/grdma/mpool_grdma_module.c
Просмотреть файл

@ -16,7 +16,7 @@
* Copyright (c) 2010 IBM Corporation. All rights reserved. * Copyright (c) 2010 IBM Corporation. All rights reserved.
* Copyright (c) 2011-2012 Los Alamos National Security, LLC. All rights * Copyright (c) 2011-2012 Los Alamos National Security, LLC. All rights
* reserved. * reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. *
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -27,7 +27,8 @@
#define OPAL_DISABLE_ENABLE_MEM_DEBUG 1 #define OPAL_DISABLE_ENABLE_MEM_DEBUG 1
#include "ompi_config.h" #include "ompi_config.h"
#include "opal/align.h" #include "opal/align.h"
#include "orca/include/rte_orca.h" #include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
#include "mpool_grdma.h" #include "mpool_grdma.h"
#include <errno.h> #include <errno.h>
#include <string.h> #include <string.h>
@ -451,7 +452,7 @@ void mca_mpool_grdma_finalize(struct mca_mpool_base_module_t *mpool)
if (true == mca_mpool_grdma_component.print_stats) { if (true == mca_mpool_grdma_component.print_stats) {
opal_output(0, "%s grdma: stats " opal_output(0, "%s grdma: stats "
"(hit/miss/found/not found/evicted): %d/%d/%d/%d/%d\n", "(hit/miss/found/not found/evicted): %d/%d/%d/%d/%d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
mpool_grdma->stat_cache_hit, mpool_grdma->stat_cache_miss, mpool_grdma->stat_cache_hit, mpool_grdma->stat_cache_miss,
mpool_grdma->stat_cache_found, mpool_grdma->stat_cache_notfound, mpool_grdma->stat_cache_found, mpool_grdma->stat_cache_notfound,
mpool_grdma->stat_evicted); mpool_grdma->stat_evicted);

6
ompi/mca/mpool/rgpusm/mpool_rgpusm_module.c
Просмотреть файл

@ -14,7 +14,6 @@
* Copyright (c) 2007 Mellanox Technologies. All rights reserved. * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
* Copyright (c) 2010 IBM Corporation. All rights reserved. * Copyright (c) 2010 IBM Corporation. All rights reserved.
* Copyright (c) 2012 NVIDIA Corporation. All rights reserved. * Copyright (c) 2012 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* *
* $COPYRIGHT$ * $COPYRIGHT$
* *
@ -78,7 +77,8 @@
#define OPAL_DISABLE_ENABLE_MEM_DEBUG 1 #define OPAL_DISABLE_ENABLE_MEM_DEBUG 1
#include "ompi_config.h" #include "ompi_config.h"
#include "opal/align.h" #include "opal/align.h"
#include "orca/include/rte_orca.h" #include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
#include "ompi/mca/mpool/rgpusm/mpool_rgpusm.h" #include "ompi/mca/mpool/rgpusm/mpool_rgpusm.h"
#include <errno.h> #include <errno.h>
#include <string.h> #include <string.h>
@ -556,7 +556,7 @@ void mca_mpool_rgpusm_finalize(struct mca_mpool_base_module_t *mpool)
if(true == mca_mpool_rgpusm_component.print_stats) { if(true == mca_mpool_rgpusm_component.print_stats) {
opal_output(0, "%s rgpusm: stats " opal_output(0, "%s rgpusm: stats "
"(hit/valid/invalid/miss/evicted): %d/%d/%d/%d/%d\n", "(hit/valid/invalid/miss/evicted): %d/%d/%d/%d/%d\n",
ORCA_NAME_PRINT(ORCA_PROC_MY_NAME), ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
mpool_rgpusm->stat_cache_hit, mpool_rgpusm->stat_cache_valid, mpool_rgpusm->stat_cache_hit, mpool_rgpusm->stat_cache_valid,
mpool_rgpusm->stat_cache_invalid, mpool_rgpusm->stat_cache_miss, mpool_rgpusm->stat_cache_invalid, mpool_rgpusm->stat_cache_miss,
mpool_rgpusm->stat_evicted); mpool_rgpusm->stat_evicted);

