2004-07-21 02:32:45 +04:00
|
|
|
/*
|
2005-11-05 22:57:48 +03:00
|
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2005 The University of Tennessee and The University
|
|
|
|
* of Tennessee Research Foundation. All rights
|
|
|
|
* reserved.
|
2004-11-28 23:09:25 +03:00
|
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
|
|
|
* University of Stuttgart. All rights reserved.
|
2005-03-24 15:43:37 +03:00
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
2004-11-22 04:38:40 +03:00
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
2004-07-21 02:32:45 +04:00
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "ompi_config.h"
|
|
|
|
|
2004-11-05 10:52:30 +03:00
|
|
|
#include <string.h>
|
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
#include "opal/class/opal_list.h"
|
2005-07-04 04:13:44 +04:00
|
|
|
#include "opal/util/argv.h"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "opal/mca/mca.h"
|
|
|
|
#include "opal/mca/base/base.h"
|
|
|
|
#include "ompi/mca/topo/topo.h"
|
|
|
|
#include "ompi/mca/topo/base/base.h"
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
#include "orte/util/output.h"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/communicator/communicator.h"
|
2004-07-21 02:32:45 +04:00
|
|
|
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
/*
|
|
|
|
* Local functions
|
|
|
|
*/
|
|
|
|
static void fill_null_pointers(mca_topo_base_module_t *module);
|
|
|
|
|
|
|
|
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
|
|
|
* This structure is needed so that we can close the modules
|
|
|
|
* which are not selected but were opened. mca_base_modules_close
|
2005-07-03 20:22:16 +04:00
|
|
|
* which does this job for us requires a opal_list_t which contains
|
2004-07-21 02:32:45 +04:00
|
|
|
* these modules
|
|
|
|
*/
|
|
|
|
struct queried_module_t {
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_item_t super;
|
2004-08-02 04:24:22 +04:00
|
|
|
mca_topo_base_component_t *om_component;
|
2004-07-21 02:32:45 +04:00
|
|
|
mca_topo_base_module_t *om_module;
|
|
|
|
};
|
|
|
|
typedef struct queried_module_t queried_module_t;
|
2005-07-03 20:22:16 +04:00
|
|
|
static OBJ_CLASS_INSTANCE(queried_module_t, opal_list_item_t, NULL, NULL);
|
2004-07-21 02:32:45 +04:00
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
|
|
|
* Only one topo module can be attached to each communicator.
|
|
|
|
*
|
2004-08-02 04:24:22 +04:00
|
|
|
* This module calls the query funtion on all the components that were
|
|
|
|
* detected by topo_base_open. This function is called on a
|
|
|
|
* per-communicator basis. This function has the following function.
|
2004-07-21 02:32:45 +04:00
|
|
|
*
|
2004-08-02 04:24:22 +04:00
|
|
|
* 1. Iterate over the list of available_components
|
|
|
|
* 2. Call the query function on each of these components.
|
2004-07-21 02:32:45 +04:00
|
|
|
* 3. query function returns the structure containing pointers
|
2004-08-02 04:24:22 +04:00
|
|
|
* to its module and its priority
|
2004-07-21 02:32:45 +04:00
|
|
|
* 4. Select the module with the highest priority
|
2004-08-02 04:24:22 +04:00
|
|
|
* 5. Call the init function on the selected module so that it does the
|
2004-07-21 02:32:45 +04:00
|
|
|
* right setup for the communicator
|
|
|
|
* 6. Call finalize on all the other modules which returned
|
2004-08-02 04:24:22 +04:00
|
|
|
* their module but were unfortunate to not get selected
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
|
|
|
|
|
|
|
int mca_topo_base_comm_select (struct ompi_communicator_t *comm,
|
2004-08-02 04:24:22 +04:00
|
|
|
struct mca_base_component_t *preferred)
|
|
|
|
{
|
2004-07-21 02:32:45 +04:00
|
|
|
int priority;
|
|
|
|
int best_priority;
|
|
|
|
char name[MPI_MAX_OBJECT_NAME+32];
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_item_t *item;
|
|
|
|
opal_list_item_t *next_item;
|
2004-08-06 00:31:14 +04:00
|
|
|
mca_base_component_priority_list_item_t *selectable_item;
|
|
|
|
char *names, **name_array;
|
|
|
|
int num_names;
|
2004-07-21 02:32:45 +04:00
|
|
|
mca_base_component_priority_list_item_t *cpli;
|
2004-08-02 04:24:22 +04:00
|
|
|
mca_topo_base_component_t *component;
|
|
|
|
mca_topo_base_component_t *best_component;
|
2004-07-21 02:32:45 +04:00
|
|
|
mca_topo_base_module_t *module;
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_t queried;
|
2004-07-21 02:32:45 +04:00
|
|
|
queried_module_t *om;
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_t *selectable;
|
2004-07-21 02:32:45 +04:00
|
|
|
char *str;
|
2004-09-23 03:58:13 +04:00
|
|
|
int err = MPI_SUCCESS;
