Implementation of the current state of MPI Continuations proposal [WIP] #9502

devreal · 2021-10-09T22:03:25Z

This PR adds an implementation of the MPI Continuations proposal as extension to Open MPI. This proposal is current under discussion in the MPI hybrid and accelerator working group. By integrating it as an extension into Open MPI, I am hoping to provide an up-to-date implementation for the community to experiment with. It reflects the current state of the proposal and will be kept in sync with the evolving proposal.

The implementation is mostly confined to the extension itself, with the exception of hooks in request test and wait functions used to allow polling on a continuation request to complete outstanding continuations. This is needed for the implementation of the "mpi_continue_poll_only" info key (see the description below).

Note: this PR is currently WIP as it relies on a workaround to define OMPI_HAVE_MPI_EXT_CONTINUE, which is used to disable the hooks in request test/wait functions if the extension is not enabled. The underlying issue is that the extensions integration currently does not support including the mpiext.h header in implementation files (needed to disable the hooks mentioned above).

Overview of MPI Continuations

Continuations provide a mechanism for attaching callbacks to outstanding operation requests. A call to MPIX_Continue takes a request, a function pointer, a user-provided data pointer, and a status object (or MPI_STATUS_IGNORE), along with a continuation request and attaches the continuation to the operation request:

MPI_Request req;
// status object has to remain valid until the callback is invoked
MPI_Status *status = malloc(sizeof(MPI_Status));
char *buf = ...;
MPI_Irecv(buf, ..., MPI_ANY_SOURCE, ... &req);
MPIX_Continue(&req, &complete_cb, buf, status, cont_req);
assert(req == MPI_REQUEST_NULL);

The ownership of non-persistent requests is returned to MPI and the pointer to the request will be set to MPI_REQUEST_NULL. The callback is passed the status pointer and the user-provided data pointer:

void complete_cb(MPI_Status *status, void *user_data) {
  printf("Send completed\n");
  char *buf = (char*)user_data;
  process_msg(buf, status->MPI_SOURCE);
  free(buf); // free the send buffer
  free(status);    // free the status
}

The status has to remain valid until the invocation of the callback and is set according to the operation before the callback is invoked.

The continuation is registered with the provided continuation request. The continuation request is a request allocated using MPIX_Continue_init:

MPIX_Continue_init(info, &cont_req);

Continuation requests may be used to test/wait for completion of all continuations registered with it using MPI_Test/Wait. Supported info keys are:

"mpi_continue_poll_only": if true, only execute continuations when MPI_Test/Wait is called on the continuation request. If false, continuations may be executed at any time inside a call into MPI (inside a callback registered with opal_progress in the implementation; default: false).
"mpi_continue_enqueue_complete": if false, the continuation is executed immediately if the operations are already complete when MPIX_Continue is called. Execution is deferred otherwise (default: false)
"mpi_continue_max_poll": the maximum number of continuations to execute when calling MPI_Test on the continuation request (default: -1, meaning unlimited)

A continuation may in turn be attached to a continuation request, in which case it will be executed once all continuations registered with the continuation request have completed.

In addition to MPIX_Continue, the proposal also includes MPIX_Continueall which attaches a continuation to a set of requests such that the continuation is executed once all operations have completed.

Signed-off-by: Joseph Schuchart [email protected]
Signed-off-by: George Bosilca [email protected]

bosilca

What is the impact of the continuations on the latency ? What if we mix normal requests and continuations in the same app ?

ompi/mpiext/continue/c/continuation.c

devreal · 2021-10-18T23:20:40Z

I pushed a bunch of changes. The biggest change is that a continuation request is now derived from ompi_request_t, which shrunk ompi_request_cont_data_t considerably and cuts down on the pointer chasing. The latter is now only used to link operation requests to the continuation object and the user-provided status object.

devreal · 2021-10-18T23:21:13Z

Oh, and I will squash everything down to one commit once we're done...

bosilca

Definitively much cleaner.

bosilca · 2021-10-19T13:57:35Z

ompi/mpiext/continue/configure.m4

+
+    # This example can always build, so we just execute $1 if it was
+    # requested.
+    AS_IF([test "$ENABLE_continue" = "1" || \


