feat: optimized async-runtime (replace for gstd) by vobradovich · Pull Request #1090 · gear-tech/sails

vobradovich · 2025-10-16T14:21:04Z

Summary

Reduce number of hashmaps to 2
Task now tracks the user future, pending reply locks, and (non-ethexe) critical hooks, letting message_loop reuse locks and enforce timeouts.
Added WakeSignals to manage reply lifecycle states—Pending, Ready, Timeout—so record_reply, poll, and timeout propagation share one typed registry, including optional reply hooks and automatic wake/cleanup
Refined Lock to store deadline and wait type

Sails `async_runtime` vs the async runtime in `gstd`

Sails reimplements the on-chain async loop, collapsing gstd's four global maps into
two, moving per-message state into the structures it belongs to (wait locks into the
Task, reply hooks into the WakeSignal), swapping the hasher, dropping waker machinery, and
adding a first-class timeout/expiry lifecycle with RAII cleanup. Same actor / wait /
wake model underneath; leaner bookkeeping and more explicit edge-case handling on top.

Global state: 4 maps -> 2

Concern	gstd	Sails
Futures	`FUTURES: HashMap<MessageId, Task>`	`tasks: IdMap<Task>`
Reply signals	`SIGNALS: HashMap<MessageId, WakeSignal>`	`signals: IdMap<WakeSignal>`
Wait locks	`LOCKS: HashMap<MessageId, BTreeMap<LockContext, Lock>>`	folded into `Task.locks`
Reply hooks	`REPLY_HOOKS: HashMap<MessageId, Box<dyn FnOnce()>>`	folded into `WakeSignal::{Pending,Expired}.reply_hook`

gstd does 2-3 independent keyed map operations per send and touches all four maps on teardown.
Sails registers a wait with one signals insert plus one push into the owning task's heap, and
a hook rides inside the signal it belongs to — fewer hash lookups and no cross-map lifetime
coordination.

Hashing

gstd: hashbrown default hasher — mixes all 32 bytes of every MessageId.
Sails: IdMap = HashMap<.., BuildHasherDefault<IdHasher>>. MessageId is already a
cryptographic hash, so IdHasher takes its first 8 bytes verbatim. No mixing -> cheaper gas on
every map op.

Lock storage & "shortest wait" selection

gstd: nested HashMap<MessageId, BTreeMap<LockContext, Lock>>. LockContext is an enum —
ReplyTo, Sleep, MxLockMonitor (the last serves gstd's Mutex/RwLock). On wait() it
iterates the inner map, filters deadline > now, and min_by. Lock::cmp compares deadline
only.
Sails: Task.locks: BinaryHeap<(Reverse<Lock>, Option<MessageId>)>. Earliest deadline is
the heap root; next_lock peeks/pops, lazily dropping locks that expired or whose signal is no
longer waited-for. Some(reply_to) = reply wait, None = sleep. No MxLockMonitor variant
(Sails doesn't ship gstd's mutexes). Its Lock::cmp breaks deadline ties UpTo < Exactly
(UpTo wins) — deterministic, where gstd leaves ties unordered.

Reply lifecycle & timeout semantics (the biggest behavioral difference)

gstd WakeSignal is a struct: { message_id, payload: Option<(Payload, ReplyCode)>, waker: Option<Waker> } — effectively two states (pending / has-payload). Timeout is not part of the
signal; it's computed externally in msg::async::poll via locks().is_timeout(...), and on
timeout the lock and signal entry are removed eagerly. Polling a missing entry panics.
Sails WakeSignal is a 3-variant enum: Pending { message_id, deadline, reply_hook } ->
Ready { payload, reply_code } | Expired { expected, now, reply_hook }. Timeout is a
first-class, retained state. Expired entries are deliberately kept so that:
- a deferred re-poll (e.g. select!/join! where another branch finished first and the
  MessageFuture is polled late) observes Err(Timeout) instead of panicking on a missing
  entry;
- a late reply can still fire a preserved reply_hook via record_reply;
- final deletion is MessageFuture::drop's job (RAII), not poll's.
The expire() transition is shared by record_timeout (natural timeout), poll (deferred
timeout), and forget_future (cancellation), and drop/cancel is symmetric with timeout
(Pending+hook -> Expired+hook). gstd has none of this retain/RAII machinery — it's covered in
Sails by a dedicated test matrix.

