# Objective
- Currently exclusive systems and applying buffers run outside of the multithreaded executor and just calls the funtions on the thread the schedule is running on. Stageless changes this to run these using tasks in a scope. Specifically It uses `spawn_on_scope` to run these. For the render thread this is incorrect as calling `spawn_on_scope` there runs tasks on the main thread. It should instead run these on the render thread and only run nonsend systems on the main thread.
## Solution
- Add another executor to `Scope` for spawning tasks on the scope. `spawn_on_scope` now always runs the task on the thread the scope is running on. `spawn_on_external` spawns onto the external executor than is optionally passed in. If None is passed `spawn_on_external` will spawn onto the scope executor.
- Eventually this new machinery will be able to be removed. This will happen once a fix for removing NonSend resources from the world lands. So this is a temporary fix to support stageless.
---
## Changelog
- add a spawn_on_external method to allow spawning on the scope's thread or an external thread
## Migration Guide
> No migration guide. The main thread executor was introduced in pipelined rendering which was merged for 0.10. spawn_on_scope now behaves the same way as on 0.9.
# Objective
- Make the internals of `RemovedComponents` clearer
## Solution
- Add a wrapper around `Entity`, used in `RemovedComponents` as `Events<RemovedComponentsEntity>`
---
## Changelog
- `RemovedComponents` now internally uses an `Events<RemovedComponentsEntity>` instead of an `Events<Entity>`
The `DoubleEndedIterator` impls produce incorrect results on subsequent calls to `iter()` if the iterator is only partially consumed.
The following code shows what happens
```rust
fn next_back_is_bad() {
let mut events = Events::<TestEvent>::default();
events.send(TestEvent { i: 0 });
events.send(TestEvent { i: 1 });
events.send(TestEvent { i: 2 });
let mut reader = events.get_reader();
let mut iter = reader.iter(&events);
assert_eq!(iter.next_back(), Some(&TestEvent { i: 2 }));
assert_eq!(iter.next(), Some(&TestEvent { i: 0 }));
let mut iter = reader.iter(&events);
// `i: 2` event is returned twice! The `i: 1` event is missed.
assert_eq!(iter.next(), Some(&TestEvent { i: 2 }));
assert_eq!(iter.next(), None);
}
```
I don't think this can be fixed without adding some very convoluted bookkeeping.
## Migration Guide
`ManualEventIterator` and `ManualEventIteratorWithId` are no longer `DoubleEndedIterator`s.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Removal events are unwieldy and require some knowledge of when to put systems that need to catch events for them, it is very easy to end up missing one and end up with memory leak-ish issues where you don't clean up after yourself.
## Solution
Consolidate removals with the benefits of `Events<...>` (such as double buffering and per system ticks for reading the events) and reduce the special casing of it, ideally I was hoping to move the removals to a `Resource` in the world, but that seems a bit more rough to implement/maintain because of double mutable borrowing issues.
This doesn't go the full length of change detection esque removal detection a la https://github.com/bevyengine/rfcs/pull/44.
Just tries to make the current workflow a bit more user friendly so detecting removals isn't such a scheduling nightmare.
---
## Changelog
- RemovedComponents<T> is now backed by an `Events<Entity>` for the benefits of double buffering.
## Migration Guide
- Add a `mut` for `removed: RemovedComponents<T>` since we are now modifying an event reader internally.
- Iterating over removed components now requires `&mut removed_components` or `removed_components.iter()` instead of `&removed_components`.
# Objective
The trait method `SystemParam::apply` allows a `SystemParam` type to defer world mutations, which is internally used to apply `Commands` at the end of the stage. Any operations that require `&mut World` access must be deferred in this way, since parallel systems do not have exclusive access to the world.
The `ExclusiveSystemParam` trait (added in #6083) has an `apply` method which serves the same purpose. However, deferring mutations in this way does not make sense for exclusive systems since they already have `&mut World` access: there is no need to wait until a hard sync point, as the system *is* a hard sync point. World mutations can and should be performed within the body of the system.
## Solution
Remove the method. There were no implementations of this method in the engine.
---
## Changelog
*Note for maintainers: this changelog makes more sense if it's placed above the one for #6919.*
- Removed the method `ExclusiveSystemParamState::apply`.
