2,651 Commits over 669 Days - 0.17cph!
Update: use volatile int instead of Volatile.Read(ref long), as the latter is an expensive CAS
Existing code was using non-volatile read of long, but I don't think it's safe
Tests: ran perf tests
Bugfix: fix extreme spillage in high-contention perf test
Turns out I needed to do an extra spin, whoops
Tests: ran unit and perf tests
Update(tests): add 4, 8, 16, 32 task scenarios (previously was just 32) for pool perf tests
Tests: ran perf tests
Update(tests): expand AllocDeallocMTShortLived to optionally simulate busywork
We're never going to be just slamming the pool(or if we do - we shouldn't as that's overhead). This makes the test closer to real world.
Tests: ran perf test
Clean(tests): consolidate pool create-spilled code into Utils.MeasureAndPrint
Tests: ran perf tests
Update: fuzzy version of pool with CircularBuffer
Fastest so far, but has a tendency to spill/allocate in during very high contention
AllocDeallocST - Avg: 0.63ms, Created: 0 Spilled: 0
AllocDeallocMTShortLived - Avg: 2.79ms, Created: 144(0.45%) Spilled: 169(0.53%)
AllocDeallocMTLongLived - Avg: 2.94ms, Created: 0 Spilled: 0
AllocDeallocMPSC - Avg: 3.26ms, Created: 0 Spilled: 0
Tests: ran unit and perf tests
Cherrypick from pool_mt/circularbuffer - test improvements
Update(tests): gather and output more data about the test
- track created and spilled (matters for next pool version)
- more profiler markers
Tests: ran perf tests
Update(tests): add a couple more perf tests
- AllocDeallocMT -> AllocDeallocMTLongLived (biases Get, then Free operations), Avg: 3.56901ms
- AllocDeallocMTShortLived - fetches and returns to pool immediately (to evenly spread pressure), Avg: 3.01974ms
- AllocDeallocMPSC - simulates multiple-producers-single-consumer (emulates network write thread under jobs 2) Avg: 12.20773ms
All cases - 32 tasks doing 1k allocs
Tests: ran perf tests
Update: rewrite pool as a circular buffer with 0 locks
Weirdly loses to ConcurrentBag in mono by 2x, but in CLR wins 4x >.> Going to poke around more
ST avg: 0.62554ms
MT avg: 9.92301ms (dafuq?!)
Tests: ran unit and perf tests
Update(tests): AllocDeallocMT perf test scope annotations
Tests: ran perf test
Bugfix: ensure Pool.Fill keeps items in use valid
Tests: ran unit tests
Optim: rewrite Pool to use ConcurrentBag(many mutexes) instead of single mutex
Fails unit tests, but not going to focus on fixing it just yet - want to make it faster first.
ST avg: 0.88781ms
MT avg: 3.1281ms
Tests: ran perf tests
Update(tests): add TestMTPressure unit test
Wrote it cause my ConcurrentBag based pool is misbehaving
Tests: ran new unit test
Update(tests): pool now has a runtime switch to disable editor-only safety logic, so that perf is more apples-to-apples
ST avg: 1.0358ms
MT avg: 13.6264ms
Tests: ran perf tests
Update(tests): add simple pool perf tests (ST & MT)
ST avg: 1.29925ms for 10k alloc-deallocs
MT avg: 21.16116ms for 1k alloc-deallocs per task, 32 tasks
Tests: ran tests
Update: move Utils Measure routines to Facepunch.System.Tests
- also make it auto-referenced
Tests: compiles
Merge: from sanitize_specatating_players
- Bugfix for NRE spam when spectate mods don't correctly setup spectating
Tests: none, trivial change
Bugfix: ServerOcclusion - add defensive checks for suspected spectate mods
The mods should be fixed, but at least it won't put servers in a weird state.
