Struct Opts

pub(crate) struct Opts {

    pub(crate) slice_us_underutil: u64,
    pub(crate) slice_us_overutil: u64,
    pub(crate) interval: f64,
    pub(crate) tune_interval: f64,
    pub(crate) load_half_life: f64,
    pub(crate) cache_level: u32,
    pub(crate) cpumasks: Vec<String>,
    pub(crate) greedy_threshold: u32,
    pub(crate) greedy_threshold_x_numa: u32,
    pub(crate) no_load_balance: bool,
    pub(crate) kthreads_local: bool,
    pub(crate) balanced_kworkers: bool,
    pub(crate) fifo_sched: bool,
    pub(crate) direct_greedy_under: f64,
    pub(crate) kick_greedy_under: f64,
    pub(crate) direct_greedy_numa: bool,
    pub(crate) partial: bool,
    pub(crate) mempolicy_affinity: bool,
    pub(crate) stats: Option<f64>,
    pub(crate) monitor: Option<f64>,
    pub(crate) exit_dump_len: u32,
    pub(crate) verbose: u8,
    pub(crate) version: bool,
    pub(crate) help_stats: bool,
    pub(crate) perf: u32,
}

scx_rusty: A multi-domain BPF / userspace hybrid scheduler

The BPF part does simple vtime or round robin scheduling in each domain while tracking average load of each domain and duty cycle of each task.

The userspace part performs two roles. First, it makes higher frequency (100ms) tuning decisions. It identifies CPUs which are not too heavily loaded and marks them so that they can pull tasks from other overloaded domains on the fly.

Second, it drives lower frequency (2s) load balancing. It determines whether load balancing is necessary by comparing domain load averages. If there are large enough load differences, it examines upto 1024 recently active tasks on the domain to determine which should be migrated.

The overhead of userspace operations is low. Load balancing is not performed frequently, but work-conservation is still maintained through tuning and greedy execution. Load balancing itself is not that expensive either. It only accesses per-domain load metrics to determine the domains that need load balancing, as well as limited number of per-task metrics for each pushing domain.

An earlier variant of this scheduler was used to balance across six domains, each representing a chiplet in a six-chiplet AMD processor, and could match the performance of production setup using CFS.

WARNING: scx_rusty currently assumes that all domains have equal processing power and at similar distances from each other. This limitation will be removed in the future.

Fields

slice_us_underutil: u64

Scheduling slice duration for under-utilized hosts, in microseconds.

slice_us_overutil: u64

Scheduling slice duration for over-utilized hosts, in microseconds.

interval: f64

Load balance interval in seconds.

tune_interval: f64

The tuner runs at a higher frequency than the load balancer to dynamically tune scheduling behavior. Tuning interval in seconds.

load_half_life: f64

The half-life of task and domain load running averages in seconds.

cache_level: u32

Build domains according to how CPUs are grouped at this cache level as determined by /sys/devices/system/cpu/cpuX/cache/indexI/id.

cpumasks: Vec<String>

Instead of using cache locality, set the cpumask for each domain manually. Provide multiple –cpumasks, one for each domain. E.g. –cpumasks 0xff_00ff –cpumasks 0xff00 will create two domains, with the corresponding CPUs belonging to each domain. Each CPU must belong to precisely one domain.

greedy_threshold: u32

When non-zero, enable greedy task stealing. When a domain is idle, a cpu will attempt to steal tasks from another domain as follows:

1. Try to consume a task from the current domain
2. Try to consume a task from another domain in the current NUMA node (or globally, if running on a single-socket system), if the domain has at least this specified number of tasks enqueued.

