2004-09-26 21:43:35 +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.
|
2006-08-23 07:32:36 +04:00
|
|
|
* Copyright (c) 2004-2006 The University of Tennessee and The University
|
2005-11-05 22:57:48 +03:00
|
|
|
* 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
|
|
|
|
*
|
2004-09-26 21:43:35 +04:00
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* @file
|
|
|
|
*
|
|
|
|
* Interface for waitpid / async notification of child death with the
|
|
|
|
* libevent runtime system.
|
|
|
|
*/
|
2005-03-14 23:57:21 +03:00
|
|
|
#ifndef ORTE_WAIT_H
|
|
|
|
#define ORTE_WAIT_H
|
2004-09-26 21:43:35 +04:00
|
|
|
|
2005-03-14 23:57:21 +03:00
|
|
|
#include "orte_config.h"
|
2004-09-26 21:43:35 +04:00
|
|
|
|
2004-10-20 05:03:09 +04:00
|
|
|
#ifdef HAVE_SYS_TYPES_H
|
2004-09-26 21:43:35 +04:00
|
|
|
#include <sys/types.h>
|
2004-10-20 05:03:09 +04:00
|
|
|
#endif
|
2008-02-28 04:57:57 +03:00
|
|
|
#ifdef HAVE_SYS_TIME_H
|
|
|
|
#include <sys/time.h>
|
2006-08-23 07:32:36 +04:00
|
|
|
#endif
|
|
|
|
|
2008-02-28 22:58:32 +03:00
|
|
|
#include "opal/dss/dss.h"
|
|
|
|
#include "opal/class/opal_list.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"
|
2008-02-28 04:57:57 +03:00
|
|
|
#include "opal/event/event.h"
|
2008-02-28 22:58:32 +03:00
|
|
|
#include "opal/runtime/opal_progress.h"
|
2008-02-28 04:57:57 +03:00
|
|
|
|
2008-02-28 22:58:32 +03:00
|
|
|
#include "orte/mca/rml/rml_types.h"
|
2008-02-28 04:57:57 +03:00
|
|
|
#include "orte/runtime/orte_globals.h"
|
|
|
|
|
|
|
|
BEGIN_C_DECLS
|
|
|
|
|
2004-09-26 21:43:35 +04:00
|
|
|
/** typedef for callback function used in \c ompi_rte_wait_cb */
|
2005-03-14 23:57:21 +03:00
|
|
|
typedef void (*orte_wait_fn_t)(pid_t wpid, int status, void *data);
|
2004-09-26 21:43:35 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* Wait for process terminiation
|
|
|
|
*
|
2005-03-14 23:57:21 +03:00
|
|
|
* Similar to \c waitpid, \c orte_waitpid utilizes the run-time
|
2004-09-26 21:43:35 +04:00
|
|
|
* event library for process terminiation notification. The \c
|
|
|
|
* WUNTRACED option is not supported, but the \c WNOHANG option is
|
|
|
|
* supported.
|
|
|
|
*
|
|
|
|
* \note A \c wpid value of \c -1 is not currently supported and will
|
|
|
|
* return an error.
|
|
|
|
*/
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC pid_t orte_waitpid(pid_t wpid, int *status, int options);
|
2004-09-26 21:43:35 +04:00
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Register a callback for process termination
|
|
|
|
*
|
|
|
|
* Register a callback for notification when \c wpid causes a SIGCHLD.
|
|
|
|
* \c waitpid() will have already been called on the process at this
|
|
|
|
* time.
|
|
|
|
*
|
|
|
|
* If a thread is already blocked in \c ompi_rte_waitpid for \c wpid,
|
2006-02-12 04:33:29 +03:00
|
|
|
* this function will return \c ORTE_ERR_EXISTS. It is illegal for
|
2005-03-18 06:43:59 +03:00
|
|
|
* multiple callbacks to be registered for a single \c wpid
|
|
|
|
* (OMPI_EXISTS will be returned in this case).
|
2004-09-26 21:43:35 +04:00
|
|
|
*
|
|
|
|
* \warning It is not legal for \c wpid to be -1 when registering a
|
|
|
|
* callback.
