Merged
Conversation
When we are in a singleton, Fence should automatically return SUCCESS while Fence_nb returns OPERATION_SUCCEEDED. Cleanup a few spots to ensure we always get the release code back to the user. Need to check because the server returns SUCCEEDED if it is locally complete on the node. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 0cca795)
The PMIx IOF functions access global storage and therefore must be threadshifted prior to execution. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit e37fa25)
The Standard requires that the PMIx_Log_nb function call its callback function after it completes if it returned success to the caller. It is to return operation_succeeded if it atomically completes the operation, and an error if it cannot perform the operation. So properly handle those cases and support the callback handshake with the server and host. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 539b8e1)
Ported from an OMPI contribution: ``` update-my-copyright.py: properly support git workspaces Don't search for a .git directory; it might not exist. Also, remove unnecessary Mercurial and Subversion support; we haven't used these for years. Signed-off-by: Jeff Squyres <jeff@squyres.com> ``` Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit e238a6e)
When someone calls PMIx_Init again and passes info objects, we shift to a function to see if that info requires notification. The shift caddy copies the info, so we need to indicate it has been copied so that the destructor will release them. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit cbac91e)
Just some minor updates Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit a80c055)
Not sure how much it will help, but let's at least try to see if stopping the progress thread (as opposed to trying to let it drain) will help the AWS situation. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 7704efa)
Silence valgrind leak reports except for some leaks in HWLOC and libevent that are outside of our control. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 75711b6)
Stop the main PMIx progress thread, but leave the event base intact so that infrastructure can be safely torn down. Protect all APIs that threadshift so that calls down into the library are properly rejected. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 2abc2bc)
Create new attribute to pass name of thread to stop. If nothing passed, then stop all progress threads. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 1eb961d)
Ensure we don't add events to the stopped progress thread while we are trying to flush it. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 5541eeb)
Mark that this version of PMIx provides the PMIx_Value_get_number function. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 2967f4a)
When group info is returned, store it as qualified values so they can be retrieved. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit c5e3c3f)
Scrub the pmix_info support (based on the OMPI model) to enable reuse by PRRTE. Ensure all cmd line options are supported and documented. Support color output, report by type, and other cute features. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit f8b75a2)
Because the show-version option of pmix_info has a complex syntax with multiple choices for arguments, add some special code for processing it in the cmd line. Ensure that pmix_info knows how to handle all the argument choices. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit ee1d6c1)
Include all the inputs. Note that a NULL topology input results in use of the PMIx libraries internal copy of the local topology. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit b21daa0)
If someone provides backing storage and request we return a static value, then return the value in their location. Note this is only available via the blocking form of PMIx_Get as there is no way to pass the address of the backing storage to the non-blocking function. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 679f695)
Ensure we report the correct cmd line. Abstract the support functions so a 3rd party can call and get the PMIx configuration, and still be able to use the basic functions to report their own config. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 23908e2)
Use the directives/ndirs fields as that is what the threadshifted execution function is expecting to see. Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit 9b222f4)
Patch provided by George Bosilca Signed-off-by: Ralph Castain <rhc@pmix.org> (cherry picked from commit f29cfc7)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update PMIx_Fence to fully conform to Standard
When we are in a singleton, Fence should automatically return SUCCESS
while Fence_nb returns OPERATION_SUCCEEDED. Cleanup a few spots to
ensure we always get the release code back to the user. Need to
check because the server returns SUCCEEDED if it is locally
complete on the node.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 0cca795)
Threadshift IOF API calls
The PMIx IOF functions access global storage and
therefore must be threadshifted prior to execution.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit e37fa25)
Update log support to conform to Standard
The Standard requires that the PMIx_Log_nb function call its
callback function after it completes if it returned success
to the caller. It is to return operation_succeeded if it
atomically completes the operation, and an error if it cannot
perform the operation. So properly handle those cases and
support the callback handshake with the server and host.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 539b8e1)
update-my-copyright.py: properly support git workspaces
Ported from an OMPI contribution:
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit e238a6e)
Seal memory leak
When someone calls PMIx_Init again and passes info objects, we
shift to a function to see if that info requires notification.
The shift caddy copies the info, so we need to indicate it has
been copied so that the destructor will release them.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit cbac91e)
Update the monitor_multi example
Just some minor updates
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit a80c055)
Stop the progress thread right away in server_finalize
Not sure how much it will help, but let's at least try
to see if stopping the progress thread (as opposed to
trying to let it drain) will help the AWS situation.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 7704efa)
Silence valgrind issues
Silence valgrind leak reports except for some leaks in HWLOC and libevent that are outside of our control.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 75711b6)
Implement new API to stop the progress thread
Stop the main PMIx progress thread, but leave the event
base intact so that infrastructure can be safely torn
down. Protect all APIs that threadshift so that calls
down into the library are properly rejected.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 2abc2bc)
Allow passing of progress thread to stop - default NULL to all
Create new attribute to pass name of thread to stop. If nothing
passed, then stop all progress threads.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 1eb961d)
Protect callbacks from threadshift when progress thread is stopped
Ensure we don't add events to the stopped progress thread while we
are trying to flush it.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 5541eeb)
Add capability: get number function available
Mark that this version of PMIx provides the PMIx_Value_get_number function.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 2967f4a)
Ensure to store group info in PMIx server
When group info is returned, store it as qualified
values so they can be retrieved.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit c5e3c3f)
Revamp the pmix_info support
Scrub the pmix_info support (based on the OMPI model) to enable
reuse by PRRTE. Ensure all cmd line options are supported and
documented. Support color output, report by type, and other
cute features.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit f8b75a2)
Fine-tune the show-version option
Because the show-version option of pmix_info has a complex
syntax with multiple choices for arguments, add some special
code for processing it in the cmd line. Ensure that pmix_info
knows how to handle all the argument choices.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit ee1d6c1)
Improve description of PMIx_Compute_distances API
Include all the inputs. Note that a NULL topology input
results in use of the PMIx libraries internal copy of
the local topology.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit b21daa0)
Fully support return of static values
If someone provides backing storage and request we return a static
value, then return the value in their location. Note this is only
available via the blocking form of PMIx_Get as there is no way
to pass the address of the backing storage to the non-blocking
function.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 679f695)
Cleanup and abstract pmix_info support
Ensure we report the correct cmd line. Abstract the
support functions so a 3rd party can call and get
the PMIx configuration, and still be able to use
the basic functions to report their own config.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 23908e2)
Correctly threadshift PMIx_IOF_push directives
Use the directives/ndirs fields as that is what the
threadshifted execution function is expecting to see.
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit 9b222f4)
Correct cflags used for check_compiler_version.m4
Patch provided by George Bosilca
Signed-off-by: Ralph Castain rhc@pmix.org
(cherry picked from commit f29cfc7)