|
2004-08-06 00:31:14 +04:00
|
|
|
int i;
|
|
|
|
bool was_selectable_constructed = false;
|
2004-07-21 02:32:45 +04:00
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
/* Announce */
|
2004-07-21 02:32:45 +04:00
|
|
|
|
|
|
|
/* ANJU:
|
|
|
|
* check for names array .... mca_base_param_ */
|
|
|
|
|
|
|
|
snprintf(name, sizeof(name), "%s (cid %d)", comm->c_name,
|
|
|
|
comm->c_contextid);
|
|
|
|
name[sizeof(name) - 1] = '\0';
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-07-21 02:32:45 +04:00
|
|
|
"topo:base:comm_select: new communicator: %s",
|
|
|
|
name);
|
|
|
|
|
2004-08-06 00:31:14 +04:00
|
|
|
|
|
|
|
/* Check and see if a preferred component was provided. If it was
|
2004-08-02 04:24:22 +04:00
|
|
|
provided then it should be used (if possible) */
|
|
|
|
|
2004-07-21 02:32:45 +04:00
|
|
|
if (NULL != preferred) {
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
/* We have a preferred component. Check if it is available
|
2004-07-21 02:32:45 +04:00
|
|
|
and if so, whether it wants to run */
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
str = &(preferred->mca_component_name[0]);
|
2004-07-21 02:32:45 +04:00
|
|
|
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-02 04:24:22 +04:00
|
|
|
"topo:base:comm_select: Checking preferred component: %s",
|
2004-07-21 02:32:45 +04:00
|
|
|
str);
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
/* query the component for its priority and get its module
|
2004-07-21 02:32:45 +04:00
|
|
|
structure. This is necessary to proceed */
|
|
|
|
|
2004-10-05 20:32:52 +04:00
|
|
|
component = (mca_topo_base_component_t *)preferred;
|
|
|
|
module = component->topom_comm_query (&priority);
|
2004-08-02 04:24:22 +04:00
|
|
|
if (NULL != module &&
|
|
|
|
NULL != module->topo_module_init &&
|
|
|
|
NULL != module->topo_graph_map &&
|
|
|
|
NULL != module->topo_cart_map) {
|
2004-07-21 02:32:45 +04:00
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
/* this query seems to have returned something legitimate
|
|
|
|
* and we can now go ahead and initialize the
|
|
|
|
* communicator with it * but first, the functions which
|
|
|
|
* are null need to be filled in */
|
2004-07-21 02:32:45 +04:00
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
fill_null_pointers (module);
|
|
|
|
comm->c_topo = module;
|
2004-10-05 20:32:52 +04:00
|
|
|
comm->c_topo_component = preferred;
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
return module->topo_module_init(comm);
|
2004-07-21 02:32:45 +04:00
|
|
|
}
|
2004-08-02 04:24:22 +04:00
|
|
|
/* His preferred component is present, but is unable to
|
|
|
|
* run. This is not a good sign. We should try selecting
|
|
|
|
* some other component We let it fall through and select
|
|
|
|
* from the list of available components
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
} /*end of selection for preferred component */
|
2004-07-21 02:32:45 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We fall till here if one of the two things happened:
|
2004-08-02 04:24:22 +04:00
|
|
|
* 1. The preferred component was provided but for some reason was
|
|
|
|
* not able to be selected
|
|
|
|
* 2. No preferred component was provided
|
2004-07-21 02:32:45 +04:00
|
|
|
*
|
2004-08-02 04:24:22 +04:00
|
|
|
* All we need to do is to go through the list of available
|
|
|
|
* components and find the one which has the highest priority and
|
|
|
|
* use that for this communicator
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
|
|
|
|
2004-08-06 00:31:14 +04:00
|
|
|
/* Check if anything was requested by means on the name parameters */
|
|
|
|
names = NULL;
|
|
|
|
mca_base_param_lookup_string (mca_topo_base_param, &names);
|
|
|
|
|
|
|
|
if (NULL != names && 0 < strlen(names)) {
|
2005-07-04 04:13:44 +04:00
|
|
|
name_array = opal_argv_split (names, ',');
|
|
|
|
num_names = opal_argv_count (name_array);
|
2004-08-06 00:31:14 +04:00
|
|
|
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-06 00:31:14 +04:00
|
|
|
"topo:base:comm_Select: Checking all available module");
|
|
|
|
|
|
|
|
/* since there are somethings which the mca requested through the
|
|
|
|
if the intersection is NULL, then we barf saying that the requested
|
|
|
|
modules are not being available */
|
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
selectable = OBJ_NEW(opal_list_t);
|
2004-08-06 00:31:14 +04:00
|
|
|
was_selectable_constructed = true;
|
|
|
|
|
|
|
|
/* go through the compoents_available list and check against the names
|
|
|
|
* to see whether this can be added or not */