Wakers

gstd: real waker plumbing — Task holds a waker_fn, each pending poll clones
cx.waker() into the signal, and record_reply calls both waker.wake_by_ref() and
exec::wake.
Sails: Context::from_waker(Waker::noop()) — no waker stored or cloned, ever. Re-entry is
driven solely by Syscall::wake(message_id). Valid because the loop drives exactly one future
per message and never needs intra-future notification; saves the per-poll waker clone.

Reply hooks

gstd: separate HooksMap; register panics on duplicate; executed in
handle_reply_with_hook after record_reply, removed in handle_signal. Decoupled from the
signal.
Sails: the hook lives inside the signal (Pending/Expired), so it travels through the
state transitions — the "late reply still fires the hook" and "cancellation preserves the hook"
behaviors fall out for free. One fewer map and lookup. send_one_way adds register_hook,
which stores an Expired-with-hook entry so fire-and-forget sends can still fire a hook from
outside message_loop.

Critical hooks

gstd: separate critical module (take_and_execute/take_hook) as global state; the loop
clears it on completion, handle_signal runs it.
Sails: critical_hook: Option<Box<dyn FnOnce(MessageId)>> is a Task field, set via
with_task_mut, taken from the removed task in handle_signal. No separate global; compiled
out under ethexe.

`message_loop` & the reentrancy/aliasing fix

gstd: holds &mut task and polls Pin::new(&mut task.future) while that borrow (and the
&mut FUTURES borrow) is live. Safe for gstd because wait/lock/hook registration goes through
different maps, never re-borrowing FUTURES during the poll. It also carries a
lock_exceeded flag, checked at loop entry, to panic on lock-ownership-time overrun.
Sails: because locks now live in the Task, registering a wait during a poll
(register_signal -> task.insert_lock) would re-borrow the tasks map while the future is
being polled. Sails fixes this structurally: Task.future is Option<...>, moved out with
.take() before polling (so the &mut Task borrow ends), and every reentrant access
(register_signal, set_critical_hook, sleep_for) goes through a short-lived
with_task_mut. Result: sound under Stacked/Tree Borrows, no Box-noalias hazard, no unsafe
— validated by a Miri test. There's no lock_exceeded flag; overrun relies on node-side
timeout plus the Expired transitions.

Sleep

gstd: LocksMap::insert_sleep/remove_sleep, keyed by wake-up block in the nested
BTreeMap under a Sleep context.
Sails: sleep_for pushes an unkeyed (reply_to = None) lock into the task heap and returns
a MessageSleepFuture that resolves when block_height >= deadline. No separate sleep context.

Smaller but real

No Send bound on reply hooks (FnOnce() + 'static) — matches gstd's hook type; the client
GstdParams::with_reply_hook dropped its old Send requirement to suit.
entry_ref/EntryRef API used in record_reply/forget_future/poll to avoid double
lookups.
No mutex/async_init/async_main — Sails serves fewer clients than gstd's runtime (it
generates its own ctor/dispatch via macros), which is why its lock model can be leaner.
Host-testable: routes through the Syscall abstraction
(with_message_id/with_block_height/with_reply_code mocks), giving the whole runtime
host-side unit tests plus Miri — gstd's runtime is largely on-chain-tested only.

Net trade-offs

Sails gains: fewer maps/lookups and a faster hasher (gas), explicit and tested
timeout/cancellation/late-reply semantics, no waker allocation, and unit/Miri testability.

Sails takes on: a subtle reentrancy invariant (future-out-of-Task during poll) that gstd
sidesteps by construction, and re-implements lock/signal/hook bookkeeping that it must keep
correct itself rather than inheriting from gstd.

