1
1
openmpi/ompi/runtime/ompi_mpi_params.c
Ralph Castain 3e55fe6f6d Fold in the revised modex scheme. Move the ompi_proc_t modex portions to the RTE level since the daemons already have that info. Provide each process with the equivalent of a "nidmap" - both a map of what nodes are in the job, and a map of which node each process is on. This enables the use of static ports, though that hasn't been turned "on" in this commit.
Update the rsh tree spawn capability so we spawn the next wave of daemons before launching our own local procs.

Add an ability to encode nodenames for large clusters with contiguous node name numbering schemes - this allows communication of all node names in a few bytes instead of tens-of-bytes/node.

This commit was SVN r18338.
2008-04-30 19:49:53 +00:00

320 строки
14 KiB
C

/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2006-2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2007 Los Alamos National Security, LLC. All rights
* reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#if HAVE_TIME_H
#include <time.h>
#endif /* HAVE_TIME_H */
#include "ompi/constants.h"
#include "ompi/runtime/mpiruntime.h"
#include "ompi/runtime/params.h"
#include "ompi/datatype/datatype.h"
#include "opal/util/output.h"
#include "opal/util/show_help.h"
#include "opal/mca/base/mca_base_param.h"
/*
* Global variables
*
* As a deviation from the norm, ompi_mpi_param_check is also
* extern'ed in src/mpi/interface/c/bindings.h because it is already
* included in all MPI function imlementation files
*
* The values below are the default values.
*/
bool ompi_mpi_param_check = true;
bool ompi_debug_show_handle_leaks = false;
int ompi_debug_show_mpi_alloc_mem_leaks = 0;
bool ompi_debug_no_free_handles = false;
bool ompi_mpi_show_mca_params = false;
char *ompi_mpi_show_mca_params_file = NULL;
bool ompi_mpi_abort_print_stack = false;
int ompi_mpi_abort_delay = 0;
bool ompi_mpi_keep_peer_hostnames = true;
bool ompi_mpi_keep_fqdn_hostnames = false;
bool ompi_mpi_leave_pinned = false;
bool ompi_mpi_leave_pinned_pipeline = false;
bool ompi_have_sparse_group_storage = OPAL_INT_TO_BOOL(OMPI_GROUP_SPARSE);
bool ompi_use_sparse_group_storage = OPAL_INT_TO_BOOL(OMPI_GROUP_SPARSE);
int ompi_mpi_register_params(void)
{
int value;
/* Whether we want MPI API function parameter checking or not */
mca_base_param_reg_int_name("mpi", "param_check",
"Whether you want MPI API parameters checked at run-time or not. Possible values are 0 (no checking) and 1 (perform checking at run-time)",
false, false, MPI_PARAM_CHECK, &value);
ompi_mpi_param_check = OPAL_INT_TO_BOOL(value);
if (ompi_mpi_param_check) {
value = 0;
if (MPI_PARAM_CHECK) {
value = 1;
}
if (0 == value) {
opal_show_help("help-mpi-runtime.txt",
"mpi-param-check-enabled-but-compiled-out",
true);
ompi_mpi_param_check = false;
}
}
/*
* opal_progress: decide whether to yield and the event library
* tick rate
*/
/* JMS: Need ORTE data here -- set this to 0 when
exactly/under-subscribed, or 1 when oversubscribed */
mca_base_param_reg_int_name("mpi", "yield_when_idle",
"Yield the processor when waiting for MPI communication (for MPI processes, will default to 1 when oversubscribing nodes)",
false, false, -1, NULL);
mca_base_param_reg_int_name("mpi", "event_tick_rate",
"How often to progress TCP communications (0 = never, otherwise specified in microseconds)",
false, false, -1, NULL);
/* Whether or not to show MPI handle leaks */
mca_base_param_reg_int_name("mpi", "show_handle_leaks",
"Whether MPI_FINALIZE shows all MPI handles that were not freed or not",
false, false,
(int) ompi_debug_show_handle_leaks, &value);
ompi_debug_show_handle_leaks = OPAL_INT_TO_BOOL(value);
/* Whether or not to free MPI handles. Useless without run-time
param checking, so implicitly set that to true if we don't want