|
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
for (item = opal_list_get_first(&mca_topo_base_components_available);
|
|
|
|
item != opal_list_get_end(&mca_topo_base_components_available);
|
|
|
|
item = opal_list_get_next(item)) {
|
|
|
|
/* convert the opal_list_item_t returned into the proper type */
|
2004-08-06 00:31:14 +04:00
|
|
|
cpli = (mca_base_component_priority_list_item_t *) item;
|
2004-11-12 19:55:41 +03:00
|
|
|
component = (mca_topo_base_component_t *) cpli->super.cli_component;
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-06 00:31:14 +04:00
|
|
|
"select: initialising %s component %s",
|
|
|
|
component->topom_version.mca_type_name,
|
|
|
|
component->topom_version.mca_component_name);
|
|
|
|
|
|
|
|
/* check if this name is present in the mca_base_params */
|
|
|
|
for (i=0; i < num_names; i++) {
|
|
|
|
if (0 == strcmp(name_array[i], component->topom_version.mca_component_name)) {
|
|
|
|
/* this is present, and should be added o the selectable list */
|
|
|
|
|
|
|
|
/* We need to create a seperate object to initialise this list with
|
|
|
|
* since we cannot have the same item in 2 lists */
|
|
|
|
|
|
|
|
selectable_item = OBJ_NEW (mca_base_component_priority_list_item_t);
|
|
|
|
*selectable_item = *cpli;
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_append (selectable, (opal_list_item_t *)selectable_item);
|
2004-08-06 00:31:14 +04:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* check for a NULL intersection between the available list and the
|
|
|
|
* list which was asked for */
|
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
if (0 == opal_list_get_size(selectable)) {
|
2004-08-06 00:31:14 +04:00
|
|
|
was_selectable_constructed = true;
|
|
|
|
OBJ_RELEASE (selectable);
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose (10, mca_topo_base_output,
|
2004-08-06 00:31:14 +04:00
|
|
|
"topo:base:comm_select: preferred modules were not available");
|
|
|
|
return OMPI_ERROR;
|
|
|
|
}
|
|
|
|
} else { /* if there was no name_array, then we need to simply initialize
|
|
|
|
selectable to mca_topo_base_components_available */
|
|
|
|
selectable = &mca_topo_base_components_available;
|
|
|
|
}
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
best_component = NULL;
|
2004-07-21 02:32:45 +04:00
|
|
|
best_priority = -1;
|
2005-07-03 20:22:16 +04:00
|
|
|
OBJ_CONSTRUCT(&queried, opal_list_t);
|
2004-07-21 02:32:45 +04:00
|
|
|
|
2005-07-03 20:22:16 +04:00
|
|
|
for (item = opal_list_get_first(selectable);
|
|
|
|
item != opal_list_get_end(selectable);
|
|
|
|
item = opal_list_get_next(item)) {
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
2005-07-03 20:22:16 +04:00
|
|
|
* convert the opal_list_item_t returned into the proper type
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
|
|
|
cpli = (mca_base_component_priority_list_item_t *) item;
|
2004-11-12 19:55:41 +03:00
|
|
|
component = (mca_topo_base_component_t *) cpli->super.cli_component;
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-02 04:24:22 +04:00
|
|
|
"select: initialising %s component %s",
|
|
|
|
component->topom_version.mca_type_name,
|
|
|
|
component->topom_version.mca_component_name);
|
2004-07-21 02:32:45 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* we can call the query function only if there is a function :-)
|
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
if (NULL == component->topom_comm_query) {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-02 04:24:22 +04:00
|
|
|
"select: no query, ignoring the component");
|
2004-07-21 02:32:45 +04:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* call the query function and see what it returns
|
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
module = component->topom_comm_query (&priority);
|
2004-07-21 02:32:45 +04:00
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
if (NULL == module ||
|
|
|
|
NULL == module->topo_module_init ||
|
|
|
|
NULL == module->topo_graph_map ||
|
|
|
|
NULL == module->topo_cart_map) {
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
|
|
|
* query did not return any action which can be used
|
|
|
|
*/
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-07-21 02:32:45 +04:00
|
|
|
"select: query returned failure");
|
|
|
|
} else {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2007-01-30 23:56:31 +03:00
|
|
|
"select: query returned priority %d",
|
2004-07-21 02:32:45 +04:00
|
|
|
priority);
|
|
|
|
/*
|
2004-08-02 04:24:22 +04:00
|
|
|
* is this the best component we have found till now?