|
|
|
|
*/
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_cb(pid_t wpid, orte_wait_fn_t callback, void *data);
|
2004-09-26 21:43:35 +04:00
|
|
|
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_cb_cancel(pid_t wpid);
|
2004-10-27 02:11:03 +04:00
|
|
|
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_cb_disable(void);
|
2004-10-27 02:11:03 +04:00
|
|
|
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_cb_enable(void);
|
2004-09-26 21:43:35 +04:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/**
|
|
|
|
* Setup to wait for an event
|
|
|
|
*
|
|
|
|
* This function is used to setup a pipe that can be used elsewhere
|
|
|
|
* in the code base where we want to wait for some event to
|
|
|
|
* happen. For example, orterun uses this function to setup an event
|
|
|
|
* that is used to notify orterun of abnormal and normal termination
|
|
|
|
* so it can wakeup and exit cleanly.
|
|
|
|
*
|
|
|
|
* The event will be defined so that a write to the provided trigger
|
|
|
|
* pipe will cause the event to trigger and callback to the provided
|
|
|
|
* function
|
|
|
|
*/
|
|
|
|
ORTE_DECLSPEC int orte_wait_event(opal_event_t **event, int *trig,
|
|
|
|
void (*cbfunc)(int, short, void*));
|
|
|
|
|
2008-02-28 22:58:32 +03:00
|
|
|
/**
|
|
|
|
* In a number of places in the code, we need to wait for something
|
|
|
|
* to complete - for example, waiting for all launched procs to
|
|
|
|
* report into the HNP. In such cases, we want to just call progress
|
|
|
|
* so that any messages get processed, but otherwise "hold" the
|
|
|
|
* program at this spot until the counter achieves the specified
|
|
|
|
* value. We also want to provide a boolean flag, though, so that
|
|
|
|
* we break out of the loop should something go wrong.
|
|
|
|
*/
|
|
|
|
#define ORTE_PROGRESSED_WAIT(failed, counter, limit) \
|
|
|
|
do { \
|
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((1, orte_debug_output, \
|
2008-02-28 22:58:32 +03:00
|
|
|
"progressed_wait: %s %d", \
|
|
|
|
__FILE__, __LINE__)); \
|
|
|
|
while (!(failed) && (counter) < (limit)) { \
|
|
|
|
opal_progress(); \
|
|
|
|
} \
|
|
|
|
} while(0); \
|
|
|
|
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/**
|
|
|
|
* Trigger a defined event
|
|
|
|
*
|
|
|
|
* This function will trigger a previously-defined event - as setup
|
|
|
|
* by orte_wait_event - by sending a message to the provided pipe
|
|
|
|
*/
|
|
|
|
ORTE_DECLSPEC void orte_trigger_event(int trig);
|
|
|
|
|
|
|
|
/**
|
2008-02-28 22:58:32 +03:00
|
|
|
* Setup an event to process a message
|
|
|
|
*
|
|
|
|
* If we are in an OOB recv callback, we frequently cannot process
|
|
|
|
* the received message until after we return from the callback to
|
|
|
|
* avoid a potential loopback situation - i.e., where processing
|
|
|
|
* the message can result in a message being sent somewhere that
|
|
|
|
* subsequently causes the recv we are in to get called again.
|
|
|
|
* This is typically the problem facing the daemons and HNP.
|
2008-02-28 04:57:57 +03:00
|
|
|
*
|
2008-02-28 22:58:32 +03:00
|
|
|
* To resolve this problem, we place the message to be processed on
|
|
|
|
* a list, and create a zero-time event that calls the function
|
|
|
|
* that will process the received message. The event library kindly
|
|
|
|
* does not trigger this event until after we return from the recv
|
|
|
|
* since the recv itself is considered an "event"! Thus, we will
|
|
|
|
* always execute the specified event cb function -after- leaving
|
|
|
|
* the recv.
