v4.1.x: Backport alltoall in place fixes #9672
Merged
Conversation
This file contains 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
The temporary buffer must be shifted by the true_extent on a per type basis (since the various datatypes might have different true_extent). Thanks Heiko Bauke for reporting this. Refs. open-mpi#9329 Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp> (cherry picked from commit 0041ce8) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
This function can be used to compute the packed size of a datatype on a target architecture. Signed-off-by: George Bosilca <bosilca@icl.utk.edu> (cherry picked from commit 74049fc) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
Dont copy the datatype into a buffer with the same extent, but instead pack it and send it to the peer as packed. Signed-off-by: George Bosilca <bosilca@icl.utk.edu> (cherry picked from commit 447b289) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
Provide optimized variant for the homogeneous case. Signed-off-by: George Bosilca <bosilca@icl.utk.edu> (cherry picked from commit dc4e2ce) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
Signed-off-by: George Bosilca <bosilca@icl.utk.edu> (cherry picked from commit b9012a3) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
Signed-off-by: Mikhail Kurnosov <mkurnosov@gmail.com> (cherry picked from commit aba6765) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
b9012a3 used the alltoallw interpretation of rdisps instead of the alltoall/alltoallv interpretation. According to the MPI standard, the byte displacement is recvbuf + rdispls[i] * extent(recvtype) for alltoall and alltoallv, but is recvbuf + rdispls[i] for alltoallw. Signed-off-by: Brian Barrett <bbarrett@amazon.com> (cherry picked from commit d8c50a5)
Without waiting for the last receive before going to the next iteration we might overwrite data if the current left neighbor become the right at the next iteration. Start with an MPI_REQUEST_NULL request. If the request is not NULL and the first rcounts is 0, the ompi_request_wait will segfault. Signed-off-by: George Bosilca <bosilca@icl.utk.edu> (cherry picked from commit 6802702) Signed-off-by: Brian Barrett <bbarrett@amazon.com>
bosilca
approved these changes
Nov 16, 2021
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Backport a couple of different patch series which started as fix for #9329, included some performance fixes, and then fixes for our fixes (including fixing #9501).