If a task is performance-critical, pick_idle_cpu() checks if the
previous core is a big core or not. If not, don't try to run on previous
core since a performance-critical task is better to run on a big core.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
A single threshold for a low watermark does not work well across systems
with various numbers of cores and core types. Instead of using a single
low watermark value, we dynamically decide the low watermark: 1) until
one little core is fully utilized or 2) until two big cores are fully
utilized. This works better across systems.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
The scx_rustland_core API has been redesigned recently, breaking the
compatibility with the past.
Considering that Rust crates should update their major version when the
previous API becomes incompatible [1], bump up the version to 2.0.0.
[1] https://semver.org/
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
We want to directly dispatch only kthreads when local_kthreads is
enabled, not all tasks that can run on a single CPU.
Fixes: 7cc1846 ("scx_bpfland: always rely on prev_cpu with single-CPU tasks")
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
When the no_freq_scaling changes during runtime in the autopilot mode,
the last target freq set would not be 1024. So the performance mode
enabled by the autopilot mode would not run in the best profile. Hence,
we set the target freq to 1024 always when no_freq_scaling is set.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
This allows scx_rustland to avoid generating excessive logs for
statistics while still allowing detailed monitoring on demand.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Add "--autopilot" option and mode. In the autopilot mode, the scheduler
dynamically changes its power mode according to system's load (cpu
utilization). When the cpu utilization is low enough (say <=5%), it
switches to the powersave mode since there is nothing to process fast so
powersaving is the primary goal. When the utilization is moderate (say
>5%, <=30%), it runs in balanced mode. When the utilization is high
enough (say >30%), it runs in performance mode.
Note that it only changes scheduler's power mode but it does not change
system's energy profile.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
When a cpu is idle for a whole interval, its idle time does not
correctlyh adds up so the utilization of such cpu tends to be higher
than the actual utilization. Now it is fixedk, so cpu utilization
becomes more accurate.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
When the power mode changes back to performance mode, we should
active/overflow cpumask to its initial state -- all big cores are in
active cpumask and all little cores are in overflow cpumask. Otherwise,
the active/overflow cpumasks will be used in the perfformance mode.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
If a task can be run only on a single cpu, we don't need to go through
all the steps in ops.select_cpu(). Instread, we simply check if a task
is still pinned on the prev_cpu and go.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
When selecting an idle for tasks that can only run on a single CPU,
always check if the previously used CPU is sill usable, instead of
trying to figure out the single allowed CPU looking at the task's
cpumask.
Apparently, single-CPU tasks can report a prev_cpu that is not in the
allowed cpumask when they rapidly change affinity.
This could lead to stalls, because we may end up dispatching the kthread
to a per-CPU DSQ that is not compatible with its allowed cpumask.
Example:
kworker/u32:2[173797] triggered exit kind 1026:
runnable task stall (kworker/2:1[70] failed to run for 7.552s)
...
R kworker/2:1[70] -7552ms
scx_state/flags=3/0x9 dsq_flags=0x1 ops_state/qseq=0/0
sticky/holding_cpu=-1/-1 dsq_id=0x8 dsq_vtime=234483011369
cpus=04
In this case kworker/2 can only run on CPU #2 (cpus=0x4), but it's
dispatched to dsq_id=0x8, that can only be consumed by CPU 8 => stall.
To prevent this, do not try to figure out the best idle CPU for tasks
that are changing affinity and just dispatch them to a global DSQ
(either priority or regular, depending on its interactive state).
Moreover, introduce an explicit error check in dispatch_direct_cpu() to
improve detection of similar issues in the future, and drop
lookup_task_ctx() in favor of try_lookup_task_ctx(), since we can now
safely handle all the cases where the task context is not found.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
It checkes the EPP (energy performance preference) peirodically and sets
the power profile of the scheduler during runtiime as a user changes its
EPP profile (from her desktop UI).
Signed-off-by: Changwoo Min <changwoo@igalia.com>
Add a per layer config for different implementations of layer growth
algorithms. Convert the existing default logic into a default layer
growth algorithm and add a linear implementation.
Signed-off-by: Daniel Hodges <hodges.daniel.scott@gmail.com>
Refactor some BPF code to make verification easier on older kernels.
This is to make it easier to maintain backports.
