Adding slot_hits + slot_displacements in RFC 005 grew SchedulerStats
from 16 to 32 bytes — exactly 2 per cache line. run_queue_len (written
on every push/pop by thread N) then false-shared with fields of thread
N+1, causing ~45-60% yield-storm regression at t≥8 visible equally in
slot-on and slot-off paths.
A thread-local Cell<Option<Pid>> per scheduler, checked before the shared
run queue. Runtime-selected via Config { wake_slot: bool }, default OFF
until the slot shootout accepts it (one binary benches both arms).
Push policy: slot-eligible iff the wake originates from actor context
(current_pid().is_some()) — the slot push replaces run_queue.push at the
tail of the unpark protocol's Parked → Queued CAS, so at-most-once-enqueued
holds verbatim as (slot ⊕ shared queue). Scheduler-context wakes (timer/IO
drain) and spawns always go shared (spawns never reach unpark_inner at all).
Displacement is Go semantics: newest wake takes the slot, occupant pushed
shared — moved, never copied.
Pop order: slot, then shared. Slot-popped actors skip reset_timeslice() and
inherit the waker's remaining slice; a handoff chain is bounded by one
slice, after which the preempt-yield re-enqueue goes shared (a yield is not
a wake) — the one-slice starvation bound, zero new counters. Idle and
AllDone are only reachable with an empty local slot by pop order; an
occupied slot elsewhere holds a Queued (live) actor, so the counter-first
termination argument is untouched.
Observability: per-thread slot_hits / slot_displacements (reset at run()
start so post-run stats() reads are per-run), SlotPush/SlotPop trace events.
Bench plan (roadmap v0.9 item 2): rq_runtime gains the slot on/off
dimension (SMARM_BENCH_SLOT, default "0 1") — ping-pong-pairs is the
target metric, yield-storm the regression guard, spawn-storm the
neutrality check. RQCSV grows a slot column; RQSLOT lines carry the
counters; bench_rq.sh aggregates both. Tests pin the push policy through
the counters (actor-context hits, spawn/join bypass, displacement,
default-off, per-run reset).
benches/README.md: catalog of the two bench families (cross-runtime
comparisons + the v0.5 run-queue shootout), how to run the shootout
(scripts/bench_rq.sh, SMARM_BENCH_* knobs, the --no-default-features dance,
summary.csv/RQCSV format), honest-numbers caveats (core count is the
experiment; striped losing at low thread counts is expected), and the
conventions for adding a bench.
tests/README.md: catalog grouped by layer (low-level units / feature areas /
regression+stress), the run matrix (debug-first — that's where the invariant
asserts live — three queue variants, release, loom, trace), why loom tests
live in src/ rather than tests/, and the house conventions: one runtime per
test, oversubscription on purpose, regression tests validated against the
reintroduced bug, odds-stacking for stochastic tests, ordering-not-duration
time assertions, and new-invariant = assert + loom model.
Also fixes a stale header in tests/mutex.rs that claimed 'loom::Mutex' —
it tests smarm::Mutex, and the old name is actively confusing now that loom
is real in this repo.
- benches/rq_micro.rs: raw-structure microbench, threads x producer:consumer
ratio sweep. Benches all three queue types in one binary (they compile in
every build; only the runtime alias is feature-selected), so no rebuild
dance. Queues sized to the op count so the occupancy contract is met.
- benches/rq_runtime.rs: whole-runtime benches with the selected variant:
yield-storm (pure queue churn), ping-pong-pairs (park/unpark latency),
spawn-storm (slab + free list + queue under churn). Scheduler-count sweep.
- scripts/bench_rq.sh: rebuilds rq_runtime per rq-* feature, runs rq_micro
once, aggregates RQCSV lines into bench_results/summary.csv.
- All knobs via SMARM_BENCH_* env vars; house table format + machine lines.
- run_queue module is now #[doc(hidden)] pub (types + push/pop/len +
MpmcRing::with_capacity) solely so the external bench binary can drive the
raw structures.
docs(roadmap): phase 4 ticked (harness done; numbers from the 20-core box).
New fast-follow per review: assert the invariants we lean on — debug_assert!
on hot paths, loud assert!/panic on cold ones, at the point of reliance;
sweep existing code during the phase-5 audit, adopt as house style.
Validated end-to-end at smoke scale on the 1-core sandbox: full driver run,
24-row summary.csv across micro (3 structures x sweeps) and runtime
(3 variants x 3 benches x thread sweep).
While running benchmarks, a hang surfaced in timer-only workloads — actors
sleeping with no IO in flight. Tracing it down, the race lives in sleep():
between the call to `timers.insert_sleep` and the subsequent `park_current`
yield, the actor is still in State::Runnable. If the timer fires in that
window, the old code's `if matches!(slot.state, State::Parked)` guard silently
drops the wakeup. The actor then parks normally and never gets re-queued —
it sleeps forever.
