Wrap world access in a SystemParam for granular access
#418
Conversation
This allows replication to run in parallel with other systems that write to non-replicated components. Benchmarks show the same performance, but we run them without `multi_threaded`.
Shatur
changed the title
SystemParam for granular access.SystemParam for granular access
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #418 +/- ##
==========================================
- Coverage 89.17% 85.12% -4.06%
==========================================
Files 53 54 +1
Lines 2808 2957 +149
==========================================
+ Hits 2504 2517 +13
- Misses 304 440 +136 ☔ View full report in Codecov by Sentry. |
src/server/replicated_world.rs
Outdated
| impl<'w> ReplicatedWorld<'w, '_> { | ||
| /// Extracts a component as [`Ptr`] and its ticks from a table or sparse set, depending on its storage type. | ||
| /// | ||
| /// # Safety | ||
| /// | ||
| /// The component must be present in this archetype, have the specified storage type, and be previously marked for replication. | ||
| pub(super) unsafe fn get_component_unchecked( |
There was a problem hiding this comment.
Doesn't this also require that the source of the ptr doesn't get reallocated? I don't think we can make that guarantee.
There was a problem hiding this comment.
We configured the access by manually implementing SystemParam to prevent other systems running in parallel from writing to replicated components. And we expose only read access with the associated runtime on Ptr, so it should be fine.
I will need a similar thing for writing part in the future.
|
Pushed a small rename to rework the receiving. But I'll make a separate PR for it. |
This allows replication to run in parallel with other systems that write to non-replicated components.
Benchmarks show the same performance, but we run them without
multi_threaded.Supersedes #417.