Tests: none, trivial change
▆ ▍▇▆▅▌▋ ▅▌▍▄ ▄▍▅▇▇▍▌▉██▇▄▋▊▅▄▍▄▄█▇▍▆▍▋█▌▌▋▆▋▊▍▋▇▋▌▊ ▅▆▌▊▆ ▇▌▍▉▊▍▌▋▄▋▉▊▅ ▊▅▊▌▄ ▄▌▅▋ ▊▅▄▆█▅▆▍▋▉▊▊ ▍▅▋▍▆▌▍█▊ ▍▇▄▆▄ ▊▍▆▆▌▌▄▉▅▌▉▇▉▊▉ ▄▍▍▍█ ▋▄▌ ▍▌▅▌▋▄ ▋▋ ▍▍██▆▋▉ ▊▌▉ █▍▍█▌▆▇ ▍▊▍█▄▋ ▇▊ ▅▆█▋▄▇█▇▄▉▇▅▅▆▇█ ▆ ▅▌ ▆▅▌▉ ▍▌█▋█▌▆
▋ ▊▊▄▊▍▆▊ ▋▊▆▆▆▊ ▊▋▉▆▉▄ ▇▅▌█▉█▅▇ ▌▆▄▄ ▉ ▆▊▊ ▊▅▊ ▆██▄▋▆▉▇▌▉▍▅█▍▄▅▉▋▆▉▅▋▍▋█▉▅▋▄▍▍▅▇▉▇▉▍▄▄▆▄▅▇▋▌ ▄▊▅█▅ ▄▌█ ▊██▉▊▊ ▆▄ ▋▇▌▉▌▌
Merge: from objectworkqueue_valuetype_allocs
- Optim: eliminate allocs from value types used in ObjectWorkQueue. Needs manual changes to be eneabled
Tests: none, trivial change. Checked IL
Optim: Apply IsValidToRun for value-type ObjectWorkQueues
- overrode BoatAIWorkQueue, UpdateSurroundingsQueue, UpdateMissionValidStateWorkQueue
Tests: none, checked IL
Optim: use virtual IsValidToRun to avoid boxing allocs in ObjectWorkQueue.RunQueue
Sad-noises-about-C#-generics, couldn't figure how to automate it
Tests: none, inspected IL, no box instructions generated
Merge: from createcorpse_triggerparent_nre
- Bugfix for TriggerParent not cleaning themselves up if a corpse forces it's way just before volume is destroyed
Tests: ran auto tests
Bugfix: Ensure TriggerParent properly tracks forced corpse
Corpses forcefully inherit trigger volumes, which doesn't properly initialize internal contents, leading to skipped cleanup when volume is destroyed. Rare bug in the wild, but consistent in auto tests.
Tests: ran all PlayerMountedToArmoredVehicle_PlayerNotTargeted auto tests
Merge: from stableobjectarray_remove_dirty
- Optim: maintain synchronized order of dependent data as we remove from StableObjectCache
Tests: unit tests and a couple 2p sessions on Craggy with disconnects, shooting
Update: codegen
Tests: none
Bugfix: fix invalid index assigned to a projectile on projectile removal from cache
Copy paste mistake, happy we got unit tests
Tests: ran ProjectileTests.TestConsistency - it passes
Clean: consolidate player-related state cache growth to AddToPlayerCache
Tests: compiles
Clean: make the code clearer around maintianing packed order
Using better names for vars, easier to understand ifs and a safety assert
Tests: compiles
Update(tests): add TestMovePlayer
- bugfix for missing to copy an extra tick/reset point in TickInterpolatorCache.MovePlayer
- added more docs to StableObjectArray
Tests: ran new unit test
Optim: replace BasePlayer.SparsePlayers with PlayerCache.Objects now that indices always match packed order
Tests: 2p on Craggy with disconnects of 2nd player
Update: repack player cache and dependent state as players are removed
Should allow us to clean up a bit of indirection and optimize things further
Tests: 2p session on Craggy with disconnects and being kicked for flying under terrain
Clean: projectiles now internally iterate over their cache rather than instance list
Tests: unit tests, 2p on craggy
Update: ensure projectile stable indices are packed
This simplifies code and opens up the door for better cache utilisation & simd processing, but currently savings are on the noise level
Tests: unit tests + shot around player with 2 clients connected
Clean: cache span from ActiveTriggers.Objects
Tests: none, trivial change
Merge: from terrainkick_nre_fix
- Bugfix for NREs caused by kicking players(invalidating iterators) while looping over StableObjectArray
Tests: kicked self through clipping in terrain - no NREs
Optim: Avoid double indirection when looking up active triggers
- enabled by use of invalidating RemoveAtSwapback
TriggerParent.StableIndex is only const during a frame, it now can change between frames
Tests: built a 3-piece boat with a ladder, jumped around it
Clean: removed FilterInvalidPlayers now that I've validated that all kicks were grouped at end
- cleaned up comments that pointed out kick locations
Tests: none, trivial change
Bugfix: avoid NRE after kicking a player while iterating over a player cache
- consolidated all kicks in ServerUpdateParallel to end of the method
Tests: got kicked for terrain violations, no NRE
Update: StableObjectArray.RemoveAtSwapback can now optionally invalidate stable indices
- added unit tests
Should help us maintain packed state of dependent arrays - will start converting current use cases next.
Tests: ran unit tests
Update: replace hardcoded constant with GamePhysics.DefaultMaxResultsPerQuery (same value)
- also submitting missing meta file
Tests: compiles
Merge: from fix_boat_no_sail
- Bugfix for boats and vehicles getting stuck in open areas, effects not playing, building appearing slow
Tests: unit and perf tests, built and sailed a boat on craggy