7
ompi/mca/mpool/sm/mpool_sm_component.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010 Los Alamos National Security, LLC. * Copyright (c) 2010 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -32,7 +31,7 @@
#include "opal/mca/base/base.h" #include "opal/mca/base/base.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#include "ompi/mca/allocator/base/base.h" #include "ompi/mca/allocator/base/base.h"
#include "mpool_sm.h" #include "mpool_sm.h"
@ -200,8 +199,8 @@ static mca_mpool_base_module_t* mca_mpool_sm_init(
/* create initial shared memory mapping */ /* create initial shared memory mapping */
len = asprintf( &file_name, "%s"OPAL_PATH_SEP"shared_mem_pool.%s", len = asprintf( &file_name, "%s"OPAL_PATH_SEP"shared_mem_pool.%s",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
orca_process_info_get_nodename() ); orte_process_info.nodename );
if ( 0 > len ) { if ( 0 > len ) {
free(mpool_module); free(mpool_module);
free(procs); free(procs);

9
ompi/mca/mpool/sm/mpool_sm_module.c
Просмотреть файл

@ -13,7 +13,6 @@
* Copyright (c) 2011-2012 Los Alamos National Security, LLC. * Copyright (c) 2011-2012 Los Alamos National Security, LLC.
* All rights reserved. * All rights reserved.
* Copyright (c) 2011 NVIDIA Corporation. All rights reserved. * Copyright (c) 2011 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -30,7 +29,7 @@
#include <unistd.h> #include <unistd.h>
#endif #endif
#include "opal/mca/hwloc/base/base.h" #include "opal/mca/hwloc/base/base.h"
#include "orca/include/rte_orca.h" #include "orte/util/proc_info.h"
#if OPAL_ENABLE_FT_CR == 1 #if OPAL_ENABLE_FT_CR == 1
#include "orte/mca/sstore/sstore.h" #include "orte/mca/sstore/sstore.h"
@ -176,14 +175,14 @@ int mca_mpool_sm_ft_event(int state) {
if(OPAL_CRS_CHECKPOINT == state) { if(OPAL_CRS_CHECKPOINT == state) {
/* Record the shared memory filename */ /* Record the shared memory filename */
asprintf( &file_name, "%s"OPAL_PATH_SEP"shared_mem_pool.%s", asprintf( &file_name, "%s"OPAL_PATH_SEP"shared_mem_pool.%s",
orca_process_info_get_job_session_dir(), orte_process_info.job_session_dir,
orca_process_info_get_nodename() ); orte_process_info.nodename );
orte_sstore.set_attr(orte_sstore_handle_current, SSTORE_METADATA_LOCAL_TOUCH, file_name); orte_sstore.set_attr(orte_sstore_handle_current, SSTORE_METADATA_LOCAL_TOUCH, file_name);
free(file_name); free(file_name);
file_name = NULL; file_name = NULL;
} }
else if(OPAL_CRS_CONTINUE == state) { else if(OPAL_CRS_CONTINUE == state) {
if(orca_info_cr_continue_like_restart()) { if(orte_cr_continue_like_restart) {
/* Find the sm module */ /* Find the sm module */
self_module = mca_mpool_base_module_lookup("sm"); self_module = mca_mpool_base_module_lookup("sm");
self_sm_module = (mca_mpool_sm_module_t*) self_module; self_sm_module = (mca_mpool_sm_module_t*) self_module;