The fix mirrors what scheduler::unpark() and the IO FdReady path already do:
when the timer fires and the slot is still Runnable, set `pending_unpark`
instead of re-queuing immediately. The upcoming Park yield sees the flag and
re-queues the actor rather than suspending it, closing the race without any
new synchronisation.
Adds a regression test: 100 actors doing pure timer sleeps across ≥2
scheduler threads. The test asserts both correctness (all actors complete)
and timeliness (wall time < 2×sleep duration), which is enough signal to
catch a stuck actor even on a single-core CI runner.
Three changes, each independently measured, landed together.
Fuse pre-resume mutex acquisitions
-----------------------------------
The schedule loop previously took the shared lock three separate times
per actor resume: once to pop the run queue, once to read the stack
pointer, and once to pop the first-resume closure. These are now a
single acquisition that returns everything needed to resume the actor.
As a side effect, pending_closures was changed from a Vec<(Pid, Closure)>
with O(n) linear scan to a Vec<Option<Closure>> indexed by slot index,
making first-closure lookup O(1).
Move stack allocation outside the shared lock
----------------------------------------------
Stack::new() (mmap + mprotect) was previously called inside with_shared,
stalling every other scheduler thread for the duration of two syscalls on
every spawn. It now runs before the lock is acquired.
Stack pool
----------
Rather than munmap-ing a stack when an actor finishes and mmap-ing a fresh
one on the next spawn, stacks are now recycled through a per-Runtime pool
(Mutex<Vec<Stack>>). finalize_actor extracts the stack from the Actor
before clearing the slot and pushes it to the pool outside the shared lock.
spawn_under pops from the pool before falling back to Stack::new().
The pool is unbounded for now (shrink policy TBD) but capped at
stack_pool_cap stacks on return, defaulting to thread_count * 4.
The cap is configurable via Config::stack_pool_cap(n).
Results (24-thread, against stored baseline)
--------------------------------------------
chained_spawn smarm 1-thread: 9763 → 261 µs (-97%)
chained_spawn smarm 24-thread: 23562 → 838 µs (-96%)
ping_pong_oneshot smarm 1-thread: 18409 → 742 µs (-96%)
ping_pong_oneshot smarm 24-thread: 44596 → 1425 µs (-97%)
catch_unwind_panics smarm 24-thread: 267812 → 124094 µs (-54%)
fan_out_compute smarm 24-thread: 2839 → 2226 µs (-22%)
Tokio regressions in the checker output are baseline measurement drift;
no tokio code was changed.
Config API changes (src/preempt.rs, src/runtime.rs):
- preempt: promote ALLOC_INTERVAL and TIMESLICE_CYCLES from bare consts to
DEFAULT_ALLOC_INTERVAL / DEFAULT_TIMESLICE_CYCLES; store active values in
thread-locals set on each actor resume so multiple runtimes can use
different settings concurrently.
- runtime: add alloc_interval / timeslice_cycles fields to Config; add
Config::alloc_interval(n) and Config::timeslice_cycles(c) builder methods;
thread the values through RuntimeInner to the reset_timeslice() call in
schedule_loop.
Bench changes:
- Add bench_cfg(threads) helper to general/tokio_favored/smarm_favored that
wraps Config::exact and reads SMARM_ALLOC_INTERVAL / SMARM_TIMESLICE_CYCLES
env vars, so the sweep script can vary knobs without recompiling.
Sweep tooling (benches/sweep.py):
- 'run': run the 3-file bench suite once; --save-baseline persists JSON
- 'regress': compare current run against baseline.json, exit 1 on any bench
that regresses >10% vs stored medians
- 'sweep': run the full SWEEP_GRID (10 points), print comparison table,
optional --save-csv; binaries pre-built so no recompile per point
Sweep results (10-point grid, 1-CPU sandbox):
- The preemption knobs have very little effect on this single-CPU machine.
Most benches move <5% across the entire grid.
- Longer timeslices (tc=600k, tc=1200k) reliably hurt spawn_storm_busy
(+11-15%) and catch_unwind_panics (+10-12%) because actors hold the
scheduler mutex longer per timeslice, stalling the storm of joinable tasks.
- Shorter timeslices (tc=150k) give a small improvement on many_timers
(-3-4%) and a wash everywhere else.
- yield_in_hot_loop and uncontended_channel are essentially flat across all
knobs — both are scheduling-dominated and call yield_now explicitly, so
the RDTSC-driven preemption path is irrelevant.
- Conclusion: the knobs matter primarily under contention (multi-core).
Re-run sweep on a multi-core machine before drawing tuning conclusions.
Hand-rolled context switching on mmap'd stacks with guard pages,
allocator-driven RDTSC preemption, unbounded MPSC channels, supervision
via per-slot Signal mailboxes, root supervisor as sentinel PID.
Lib + tests + benches clean check/clippy. All 29 tests pass.
Bench: smarm 3.4% over serial baseline, within 160us of tokio
current-thread on prime-counting fan-out.
smarm