github-actions · 2025-10-22T16:52:01Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	450_513_679_484	450_514_036_959	-357_475	-0.00%	✅
alloc-0	563_780_574	564_376_739	-596_165	-0.11%	✅
alloc-12	567_287_166	567_864_696	-577_530	-0.10%	✅
alloc-143	726_655_761	727_298_127	-642_366	-0.09%	✅
alloc-986	827_641_528	828_290_311	-648_783	-0.08%	✅
alloc-10945	2_018_406_054	2_019_067_671	-661_617	-0.03%	✅
alloc-46367	6_403_667_537	6_404_351_520	-683_983	-0.01%	✅
alloc-121392	16_858_359_295	16_859_084_038	-724_743	-0.00%	✅
alloc-317810	43_146_678_632	43_147_409_792	-731_160	-0.00%	✅
counter_sync	637_172_916	678_859_589	-41_686_673	-6.14%	🚀
counter_async	803_743_595	851_480_134	-47_736_539	-5.61%	🚀
cross_program	2_238_138_550	2_436_719_419	-198_580_869	-8.15%	🚀
redirect	3_194_681_197	3_474_613_683	-279_932_486	-8.06%	🚀
stack_0	700_149_834	799_115_685	-98_965_851	-12.38%	🚀
stack_1	3_361_996_318	3_769_584_063	-407_587_745	-10.81%	🚀
stack_5	14_040_202_619	15_693_086_906	-1_652_884_287	-10.53%	🚀
stack_10	26_229_111_691	29_592_077_575	-3_362_965_884	-11.36%	🚀
stack_20	51_819_675_943	63_989_367_411	-12_169_691_468	-19.02%	🚀

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at aeaccca.}

github-actions · 2025-10-23T11:37:03Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	450_513_679_484	450_514_036_959	-357_475	-0.00%	✅
alloc-0	563_780_574	564_376_739	-596_165	-0.11%	✅
alloc-12	567_287_166	567_864_696	-577_530	-0.10%	✅
alloc-143	726_655_761	727_298_127	-642_366	-0.09%	✅
alloc-986	827_641_528	828_290_311	-648_783	-0.08%	✅
alloc-10945	2_018_406_054	2_019_067_671	-661_617	-0.03%	✅
alloc-46367	6_403_667_537	6_404_351_520	-683_983	-0.01%	✅
alloc-121392	16_858_359_295	16_859_084_038	-724_743	-0.00%	✅
alloc-317810	43_146_678_632	43_147_409_792	-731_160	-0.00%	✅
counter_sync	637_172_916	678_859_589	-41_686_673	-6.14%	🚀
counter_async	803_743_595	851_480_134	-47_736_539	-5.61%	🚀
cross_program	2_238_936_989	2_436_719_419	-197_782_430	-8.12%	🚀
redirect	3_194_681_197	3_474_613_683	-279_932_486	-8.06%	🚀
stack_0	700_149_834	799_115_685	-98_965_851	-12.38%	🚀
stack_1	3_361_996_318	3_769_584_063	-407_587_745	-10.81%	🚀
stack_5	14_040_202_619	15_693_086_906	-1_652_884_287	-10.53%	🚀
stack_10	26_229_111_691	29_592_077_575	-3_362_965_884	-11.36%	🚀
stack_20	51_819_675_943	63_989_367_411	-12_169_691_468	-19.02%	🚀

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at 5413c5a.}

github-actions · 2026-05-15T11:48:38Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_553_267	465_521_292_356	+2_260_911	+0.00%	✅
alloc-0	598_194_209	596_089_512	+2_104_697	+0.35%	✅
alloc-12	601_635_117	599_460_876	+2_174_241	+0.36%	✅
alloc-143	761_411_968	759_364_306	+2_047_662	+0.27%	✅
alloc-986	864_097_354	862_362_378	+1_734_976	+0.20%	✅
alloc-10945	2_214_636_039	2_216_616_777	-1_980_738	-0.09%	✅
alloc-46367	6_544_099_901	6_541_997_330	+2_102_571	+0.03%	✅
alloc-121392	17_170_526_475	17_168_433_902	+2_092_573	+0.01%	✅
alloc-317810	43_907_840_502	43_905_750_055	+2_090_447	+0.00%	✅
counter_sync	665_785_492	693_006_121	-27_220_629	-3.93%	👍
counter_async	829_606_982	863_104_002	-33_497_020	-3.88%	👍
cross_program	2_215_441_187	2_397_009_996	-181_568_809	-7.57%	🚀
redirect	3_197_138_740	3_471_987_506	-274_848_766	-7.92%	🚀
stack_0	705_592_615	799_415_299	-93_822_684	-11.74%	🚀
stack_1	3_392_136_301	3_778_945_524	-386_809_223	-10.24%	🚀
stack_5	14_168_220_708	15_738_223_846	-1_570_003_138	-9.98%	🚀
stack_10	26_478_961_103	29_542_694_740	-3_063_733_637	-10.37%	🚀
stack_20	54_514_692_122	66_032_420_478	-11_517_728_356	-17.44%	🚀

