Commit Graph
11 Commits
Author SHA1 Message Date
smarm-dev d9a6520a24 feat(cancel): cooperative cancellation via sentinel unwind
request_stop(pid) sets a per-actor flag and wakes a parked target. The
actor realizes the stop as a controlled unwind at its next observation
point (check!()/alloc via maybe_preempt, or the wakeup side of any
blocking park): a StopSentinel panic tears the stack down via the
trampoline's existing catch_unwind, running Drop, and is reported as the
new Outcome::Stopped (distinct from a user Panic).

Death surface kept distinct from Exit: Signal::Stopped(pid) +
DownReason::Stopped, so the supervisor await logic to come can tell a
requested stop apart from a self-termination.

Flag lives on Actor behind Arc<AtomicBool>; the scheduler resume path
takes a raw pointer into it (no per-resume refcount traffic), keeping
yield throughput at baseline.
2026-06-07 21:51:56 +00:00
smarm-dev 09236b8bf4 feat(supervisor): one-for-one supervisor with restart policies + intensity cap
Builds on the existing supervisor_channel funnel (one mailbox per
supervisor; better than N monitor channels given there is no select).

- supervisor.rs: Restart{Permanent,Transient,Temporary}, ChildSpec (an
  Fn factory so children can be re-instantiated), and OneForOne with a
  builder (child/intensity) and run() supervision loop. Restart decision
  keys off whether the Signal was a panic; a sliding-window intensity cap
  stops crash loops and returns instead of spinning.
- Does NOT forcibly terminate children: shared-heap + Drop make async
  teardown of a peer unsound, so one_for_all/rest_for_one and true links
  wait on a cooperative-cancellation primitive. Cap-trip just stops
  restarting; live children are left as-is.

tests/supervisor.rs: transient restart-then-settle, transient/temporary
no-restart, permanent crash-loop hitting the cap, and one-for-one
isolation across two children. Full suite green.
2026-06-07 21:51:56 +00:00
smarm-dev 461de2c451 feat(monitor): unidirectional one-shot process monitors
`monitor(pid) -> Receiver<Down>` lets any actor watch any other and
receive a single Down{pid, reason} when it terminates. Generalizes the
existing single supervisor_channel (which is just a hard-wired monitor
the parent installs at spawn).

- src/monitor.rs: Down / DownReason{Exit,Panic,NoProc} + monitor().
  Payload-free: a panic payload has one owner and goes to the joiner via
  JoinError, so monitors learn only that it panicked. Monitoring an
  already-gone pid yields NoProc immediately.
- runtime: Slot grows `monitors: Vec<Sender<Down>>`; finalize_actor
  drains and notifies them outside the shared lock (send can unpark a
  receiver, which re-takes the non-reentrant mutex). Cleared in
  reclaim_slot and on slot reuse at spawn.
- Registration and finalize both serialize on the shared mutex, so a
  target alive at registration always delivers a real Down (no race
  between liveness check and registration).

tests/monitor.rs: Exit, Panic, NoProc-on-dead-target, and fan-out to
multiple monitors. Full suite green.
2026-06-07 21:51:56 +00:00
Claude 978678a46e feat: full runtime redesign (v0.6)
Complete rewrite with improved architecture & correctness:
- src/runtime.rs: Simplified task scheduling with proper state transitions
- src/scheduler.rs: Decoupled from runtime, pure task queue logic
- src/io.rs, src/mutex.rs: Refactored for clarity & performance
- New actor model framework (src/actor.rs, src/context.rs)
- Channel primitives (src/channel.rs) & process IDs (src/pid.rs)
- Preemption framework (src/preempt.rs) for fair timeslicing
- Expanded benchmarks & tests (multi_scheduler, primes, runtime)
2026-05-23 16:09:35 +00:00
Claude 078447539c chore: reset working tree (v0.5)
Temporary commit clearing working tree for v0.6 rebuild
2026-05-23 16:09:35 +00:00
Claude e9fdbb1160 refactor: centralize runtime logic (v0.4)
Extract scheduler responsibilities into a dedicated Runtime component:
- src/runtime.rs: New centralized control flow (669 lines)
- src/scheduler.rs: Simplified to task queue & preemption management
- tests/runtime.rs: Comprehensive runtime test suite
- benches/multi_scheduler.rs: Multi-runtime scheduling benchmarks
- Improves modularity and enables per-runtime configuration
2026-05-23 16:09:32 +00:00
Claude 8cbef1dfc1 feat: I/O and mutex support (v0.3)
Add epoll-based non-blocking I/O and kernel-like mutexes:
- src/io.rs: Complete epoll backend with timeout & error handling
- src/mutex.rs: Fair mutex with waiter queues & parking integration
- Enhanced scheduler to support synchronous I/O blocking
- Comprehensive test suites for I/O (epoll) and mutex behavior
- Documentation: LOOM.md concurrency model & README
2026-05-23 16:09:29 +00:00
Claude d3ab81b833 preempt: explicit check!() macro for no-alloc loops
Stable Rust emits stack probes inline (subq/movq/jne loop) rather than
calling __rust_probestack, so there's no transparent hook for stack-
frame preemption. Override of __rust_probestack links cleanly but never
runs. Falling back to an explicit check!() that users drop into hot
compute loops.

check!() decrements the same ALLOC_COUNT counter as the heap path, so
both event sources fire timeslice checks at the same rate. Documents
the prep-to-park invariant on maybe_preempt — library code that
registers a wakeup and then parks must keep that window alloc-free and
check-free, or a preemption-driven yield in the middle would lose the
wakeup.
2026-05-22 05:37:04 +00:00
Claude 51bfccc3c2 feat: I/O and mutex support (v0.3)
Add epoll-based non-blocking I/O and kernel-like mutexes:
- src/io.rs: Complete epoll backend with timeout & error handling
- src/mutex.rs: Fair mutex with waiter queues & parking integration
- Enhanced scheduler to support synchronous I/O blocking
- Comprehensive test suites for I/O (epoll) and mutex behavior
- Documentation: LOOM.md concurrency model & README
2026-05-22 05:32:24 +00:00
Claude 2cf75febdc timer: sleep(duration) via min-heap of (deadline, pid)
Adds a BinaryHeap of timer entries on SchedulerState. sleep() inserts
an entry and parks; schedule_loop pops due entries each iteration and
unparks them. When the run queue is empty but timers are pending, the
OS thread sleeps until the soonest deadline.

Single-threaded only; thread::sleep is fine because no other thread
can wake us. The IO thread coming next will need a Condvar or pipe
wakeup to break this OS-sleep early.
2026-05-22 05:22:55 +00:00
Claude 0e9d9d7d5f v0.1: green-thread actors, supervision, channels, benchmark
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.
2026-05-22 05:01:51 +00:00