Signed-off-by: Daniel Hodges <hodges.daniel.scott@gmail.com>
Aggressively try to keep tasks running on the same CPU / cache / domain,
to achieve higher performance when the system is not over commissioned.
This is done by giving a second chance in ops.enqueue(), in addition to
ops.select_cpu(), to find an idle CPU close to the previously used CPU.
Moreover, even if the task is dispatched to the global DSQs, always try
to check if there is an idle CPU in the primary domain that can
immediately consume the task.
= Results =
This change seems to provide a minor, but consistent, boost of
performance with the CPU-intensive benchmarks from the CachyOS
benchmarks selection [1].
Similar results can also be noticed with some WebGL benchmarks [2], when
system usage is close to its maximum capacity.
Test:
- cachyos-benchmarker
System:
- AMD Ryzen 7 5800X 8-Core Processor
Metrics:
- total time: elapsed time of all benchmarks
- total score: geometric mean of all benchmarks
NOTE: total time is the most relevant, since it gives a measure of the
aggregate performance, while the total score emphasizes more on
performance consistency across all benchmarks.
== Results: summary ==
+-------------------------+---------------------+---------------------+
| Scheduler | Total Time | Total Score |
| | (less = better) | (less = better) |
+-------------------------+---------------------+---------------------+
| EEVDF | 624.44 sec | 123.68 |
| bpfland | 625.34 sec | 122.21 |
| bpfland-task-affinity | 623.67 sec | 122.27 |
+-------------------------+---------------------+---------------------+
== Conclusion ==
With this patch applied, bpfland shows both a better performance and
consistency. Although the gains are small (less than 1%), they are still
significant for this type of benchmark and consistently appear across
multiple runs.
[1] https://github.com/CachyOS/cachyos-benchmarker
[2] https://webglsamples.org/aquarium/aquarium.html
Tested-by: Piotr Gorski < piotr.gorski@cachyos.org >
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
When iterating neighbors, the existing code unnecessarily iterates all
the neighbors to the maximum even if there is no neighors. So the fix
escapes early when there is no neighbors.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
Ctrl-c wasn't properly handled in the monitoring mode
(`--monitor-sched-samples`), so the scheduler could not be terminated by
pressing ctrl-c. The missing ctrl-c handling is added to the monitor
thread.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
Rely on scx_utils::Topology to classify Big, Little and Turbo CPUs.
Moreover, support the special keyword "all" with --primary-domain to
include all the CPUs in the system (default).
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Integrate the logic used by scx_bpfland to detect turbo-boosted cores in
Topology.
Also change the logic to detect Big/Little cores in function of
base_frequency, instead of scaling_max_freq, otherwise turbo-boosted
cores in homogeneous systems may be incorrectly classified as Big.
Moreover, introduce the following new methods to Cpu to check for the
core type:
- is_turbo(): return true if the CPU is Turbo, false otherwise
- is_big(): return true if the CPU is either Turbo or Big
- is_little(): return true if the CPU is Little
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
When creating the turbo boost scheduling domain, we might use a full CPU
mask (selecting all possible CPUs) to indicate "do not prioritize turbo
boost CPUs" or when all CPUs have the same maximum frequency.
This approach works when the primary domain also contains all the CPUs,
as the complete overlap allows the CPU selection logic to ignore the
turbo boost domain and start picking CPUs directly from the primary
domain.
However, if the primary domain doesn't include all CPUs, the two domains
won't fully overlap, which can lead to the turbo boost domain
incorrectly including all CPUs, thereby negating the restrictions set by
the primary scheduling domain.
To resolve this, an empty CPU mask should be used for the turbo boost
domain when turbo boost CPUs aren't prioritized. If the turbo boost
domain is empty, it should be entirely bypassed, and the selection
should proceed directly to the primary domain.
Reported-by: Changwoo Min <changwoo@igalia.com>
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
With an ill combination of old kernel and old LLVM, the BPF verifier
incorrectly detects an infinite loop. After changing the loop with a
constant end, the old verifier can pass the code.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
Refactor the code design to make it more suitable as a template for
implementing advanced scheduling policies.
In particular, create separate loops for task consumption and task
dispatching. This will make the scheduler easier to adapt as a
foundation for implementing more complex scheduling policies.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Now it checks an active cpumask within a previous core's compute domain
before checking the full active CPUs.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
The BPF verifier in the old kernel gives up to analysis the nested loop
in the consume_task(). We reduce the loop less complex by reducing
LAVD_CPDOM_MAX_DIST from 6 to 4 in order to make the verifier happy.