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at 897832b.}

github-actions · 2026-05-28T12:27:38Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_525_935_187	465_521_292_356	+4_642_831	+0.00%	✅
alloc-0	600_689_965	596_089_512	+4_600_453	+0.77%	✅
alloc-12	604_170_814	599_460_876	+4_709_938	+0.79%	✅
alloc-143	763_943_813	759_364_306	+4_579_507	+0.60%	✅
alloc-986	866_629_199	862_362_378	+4_266_821	+0.49%	✅
alloc-10945	2_217_167_884	2_216_616_777	+551_107	+0.02%	✅
alloc-46367	6_546_631_746	6_541_997_330	+4_634_416	+0.07%	✅
alloc-121392	17_173_094_409	17_168_433_902	+4_660_507	+0.03%	✅
alloc-317810	43_910_408_436	43_905_750_055	+4_658_381	+0.01%	✅
counter_sync	667_803_711	693_006_121	-25_202_410	-3.64%	👍
counter_async	831_625_201	863_104_002	-31_478_801	-3.65%	👍
cross_program	2_218_461_812	2_397_009_996	-178_548_184	-7.45%	🚀
redirect	3_199_897_684	3_471_987_506	-272_089_822	-7.84%	🚀
stack_0	705_318_790	799_415_299	-94_096_509	-11.77%	🚀
stack_1	3_391_069_123	3_778_945_524	-387_876_401	-10.26%	🚀
stack_5	14_163_980_118	15_738_223_846	-1_574_243_728	-10.00%	🚀
stack_10	26_470_753_748	29_542_694_740	-3_071_940_992	-10.40%	🚀
stack_20	54_498_551_237	66_032_420_478	-11_533_869_241	-17.47%	🚀

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at d1e81d4.}

…ut a reply

…_bytes

github-actions · 2026-05-29T14:36:10Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_945_542	465_521_292_356	+2_653_186	+0.00%	✅
alloc-0	598_700_320	596_089_512	+2_610_808	+0.44%	✅
alloc-12	602_141_228	599_460_876	+2_680_352	+0.45%	✅
alloc-143	761_918_079	759_364_306	+2_553_773	+0.34%	✅
alloc-986	864_603_465	862_362_378	+2_241_087	+0.26%	✅
alloc-10945	2_215_142_150	2_216_616_777	-1_474_627	-0.07%	✅
alloc-46367	6_544_606_012	6_541_997_330	+2_608_682	+0.04%	✅
alloc-121392	17_171_032_586	17_168_433_902	+2_598_684	+0.02%	✅
alloc-317810	43_908_346_613	43_905_750_055	+2_596_558	+0.01%	✅
counter_sync	653_485_763	693_006_121	-39_520_358	-5.70%	🚀
counter_async	818_743_175	863_104_002	-44_360_827	-5.14%	🚀
cross_program	2_198_786_808	2_397_009_996	-198_223_188	-8.27%	🚀
redirect	3_182_922_366	3_471_987_506	-289_065_140	-8.33%	🚀
stack_0	696_996_324	799_415_299	-102_418_975	-12.81%	🚀
stack_1	3_357_497_507	3_778_945_524	-421_448_017	-11.15%	🚀
stack_5	14_036_020_145	15_738_223_846	-1_702_203_701	-10.82%	🚀
stack_10	26_233_598_416	29_542_694_740	-3_309_096_324	-11.20%	🚀
stack_20	51_950_481_451	66_032_420_478	-14_081_939_027	-21.33%	🚀
noop_baseline	363_294_902	0	+363_294_902	+inf%	❌

