-
Notifications
You must be signed in to change notification settings - Fork 25
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
sysfs, cpuallocator: topology discovery fixes and improvements. #206
sysfs, cpuallocator: topology discovery fixes and improvements. #206
Conversation
49a0016
to
79015e6
Compare
79015e6
to
06663c7
Compare
06663c7
to
8820726
Compare
c98ff97
to
73963dc
Compare
c6a9e24
to
e7085e6
Compare
894fb23
to
41bb78f
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
41bb78f
to
bd6a8d3
Compare
bd6a8d3
to
cce7adb
Compare
84ff7f7
to
5fbc25a
Compare
e915a21
to
a61f7eb
Compare
Implement missing cache discovery bits. Add CPU interface for querying caches. Enable cache discovery by default. Signed-off-by: Krisztian Litkey <krisztian.litkey@intel.com>
Implement DieClusterIDs() and DieClusterCPUSet() for querying CPU cluster IDs and CPUs per cluster (and per die). There are two other variants (LogicalDieCluster{IDs,CPUSet}() provided. These filter out any reported clusters which are hyperthreads and replace them with a single cluster which contains all the hyperthreads and has the cluster ID of the lowest CPU. Signed-off-by: Krisztian Litkey <krisztian.litkey@intel.com>
Clean up a bit detection of online, offline and isolated CPU sets, and add more consistent interface names. Add support for querying possible, and present CPUs as well. Signed-off-by: Krisztian Litkey <krisztian.litkey@intel.com>
Try to satisfy satisfy allocation requests consuming one or more full CPU clusters before falling back to smaller blocks of allocation. Try to squeeze the full allocation into a single die or a single package. Don't attempt any cluster-based allocation if neither of these is possible. Satisfy low-prio allocations using E-core clusters only. Use P-core clusters only to satisfy priority preferences above low-prio. Signed-off-by: Krisztian Litkey <krisztian.litkey@intel.com>
0626580
to
774c205
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm
Notes: this patch series is stacked on top of #294.
This patch series improves hardware topology discovery and CPU allocation. The improvements include: