Hywan:feat-cross-process-lock-poisoning - Branch - matrix-org/matrix-rust-sdk

feat(common): Add cross-process lock dirtying

#5672

Comparing

Hywan:feat-cross-process-lock-poisoning

(

985032c

) with

main

(

d6d720c

)

-1%

Untouched: 50

Benchmarks

Passed

Restore session [SQLite][encrypted]

benchmarks/benches/store_bench.rs::benches::restore_session::Client reload

+5%

1.4 s1.4 s

Linked chunk lazy loader[memory store][100]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

+4%

195 µs187.5 µs

Devices collecting [memory][2003 devices]

benchmarks/benches/crypto_bench.rs::benches::devices_missing_sessions_collecting::Devices missing sessions collecting

+4%

11.1 ms10.7 ms

Linked chunk lazy loader[memory store][10]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

+4%

34.7 µs33.5 µs

Linked chunk writing [memory store][100]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

+1%

372.3 µs368.9 µs

Device keys query [SQLite][153 device and cross signing keys]

benchmarks/benches/crypto_bench.rs::benches::keys_query::Keys querying

+1%

40.3 ms40 ms

Linked chunk metadata loader[memory store][1000]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

741.9 µs738.6 µs

Room key sharing [SQLite][163 devices]

benchmarks/benches/crypto_bench.rs::benches::room_key_sharing::Room key sharing

57.4 ms57.2 ms

Event cache find_event_relations[memory][100 events, edits filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

46.9 ms46.8 ms

Linked chunk metadata loader[sqlite store][100]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

244.6 µs243.9 µs

Restore session [memory store]

benchmarks/benches/store_bench.rs::benches::restore_session::Client reload

338.3 ms337.3 ms

Linked chunk writing [none][10]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

15.9 µs15.8 µs

Event cache room updates[memory][room count: 10]

benchmarks/benches/event_cache.rs::event_cache::handle_room_updates::Event cache room updates

145.5 ms145.2 ms

Room key sharing [memory][163 devices]

benchmarks/benches/crypto_bench.rs::benches::room_key_sharing::Room key sharing

46.8 ms46.7 ms

Linked chunk metadata loader[sqlite store][10]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

171.4 µs171.2 µs

Linked chunk writing [sqlite store][1000]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

27.7 ms27.6 ms

Event cache find_event_relations[memory][10 events, edits filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

3.3 ms3.3 ms

Linked chunk lazy loader[memory store][1000]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

2.5 ms2.5 ms

Event cache find_event_relations[memory][100 events, #no filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

46.8 ms46.8 ms

Event cache find_event_relations[memory][1000 events, #no filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

2.2 s2.2 s

Event cache find_event_relations[memory][1000 events, edits filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

2.2 s2.2 s

Event cache find_event_relations[memory][10 events, #no filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

3.3 ms3.3 ms

Create a timeline with initial events[10000 events]

benchmarks/benches/timeline.rs::room::create_timeline_with_initial_events::Create a timeline

731.6 ms731.4 ms

Devices collecting [SQLite][2003 devices]

benchmarks/benches/crypto_bench.rs::benches::devices_missing_sessions_collecting::Devices missing sessions collecting

112.9 ms112.9 ms

Restore session [SQLite][clear]

benchmarks/benches/store_bench.rs::benches::restore_session::Client reload

749.3 ms749.2 ms

Linked chunk writing [sqlite store][10]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

552.4 µs552.8 µs

Linked chunk writing [none][1000]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

831.8 µs832.4 µs

Device keys query [memory][153 device and cross signing keys]

benchmarks/benches/crypto_bench.rs::benches::keys_query::Keys querying

27.6 ms27.7 ms

Linked chunk metadata loader[sqlite store][1000]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

1.2 ms1.2 ms

Create[1000 rooms × 1000 events]

benchmarks/benches/room_list.rs::room_list::create::RoomList

30.7 ms30.7 ms

Linked chunk writing [sqlite store][100]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

2.6 ms2.6 ms

Event cache room updates[SQLite][room count: 100]

benchmarks/benches/event_cache.rs::event_cache::handle_room_updates::Event cache room updates

4.2 s4.2 s

Event cache room updates[SQLite][room count: 1]

benchmarks/benches/event_cache.rs::event_cache::handle_room_updates::Event cache room updates

37.6 ms37.7 ms

Linked chunk lazy loader[sqlite store][1000]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

15.5 ms15.5 ms

Linked chunk lazy loader[sqlite store][10]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

428.4 µs430 µs

Event cache room updates[memory][room count: 100]

benchmarks/benches/event_cache.rs::event_cache::handle_room_updates::Event cache room updates

1.5 s1.5 s

Event cache room updates[SQLite][room count: 10]

benchmarks/benches/event_cache.rs::event_cache::handle_room_updates::Event cache room updates

396.8 ms398.5 ms

Linked chunk writing [none][100]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

88.4 µs88.7 µs

Linked chunk lazy loader[sqlite store][100]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

1.7 ms1.7 ms

Event cache room updates[memory][room count: 1]

benchmarks/benches/event_cache.rs::event_cache::handle_room_updates::Event cache room updates

-1%

14.3 ms14.5 ms

Event cache find_event_relations[SQLite][100 events, edits filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

-2%

8.6 ms8.8 ms

Event cache find_event_relations[SQLite][100 events, #no filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

-2%

8.2 ms8.3 ms

Event cache find_event_relations[SQLite][1000 events, edits filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

-2%

83.5 ms85 ms

Linked chunk metadata loader[memory store][100]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

-2%

20.3 µs20.7 µs

Event cache find_event_relations[SQLite][1000 events, #no filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

-2%

78.4 ms80.3 ms

Linked chunk writing [memory store][10]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

-3%

52.5 µs53.9 µs

Linked chunk writing [memory store][1000]

benchmarks/benches/linked_chunk.rs::event_cache::writing::Linked chunk writing

-3%

3.6 ms3.7 ms

Event cache find_event_relations[SQLite][10 events, edits filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

-8%

1.2 ms1.3 ms

Linked chunk metadata loader[memory store][10]

benchmarks/benches/linked_chunk.rs::event_cache::reading::Linked chunk reading

-8%

5.4 µs5.8 µs

Event cache find_event_relations[SQLite][10 events, #no filter]

benchmarks/benches/event_cache.rs::event_cache::find_event_relations::Event cache room updates

-8%

1.2 ms1.3 ms

Commits

Click on a commit to change the comparison range

Base

main

d6d720c

-0.51%

feat(common): Detect when the cross-process lock has been dirtied. This patch detects when the cross-process lock has been dirtied. A new `CrossProcessLockResult` enum is introduced to simplify the returned value of `try_lock_once` and `spin_lock`. It flattens the previous `Result<Option<_>>` by providing 3 variants: `Clean`, `Dirty` and `Unobtained`.

985032c

26 days ago

by Hywan

Home Terms Privacy Docs