## Migration Guide
*Note for maintainers: this migration guide makes more sense if it's placed above the one for #6919.*
The trait method `ExclusiveSystemParamState::apply` has been removed. If you have an exclusive system with buffers that must be applied, you should apply them within the body of the exclusive system.
# Objective
Fixes#7434.
This is my first time contributing to a Rust project, so please let me know if this wasn't the change intended by the linked issue.
## Solution
Adds a test with a system that panics to `bevy_ecs`.
I'm not sure if this is the intended panic message, but this is what the test currently results in:
```
thread 'system::tests::panic_inside_system' panicked at 'called `Option::unwrap()` on a `None` value', /Users/bjorn/workplace/bevy/crates/bevy_tasks/src/task_pool.rs:354:49
```
# Objective
Fix#7447.
The `SystemParam` derive uses the wrong lifetimes for ignored fields.
## Solution
Use type inference instead of explicitly naming the types of ignored fields. This allows the compiler to automatically use the correct lifetime.
# Objective
- Fix panic_when_hierachy_cycle test hanging
- The problem is that the scope only awaits one task at a time in get_results. In stageless this task is the multithreaded executor. That tasks hangs when a system panics and cannot make anymore progress. This wasn't a problem before because the executor was spawned after all the system tasks had been spawned. But in stageless the executor is spawned before all the system tasks are spawned.
## Solution
- We can catch unwind on each system and close the finish channel if one panics. This then causes the receiver end of the finish channel to panic too.
- this might have a small perf impact, but when running many_foxes it seems to be within the noise. So less than 40us.
## Other possible solutions
- It might be possible to fairly poll all the tasks in get_results in the scope. If we could do that then the scope could panic whenever one of tasks panics. It would require a data structure that we could both poll the futures through a shared ref and also push to it. I tried FuturesUnordered, but it requires an exclusive ref to poll it.
- The catch unwind could be moved onto when we create the tasks for scope instead. We would then need something like a oneshot async channel to inform get_results if a task panics.
# Objective
Ability to use `ReflectComponent` methods in dynamic type contexts with no access to `&World`.
This problem occurred to me when wanting to apply reflected types to an entity where the `&World` reference was already consumed by query iterator leaving only `EntityMut`.
## Solution
- Remove redundant `EntityMut` or `EntityRef` lookup from `World` and `Entity` in favor of taking `EntityMut` directly in `ReflectComponentFns`.
- Added `RefectComponent::contains` to determine without panic whether `apply` can be used.
## Changelog
- Changed function signatures of `ReflectComponent` methods, `apply`, `remove`, `contains`, and `reflect`.
## Migration Guide
- Call `World::entity` before calling into the changed `ReflectComponent` methods, most likely user already has a `EntityRef` or `EntityMut` which was being queried redundantly.
# Objective
- Trying to move some of the fixes from https://github.com/bevyengine/bevy/pull/7267 to make that one easier to review
- The MainThreadExecutor is how the render world runs nonsend systems on the main thread for pipelined rendering.
- The multithread executor for stageless wasn't using the MainThreadExecutor.
- MainThreadExecutor was declared in the old executor_parallel module that is getting deleted.
- The way the MainThreadExecutor was getting passed to the scope was actually unsound as the resource could be dropped from the World while the schedule was running
## Solution
- Move MainThreadExecutor to the new multithreaded_executor's file.
- Make the multithreaded executor use the MainThreadExecutor
- Clone the MainThreadExecutor onto the stack and pass that ref in
## Changelog
- Move MainThreadExecutor for stageless migration.
# Objective
- After the multithreaded executor finishes running all the systems, we apply the buffers for any system that hasn't applied it's buffers. This is a courtesy apply for users who forget to order their systems before a apply_system_buffers. When checking stageless, it was found that this apply_system_buffers was running on the executor thread instead of the world's thread. This is a problem because anything with world access should be able to access nonsend resources.
## Solution
- Move the final apply_system_buffers outside of the executor and outside of the scope, so it runs on the same thread that schedule.run is called on.
# Objective
- The stageless executor keeps track of systems that have run, but have not applied their system buffers. The bitset for that was being cloned into apply_system_buffers and cleared in that function, but we need to clear the original version instead of the cloned version
## Solution
- move the clear out of the apply_system_buffers function.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
Clearing the reader doesn't require iterating the events. Updating the `last_event_count` of the reader is enough.