See greedy_threshold_x_numa to enable task stealing across NUMA nodes. Tasks stolen in this manner are not permanently stolen from their domain.

greedy_threshold_x_numa: u32

When non-zero, enable greedy task stealing across NUMA nodes. The order of greedy task stealing follows greedy-threshold as described above, and greedy-threshold must be nonzero to enable task stealing across NUMA nodes.

no_load_balance: bool

Disable load balancing. Unless disabled, userspace will periodically calculate the load factor of each domain and instruct BPF which processes to move.

kthreads_local: bool

Put per-cpu kthreads directly into local dsq’s.

balanced_kworkers: bool

In recent kernels (>=v6.6), the kernel is responsible for balancing kworkers across L3 cache domains. Exclude them from load-balancing to avoid conflicting operations. Greedy executions still apply.

fifo_sched: bool

Use FIFO scheduling instead of weighted vtime scheduling.

direct_greedy_under: f64

Idle CPUs with utilization lower than this will get remote tasks directly pushed onto them. 0 disables, 100 always enables.

kick_greedy_under: f64

Idle CPUs with utilization lower than this may get kicked to accelerate stealing when a task is queued on a saturated remote domain. 0 disables, 100 enables always.

direct_greedy_numa: bool

Whether tasks can be pushed directly to idle CPUs on NUMA nodes different than their domain’s node. If direct-greedy-under is disabled, this option is a no-op. Otherwise, if this option is set to false (default), tasks will only be directly pushed to idle CPUs if they reside on the same NUMA node as the task’s domain.

partial: bool

If specified, only tasks which have their scheduling policy set to SCHED_EXT using sched_setscheduler(2) are switched. Otherwise, all tasks are switched.

mempolicy_affinity: bool

Enables soft NUMA affinity for tasks that use set_mempolicy. This may improve performance in some scenarios when using mempolicies.

stats: Option<f64>

Enable stats monitoring with the specified interval.

monitor: Option<f64>

Run in stats monitoring mode with the specified interval. The scheduler is not launched.

exit_dump_len: u32

Exit debug dump buffer length. 0 indicates default.

verbose: u8

Enable verbose output, including libbpf details. Specify multiple times to increase verbosity.

version: bool

Print version and exit.

help_stats: bool

Show descriptions for statistics.

perf: u32

Tunable for prioritizing CPU performance by configuring the CPU frequency governor. Valid values are [0, 1024]. Higher values prioritize performance, lower values prioritize energy efficiency. When in doubt, use 0 or 1024.

Trait Implementations

impl Args for Opts

fn group_id() -> Option<Id>

Report the [ArgGroup::id][crate::ArgGroup::id] for this set of arguments

fn augment_args<'b>(__clap_app: Command) -> Command

Append to [Command] so it can instantiate Self via [FromArgMatches::from_arg_matches_mut]

fn augment_args_for_update<'b>(__clap_app: Command) -> Command

Append to [Command] so it can instantiate self via [FromArgMatches::update_from_arg_matches_mut]

impl CommandFactory for Opts

fn command<'b>() -> Command

Build a [Command] that can instantiate Self.

fn command_for_update<'b>() -> Command

Build a [Command] that can update self.

impl Debug for Opts

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl FromArgMatches for Opts

fn from_arg_matches(__clap_arg_matches: &ArgMatches) -> Result<Self, Error>

Instantiate Self from [ArgMatches], parsing the arguments as needed.

fn from_arg_matches_mut( __clap_arg_matches: &mut ArgMatches, ) -> Result<Self, Error>

Instantiate Self from [ArgMatches], parsing the arguments as needed.

fn update_from_arg_matches( &mut self, __clap_arg_matches: &ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.

fn update_from_arg_matches_mut( &mut self, __clap_arg_matches: &mut ArgMatches, ) -> Result<(), Error>

Assign values from ArgMatches to self.

impl Parser for Opts

fn parse() -> Self

Parse from std::env::args_os(), [exit][Error::exit] on error.

fn try_parse() -> Result<Self, Error>

Parse from std::env::args_os(), return Err on error.

fn parse_from<I, T>(itr: I) -> Self

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Parse from iterator, [exit][Error::exit] on error.

fn try_parse_from<I, T>(itr: I) -> Result<Self, Error>

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Parse from iterator, return Err on error.

fn update_from<I, T>(&mut self, itr: I)

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Update from iterator, [exit][Error::exit] on error.

fn try_update_from<I, T>(&mut self, itr: I) -> Result<(), Error>

where I: IntoIterator<Item = T>, T: Into<OsString> + Clone,

Update from iterator, return Err on error.