JakeHillion/scx
1
0
Fork 0
mirror of https://github.com/JakeHillion/scx.git synced 2024-11-29 12:40:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

scx_rustland: aggressively prioritize interactive tasks

Browse Source 
scx_rustland was originally designed as a PoC to showcase the benefits
of implementing specialized schedulers via sched_ext, focusing on a very
specific use case: prioritize game responsiveness regardless of what
runs in the background.

Its original design was subsequently modified to better serve as a
general-purpose scheduler, balancing the prioritization of interactive
tasks with CPU-intensive ones to prevent over-prioritization.

With scx_bpfland serving as a more "general-purpose" scheduler, it makes
sense to revisit scx_rustland's original goal and make it  much more
aggressive at prioritizing interactive tasks, determined in function of
their average amount of context switches.

This change makes scx_rustland again a really good PoC to showcase the
benefits of having specialized schedulers, by focusing only at a very
specific use case: provide a high and stable frames-per-second (fps)
while a kernel build is running in the background.

= Results =

 - Test: Run a WebGL application [1] while building the kernel (make -j32)
 - Hardware: 8-cores Intel 11th Gen Intel(R) Core(TM) i7-1195G7 @ 2.90GHz

  +----------------------+--------+--------+
  |      Scheduler       | avg fps|  stdev |
  +----------------------+--------+--------+
  |               EEVDF  |   28   |  4.00  |
  | scx_rustland-before  |   43   |  1.25  |
  |  scx_rustland-after  |   60   |  0.25  |
  +----------------------+--------+--------+

[1] https://webglsamples.org/aquarium/aquarium.html

Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
This commit is contained in:
Andrea Righi 2024-09-02 10:43:59 +02:00
parent 172fe1efc6
commit d0fb29a0f7
1 changed files with 41 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