I rewrote part of the documentation as some of it was incorrect or harder to understand than necessary.
## Changelog
Added `ManualEventReader::clear()`
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Found while working on #7385.
The struct `EntityMut` has the safety invariant that it's cached `EntityLocation` must always accurately specify where the entity is stored. Thus, any time its location might be invalidated (such as by calling `EntityMut::world_mut` and moving archetypes), the cached location *must* be updated by calling `EntityMut::update_location`.
The method `world_scope` encapsulates this pattern in safe API by requiring world mutations to be done in a closure, after which `update_location` will automatically be called. However, this method has a soundness hole: if a panic occurs within the closure, then `update_location` will never get called. If the panic is caught in an outer scope, then the `EntityMut` will be left with an outdated location, which is undefined behavior.
An example of this can be seen in the unit test `entity_mut_world_scope_panic`, which has been added to this PR as a regression test. Without the other changes in this PR, that test will invoke undefined behavior in safe code.
## Solution
Call `EntityMut::update_location()` from within a `Drop` impl, which ensures that it will get executed even if `EntityMut::world_scope` unwinds.
# Objective
The function `EntityMut::world_scope` is a safe abstraction that allows you to temporarily get mutable access to the underlying `World` of an `EntityMut`. This function is purely stateful, meaning it is not easily possible to return a value from it.
## Solution
Allow returning a computed value from the closure. This is similar to how `World::resource_scope` works.
---
## Changelog
- The function `EntityMut::world_scope` now allows returning a value from the immediately-computed closure.
# Objective
I found several words in code and docs are incorrect. This should be fixed.
## Solution
- Fix several minor typos
Co-authored-by: Chris Ohk <utilforever@gmail.com>
alternative to #5922, implements #5956
builds on top of https://github.com/bevyengine/bevy/pull/6402
# Objective
https://github.com/bevyengine/bevy/issues/5956 goes into more detail, but the TLDR is:
- bevy systems ensure disjoint accesses to resources and components, and for that to work there are methods `World::get_resource_unchecked_mut(&self)`, ..., `EntityRef::get_mut_unchecked(&self)` etc.
- we don't have these unchecked methods for `by_id` variants, so third-party crate authors cannot build their own safe disjoint-access abstractions with these
- having `_unchecked_mut` methods is not great, because in their presence safe code can accidentally violate subtle invariants. Having to go through `world.as_unsafe_world_cell().unsafe_method()` forces you to stop and think about what you want to write in your `// SAFETY` comment.
The alternative is to keep exposing `_unchecked_mut` variants for every operation that we want third-party crates to build upon, but we'd prefer to avoid using these methods alltogether: https://github.com/bevyengine/bevy/pull/5922#issuecomment-1241954543
Also, this is something that **cannot be implemented outside of bevy**, so having either this PR or #5922 as an escape hatch with lots of discouraging comments would be great.
## Solution
- add `UnsafeWorldCell` with `unsafe fn get_resource(&self)`, `unsafe fn get_resource_mut(&self)`
- add `fn World::as_unsafe_world_cell(&mut self) -> UnsafeWorldCell<'_>` (and `as_unsafe_world_cell_readonly(&self)`)
- add `UnsafeWorldCellEntityRef` with `unsafe fn get`, `unsafe fn get_mut` and the other utilities on `EntityRef` (no methods for spawning, despawning, insertion)
- use the `UnsafeWorldCell` abstraction in `ReflectComponent`, `ReflectResource` and `ReflectAsset`, so these APIs are easier to reason about
- remove `World::get_resource_mut_unchecked`, `EntityRef::get_mut_unchecked` and use `unsafe { world.as_unsafe_world_cell().get_mut() }` and `unsafe { world.as_unsafe_world_cell().get_entity(entity)?.get_mut() }` instead
This PR does **not** make use of `UnsafeWorldCell` for anywhere else in `bevy_ecs` such as `SystemParam` or `Query`. That is a much larger change, and I am convinced that having `UnsafeWorldCell` is already useful for third-party crates.