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at 9b98f77.}

github-actions · 2026-06-01T13:22:27Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_945_542	465_525_077_223	-1_131_681	-0.00%	✅
alloc-0	598_700_320	599_843_943	-1_143_623	-0.19%	✅
alloc-12	602_141_228	603_255_248	-1_114_020	-0.18%	✅
alloc-143	761_918_079	763_104_169	-1_186_090	-0.16%	✅
alloc-986	864_603_465	865_787_429	-1_183_964	-0.14%	✅
alloc-10945	2_215_142_150	2_216_328_240	-1_186_090	-0.05%	✅
alloc-46367	6_544_606_012	6_545_787_850	-1_181_838	-0.02%	✅
alloc-121392	17_171_032_586	17_172_248_387	-1_215_801	-0.01%	✅
alloc-317810	43_908_346_613	43_909_564_540	-1_217_927	-0.00%	✅
counter_sync	653_512_086	695_201_799	-41_689_713	-6.00%	🚀
counter_async	818_768_695	865_267_022	-46_498_327	-5.37%	🚀
cross_program	2_198_807_101	2_391_505_600	-192_698_499	-8.06%	🚀
redirect	3_182_972_292	3_464_158_516	-281_186_224	-8.12%	🚀
stack_0	696_643_269	792_932_822	-96_289_553	-12.14%	🚀
stack_1	3_356_086_077	3_752_847_200	-396_761_123	-10.57%	🚀
stack_5	14_030_375_215	15_633_706_718	-1_603_331_503	-10.26%	🚀
stack_10	26_222_661_611	29_340_040_097	-3_117_378_486	-10.62%	🚀
stack_20	51_928_960_896	65_942_391_269	-14_013_430_373	-21.25%	🚀
noop_baseline	363_294_902	363_446_797	-151_895	-0.04%	✅

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at 504ec80.}

github-actions · 2026-06-01T15:54:51Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_945_542	465_525_077_223	-1_131_681	-0.00%	✅
alloc-0	598_700_320	599_843_943	-1_143_623	-0.19%	✅
alloc-12	602_141_228	603_255_248	-1_114_020	-0.18%	✅
alloc-143	761_918_079	763_104_169	-1_186_090	-0.16%	✅
alloc-986	864_603_465	865_787_429	-1_183_964	-0.14%	✅
alloc-10945	2_215_142_150	2_216_328_240	-1_186_090	-0.05%	✅
alloc-46367	6_544_606_012	6_545_787_850	-1_181_838	-0.02%	✅
alloc-121392	17_171_032_586	17_172_248_387	-1_215_801	-0.01%	✅
alloc-317810	43_908_346_613	43_909_564_540	-1_217_927	-0.00%	✅
counter_sync	653_730_393	695_201_799	-41_471_406	-5.97%	🚀
counter_async	818_987_002	865_267_022	-46_280_020	-5.35%	🚀
cross_program	2_199_140_871	2_391_505_600	-192_364_729	-8.04%	🚀
redirect	3_183_501_094	3_464_158_516	-280_657_422	-8.10%	🚀
stack_0	697_568_275	792_932_822	-95_364_547	-12.03%	🚀
stack_1	3_360_177_351	3_752_847_200	-392_669_849	-10.46%	🚀
stack_5	14_047_131_561	15_633_706_718	-1_586_575_157	-10.15%	🚀
stack_10	26_255_249_297	29_340_040_097	-3_084_790_800	-10.51%	🚀
stack_20	51_993_211_262	65_942_391_269	-13_949_180_007	-21.15%	🚀
noop_baseline	363_294_902	363_446_797	-151_895	-0.04%	✅

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at a6a2fe1.}

github-actions · 2026-06-01T16:04:45Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_945_542	465_525_077_223	-1_131_681	-0.00%	✅
alloc-0	598_700_320	599_843_943	-1_143_623	-0.19%	✅
alloc-12	602_141_228	603_255_248	-1_114_020	-0.18%	✅
alloc-143	761_918_079	763_104_169	-1_186_090	-0.16%	✅
alloc-986	864_603_465	865_787_429	-1_183_964	-0.14%	✅
alloc-10945	2_215_142_150	2_216_328_240	-1_186_090	-0.05%	✅
alloc-46367	6_544_606_012	6_545_787_850	-1_181_838	-0.02%	✅
alloc-121392	17_171_032_586	17_172_248_387	-1_215_801	-0.01%	✅
alloc-317810	43_908_346_613	43_909_564_540	-1_217_927	-0.00%	✅
counter_sync	653_730_393	695_201_799	-41_471_406	-5.97%	🚀
counter_async	818_987_002	865_267_022	-46_280_020	-5.35%	🚀
cross_program	2_199_139_165	2_391_505_600	-192_366_435	-8.04%	🚀
redirect	3_183_501_094	3_464_158_516	-280_657_422	-8.10%	🚀
stack_0	697_569_981	792_932_822	-95_362_841	-12.03%	🚀
stack_1	3_360_184_175	3_752_847_200	-392_663_025	-10.46%	🚀
stack_5	14_047_158_857	15_633_706_718	-1_586_547_861	-10.15%	🚀
stack_10	26_255_302_183	29_340_040_097	-3_084_737_914	-10.51%	🚀
stack_20	51_993_315_328	65_942_391_269	-13_949_075_941	-21.15%	🚀
noop_baseline	363_294_902	363_446_797	-151_895	-0.04%	✅