These IFs looks similar enough to be merged.

ompi/request/req_test.c

devreal · 2021-11-09T15:15:00Z

@bosilca I pushed a few changes:

Now using the new OMPI_MPIEXT_*_POST_CONFIG hook. I hope this is how it is intended to be used.
I had to make sure that the state of continuation requests is not changed to inactive in test/wait. They are always active and never explicitly (re)started by the user.
Continuations that are poll-only (not executed as part of global progress) will now only be executed by the thread testing/waiting for their completion. We don't merge the local list into the global list anymore but instead allow the waiting thread to execute them if not blocked on the sync. Otherwise they will be executed at the end of the wait, once all continuations are eligible for execution. I think this change makes sense as this avoids other threads from being disturbed by poll-only continuations.

bosilca · 2021-11-10T06:13:58Z

ompi/mpiext/continue/c/continuation.h

+int ompi_continue_progress_request(ompi_request_t *cont_req);
+
+/**
+ * Attach a continuation to a set of operations represented by \c requests.


what is this \c ?

bosilca · 2021-11-10T06:25:15Z

ompi/mpiext/continue/c/continuation.c

+
+/**
+ * Register the provided continuation request to be included in the
+ * global progress loop (used while a thread is waiting for the contnuation


typo. contnuation

bosilca · 2021-11-10T06:27:29Z

ompi/mpiext/continue/c/continuation.c

+
+    opal_atomic_unlock(&cont_req->cont_lock);
+
+    if (NULL == thread_progress_list) {


Why is this global object not protected for concurrent accesses ?

That's a thread_local variable that stores the continuation requests the current thread is waiting on

github-actions · 2022-11-03T17:52:53Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

d80d3ca: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2022-11-22T15:05:32Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2022-11-29T20:39:50Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2022-12-12T13:48:47Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2022-12-16T10:52:19Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2022-12-19T15:34:27Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2023-01-31T00:33:37Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2023-03-02T21:40:41Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2023-08-21T17:22:41Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2023-09-07T17:58:47Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2024-02-06T18:37:50Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

f3babb5: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2024-04-02T19:45:23Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

a32e7e3: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2024-04-02T20:48:23Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

a32e7e3: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

github-actions · 2024-04-02T20:51:50Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

a32e7e3: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

Signed-off-by: Joseph Schuchart <[email protected]>

Explicit global progress instead of global progress when progressing a request, to avoid recursive progress calls. Signed-off-by: Joseph Schuchart <[email protected]>

Signed-off-by: Joseph Schuchart <[email protected]>

Also: always store continuations directly in poll-only CRs. Signed-off-by: Joseph Schuchart <[email protected]>

Signed-off-by: Joseph Schuchart <[email protected]>

- Swap flags and max_poll parameter in MPI_Continue_init - Swap cont_req and info parameter in MPI_Continue_init - Add new flags and remove support for info keys mpi_continue_poll_only, mpi_continue_max_poll, and mpi_continue_enqueue_complete Signed-off-by: Joseph Schuchart <[email protected]>

This was a temporary bandaid to fix the C11 sequentially consistent store... Signed-off-by: Joseph Schuchart <[email protected]>

Continuation requests may be marked as complete before another continuation is registered. In that case, we should mark the continuation request as pending again, unless a sync object has been registered already. Signed-off-by: Joseph Schuchart <[email protected]>

Signed-off-by: Joseph Schuchart <[email protected]>

The cont->cont_req field was reset *after* it was returned to the freelist... Signed-off-by: Joseph Schuchart <[email protected]>

We must protect the activation of the completion of a continuation request to ensure that we don't miss any updates. Signed-off-by: Joseph Schuchart <[email protected]>

Runnable continuations should be added to cont_complete_defer_list if the continuation request is inactive, so that it can be enqueued for execution once the request is enabled. Signed-off-by: Joseph Schuchart <[email protected]>

Signed-off-by: Joseph Schuchart <[email protected]>

This was a temporary idea that has been removed again. Signed-off-by: Joseph Schuchart <[email protected]>

MPIX_CONT_PERSISTENT will not be part of the first proposal and has been removed. Signed-off-by: Joseph Schuchart <[email protected]>