Note that the theoretical maximum distance is 6 (numa > llc > core type)
but there is no such hardware today, hence reducing it to 6 should be
okay in next few years, when hopefully the verifier becomes smarter.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
Updating nr_queue_task every runqueue operation is expensive and
unnecessary. So we do update every system state update interval and use
moving average, which is accurate enough.
Signed-off-by: Changwoo Min <changwoo@igalia.com>
Avoid to periodically read the current performance profile from
/sys/devices/system/cpu/cpufreq/policy0/energy_performance_preference if
it's not available (i.e., with older CPUs or kernels without cpufreq).
This fixes issue #560.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
meson build script was building each rust sub-project under rust/ and
scheds/rust/ separately. This means that each rust project is built
independently which leads to a couple problems - 1. There are a lot of
shared dependencies but they have to be built over and over again for each
proejct. 2. Concurrency management becomes sad - we either have to unleash
multiple cargo builds at the same time possibly thrashing the system or
build one by one.
We've been trying to solve this from meson side in vain. Thankfully, in
issue #546, @vimproved suggested using cargo workspace which makes the
sub-projects share the same target directory and built together by the same
cargo instance while still allowing each project to behave independently for
development and publishing purposes.
Make the following changes:
- Create two cargo workspaces - one under rust/, the other under
scheds/rust/. Each contains all rust projects underneath it.
- Don't let meson descend into rust/. These are libraries used by the rust
schedulers. No need to build them from meson. Cargo will build them as
needed.
- Change the rust_scheds build target to invoke `cargo build` in
scheds/rust/ and let cargo do its thing.
- Remove per-scheduler meson.build files and instead generate custom_targets
in scheds/rust/meson.build which invokes `cargo build -p $SCHED`.
- This changes rust binary directory. Update README and
meson-scripts/install_rust_user_scheds accordingly.
- Remove per-scheduler Cargo.lock as scheds/rust/Cargo.lock is shared by all
schedulers now.
- Unify .gitignore handling.
The followings are build times on Ryzen 3975W:
Before:
________________________________________________________
Executed in 165.93 secs fish external
usr time 40.55 mins 2.71 millis 40.55 mins
sys time 3.34 mins 36.40 millis 3.34 mins
After:
________________________________________________________
Executed in 36.04 secs fish external
usr time 336.42 secs 0.00 millis 336.42 secs
sys time 36.65 secs 43.95 millis 36.61 secs
Wallclock time is reduced 5x and CPU time 7x.
Refactor the code to hide the shutdown handling inside BpfScheduler and
simply use the exited() method to check when the scheduler is stopped.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Three of the reported stats are cumulative. While they obviously can be
processed into delta values, that holds for the other direction too and the
cumulative values are difficult to make intutive sense of. Report interval
delta values instead.
Note that a stats client can reliably build back cumulative values even
under heavy system contention - the delta values reported between two
consecutive reads are guaranteed to be correct regardless of the duration of
the interval.
Use scx_stats instead of prometheus for stats reporting. This has a few
advantages:
- Stats metadata can be defined more succinctly.
- Natural support for nesting statistics which will be useful in making
scheduler components composable.
- Support for multiple programmable readers where each reader can use their
own reading interval.
- Built-in stats help message generation.
- Openmetrics integration is still available through
scx_stats/scripts/scxstats_to_openmetrics.py.
Let's make it a bit easier to use:
- Shorten exported names by changing the prefix from ScxStats to Stats. This
should be distinctive enough and more inline with how most libraries name
their exports.
- Importing the right set of traits can be tricky. Introduce prelude module
so that importing is a bit less painful.
There is no reason to have two separate options for "verbose" and
"debug" mode. Just merge the two and always use "debug". If enabled,
increase verbosity to stdout and enable reporting BPF scheduling events
in debugfs (e.g., /sys/kernel/debug/tracing/trace_pipe).
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Since scx_rustland_core enables setting a time slice on a per-task basis
during task dispatch, there's no need to maintain a global time slice in
the BPF component. Instead, a global time slice can simply be managed in
user-space, achieving the same outcome.