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at e06176a.}

github-actions · 2026-06-01T16:41:16Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_945_542	465_525_077_223	-1_131_681	-0.00%	✅
alloc-0	598_700_320	599_843_943	-1_143_623	-0.19%	✅
alloc-12	602_141_228	603_255_248	-1_114_020	-0.18%	✅
alloc-143	761_918_079	763_104_169	-1_186_090	-0.16%	✅
alloc-986	864_603_465	865_787_429	-1_183_964	-0.14%	✅
alloc-10945	2_215_142_150	2_216_328_240	-1_186_090	-0.05%	✅
alloc-46367	6_544_606_012	6_545_787_850	-1_181_838	-0.02%	✅
alloc-121392	17_171_032_586	17_172_248_387	-1_215_801	-0.01%	✅
alloc-317810	43_908_346_613	43_909_564_540	-1_217_927	-0.00%	✅
counter_sync	653_730_393	695_201_799	-41_471_406	-5.97%	🚀
counter_async	818_987_002	865_267_022	-46_280_020	-5.35%	🚀
cross_program	2_199_139_165	2_391_505_600	-192_366_435	-8.04%	🚀
redirect	3_183_501_094	3_464_158_516	-280_657_422	-8.10%	🚀
stack_0	697_569_981	792_932_822	-95_362_841	-12.03%	🚀
stack_1	3_360_184_175	3_752_847_200	-392_663_025	-10.46%	🚀
stack_5	14_047_158_857	15_633_706_718	-1_586_547_861	-10.15%	🚀
stack_10	26_255_302_183	29_340_040_097	-3_084_737_914	-10.51%	🚀
stack_20	51_993_315_328	65_942_391_269	-13_949_075_941	-21.15%	🚀
noop_baseline	363_294_902	363_446_797	-151_895	-0.04%	✅

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at f59aa21.}

github-actions · 2026-06-02T15:37:37Z

🔬 Benchmark Comparison

Benchmark	Current	Baseline	Change	Change %	Status
compute	465_523_945_542	465_525_077_223	-1_131_681	-0.00%	✅
alloc-0	598_700_320	599_843_943	-1_143_623	-0.19%	✅
alloc-12	602_141_228	603_255_248	-1_114_020	-0.18%	✅
alloc-143	761_918_079	763_104_169	-1_186_090	-0.16%	✅
alloc-986	864_603_465	865_787_429	-1_183_964	-0.14%	✅
alloc-10945	2_215_142_150	2_216_328_240	-1_186_090	-0.05%	✅
alloc-46367	6_544_606_012	6_545_787_850	-1_181_838	-0.02%	✅
alloc-121392	17_171_032_586	17_172_248_387	-1_215_801	-0.01%	✅
alloc-317810	43_908_346_613	43_909_564_540	-1_217_927	-0.00%	✅
counter_sync	655_716_261	695_201_799	-39_485_538	-5.68%	🚀
counter_async	821_112_641	865_267_022	-44_154_381	-5.10%	🚀
cross_program	2_207_576_341	2_391_505_600	-183_929_259	-7.69%	🚀
redirect	3_198_180_511	3_464_158_516	-265_978_005	-7.68%	🚀
stack_0	701_198_340	792_932_822	-91_734_482	-11.57%	🚀
stack_1	3_378_614_729	3_752_847_200	-374_232_471	-9.97%	🚀
stack_5	14_123_002_376	15_633_706_718	-1_510_704_342	-9.66%	🚀
stack_10	26_406_956_141	29_340_040_097	-2_933_083_956	-10.00%	🚀
stack_20	52_286_821_567	65_942_391_269	-13_655_569_702	-20.71%	🚀
noop_baseline	363_294_902	363_446_797	-151_895	-0.04%	✅

Legend

🚀 Significant improvement (>5% reduction)
👍 Minor improvement (<5% reduction)
✅ No significant change
⚠️ Minor regression (<5% increase)
❌ Significant regression (>5% increase)