Signed-off-by: Joseph Schuchart <[email protected]>

github-actions · 2024-07-18T13:55:12Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

d77ca1c: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

It is not sufficient to just set one value, we need proper initialization through OMPI_REQUEST_INIT. Signed-off-by: Joseph Schuchart <[email protected]>

github-actions · 2024-08-08T21:01:49Z

Hello! The Git Commit Checker CI bot found a few problems with this PR:

d77ca1c: Remove re-iteration of continuation requests in te...

check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

devreal added ⚠️ WIP-DNM! Target: main labels Oct 9, 2021

devreal requested a review from bosilca October 9, 2021 22:03

bosilca reviewed Oct 13, 2021

View reviewed changes

bosilca reviewed Oct 19, 2021

View reviewed changes

devreal force-pushed the mpi-continue-master branch from 2f49fa2 to fa149c2 Compare November 9, 2021 04:40

bosilca reviewed Nov 10, 2021

View reviewed changes

devreal force-pushed the mpi-continue-master branch from 0fffb9c to e014717 Compare April 18, 2022 13:22

devreal marked this pull request as draft October 6, 2022 15:23

devreal force-pushed the mpi-continue-master branch from 03c36a5 to e2979db Compare November 22, 2022 15:04

devreal force-pushed the mpi-continue-master branch from df79651 to c8c0437 Compare March 2, 2023 21:40

devreal force-pushed the mpi-continue-master branch from a02d677 to b6440c0 Compare April 2, 2024 19:44

devreal added 26 commits July 18, 2024 09:54

Reduce atomic operations by checking for empty list

3f87912

Signed-off-by: Joseph Schuchart <[email protected]>

Reorder progress of individual requests

e9caa09

Explicit global progress instead of global progress when progressing a request, to avoid recursive progress calls. Signed-off-by: Joseph Schuchart <[email protected]>

Avoid atomic operation if all requests have completed

57d9c50

Signed-off-by: Joseph Schuchart <[email protected]>

Let the test and wait calls set CRs to inactive

d1a4206

Also: always store continuations directly in poll-only CRs. Signed-off-by: Joseph Schuchart <[email protected]>

Progress CR in ompi_request_wait_completion before creating a sync

9e8505c

Signed-off-by: Joseph Schuchart <[email protected]>

Continuation tests: test for delayed start

e26a9ba

Signed-off-by: Joseph Schuchart <[email protected]>

Infrastructure to test for MPI object error handler, not used yet

8e3be49

Signed-off-by: Joseph Schuchart <[email protected]>

Remove an atomic op from wait_sync_update

488909e

Signed-off-by: Joseph Schuchart <[email protected]>

Outline constructor for thread-local data

0bb7352

Signed-off-by: Joseph Schuchart <[email protected]>

Remove left over use of OPAL_ATOMIC_RELAXED_STORE

27736e0

This was a temporary bandaid to fix the C11 sequentially consistent store... Signed-off-by: Joseph Schuchart <[email protected]>

Don't set persistent requests to MPI_REQUEST_NULL

0984ecd

Signed-off-by: Joseph Schuchart <[email protected]>

Fix wrong locking when freeing a continuation request

d90b4e1

Signed-off-by: Joseph Schuchart <[email protected]>

Continuations: Fix potential race condition in debug mode

1b042c5

The cont->cont_req field was reset *after* it was returned to the freelist... Signed-off-by: Joseph Schuchart <[email protected]>

Hotfix for lost callbacks

a8ef1ce

We must protect the activation of the completion of a continuation request to ensure that we don't miss any updates. Signed-off-by: Joseph Schuchart <[email protected]>

Properly defer continuations if the cont_req is inactive

d88917e

Runnable continuations should be added to cont_complete_defer_list if the continuation request is inactive, so that it can be enqueued for execution once the request is enabled. Signed-off-by: Joseph Schuchart <[email protected]>

Slight code cleanup and remove locks if threads are not enabled

e605570

Signed-off-by: Joseph Schuchart <[email protected]>

Docs: don't mention MPI_UNDEFINED

7cf2ffd

This was a temporary idea that has been removed again. Signed-off-by: Joseph Schuchart <[email protected]>