to free the handles. */
mca_base_param_reg_int_name("mpi", "no_free_handles",
"Whether to actually free MPI objects when their handles are freed",
false, false,
(int) ompi_debug_no_free_handles, &value);
ompi_debug_no_free_handles = OPAL_INT_TO_BOOL(value);
if (ompi_debug_no_free_handles) {
ompi_mpi_param_check = true;
value = 0;
if (MPI_PARAM_CHECK) {
value = 1;
}
if (0 == value) {
opal_output(0, "WARNING: MCA parameter mpi_no_free_handles set to true, but MPI");
opal_output(0, "WARNING: parameter checking has been compiled out of Open MPI.");
opal_output(0, "WARNING: mpi_no_free_handles is therefore only partially effective!");
}
}
/* Whether or not to show MPI_ALLOC_MEM leaks */
mca_base_param_reg_int_name("mpi", "show_mpi_alloc_mem_leaks",
"If >0, MPI_FINALIZE will show up to this many instances of memory allocated by MPI_ALLOC_MEM that was not freed by MPI_FREE_MEM",
false, false,
ompi_debug_show_mpi_alloc_mem_leaks,
&ompi_debug_show_mpi_alloc_mem_leaks);
/* Whether or not to print all MCA parameters in MPI_INIT */
mca_base_param_reg_int_name("mpi", "show_mca_params",
"Whether to show all MCA parameter value during MPI_INIT or not (good for reproducability of MPI jobs)",
false, false,
(int) ompi_mpi_show_mca_params, &value);
ompi_mpi_show_mca_params = OPAL_INT_TO_BOOL(value);
/* File to use when dumping the parameters */
mca_base_param_reg_string_name("mpi", "show_mca_params_file",
"If mpi_show_mca_params is true, setting this string to a valid filename tells Open MPI to dump all the MCA parameter values into a file suitable for reading via the mca_param_files parameter (good for reproducability of MPI jobs)",
false, false,
"", &ompi_mpi_show_mca_params_file);
/* User-level process pinning controls */
/* Do we want to save hostnames for debugging messages? This can
eat quite a bit of memory... */
mca_base_param_reg_int_name("mpi", "keep_peer_hostnames",
"If nonzero, save the string hostnames of all MPI peer processes (mostly for error / debugging output messages). This can add quite a bit of memory usage to each MPI process.",
false, false, 1, &value);
ompi_mpi_keep_peer_hostnames = OPAL_INT_TO_BOOL(value);
/* MPI_ABORT controls */
mca_base_param_reg_int_name("mpi", "abort_delay",
"If nonzero, print out an identifying message when MPI_ABORT is invoked (hostname, PID of the process that called MPI_ABORT) and delay for that many seconds before exiting (a negative delay value means to never abort). This allows attaching of a debugger before quitting the job.",
false, false,
ompi_mpi_abort_delay,
&ompi_mpi_abort_delay);
mca_base_param_reg_int_name("mpi", "abort_print_stack",
"If nonzero, print out a stack trace when MPI_ABORT is invoked",
false,
/* If we do not have stack trace
capability, make this a read-only
MCA param */
#if OMPI_WANT_PRETTY_PRINT_STACKTRACE && ! defined(__WINDOWS__) && defined(HAVE_BACKTRACE)
false,
#else
true,
#endif
(int) ompi_mpi_abort_print_stack,
&value);
#if OMPI_WANT_PRETTY_PRINT_STACKTRACE && ! defined(__WINDOWS__) && defined(HAVE_BACKTRACE)
/* Only take the value if we have stack trace capability */
ompi_mpi_abort_print_stack = OPAL_INT_TO_BOOL(value);
#else
/* If we do not have stack trace capability, ensure that this is
hard-coded to false */
ompi_mpi_abort_print_stack = false;
#endif
mca_base_param_reg_int_name("mpi", "preconnect_all",
"Whether to force MPI processes to create OOB "
"and MPI connections with *all* peers during "
"MPI_INIT (vs. making connections lazily -- "
"upon the first MPI traffic between each "
"process peer pair)",
false, false, 0, NULL);
mca_base_param_reg_int_name("mpi", "preconnect_mpi",
"Whether to force MPI processes to fully "
"wire-up the MPI connections between MPI "
"processes.",
false, false, 0, NULL);