26
ompi/mca/mtl/mxm/mtl_mxm.c
Просмотреть файл

@ -1,6 +1,5 @@
/* /*
* Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED. * Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -11,15 +10,14 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "orte/util/proc_info.h"
#include "ompi/mca/mtl/mtl.h" #include "ompi/mca/mtl/mtl.h"
#include "ompi/runtime/ompi_module_exchange.h" #include "ompi/runtime/ompi_module_exchange.h"
#include "ompi/mca/mtl/base/mtl_base_datatype.h" #include "ompi/mca/mtl/base/mtl_base_datatype.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
#include "orte/mca/ess/ess.h"
#include "orca/include/rte_orca.h" #include "orte/runtime/orte_globals.h"
#include "ompi/communicator/communicator.h" #include "ompi/communicator/communicator.h"
#include "mtl_mxm.h" #include "mtl_mxm.h"
@ -64,9 +62,9 @@ static uint32_t ompi_mtl_mxm_get_job_id(void)
memset(uu, 0, sizeof(unique_job_key)); memset(uu, 0, sizeof(unique_job_key));
if (!generated_key || (strlen(generated_key) != 33) || sscanf(generated_key, "%016llx-%016llx", &uu[0], &uu[1]) != 2) { if (!generated_key || (strlen(generated_key) != 33) || sscanf(generated_key, "%016llx-%016llx", &uu[0], &uu[1]) != 2) {
orca_show_help("help-mtl-mxm.txt", "no uuid present", true, orte_show_help("help-mtl-mxm.txt", "no uuid present", true,
generated_key ? "could not be parsed from" : generated_key ? "could not be parsed from" :
"not present in", orca_process_info_get_nodename()); "not present in", orte_process_info.nodename);
return 0; return 0;
} }
@ -94,7 +92,7 @@ static int ompi_mtl_mxm_get_ep_address(ompi_mtl_mxm_ep_conn_info_t *ep_info, mxm
err = mxm_ep_address(ompi_mtl_mxm.ep, ptlid, err = mxm_ep_address(ompi_mtl_mxm.ep, ptlid,
(struct sockaddr *) &ep_info->ptl_addr[ptlid], &addrlen); (struct sockaddr *) &ep_info->ptl_addr[ptlid], &addrlen);
if (MXM_OK != err) { if (MXM_OK != err) {
orca_show_help("help-mtl-mxm.txt", "unable to extract endpoint address", orte_show_help("help-mtl-mxm.txt", "unable to extract endpoint address",
true, (int)ptlid, mxm_error_string(err)); true, (int)ptlid, mxm_error_string(err));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -150,13 +148,13 @@ int ompi_mtl_mxm_module_init(void)
} }
MXM_VERBOSE(1, "MXM support enabled"); MXM_VERBOSE(1, "MXM support enabled");
if ((lr = orca_node_info_get_rank(ORCA_PROC_MY_NAME)) == ORCA_NODE_RANK_INVALID) { if ((lr = orte_ess.get_node_rank(ORTE_PROC_MY_NAME)) == ORTE_NODE_RANK_INVALID) {
MXM_ERROR("Unable to obtain local node rank"); MXM_ERROR("Unable to obtain local node rank");
return OMPI_ERROR; return OMPI_ERROR;
} }
for (proc = 0; proc < totps; proc++) { for (proc = 0; proc < totps; proc++) {
if(OPAL_PROC_ON_LOCAL_NODE(orca_process_get_locality(&procs[proc]->proc_name))) { if(OPAL_PROC_ON_LOCAL_NODE(orte_ess.proc_get_locality(&procs[proc]->proc_name))) {
mxlr = max(mxlr, procs[proc]->proc_name.vpid); mxlr = max(mxlr, procs[proc]->proc_name.vpid);
nlps++; nlps++;
} }
@ -193,7 +191,7 @@ int ompi_mtl_mxm_module_init(void)
/* Open MXM endpoint */ /* Open MXM endpoint */
err = mxm_ep_create(ompi_mtl_mxm.mxm_context, &ep_opt, &ompi_mtl_mxm.ep); err = mxm_ep_create(ompi_mtl_mxm.mxm_context, &ep_opt, &ompi_mtl_mxm.ep);
if (MXM_OK != err) { if (MXM_OK != err) {
orca_show_help("help-mtl-mxm.txt", "unable to create endpoint", true, orte_show_help("help-mtl-mxm.txt", "unable to create endpoint", true,
mxm_error_string(err)); mxm_error_string(err));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -380,7 +378,7 @@ int ompi_mtl_mxm_add_comm(struct mca_mtl_base_module_t *mtl,
err = mxm_mq_create(ompi_mtl_mxm.mxm_context, comm->c_contextid, &mq); err = mxm_mq_create(ompi_mtl_mxm.mxm_context, comm->c_contextid, &mq);
if (MXM_OK != err) { if (MXM_OK != err) {
orca_show_help("help-mtl-mxm.txt", "mxm mq create", true, mxm_error_string(err)); orte_show_help("help-mtl-mxm.txt", "mxm mq create", true, mxm_error_string(err));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -404,7 +402,7 @@ int ompi_mtl_mxm_progress(void)
err = mxm_progress(ompi_mtl_mxm.mxm_context); err = mxm_progress(ompi_mtl_mxm.mxm_context);
if ((MXM_OK != err) && (MXM_ERR_NO_PROGRESS != err) ) { if ((MXM_OK != err) && (MXM_ERR_NO_PROGRESS != err) ) {
orca_show_help("help-mtl-mxm.txt", "errors during mxm_progress", true, mxm_error_string(err)); orte_show_help("help-mtl-mxm.txt", "errors during mxm_progress", true, mxm_error_string(err));
} }
return 1; return 1;
} }