Implemented API:
```rust
struct World { .. }
impl World {
fn as_unsafe_world_cell(&self) -> UnsafeWorldCell<'_>;
}
struct UnsafeWorldCell<'w>(&'w World);
impl<'w> UnsafeWorldCell {
unsafe fn world(&self) -> &World;
fn get_entity(&self) -> UnsafeWorldCellEntityRef<'w>; // returns 'w which is `'self` of the `World::as_unsafe_world_cell(&'w self)`
unsafe fn get_resource<T>(&self) -> Option<&'w T>;
unsafe fn get_resource_by_id(&self, ComponentId) -> Option<&'w T>;
unsafe fn get_resource_mut<T>(&self) -> Option<Mut<'w, T>>;
unsafe fn get_resource_mut_by_id(&self) -> Option<MutUntyped<'w>>;
unsafe fn get_non_send_resource<T>(&self) -> Option<&'w T>;
unsafe fn get_non_send_resource_mut<T>(&self) -> Option<Mut<'w, T>>>;
// not included: remove, remove_resource, despawn, anything that might change archetypes
}
struct UnsafeWorldCellEntityRef<'w> { .. }
impl UnsafeWorldCellEntityRef<'w> {
unsafe fn get<T>(&self, Entity) -> Option<&'w T>;
unsafe fn get_by_id(&self, Entity, ComponentId) -> Option<Ptr<'w>>;
unsafe fn get_mut<T>(&self, Entity) -> Option<Mut<'w, T>>;
unsafe fn get_mut_by_id(&self, Entity, ComponentId) -> Option<MutUntyped<'w>>;
unsafe fn get_change_ticks<T>(&self, Entity) -> Option<Mut<'w, T>>;
// fn id, archetype, contains, contains_id, containts_type_id
}
```
<details>
<summary>UnsafeWorldCell docs</summary>
Variant of the [`World`] where resource and component accesses takes a `&World`, and the responsibility to avoid
aliasing violations are given to the caller instead of being checked at compile-time by rust's unique XOR shared rule.
### Rationale
In rust, having a `&mut World` means that there are absolutely no other references to the safe world alive at the same time,
without exceptions. Not even unsafe code can change this.
But there are situations where careful shared mutable access through a type is possible and safe. For this, rust provides the [`UnsafeCell`](std::cell::UnsafeCell)
escape hatch, which allows you to get a `*mut T` from a `&UnsafeCell<T>` and around which safe abstractions can be built.
Access to resources and components can be done uniquely using [`World::resource_mut`] and [`World::entity_mut`], and shared using [`World::resource`] and [`World::entity`].
These methods use lifetimes to check at compile time that no aliasing rules are being broken.
This alone is not enough to implement bevy systems where multiple systems can access *disjoint* parts of the world concurrently. For this, bevy stores all values of
resources and components (and [`ComponentTicks`](crate::component::ComponentTicks)) in [`UnsafeCell`](std::cell::UnsafeCell)s, and carefully validates disjoint access patterns using
APIs like [`System::component_access`](crate::system::System::component_access).
A system then can be executed using [`System::run_unsafe`](crate::system::System::run_unsafe) with a `&World` and use methods with interior mutability to access resource values.
access resource values.
### Example Usage
[`UnsafeWorldCell`] can be used as a building block for writing APIs that safely allow disjoint access into the world.
In the following example, the world is split into a resource access half and a component access half, where each one can
safely hand out mutable references.
```rust
use bevy_ecs::world::World;
use bevy_ecs::change_detection::Mut;
use bevy_ecs::system::Resource;
use bevy_ecs::world::unsafe_world_cell_world::UnsafeWorldCell;
// INVARIANT: existance of this struct means that users of it are the only ones being able to access resources in the world
struct OnlyResourceAccessWorld<'w>(UnsafeWorldCell<'w>);
// INVARIANT: existance of this struct means that users of it are the only ones being able to access components in the world
struct OnlyComponentAccessWorld<'w>(UnsafeWorldCell<'w>);
impl<'w> OnlyResourceAccessWorld<'w> {
fn get_resource_mut<T: Resource>(&mut self) -> Option<Mut<'w, T>> {
// SAFETY: resource access is allowed through this UnsafeWorldCell
unsafe { self.0.get_resource_mut::<T>() }
}
}
// impl<'w> OnlyComponentAccessWorld<'w> {
// ...
