scx_layered: Use topology for core selection #510
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
hodgesds
commented
Aug 19, 2024
scheds/rust/scx_layered/src/main.rs
Outdated
| @@ -1030,16 +1048,43 @@ impl Layer { | |||
| } | |||
| } | |||
|
|
|||
| let mut nodes = topo.nodes().iter().collect::<Vec<_>>().clone(); | |||
| let num_nodes = nodes.len(); | |||
| nodes.rotate_left(if idx <= num_nodes { idx } else { idx % num_nodes }); | |||
There was a problem hiding this comment.
This is the main logic, it makes use of rotate_left, which annoyingly will panic if the rotate size is too large. To get around that it mods by the number of nodes, llcs, or cores if the number of layers is larger than any of those values.
hodgesds
force-pushed
the
layered-core-topo-selection
branch
from
b130a8c
to
7f01ad3
Compare
hodgesds
commented
Aug 19, 2024
| let available_cores = self.cpus_to_cores(&available_cpus).ok()?; | ||
|
|
||
| for alloc_core in layer.core_alloc_order() { | ||
| match available_cores.get(*alloc_core) { |
There was a problem hiding this comment.
this goes from the core id to the bitmask bit for checking if the core is allowed to be used
htejun
approved these changes
Aug 19, 2024
scheds/rust/scx_layered/src/main.rs
Outdated
| let num_nodes = nodes.len(); | ||
| let is_left = idx % 2 == 0; | ||
| if is_left { | ||
| nodes.rotate_left(if idx <= num_nodes { idx } else { idx % num_nodes }); |
There was a problem hiding this comment.
Maybe sth like let rot_by |idx, len| if idx <= len { idx } else { idx % len }; would help? Hmm... isn't this the same as idx % num_nodes?
hodgesds
force-pushed
the
layered-core-topo-selection
branch
from
7f01ad3
to
3e2ff83
Compare
Currently the core selection logic in scx_layered uses the first available core in the bitmask. This is suboptimal when the scheduler is configured with specific NUMA/LLC restrictions. The ideal core selection logic should try to find the least used cores within the preferred scheduling domain and allocate new cpus from shared cores within that domain. Signed-off-by: Daniel Hodges <[email protected]>
hodgesds
force-pushed
the
layered-core-topo-selection
branch
from
3e2ff83
to
b3793e0
Compare
hodgesds
added a commit
to hodgesds/scx
that referenced
this pull request
Aug 21, 2024
Fix a bug introduced in sched-ext#510 where it assumed core ids are incremental. This refactors the core ordering for layers to be far more simple and provide some space for layer core isolation in low utilization. Signed-off-by: Daniel Hodges <[email protected]>
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.
Currently the core selection logic in scx_layered uses the first available core in the bitmask. This is suboptimal when the scheduler is configured with specific NUMA/LLC restrictions and with multiple layers. This refactors the core selection to use a preferred order of cores based on the layer index.