|
2008-02-28 04:57:57 +03:00
|
|
|
*/
|
2008-02-28 22:58:32 +03:00
|
|
|
typedef struct {
|
|
|
|
opal_object_t super;
|
|
|
|
orte_process_name_t sender;
|
|
|
|
opal_buffer_t *buffer;
|
|
|
|
orte_rml_tag_t tag;
|
2008-03-03 19:06:47 +03:00
|
|
|
#if OMPI_ENABLE_DEBUG
|
2008-02-29 23:10:31 +03:00
|
|
|
char *file;
|
|
|
|
int line;
|
2008-03-03 19:06:47 +03:00
|
|
|
#endif
|
2008-02-28 22:58:32 +03:00
|
|
|
} orte_message_event_t;
|
|
|
|
ORTE_DECLSPEC OBJ_CLASS_DECLARATION(orte_message_event_t);
|
|
|
|
|
2008-02-29 23:10:31 +03:00
|
|
|
#if OMPI_ENABLE_DEBUG
|
|
|
|
|
2008-02-28 22:58:32 +03:00
|
|
|
#define ORTE_MESSAGE_EVENT(sndr, buf, tg, cbfunc) \
|
|
|
|
do { \
|
|
|
|
orte_message_event_t *mev; \
|
|
|
|
struct timeval now; \
|
|
|
|
opal_event_t *tmp; \
|
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((1, orte_debug_output, \
|
2008-02-28 22:58:32 +03:00
|
|
|
"defining message event: %s %d", \
|
|
|
|
__FILE__, __LINE__)); \
|
|
|
|
tmp = (opal_event_t*)malloc(sizeof(opal_event_t)); \
|
|
|
|
mev = OBJ_NEW(orte_message_event_t); \
|
|
|
|
mev->sender.jobid = (sndr)->jobid; \
|
|
|
|
mev->sender.vpid = (sndr)->vpid; \
|
|
|
|
opal_dss.copy_payload(mev->buffer, (buf)); \
|
|
|
|
mev->tag = (tg); \
|
2008-02-29 23:10:31 +03:00
|
|
|
mev->file = strdup((buf)->parent.cls_init_file_name); \
|
|
|
|
mev->line = (buf)->parent.cls_init_lineno; \
|
2008-02-28 22:58:32 +03:00
|
|
|
opal_evtimer_set(tmp, (cbfunc), mev); \
|
|
|
|
now.tv_sec = 0; \
|
|
|
|
now.tv_usec = 0; \
|
|
|
|
opal_evtimer_add(tmp, &now); \
|
|
|
|
} while(0);
|
2008-02-29 23:10:31 +03:00
|
|
|
#else
|
|
|
|
|
|
|
|
#define ORTE_MESSAGE_EVENT(sndr, buf, tg, cbfunc) \
|
|
|
|
do { \
|
|
|
|
orte_message_event_t *mev; \
|
|
|
|
struct timeval now; \
|
|
|
|
opal_event_t *tmp; \
|
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((1, orte_debug_output, \
|
2008-02-29 23:10:31 +03:00
|
|
|
"defining message event: %s %d", \
|
|
|
|
__FILE__, __LINE__)); \
|
|
|
|
tmp = (opal_event_t*)malloc(sizeof(opal_event_t)); \
|
|
|
|
mev = OBJ_NEW(orte_message_event_t); \
|
|
|
|
mev->sender.jobid = (sndr)->jobid; \
|
|
|
|
mev->sender.vpid = (sndr)->vpid; \
|
|
|
|
opal_dss.copy_payload(mev->buffer, (buf)); \
|
|
|
|
mev->tag = (tg); \
|
|
|
|
opal_evtimer_set(tmp, (cbfunc), mev); \
|
|
|
|
now.tv_sec = 0; \
|
|
|
|
now.tv_usec = 0; \
|
|
|
|
opal_evtimer_add(tmp, &now); \
|
|
|
|
} while(0);
|
|
|
|
|
|
|
|
#endif
|
2008-02-28 22:58:32 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
|
|
|
|
/**
|
|
|
|
* In a number of places within the code, we want to setup a timer
|
|
|
|
* to detect when some procedure failed to complete. For example,
|
|
|
|
* when we launch the daemons, we frequently have no way to directly
|
|
|
|
* detect that a daemon failed to launch. Setting a timer allows us
|
|
|
|
* to automatically fail out of the launch if we don't hear from a
|
|
|
|
* daemon in some specified time window.