// }
// the two interior mutable worlds borrow from the `&mut World`, so it cannot be accessed while they are live
fn split_world_access(world: &mut World) -> (OnlyResourceAccessWorld<'_>, OnlyComponentAccessWorld<'_>) {
let resource_access = OnlyResourceAccessWorld(unsafe { world.as_unsafe_world_cell() });
let component_access = OnlyComponentAccessWorld(unsafe { world.as_unsafe_world_cell() });
(resource_access, component_access)
}
```
</details>
# Objective
`add_system(system)` without any `.in_set` configuration should land in `CoreSet::Update`.
We check that the sets are empty, but for systems there is always the `SystemTypeset`.
## Solution
- instead of `is_empty()`, check that the only set it the `SystemTypeSet`
# Objective
`bevy_ecs/system_param.rs` contains many seemingly-arbitrary struct definitions which serve as compile tests.
## Solution
Add a comment to each one, linking the issue or PR that motivated its addition.
# Objective
- `Components::resource_id` doesn't exist. Like `Components::component_id` but for resources.
## Solution
- Created `Components::resource_id` and added some docs.
---
## Changelog
- Added `Components::resource_id`.
- Changed `World::init_resource` to return the generated `ComponentId`.
- Changed `World::init_non_send_resource` to return the generated `ComponentId`.
# Objective
Fixes#3184. Fixes#6640. Fixes#4798. Using `Query::par_for_each(_mut)` currently requires a `batch_size` parameter, which affects how it chunks up large archetypes and tables into smaller chunks to run in parallel. Tuning this value is difficult, as the performance characteristics entirely depends on the state of the `World` it's being run on. Typically, users will just use a flat constant and just tune it by hand until it performs well in some benchmarks. However, this is both error prone and risks overfitting the tuning on that benchmark.
This PR proposes a naive automatic batch-size computation based on the current state of the `World`.
## Background
`Query::par_for_each(_mut)` schedules a new Task for every archetype or table that it matches. Archetypes/tables larger than the batch size are chunked into smaller tasks. Assuming every entity matched by the query has an identical workload, this makes the worst case scenario involve using a batch size equal to the size of the largest matched archetype or table. Conversely, a batch size of `max {archetype, table} size / thread count * COUNT_PER_THREAD` is likely the sweetspot where the overhead of scheduling tasks is minimized, at least not without grouping small archetypes/tables together.
There is also likely a strict minimum batch size below which the overhead of scheduling these tasks is heavier than running the entire thing single-threaded.
## Solution
- [x] Remove the `batch_size` from `Query(State)::par_for_each` and friends.
- [x] Add a check to compute `batch_size = max {archeytpe/table} size / thread count * COUNT_PER_THREAD`
- [x] ~~Panic if thread count is 0.~~ Defer to `for_each` if the thread count is 1 or less.
- [x] Early return if there is no matched table/archetype.
- [x] Add override option for users have queries that strongly violate the initial assumption that all iterated entities have an equal workload.
---
## Changelog
Changed: `Query::par_for_each(_mut)` has been changed to `Query::par_iter(_mut)` and will now automatically try to produce a batch size for callers based on the current `World` state.
## Migration Guide
The `batch_size` parameter for `Query(State)::par_for_each(_mut)` has been removed. These calls will automatically compute a batch size for you. Remove these parameters from all calls to these functions.
Before:
```rust
fn parallel_system(query: Query<&MyComponent>) {
query.par_for_each(32, |comp| {
...
});
}
```
After:
```rust
fn parallel_system(query: Query<&MyComponent>) {
query.par_iter().for_each(|comp| {
...
});
}
```
Co-authored-by: Arnav Choubey <56453634+x-52@users.noreply.github.com>
Co-authored-by: Robert Swain <robert.swain@gmail.com>
Co-authored-by: François <mockersf@gmail.com>
Co-authored-by: Corey Farwell <coreyf@rwell.org>
Co-authored-by: Aevyrie <aevyrie@gmail.com>
# Objective
- Implement pipelined rendering
- Fixes#5082
- Fixes#4718
## User Facing Description
Bevy now implements piplelined rendering! Pipelined rendering allows the app logic and rendering logic to run on different threads leading to large gains in performance.
*tracy capture of many_foxes example*
To use pipelined rendering, you just need to add the `PipelinedRenderingPlugin`. If you're using `DefaultPlugins` then it will automatically be added for you on all platforms except wasm. Bevy does not currently support multithreading on wasm which is needed for this feature to work. If you aren't using `DefaultPlugins` you can add the plugin manually.