_{🤖 This comment was automatically generated by the benchmark comparison workflow at 9af44cc.}

vobradovich added 30 commits August 27, 2025 09:35

wip: Gear Env Client

14282b7

wip: env impl

a2c5015

wip: env impl

c83671d

wip: env

e697db3

wip: envs

5c63597

wip: Client Gen

62a821a

wip: cleint gen, events, mocks

c190cf6

wip: rmrk example

32f4da0

wip: ping pong

6816f14

wip: env deploy, DefaultEnv

1f01c5b

wip: remove MockEnv

cb1f409

wip: gclient tests

07eb92d

wip: event listener, names

c40c98e

Merge branch 'master' into vo/client-v2-draft

4c98831

wip: fix clippy, cleanup

1a393e1

wip: benchmarks refactoring

374260f

wip: gstd future with redirect

f2803c5

wip: bench redirect

e0fb024

Merge branch 'master' into vo/client-v2-draft

1db71f4

wip: cleanup, fix clippy

57fbccd

fix: client gen full path

501708e

fix: cli program template

5e50a9b

chore: optimizations

02b778a

wip: opt, code remove

88ba3d1

wip: redirect opt

4044c93

fix: logs

4ad6e98

wip: gtest env query

227504e

wip: gclient, gtest funcs, remove old code

94ca161

chore: refactor

f74a505

perf: remove params & payload from GtsdFuture::MessageWithRedirect

f1d852f

vobradovich requested a review from techraed October 20, 2025 12:06

vobradovich changed the title ~~feat: optimezed async-runtime (replace for gstd)~~ feat: optimized async-runtime (replace for gstd) Oct 21, 2025

fix: cfg features

cf08be8

fix: remove entrye before hook exec on WakeSignal::Expired

427c8cc

m62624 approved these changes Oct 24, 2025

View reviewed changes

Merge branch 'master' into vo/async-runtime

883ebd8

Merge branch 'master' into vo/async-runtime

3df3ad4

vobradovich added 3 commits May 28, 2026 16:19

fix: Memory leak: WakeSignal::Expired entries never collected witho…

df0bc4f

…ut a reply

perf: hasher for MessageId, avoids zeroing N bytes in Syscall::read…

78e188b

…_bytes

perf: noop baseline bench, runtime .expect opt

350c14f

vobradovich added 2 commits June 1, 2026 12:58

wip: refine

2d044d8

Merge branch 'master' into vo/async-runtime

63e1c70

vobradovich added 3 commits June 1, 2026 16:45

wip: rever read_bytes() opt, propagate reply_deposit error

22c7437

wip: async runtime unit tests, wasm tests

76eb89f

fmt

2105663

clippy

872a255

fix: reentrancy/aliasing UB in the gstd async runtime

2401de0

vobradovich requested review from StackOverflowExcept1on and ukint-vs and removed request for techraed June 2, 2026 17:22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: optimized async-runtime (replace for gstd)#1090

feat: optimized async-runtime (replace for gstd)#1090
vobradovich wants to merge 83 commits into
masterfrom
vo/async-runtime

vobradovich commented Oct 16, 2025 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2025

Uh oh!

github-actions Bot commented Oct 23, 2025

Uh oh!

github-actions Bot commented May 15, 2026

Uh oh!

github-actions Bot commented May 28, 2026

Uh oh!

github-actions Bot commented May 29, 2026

Uh oh!

github-actions Bot commented Jun 1, 2026

Uh oh!

github-actions Bot commented Jun 1, 2026

Uh oh!

github-actions Bot commented Jun 1, 2026

Uh oh!

github-actions Bot commented Jun 1, 2026

Uh oh!

github-actions Bot commented Jun 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

vobradovich commented Oct 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Sails async_runtime vs the async runtime in gstd

Global state: 4 maps -> 2

Hashing

Lock storage & "shortest wait" selection

Reply lifecycle & timeout semantics (the biggest behavioral difference)

Wakers

Reply hooks

Critical hooks

message_loop & the reentrancy/aliasing fix

Sleep

Smaller but real

Net trade-offs

Uh oh!

github-actions Bot commented Oct 22, 2025

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented Oct 23, 2025

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented May 15, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented May 28, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented May 29, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented Jun 1, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented Jun 1, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented Jun 1, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented Jun 1, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

github-actions Bot commented Jun 2, 2026

🔬 Benchmark Comparison

Legend

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

vobradovich commented Oct 16, 2025 •

edited

Loading

Sails `async_runtime` vs the async runtime in `gstd`

`message_loop` & the reentrancy/aliasing fix