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
|
|
|
if (priority > best_priority) {
|
|
|
|
best_priority = priority;
|
2004-08-02 04:24:22 +04:00
|
|
|
best_component = component;
|
2004-07-21 02:32:45 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
om = OBJ_NEW(queried_module_t);
|
|
|
|
/*
|
|
|
|
* check if we have run out of space
|
|
|
|
*/
|
|
|
|
if (NULL == om) {
|
|
|
|
OBJ_DESTRUCT(&queried);
|
|
|
|
return OMPI_ERR_OUT_OF_RESOURCE;
|
|
|
|
}
|
2004-08-02 04:24:22 +04:00
|
|
|
om->om_component = component;
|
|
|
|
om->om_module = module;
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_append(&queried, (opal_list_item_t *)om);
|
2004-08-02 04:24:22 +04:00
|
|
|
} /* end else of if (NULL == module) */
|
|
|
|
} /* end else of if (NULL == component->topom_init) */
|
2004-07-21 02:32:45 +04:00
|
|
|
} /* end for ... end of traversal */
|
|
|
|
|
2004-08-06 00:31:14 +04:00
|
|
|
/* We have to remove empty out the selectable list if the selectable
|
|
|
|
* list was constructed as a duplicate and not as a pointer to the
|
|
|
|
* mca_base_components_available list. So, check and destroy */
|
|
|
|
|
|
|
|
if (was_selectable_constructed) {
|
|
|
|
|
|
|
|
/* remove all the items first */
|
2005-07-03 20:22:16 +04:00
|
|
|
for (item = opal_list_get_first(&mca_topo_base_components_available);
|
|
|
|
item != opal_list_get_end(&mca_topo_base_components_available);
|
2004-08-06 00:31:14 +04:00
|
|
|
item = next_item) {
|
2005-07-03 20:22:16 +04:00
|
|
|
next_item = opal_list_get_next(item);
|
2004-08-06 00:31:14 +04:00
|
|
|
OBJ_RELEASE (item);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* release the list itself */
|
|
|
|
OBJ_RELEASE (selectable);
|
|
|
|
was_selectable_constructed = false;
|
|
|
|
}
|
|
|
|
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
2004-08-02 04:24:22 +04:00
|
|
|
* Now we have alist of components which successfully returned
|
|
|
|
* their module struct. One of these components has the best
|
|
|
|
* priority. The rest have to be comm_unqueried to counter the
|
|
|
|
* effects of comm_query'ing them. Finalize happens only on
|
|
|
|
* components which should are initialized.
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
if (NULL == best_component) {
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
2004-08-02 04:24:22 +04:00
|
|
|
* This typically means that there was no component which was
|
|
|
|
* able to run properly this time. So, we need to abort
|
2004-07-21 02:32:45 +04:00
|
|
|
* JMS replace with show_help
|
|
|
|
*/
|
|
|
|
OBJ_DESTRUCT(&queried);
|
|
|
|
return OMPI_ERROR;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2004-08-02 04:24:22 +04:00
|
|
|
* We now have a list of components which have successfully
|
|
|
|
* returned their priorities from the query. We now have to
|
|
|
|
* unquery() those components which have not been selected and
|
|
|
|
* init() the component which was selected
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
2005-07-03 20:22:16 +04:00
|
|
|
for (item = opal_list_remove_first(&queried);
|
2004-07-21 02:32:45 +04:00
|
|
|
NULL != item;
|
2005-07-03 20:22:16 +04:00
|
|
|
item = opal_list_remove_first(&queried)) {
|
2004-07-21 02:32:45 +04:00
|
|
|
om = (queried_module_t *) item;
|
2004-08-02 04:24:22 +04:00
|
|
|
if (om->om_component == best_component) {
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
2004-08-02 04:24:22 +04:00
|
|
|
* this is the chosen component, we have to initialise the
|
|
|
|
* module of this component.