```rust
use bevy::prelude::*;
use bevy::render::pipelined_rendering::PipelinedRenderingPlugin;
fn main() {
App::new()
// whatever other plugins you need
.add_plugin(RenderPlugin)
// needs to be added after RenderPlugin
.add_plugin(PipelinedRenderingPlugin)
.run();
}
```
If for some reason pipelined rendering needs to be removed. You can also disable the plugin the normal way.
```rust
use bevy::prelude::*;
use bevy::render::pipelined_rendering::PipelinedRenderingPlugin;
fn main() {
App::new.add_plugins(DefaultPlugins.build().disable::<PipelinedRenderingPlugin>());
}
```
### A setup function was added to plugins
A optional plugin lifecycle function was added to the `Plugin trait`. This function is called after all plugins have been built, but before the app runner is called. This allows for some final setup to be done. In the case of pipelined rendering, the function removes the sub app from the main app and sends it to the render thread.
```rust
struct MyPlugin;
impl Plugin for MyPlugin {
fn build(&self, app: &mut App) {
}
// optional function
fn setup(&self, app: &mut App) {
// do some final setup before runner is called
}
}
```
### A Stage for Frame Pacing
In the `RenderExtractApp` there is a stage labelled `BeforeIoAfterRenderStart` that systems can be added to. The specific use case for this stage is for a frame pacing system that can delay the start of main app processing in render bound apps to reduce input latency i.e. "frame pacing". This is not currently built into bevy, but exists as `bevy`
```text
|-------------------------------------------------------------------|
| | BeforeIoAfterRenderStart | winit events | main schedule |
| extract |---------------------------------------------------------|
| | extract commands | rendering schedule |
|-------------------------------------------------------------------|
```
### Small API additions
* `Schedule::remove_stage`
* `App::insert_sub_app`
* `App::remove_sub_app`
* `TaskPool::scope_with_executor`
## Problems and Solutions
### Moving render app to another thread
Most of the hard bits for this were done with the render redo. This PR just sends the render app back and forth through channels which seems to work ok. I originally experimented with using a scope to run the render task. It was cuter, but that approach didn't allow render to start before i/o processing. So I switched to using channels. There is much complexity in the coordination that needs to be done, but it's worth it. By moving rendering during i/o processing the frame times should be much more consistent in render bound apps. See https://github.com/bevyengine/bevy/issues/4691.
### Unsoundness with Sending World with NonSend resources
Dropping !Send things on threads other than the thread they were spawned on is considered unsound. The render world doesn't have any nonsend resources. So if we tell the users to "pretty please don't spawn nonsend resource on the render world", we can avoid this problem.
More seriously there is this https://github.com/bevyengine/bevy/pull/6534 pr, which patches the unsoundness by aborting the app if a nonsend resource is dropped on the wrong thread. ~~That PR should probably be merged before this one.~~ For a longer term solution we have this discussion going https://github.com/bevyengine/bevy/discussions/6552.
### NonSend Systems in render world
The render world doesn't have any !Send resources, but it does have a non send system. While Window is Send, winit does have some API's that can only be accessed on the main thread. `prepare_windows` in the render schedule thus needs to be scheduled on the main thread. Currently we run nonsend systems by running them on the thread the TaskPool::scope runs on. When we move render to another thread this no longer works.
To fix this, a new `scope_with_executor` method was added that takes a optional `TheadExecutor` that can only be ticked on the thread it was initialized on. The render world then holds a `MainThreadExecutor` resource which can be passed to the scope in the parallel executor that it uses to spawn it's non send systems on.
### Scopes executors between render and main should not share tasks
Since the render world and the app world share the `ComputeTaskPool`. Because `scope` has executors for the ComputeTaskPool a system from the main world could run on the render thread or a render system could run on the main thread. This can cause performance problems because it can delay a stage from finishing. See https://github.com/bevyengine/bevy/pull/6503#issuecomment-1309791442 for more details.
To avoid this problem, `TaskPool::scope` has been changed to not tick the ComputeTaskPool when it's used by the parallel executor. In the future when we move closer to the 1 thread to 1 logical core model we may want to overprovide threads, because the render and main app threads don't do much when executing the schedule.