Fixup documentation and remove mention of MPIX_CONT_PERSISTENT

7066d62

MPIX_CONT_PERSISTENT will not be part of the first proposal and has been removed. Signed-off-by: Joseph Schuchart <[email protected]>

Continuations: Cleanup dox and implement invoke-failed

ffe1341

Signed-off-by: Joseph Schuchart <[email protected]>

Put a continuation back into the cont_incomplete_list if rechecking

d6e372b

Signed-off-by: Joseph Schuchart <[email protected]>

Add write memory barrier before invoking request callback

b5596a8

Signed-off-by: Joseph Schuchart <[email protected]>

Add REQUEST_CB_PENDING that got lost during rebase

63b25f4

Signed-off-by: Joseph Schuchart <[email protected]>

Fix compiler errors in continuation.c

3f33c7d

Signed-off-by: Joseph Schuchart <[email protected]>

Don't cast function pointers to void*

68b887c

Signed-off-by: Joseph Schuchart <[email protected]>

devreal force-pushed the mpi-continue-master branch from 790a7b6 to 68b887c Compare July 18, 2024 13:54

pml/ucx: Properly reinitialize requests

dcb9df5

It is not sufficient to just set one value, we need proper initialization through OMPI_REQUEST_INIT. Signed-off-by: Joseph Schuchart <[email protected]>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implementation of the current state of MPI Continuations proposal [WIP] #9502

Implementation of the current state of MPI Continuations proposal [WIP] #9502

devreal commented Oct 9, 2021

bosilca left a comment

devreal commented Oct 18, 2021

devreal commented Oct 18, 2021

bosilca left a comment

bosilca Oct 19, 2021

devreal commented Nov 9, 2021

bosilca Nov 10, 2021

bosilca Nov 10, 2021

bosilca Nov 10, 2021

devreal Nov 10, 2021

github-actions bot commented Nov 3, 2022

github-actions bot commented Nov 22, 2022

github-actions bot commented Nov 29, 2022

github-actions bot commented Dec 12, 2022

github-actions bot commented Dec 16, 2022

github-actions bot commented Dec 19, 2022

github-actions bot commented Jan 31, 2023

github-actions bot commented Mar 2, 2023

github-actions bot commented Aug 21, 2023

github-actions bot commented Sep 7, 2023

github-actions bot commented Feb 6, 2024

github-actions bot commented Apr 2, 2024

github-actions bot commented Apr 2, 2024

github-actions bot commented Apr 2, 2024

github-actions bot commented Jul 18, 2024

github-actions bot commented Aug 8, 2024


		opal_atomic_unlock(&cont_req->cont_lock);

		if (NULL == thread_progress_list) {

Implementation of the current state of MPI Continuations proposal [WIP] #9502

Are you sure you want to change the base?

Implementation of the current state of MPI Continuations proposal [WIP] #9502

Conversation

devreal commented Oct 9, 2021

Overview of MPI Continuations

bosilca left a comment

Choose a reason for hiding this comment

devreal commented Oct 18, 2021

devreal commented Oct 18, 2021

bosilca left a comment

Choose a reason for hiding this comment

bosilca Oct 19, 2021

Choose a reason for hiding this comment

devreal commented Nov 9, 2021

bosilca Nov 10, 2021

Choose a reason for hiding this comment

bosilca Nov 10, 2021

Choose a reason for hiding this comment

bosilca Nov 10, 2021

Choose a reason for hiding this comment

devreal Nov 10, 2021

Choose a reason for hiding this comment

github-actions bot commented Nov 3, 2022

github-actions bot commented Nov 22, 2022

github-actions bot commented Nov 29, 2022

github-actions bot commented Dec 12, 2022

github-actions bot commented Dec 16, 2022

github-actions bot commented Dec 19, 2022

github-actions bot commented Jan 31, 2023

github-actions bot commented Mar 2, 2023

github-actions bot commented Aug 21, 2023

github-actions bot commented Sep 7, 2023

github-actions bot commented Feb 6, 2024

github-actions bot commented Apr 2, 2024

github-actions bot commented Apr 2, 2024

github-actions bot commented Apr 2, 2024

github-actions bot commented Jul 18, 2024

github-actions bot commented Aug 8, 2024