69
scheds/rust/scx_rustland/src/main.rs
View File

@ -118,10 +118,13 @@ struct Opts {
const NSEC_PER_USEC: u64 = 1_000;
const NSEC_PER_SEC: u64 = 1_000_000_000;
// Maximum factor used to scale down the task's vruntime in function of its average amount of
// context switches per second.
const RL_MAX_VTIME_FACTOR: u64 = 4;
#[derive(Debug, Clone)]
struct TaskStat {
pid: i32,
comm: String,
nvcsw: u64,
}
@ -129,13 +132,19 @@ fn parse_proc_pid_stat(pid: i32) -> std::io::Result<TaskStat> {
let path = format!("/proc/{}/status", pid);
let content = std::fs::read_to_string(&path)?;
let mut comm = String::new();
let mut nvcsw = 0;
for line in content.lines() {
if line.starts_with("Name:") {
comm = line.split_whitespace().nth(1).unwrap_or("").to_string();
} else if line.starts_with("voluntary_ctxt_switches:") {
let name = line.split_whitespace().nth(1).unwrap_or("");
// Prioritize kthreads by assigning max nvcsw rate.
if name.starts_with('[') && name.ends_with(']') {
nvcsw = u64::MAX;
break;
}
}
if line.starts_with("voluntary_ctxt_switches:") {
nvcsw = line
.split_whitespace()
.nth(1)
@ -145,7 +154,7 @@ fn parse_proc_pid_stat(pid: i32) -> std::io::Result<TaskStat> {
}
}
Ok(TaskStat { pid, comm, nvcsw })
Ok(TaskStat { pid, nvcsw })
}
fn get_all_pids() -> std::io::Result<Vec<i32>> {
@ -196,7 +205,6 @@ struct Task {
qtask: QueuedTask, // queued task
vruntime: u64, // total vruntime (that determines the order how tasks are dispatched)
timestamp: u64, // task enqueue timestamp
is_interactive: bool, // task is interactive
}
// Sort tasks by their interactive status first (interactive tasks are always scheduled before
@ -204,10 +212,8 @@ struct Task {
// pid.
impl Ord for Task {
fn cmp(&self, other: &Self) -> std::cmp::Ordering {
other
.is_interactive
.cmp(&self.is_interactive)
.then_with(|| self.vruntime.cmp(&other.vruntime))
self.vruntime
.cmp(&other.vruntime)
.then_with(|| self.timestamp.cmp(&other.timestamp))
.then_with(|| self.qtask.pid.cmp(&other.qtask.pid))
}
@ -260,7 +266,7 @@ struct Scheduler<'a> {
task_pool: TaskTree, // tasks ordered by vruntime
task_map: TaskInfoMap, // map pids to the corresponding task information
proc_stats: HashMap<i32, u64>, // Task statistics from procfs
interactive_pids: Vec<i32>, // List of interactive tasks
interactive_pids: HashMap<i32, u64>, // Map of task PIDs to their avg_nvcsw metric
min_vruntime: u64, // Keep track of the minimum vruntime across all tasks
init_page_faults: u64, // Initial page faults counter
slice_ns: u64, // Default time slice (in ns)
@ -282,7 +288,7 @@ impl<'a> Scheduler<'a> {
task_pool: TaskTree::new(),
task_map: TaskInfoMap::new(),
proc_stats: HashMap::new(),
interactive_pids: Vec::new(),
interactive_pids: HashMap::new(),
min_vruntime: 0,
init_page_faults: 0,
slice_ns: opts.slice_us * NSEC_PER_USEC,
@ -363,8 +369,25 @@ impl<'a> Scheduler<'a> {
// Update total task cputime.
task_info.sum_exec_runtime = task.sum_exec_runtime;
// Evaluate the target vruntime for the task.
let mut vruntime = task_info.vruntime;
// Strongly prioritize interactive tasks by scaling their vruntime in function of their
// average amount of context switches.
//
// This essentially creates multiple priority lanes (up to RL_MAX_VTIME_FACTOR + 1), where
// tasks with a higher rate of average context are scheduled before others. Tasks with the
// same average context switch rate are still ordered by their weighted used time slice.
let avg_nvcsw = self.interactive_pids.get(&task.pid).copied().unwrap_or(0);
vruntime = vruntime / (avg_nvcsw.min(RL_MAX_VTIME_FACTOR) + 1) + slice * 100 / task.weight;
// Make sure to not de-prioritize tasks too much to prevent starvation.
if vruntime > self.min_vruntime + self.slice_ns {
vruntime = self.min_vruntime + self.slice_ns;
}
// Return the task vruntime.
task_info.vruntime
vruntime
}
// Drain all the tasks from the queued list, update their vruntime (Self::update_enqueued()),
@ -376,14 +399,12 @@ impl<'a> Scheduler<'a> {
// Update task information and determine vruntime.
let vruntime = self.update_enqueued(&task);
let timestamp = Self::now();
let is_interactive = self.interactive_pids.contains(&task.pid);
// Insert task in the task pool (ordered by vruntime).
self.task_pool.push(Task {
qtask: task,
vruntime,
timestamp,
is_interactive,
});
}
Ok(None) => {
@ -406,7 +427,7 @@ impl<'a> Scheduler<'a> {
}
// Dispatch the first task from the task pool (sending them to the BPF dispatcher).
fn dispatch_task(&mut self) {
fn dispatch_tasks(&mut self) {
match self.task_pool.pop() {
Some(task) => {
// Update global minimum vruntime.
@ -423,7 +444,7 @@ impl<'a> Scheduler<'a> {
// Create a new task to dispatch.
let mut dispatched_task = DispatchedTask::new(&task.qtask);
// Assign the time slice to the task.
// Assign the time slice to the task and propagate the vruntime.
dispatched_task.slice_ns = slice_ns;
// Try to pick an idle CPU for the task.
@ -457,7 +478,7 @@ impl<'a> Scheduler<'a> {
// and dispatch them to the BPF part via the dispatched list).
fn schedule(&mut self) {
self.drain_queued_tasks();
self.dispatch_task();
self.dispatch_tasks();
// Notify the dispatcher if there are still peding tasks to be processed,
self.bpf.notify_complete(self.task_pool.tasks.len() as u64);
@ -487,16 +508,9 @@ impl<'a> Scheduler<'a> {
}
fn sync_interactive_tasks(&mut self, stats: &[TaskStat]) {
self.interactive_pids.clear();
for i in 0..stats.len() {
let stat = &stats[i];
// At least 10 context switches per sec are required to consider the
// task as interactive.
if stat.nvcsw >= 10 {
self.interactive_pids.push(stat.pid);
}
self.interactive_pids.insert(stat.pid, stat.nvcsw);
}
}
@ -506,7 +520,7 @@ impl<'a> Scheduler<'a> {
for pid in get_all_pids()? {
if let Ok(stat) = parse_proc_pid_stat(pid) {
// Retrieve the previous nvcsw value, or 0 if not present.
let prev_nvcsw = self.proc_stats.get(&stat.pid).copied().unwrap_or_default();
let prev_nvcsw = self.proc_stats.get(&stat.pid).copied().unwrap_or(0);
// Update the proc_stats entry with the new nvcsw.
self.proc_stats.insert(stat.pid, stat.nvcsw);
@ -518,7 +532,6 @@ impl<'a> Scheduler<'a> {
let delta_nvcsw = stat.nvcsw.saturating_sub(prev_nvcsw);
new_stats.push(TaskStat {
pid: stat.pid,
comm: stat.comm,
nvcsw: delta_nvcsw,
});
}