|
2004-07-21 02:32:45 +04:00
|
|
|
*
|
2004-08-02 04:24:22 +04:00
|
|
|
* ANJU: a component might not have all the functions
|
|
|
|
* defined. Whereever a function pointer is null in the
|
|
|
|
* module structure we need to fill it in with the base
|
|
|
|
* structure function pointers. This is yet to be done
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
2004-08-02 04:24:22 +04:00
|
|
|
* We don return here coz we still need to go through and
|
|
|
|
* elease the other objects
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
fill_null_pointers (om->om_module);
|
|
|
|
comm->c_topo = om->om_module;
|
2004-09-23 03:58:13 +04:00
|
|
|
err = om->om_module->topo_module_init(comm);
|
2004-10-05 20:32:52 +04:00
|
|
|
comm->c_topo_component = (mca_base_component_t *)best_component;
|
2004-07-21 02:32:45 +04:00
|
|
|
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* this is not the "choosen one", finalize
|
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
if (NULL != om->om_component->topom_comm_unquery) {
|
|
|
|
/* unquery the component only if they have some clean
|
|
|
|
* up job to do. Components which are queried but do
|
|
|
|
* not actually do anything typically do not have a
|
|
|
|
* unquery. Hence this check is necessary
|
2004-07-21 02:32:45 +04:00
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
(void) om->om_component->topom_comm_unquery(comm);
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-02 04:24:22 +04:00
|
|
|
"select: component %s is not selected",
|
|
|
|
om->om_component->topom_version.mca_component_name);
|
2004-07-21 02:32:45 +04:00
|
|
|
} /* end if */
|
2004-08-02 04:24:22 +04:00
|
|
|
} /* if not best component */
|
2004-07-21 02:32:45 +04:00
|
|
|
OBJ_RELEASE(om);
|
|
|
|
} /* traversing through the entire list */
|
|
|
|
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_output_verbose(10, mca_topo_base_output,
|
2004-08-02 04:24:22 +04:00
|
|
|
"select: component %s selected",
|
2004-09-23 03:58:13 +04:00
|
|
|
best_component->topom_version.mca_component_name);
|
2004-07-21 02:32:45 +04:00
|
|
|
|
|
|
|
OBJ_DESTRUCT(&queried);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2004-08-02 04:24:22 +04:00
|
|
|
|
2004-07-21 02:32:45 +04:00
|
|
|
/*
|
|
|
|
* This function fills in the null function pointers, in other words,
|
2004-08-02 04:24:22 +04:00
|
|
|
* those functions which are not implemented by the module with the
|
|
|
|
* pointers from the base function. Somewhere, I need to incoroporate
|
|
|
|
* a check for the common minimum funtions being implemented by the
|
2004-07-21 02:32:45 +04:00
|
|
|
* module atleast. If not, this module cannot be considered.
|
|
|
|
*/
|
2004-08-02 04:24:22 +04:00
|
|
|
static void fill_null_pointers(mca_topo_base_module_t *module)
|
|
|
|
{
|
2004-07-21 02:32:45 +04:00
|
|
|
#define CHECK_FOR_NULL_FUNCTION_POINTER(name) \
|
2004-08-02 04:24:22 +04:00
|
|
|
if (NULL == module->topo_##name) { \
|
|
|
|
module->topo_##name = mca_topo_base_##name; \
|
2004-07-21 02:32:45 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cart_coords);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cart_create);
|
2004-10-05 20:32:52 +04:00
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cart_get);
|
2004-07-21 02:32:45 +04:00
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cartdim_get);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cart_rank);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cart_shift);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(cart_sub);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(graph_create);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(graph_get);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(graphdims_get);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(graph_neighbors);
|
|
|
|
CHECK_FOR_NULL_FUNCTION_POINTER(graph_neighbors_count);
|
|
|
|
|
|
|
|
#undef CHECK_FOR_NULL_FUNCTION_POINTER
|
|
|
|
}
|