Therefore, drop the global slice_us property from BpfScheduler to
simplify the API.
NOTE: if a time slice is not specified for a task, SCX_SLICE_DFL will be
used by default.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Add more comments to make the source code more understandable, so that
it can be easily used as a template for implementing more complex
scheduling policies.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Scheduling sample reporting is switched to use scx_stats. This makes the
scheduler run without making too much noise while still allowing monitoring
on demand. It can also make introspection more dynamic - e.g. it shouldn't
be difficult to add other monitoring commands which take scheduling samples
based on different criteria or add other types of staisitcs.
--nr_sched-samples is replaced with --monitor-nr-samples.
The update_tasks() API is somewhat confusing, so replace it with a
clearer API, notify_complete().
This new API will return control to the BPF component and inform it
about the number of tasks still pending in the user-space scheduler.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
The low-power API is a bit of a hack implemented purely in the BPF
layer, this should be better re-implemented with some concepts of
topology awareness.
Therefore, get rid of this API for now.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
The current API used to notify the user-space scheduler when a task
exits is really confusing (setting a negative value in
queued_task_ctx.cpu), and it's also possible to detect task exiting
events from user-space (or check in procfs, even if it's slower).
In any case, a better API should be provided for this, so drop the
current one for now.
NOTE: this will cause additional memory usage for scx_rustland, but it
can be fixed/addressed later in a separate commit (i.e., providing a
periodic garbage collector for the unused task entries).
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Allow user-space scheduler to pick an idle CPU via
self.bpf.select_cpu(pid, prev_task, flags), mimicking the BPF's
select_cpu() iterface.
Also remove the full_user option and always rely on the idle selection
logic from user-space.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Keep evaluating the average number of voluntary context switches for
each task when lowlatency mode is enabled, even when interactive tasks
classification is disabled (via `-c 0`).
The average nvcsw is also used in lowlatency mode to evaluate the
proportional bonus to the tasks' deadline and it shouldn't be ignored
when interactive tasks classification is disabled. Moreover, make sure
that such bonus never exceeds the starvation threshold.
Keep in mind that it is still possible to disable the periodic average
nvcsw evaluation with `-c 0`, without specifying `--lowlatency`.
Fixes: 6a22853 ("scx_bpfland: introduce --lowlatency option")
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
A lot of scx_lavd's options do not clearly explain what they do. Add
some short explanations, clean up the existing ones, and direct the user
to read the in-code documentation for more info.
And move related ops into it. This is a bit more natural and will also allow
doing other operaitons (e.g. describing stats) without launching the server.
Make `--primar-domain auto` aware of turbo boosted CPUs and prioritize
them over the primary scheduling domain when the energy model
`balance_power` is used (typically when running on battery power with
the "balanced" profile).
With this change the scheduling hierarchy becomes the following:
1) CPUs in the turbo scheduling domain
2) CPUs in the primary scheduling domain
3) full-idle SMT CPUs
4) CPUs in the same L2 cache
5) CPUs in the same L3 cache
6) CPUs in the task's allowed domain
And the idle selection logic is modified as following:
- In the turbo scheduling domain:
- pick same full-idle SMT CPU
- pick any other full-idle SMT CPU sharing the same L2 cache
- pick any other full-idle SMT CPU sharing the same L3 cache
- pick any other full-idle SMT CPU
- pick same idle CPU
- pick any other idle CPU sharing the same L2 cache
- pick any other idle CPU sharing the same L3 cache
- pick any other idle SMT CPU
- In the primary scheduling domain:
- pick same full-idle SMT CPU
- pick any other full-idle SMT CPU sharing the same L2 cache
- pick any other full-idle SMT CPU sharing the same L3 cache
- pick any other full-idle SMT CPU
- pick same idle CPU
- pick any other idle CPU sharing the same L2 cache
- pick any other idle CPU sharing the same L3 cache
- pick any other idle SMT CPU
- In the entire task domain:
- pick any other idle CPU
Keep in mind that the turbo domain will be evaluated only when the
scheduler is started with `--primary-domain auto` and only when the
`balance_power` energy profile is used.
The turbo domain is always made using the subset of CPUs in the system
with the highest max frequency. If such subset can't be determined (for
example if all the CPUs in the primary domain have all the same
frequency), the turbo domain will be ignored.
