mpi: manually heap-allocate payloads for local messages #378
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
csegarragonz
force-pushed
the
local-fastpath
branch
from
8375d31
to
9ece9e1
Compare
|
TODO:
|
csegarragonz
force-pushed
the
local-fastpath
branch
from
9ece9e1
to
715e186
Compare
csegarragonz
commented
Feb 23, 2024
| @@ -242,21 +242,6 @@ TEST_CASE_METHOD(MpiTestFixture, "Test send and recv on same host", "[mpi]") | |||
| world.send( | |||
| rankA1, rankA2, BYTES(messageData.data()), MPI_INT, messageData.size()); | |||
|
|
|||
| SECTION("Test queueing") | |||
There was a problem hiding this comment.
This does not apply as the message buffer is not in the queue any more.
csegarragonz
changed the title
csegarragonz
force-pushed
the
local-fastpath
branch
2 times, most recently
from
7ec2294
to
65cd6af
Compare
csegarragonz
force-pushed
the
local-fastpath
branch
from
65cd6af
to
c95f559
Compare
csegarragonz
force-pushed
the
local-fastpath
branch
from
6fd483a
to
c63f81c
Compare
csegarragonz
force-pushed
the
local-fastpath
branch
from
c63f81c
to
a8f9a68
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #378 +/- ##
==========================================
- Coverage 81.94% 80.86% -1.08%
==========================================
Files 106 107 +1
Lines 8013 7222 -791
==========================================
- Hits 6566 5840 -726
+ Misses 1447 1382 -65 ☔ View full report in Codecov by Sentry. |
csegarragonz
added a commit
that referenced
this pull request
Feb 28, 2024
* mpi: make local fast path faster * util/memory: use mimalloc * nits: run clang-format * nits: some self-review * mpi: fix detection of local messages in receiver * mpi: only memcpy the required bytes * tests: make lsan happy with the malloc/free optimization * nits: run clang format
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.
This PR makes sure that, when sending a local MPI message, we make the minimum number of copies (of the message payload).
The minimum number of copies is two: (i) from the sender-owned buffer, to a runtime-owned buffer, and (ii) from the runtime-owned buffer to the receiver-owned buffer.
In the future, we could consider using something different than
std::malloc.