|
|
|
|
*
|
|
|
|
* Computing the amount of time to wait takes a few lines of code, but
|
|
|
|
* this macro encapsulates those lines along with the timer event
|
|
|
|
* definition just as a convenience. It also centralizes the
|
|
|
|
* necessary checks to ensure that the microsecond field is always
|
|
|
|
* less than 1M since some systems care about that, and to ensure
|
|
|
|
* that the computed wait time doesn't exceed the desired max
|
|
|
|
* wait
|
|
|
|
*/
|
|
|
|
#define ORTE_DETECT_TIMEOUT(event, n, deltat, maxwait, cbfunc) \
|
|
|
|
do { \
|
|
|
|
struct timeval now; \
|
|
|
|
opal_event_t *tmp; \
|
|
|
|
tmp = (opal_event_t*)malloc(sizeof(opal_event_t)); \
|
|
|
|
opal_evtimer_set(tmp, (cbfunc), NULL); \
|
|
|
|
now.tv_sec = 0; \
|
|
|
|
now.tv_usec = (float)(deltat) * (float)(n); \
|
2008-06-03 01:46:34 +04:00
|
|
|
if (maxwait > 0) { \
|
|
|
|
if (now.tv_usec > (maxwait)) { \
|
|
|
|
now.tv_usec = (maxwait); \
|
|
|
|
} \
|
2008-02-28 04:57:57 +03:00
|
|
|
} \
|
|
|
|
if (now.tv_usec > 1000000.0) { \
|
|
|
|
now.tv_sec = (float)((int)(now.tv_usec/1000000.0)); \
|
|
|
|
now.tv_usec = now.tv_usec - 1000000.0*now.tv_sec; \
|
|
|
|
} \
|
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((1, orte_debug_output, \
|
2008-02-28 04:57:57 +03:00
|
|
|
"defining timeout: %ld sec %ld usec", \
|
|
|
|
(long)now.tv_sec, (long)now.tv_usec)); \
|
|
|
|
opal_evtimer_add(tmp, &now); \
|
|
|
|
*(event) = tmp; \
|
|
|
|
}while(0); \
|
|
|
|
|
2008-03-05 16:51:32 +03:00
|
|
|
|
|
|
|
/**
|
|
|
|
* There are places in the code where we just want to periodically
|
|
|
|
* wakeup to do something, and then go back to sleep again. Setting
|
|
|
|
* a timer allows us to do this
|
|
|
|
*/
|
|
|
|
#define ORTE_TIMER_EVENT(time, cbfunc) \
|
|
|
|
do { \
|
|
|
|
struct timeval now; \
|
|
|
|
opal_event_t *tmp; \
|
|
|
|
tmp = (opal_event_t*)malloc(sizeof(opal_event_t)); \
|
|
|
|
opal_evtimer_set(tmp, (cbfunc), tmp); \
|
|
|
|
now.tv_sec = (time); \
|
|
|
|
now.tv_usec = 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((1, orte_debug_output, \
|
2008-03-05 16:51:32 +03:00
|
|
|
"defining timer event: %ld sec", \
|
|
|
|
(long)now.tv_sec)); \
|
|
|
|
opal_evtimer_add(tmp, &now); \
|
|
|
|
}while(0); \
|
|
|
|
|
|
|
|
|
2004-09-26 21:43:35 +04:00
|
|
|
/**
|
|
|
|
* \internal
|
|
|
|
*
|
|
|
|
* Initialize the wait system (allocate mutexes, etc.)
|
|
|
|
*/
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_init(void);
|
2004-09-26 21:43:35 +04:00
|
|
|
|
2005-03-14 23:57:21 +03:00
|
|
|
/**
|
|
|
|
* Kill all processes we are waiting on.
|
|
|
|
*/
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_kill(int sig);
|
2004-09-26 21:43:35 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* \internal
|
|
|
|
*
|
|
|
|
* Finalize the wait system (deallocate mutexes, etc.)
|
|
|
|
*/
|
2006-08-23 07:32:36 +04:00
|
|
|
ORTE_DECLSPEC int orte_wait_finalize(void);
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
END_C_DECLS
|
2004-09-26 21:43:35 +04:00
|
|
|
|
2005-03-14 23:57:21 +03:00
|
|
|
#endif /* #ifndef ORTE_WAIT_H */
|