Improve thread scheduling for lockhammer #18
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
Add an option which controls the order in which logical core numbers are populated. Lockhammer currently specializes the order in which it schedules hardware threads for systems where 2 threads share cpu resources (same physical core, shared cache, etc). Since several configurations exist from unique resources per thread to 4 or more threads per core add the option to arbitrarily select the interleave between logical core numbers which share physical resources. Fixes ARM-software#17 Change-Id: I90a04082750f7ebdc5121294e2a214195896c373 Signed-off-by: Lucas Crowthers <lucasc.qdt@qualcommdatacenter.com>
|
Can one of the admins verify this patch? |
|
Let's rethink this one since it didn't fix the issue Cavium raised. The interleave option is still useful for other reasons but I'd rather not change the affinity setting unless we have a good reason to do so. |
lucasclucasdo
force-pushed
the
lh-threadscheduling
branch
from
aee8ca9 to
12b45f8
Compare
|
I have updated this PR to only include the interleave command line option. We can address further thread scheduling issues in subsequent PRs after further experimentation. |
geoffreyblake
suggested changes
May 22, 2018
| @@ -265,7 +278,7 @@ void* hmr(void *ptr) | |||
| } | |||
| else { | |||
| /* Calculate affinity mask for my core and set affinity */ | |||
| CPU_SET(((mycore >> 1)) + ((ncores >> 1) * (mycore & 1)), &affin_mask); | |||
| CPU_SET(((mycore * ncores / ileave) % ncores + (mycore / ileave)), &affin_mask); | |||
There was a problem hiding this comment.
Hi Lucas,
Can you add a comment here about the interleaving assumption is 0 - (numcores / numthreads) is thread 0 on each physical core, (numcores / numthreads) - 2 * (numcores / numthreads) is thread 1 on each core etc?
Add some documentation to the source code which describes the relationship between threads, physical cores, and logical core numberings and how the interleave option can be used to select which will be populated first with increasing core counts. Change-Id: I5a5eee67a4e3f1579690cd76a93deccbccaf4cfd Signed-off-by: Lucas Crowthers <lucasc.qdt@qualcommdatacenter.com>
lucasclucasdo
force-pushed
the
lh-threadscheduling
branch
from
817aba4 to
abc16f7
Compare
geoffreyblake
approved these changes
May 23, 2018
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.
Add a core interleave argument to lockhammer which determines the order in which logical core numbers are populated.
Fixes #17