Prioritizing turbo boosted CPUs can help to improve performance by
forcing the governor to scale up their frequency, without increasing too
much power consumption, due to the fact that tasks will be preferably
confined into a reduced amount of cores.
This change seems to improve performance, without increasing much
power consuption, on Intel laptops while using the `balanced_power`
energy profile.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Introduce the new option `--primary-domain auto`. With this option the
scheduler will dynamically adjusts the primary scheduling domain at
run-time, in function of the current energy profile reported in
/sys/devices/system/cpu/cpufreq/policy0/energy_performance_preference.
When the `power` energy profile is selected, the primary scheduling
domain will prioritize E-cores. Alternatively, when the `performance`
profile is selected, it will prioritize P-cores. For all the other
energy profiles, all the CPUs in the system will be used.
Note that this option is only relevant on hybrid architectures with
P-cores and E-cores.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Introduce the new `--lowlatency` option, which enables switching between
the default pure vruntime-based scheduling (more optimized for server
workloads) and a deadline-based scheduling (better suited for
low-latency workloads).
When the low-latency mode is activated, a task's deadline is calculated
as its vruntime, adjusted by a bonus proportional to the task's average
number of voluntary context switches (the more voluntary context
switches, the shorter the deadline).
This feature enhances the prioritization of interactive tasks even more,
proportionally to their average voluntary context switches, also within
the two main global queues (priority / shared) and it helps to maintain
interactive workloads always responsive, even in presence of heavy
non-interactive background work.
Low-latency mode allows to prevent audio cracking even in presence of a
large amount of short-lived tasks with pseudo-interactive behavior (i.e,
hackbench) and it enables achieving approximately a +33% average
frames-per-second (FPS) in the typical "gaming while building the
kernel" benchmark.
However, it can also amplify the de-prioritization of CPU-intensive
tasks, making this option more suitable for specific low-latency
scenarios. Therefore the low-latency mode is disabled by default and it
can only be enabled via the `--lowlatency` option.
Tested-by: Piotr Gorski (piotrgorski@cachyos.org)
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Explicitly replenish the task's time slice from ops.dispatch() if the
task still wants to run and no other task is selected. In this way the
sched_ext core won't automatically re-schedule the task on the same CPU,
implicitly assigning a time slice of SCX_SLICE_DFL.
Moreover, instead of determining the task time slice in ops.enqueue(),
refresh the time slice immediately before the task is started on its
assigned CPU in ops.running().
This allows to use a more precise time slice, adjusted based on the
actual amount of tasks that are currently waiting to be scheduled.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
The meaning of SCX_OPS_ENQ_LAST will change with future kernel updates and
enqueueing on local DSQ will no longer be sufficient to avoid stalls. No
reason to do it anyway. Just drop it.
With the global scx_utils::NR_CPU_IDS we don't need Topology anymore in
init_primary_domain(), so drop the variable to fix the following build
warning:
warning: unused variable: `topo`
--> src/main.rs:385:9
|
385 | topo: &Topology,
| ^^^^ help: if this is intentional, prefix it with an underscore: `_topo`
|
= note: `#[warn(unused_variables)]` on by default
Fixes: 1da249f ("scx_utils::topology: Always use NR_CPU_IDS and NR_CPUS_POSSIBLE")
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Use the base frequency, instead of maximum frequency, to classify fast
and slow CPUs. This ensures accurate distinction between Intel Turbo
Boost CPUs and genuinely faster CPUs when auto-detecting the primary
scheduling domain.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
- Update scx_utils/build.rs so that 12 char SHA1 is generated instead of
full one.
- Add --version to scx_rusty. Use custom one as we don't want to use the
default cargo version one.
Tasks enqueued with SCX_ENQ_WAKEUP are immediately classified as
interactive. However, if interactive tasks classification is disabled
(via `-c 0`), we should avoid promoting them as interactive.
This is particularly important because, with the nvcsw logic disabled,
tasks can remain classified as interactive indefinitely and they will
never be demoted to regular tasks.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Rely on scx_utils::Cpumask instead of re-implementing a custom struct to
parse and manage CPU masks.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
Fix a bug introduced in #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 <hodges.daniel.scott@gmail.com>
Rely on scx_utils::Topology to get CPU and cache information, instead of
re-implementing custom methods.
Signed-off-by: Andrea Righi <andrea.righi@linux.dev>
