Add multi-thread support to mpiP #13
Merged
Conversation
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
Closed
Contributor
Author
|
I performed the initial performance measurements. |
Contributor
Author
|
After reducing the number of TLS accesses (c2c1ccf) the performance of the profile-free MPI seems to be comparable to the case where mpiP is running with |
Contributor
Author
|
I am now working through the test suite coming with mpiP to make sure that it's not broken. |
artpol84
force-pushed
the
topic/mt
branch
6 times, most recently
from
252b093 to
8d9ec1b
Compare
Contributor
Author
|
All is cleared now |
1. Add arch directory with required atomics implementations (for ARM64, PPC and x86_64). TODO: add other architectures later. 2. Add a thread-safe list that will be used for tracking TLS objects. 3. Update Makefile.in to build thread-safe list 4. Add stubs for mpiP threads support Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Introduce a new "mt" (Multi-Threaded) layer for statistics to abstract multi-threaded isolation logic from the upper layers as well as from the single-threaded statistics collection. NOTE: this commit only introduces the layer and internally a 1-to-1 bypass to a single-threaded layer is performed. Signed-off-by: Artem Polyakov <artpol84@gmail.com>
The nested calls flag is located at the single-thread level. Move the check to the same level as the flag. Signed-off-by: Artem Polyakov <artpol84@gmail.com>
This functionality may be used on different layers. Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Signed-off-by: Artem Polyakov <artpol84@gmail.com>
Contributor
Author
|
@cchambreau, rebased |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Status
Details:
Test no. 1: Multi-threaded point-to-point
mpirun -np 2 -x LD_PRELOAD ./mt_test1 -m -t 2 -n 1000 -v p2p-sb-rb-menables MPI Thread multiple-t 2defines 2 threads to use on each rank-n 1000- number of sends/receives each thread is performing-v p2p-sb-rb- test type: each thread of rank=0 performs 1000 blocking sends (using integer thread id as the tag). rank=1 performs matching receives.Results (mpiP/master)
Number of Send and Recv invocations should be 2000 (1000 iteration * 2 threads),
App time is incorrectly calculated.
Results (mpiP/PR#13)
The number of Send and Recv is correct, application % is not screwed.