5
ompi/mca/mtl/mxm/mtl_mxm_component.c
Просмотреть файл

@ -1,6 +1,5 @@
/* /*
* Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED. * Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -10,7 +9,7 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "opal/util/output.h" #include "opal/util/output.h"
#include "opal/mca/base/mca_base_param.h" #include "opal/mca/base/mca_base_param.h"
#include "ompi/proc/proc.h" #include "ompi/proc/proc.h"
@ -94,7 +93,7 @@ static int ompi_mtl_mxm_component_open(void)
if (MXM_ERR_NO_DEVICE == err) { if (MXM_ERR_NO_DEVICE == err) {
MXM_VERBOSE(1, "No supported device found, disqualifying mxm"); MXM_VERBOSE(1, "No supported device found, disqualifying mxm");
} else { } else {
orca_show_help("help-mtl-mxm.txt", "mxm init", true, orte_show_help("help-mtl-mxm.txt", "mxm init", true,
mxm_error_string(err)); mxm_error_string(err));
} }
return OPAL_ERR_NOT_AVAILABLE; return OPAL_ERR_NOT_AVAILABLE;

5
ompi/mca/mtl/mxm/mtl_mxm_recv.c
Просмотреть файл

@ -1,6 +1,5 @@
/* /*
* Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED. * Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -12,7 +11,7 @@
#include "opal/datatype/opal_convertor.h" #include "opal/datatype/opal_convertor.h"
#include "ompi/mca/mtl/base/mtl_base_datatype.h" #include "ompi/mca/mtl/base/mtl_base_datatype.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "mtl_mxm.h" #include "mtl_mxm.h"
#include "mtl_mxm_types.h" #include "mtl_mxm_types.h"
@ -77,7 +76,7 @@ int ompi_mtl_mxm_irecv(struct mca_mtl_base_module_t* mtl,
/* post-recv */ /* post-recv */
err = mxm_req_recv(mxm_recv_req); err = mxm_req_recv(mxm_recv_req);
if (MXM_OK != err) { if (MXM_OK != err) {
orca_show_help("help-mtl-mxm.txt", "error posting receive", true, orte_show_help("help-mtl-mxm.txt", "error posting receive", true,
mxm_error_string(err), mtl_mxm_request->buf, mtl_mxm_request->length); mxm_error_string(err), mtl_mxm_request->buf, mtl_mxm_request->length);
return OMPI_ERROR; return OMPI_ERROR;
} }

10
ompi/mca/mtl/mxm/mtl_mxm_send.c
Просмотреть файл

@ -1,6 +1,4 @@
/* /* * Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED.
* Copyright (C) Mellanox Technologies Ltd. 2001-2011. ALL RIGHTS RESERVED.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -11,7 +9,7 @@
#include "ompi_config.h" #include "ompi_config.h"
#include "ompi/mca/pml/pml.h" #include "ompi/mca/pml/pml.h"
#include "opal/datatype/opal_convertor.h" #include "opal/datatype/opal_convertor.h"
#include "orca/include/rte_orca.h" #include "orte/util/show_help.h"
#include "mtl_mxm.h" #include "mtl_mxm.h"
#include "mtl_mxm_types.h" #include "mtl_mxm_types.h"
@ -76,7 +74,7 @@ int ompi_mtl_mxm_send(struct mca_mtl_base_module_t* mtl,
/* post-send */ /* post-send */
err = mxm_req_send(&mxm_send_req); err = mxm_req_send(&mxm_send_req);
if (MXM_OK != err) { if (MXM_OK != err) {
orca_show_help("help-mtl-mxm.txt", "error posting send", true, 0, mxm_error_string(err)); orte_show_help("help-mtl-mxm.txt", "error posting send", true, 0, mxm_error_string(err));
return OMPI_ERROR; return OMPI_ERROR;
} }
@ -136,7 +134,7 @@ int ompi_mtl_mxm_isend(struct mca_mtl_base_module_t* mtl,
/* post-send */ /* post-send */
err = mxm_req_send(mxm_send_req); err = mxm_req_send(mxm_send_req);
if (MXM_OK != err) { if (MXM_OK != err) {
orca_show_help("help-mtl-mxm.txt", "error posting send", true, 1, mxm_error_string(err)); orte_show_help("help-mtl-mxm.txt", "error posting send", true, 1, mxm_error_string(err));
return OMPI_ERROR; return OMPI_ERROR;
} }

Некоторые файлы не были показаны из-за слишком большого количества измененных файлов Показать больше