mca_base_param_reg_int_name("mpi", "preconnect_oob",
"Whether to force MPI processes to fully "
"wire-up the OOB system between MPI processes.",
false, false, 0, NULL);
mca_base_param_reg_int_name("mpi", "preconnect_oob_simultaneous",
"Number of simultaneous outstanding "
"OOB connections to allow during preconnect.",
false, false, 4, NULL);
/* Leave pinned parameter */
mca_base_param_reg_int_name("mpi", "leave_pinned",
"Whether to use the \"leave pinned\" protocol or not. Enabling this setting can help bandwidth performance when repeatedly sending and receiving large messages with the same buffers over RDMA-based networks.",
false, false,
(int) ompi_mpi_leave_pinned, &value);
ompi_mpi_leave_pinned = OPAL_INT_TO_BOOL(value);
mca_base_param_reg_int_name("mpi", "leave_pinned_pipeline",
"Whether to use the \"leave pinned pipeline\" protocol or not.",
false, false,
(int) ompi_mpi_leave_pinned_pipeline, &value);
ompi_mpi_leave_pinned_pipeline = OPAL_INT_TO_BOOL(value);
if (ompi_mpi_leave_pinned && ompi_mpi_leave_pinned_pipeline) {
ompi_mpi_leave_pinned_pipeline = 0;
opal_show_help("help-mpi-runtime.txt",
"mpi-params:leave-pinned-and-pipeline-selected",
true);
}
/* Sparse group storage support */
mca_base_param_reg_int_name("mpi", "have_sparse_group_storage",
"Whether this Open MPI installation supports storing of data in MPI groups in \"sparse\" formats (good for extremely large process count MPI jobs that create many communicators/groups)",
false, true, (int) OMPI_GROUP_SPARSE, NULL);
mca_base_param_reg_int_name("mpi", "use_sparse_group_storage",
"Whether to use \"sparse\" storage formats for MPI groups (only relevant if mpi_have_sparse_group_storage is 1)",
false, false, OMPI_GROUP_SPARSE, &value);
ompi_use_sparse_group_storage = OPAL_INT_TO_BOOL(value);
if (ompi_use_sparse_group_storage) {
value = 0;
if (OMPI_GROUP_SPARSE) {
value = 1;
}
if (0 == value) {
opal_show_help("help-mpi-runtime.txt",
"sparse groups enabled but compiled out",
true);
ompi_use_sparse_group_storage = false;
}
}
/* The ddt engine has a few parameters */
return ompi_ddt_register_params();
}
int ompi_show_all_mca_params(int32_t rank, int requested, char *nodename) {
opal_list_t *info;
opal_list_item_t *i;
mca_base_param_info_t *item;
char *value_string;
int value_int;
FILE *fp = NULL;
time_t timestamp;
if (rank != 0) {
return OMPI_SUCCESS;
}
timestamp = time(NULL);
/* Open the file if one is specified */
if (0 != strlen(ompi_mpi_show_mca_params_file)) {
if ( NULL == (fp = fopen(ompi_mpi_show_mca_params_file, "w")) ) {
opal_output(0, "Unable to open file <%s> to write MCA parameters", ompi_mpi_show_mca_params_file);
return OMPI_ERR_FILE_OPEN_FAILURE;
}
fprintf(fp, "#\n");
fprintf(fp, "# This file was automatically generated on %s", ctime(&timestamp));
fprintf(fp, "# by MPI_COMM_WORLD rank %d (out of a total of %d) on %s\n", rank, requested, nodename );
fprintf(fp, "#\n");
}
mca_base_param_dump(&info, false);
for (i = opal_list_get_first(info);
i != opal_list_get_last(info);
i = opal_list_get_next(i)) {
item = (mca_base_param_info_t*) i;
/* Get the parameter name, and convert it to a printable string */
if (MCA_BASE_PARAM_TYPE_STRING == item->mbpp_type) {
mca_base_param_lookup_string(item->mbpp_index, &value_string);
if (NULL == value_string) {
value_string = strdup("");
}
} else {
mca_base_param_lookup_int(item->mbpp_index, &value_int);
asprintf(&value_string, "%d", value_int);
}
/* Print the parameter */
if (0 != strlen(ompi_mpi_show_mca_params_file)) {
fprintf(fp, "%s=%s\n", item->mbpp_full_name, value_string);
} else {
opal_output(0, "%s=%s", item->mbpp_full_name, value_string);
}
free(value_string);
}
/* Close file, cleanup allocated memory*/
if (0 != strlen(ompi_mpi_show_mca_params_file)) {
fclose(fp);
}
mca_base_param_dump_release(info);
return OMPI_SUCCESS;
}