## Performance
My machine is Windows 11, AMD Ryzen 5600x, RX 6600
### Examples
#### This PR with pipelining vs Main
> Note that these were run on an older version of main and the performance profile has probably changed due to optimizations
Seeing a perf gain from 29% on many lights to 7% on many sprites.
<html>
<body>
<!--StartFragment--><google-sheets-html-origin>
| percent | | | Diff | | | Main | | | PR | |
-- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | --
tracy frame time | mean | median | sigma | mean | median | sigma | mean | median | sigma | mean | median | sigma
many foxes | 27.01% | 27.34% | -47.09% | 1.58 | 1.55 | -1.78 | 5.85 | 5.67 | 3.78 | 4.27 | 4.12 | 5.56
many lights | 29.35% | 29.94% | -10.84% | 3.02 | 3.03 | -0.57 | 10.29 | 10.12 | 5.26 | 7.27 | 7.09 | 5.83
many animated sprites | 13.97% | 15.69% | 14.20% | 3.79 | 4.17 | 1.41 | 27.12 | 26.57 | 9.93 | 23.33 | 22.4 | 8.52
3d scene | 25.79% | 26.78% | 7.46% | 0.49 | 0.49 | 0.15 | 1.9 | 1.83 | 2.01 | 1.41 | 1.34 | 1.86
many cubes | 11.97% | 11.28% | 14.51% | 1.93 | 1.78 | 1.31 | 16.13 | 15.78 | 9.03 | 14.2 | 14 | 7.72
many sprites | 7.14% | 9.42% | -85.42% | 1.72 | 2.23 | -6.15 | 24.09 | 23.68 | 7.2 | 22.37 | 21.45 | 13.35
<!--EndFragment-->
</body>
</html>
#### This PR with pipelining disabled vs Main
Mostly regressions here. I don't think this should be a problem as users that are disabling pipelined rendering are probably running single threaded and not using the parallel executor. The regression is probably mostly due to the switch to use `async_executor::run` instead of `try_tick` and also having one less thread to run systems on. I'll do a writeup on why switching to `run` causes regressions, so we can try to eventually fix it. Using try_tick causes issues when pipeline rendering is enable as seen [here](https://github.com/bevyengine/bevy/pull/6503#issuecomment-1380803518)
<html>
<body>
<!--StartFragment--><google-sheets-html-origin>
| percent | | | Diff | | | Main | | | PR no pipelining | |
-- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | --
tracy frame time | mean | median | sigma | mean | median | sigma | mean | median | sigma | mean | median | sigma
many foxes | -3.72% | -4.42% | -1.07% | -0.21 | -0.24 | -0.04 | 5.64 | 5.43 | 3.74 | 5.85 | 5.67 | 3.78
many lights | 0.29% | -0.30% | 4.75% | 0.03 | -0.03 | 0.25 | 10.29 | 10.12 | 5.26 | 10.26 | 10.15 | 5.01
many animated sprites | 0.22% | 1.81% | -2.72% | 0.06 | 0.48 | -0.27 | 27.12 | 26.57 | 9.93 | 27.06 | 26.09 | 10.2
3d scene | -15.79% | -14.75% | -31.34% | -0.3 | -0.27 | -0.63 | 1.9 | 1.83 | 2.01 | 2.2 | 2.1 | 2.64
many cubes | -2.85% | -3.30% | 0.00% | -0.46 | -0.52 | 0 | 16.13 | 15.78 | 9.03 | 16.59 | 16.3 | 9.03
many sprites | 2.49% | 2.41% | 0.69% | 0.6 | 0.57 | 0.05 | 24.09 | 23.68 | 7.2 | 23.49 | 23.11 | 7.15
<!--EndFragment-->
</body>
</html>
### Benchmarks
Mostly the same except empty_systems has got a touch slower. The maybe_pipelining+1 column has the compute task pool with an extra thread over default added. This is because pipelining loses one thread over main to execute systems on, since the main thread no longer runs normal systems.
<details>
<summary>Click Me</summary>
```text
group main maybe-pipelining+1
----- ------------------------- ------------------
busy_systems/01x_entities_03_systems 1.07 30.7±1.32µs ? ?/sec 1.00 28.6±1.35µs ? ?/sec
busy_systems/01x_entities_06_systems 1.10 52.1±1.10µs ? ?/sec 1.00 47.2±1.08µs ? ?/sec
busy_systems/01x_entities_09_systems 1.00 74.6±1.36µs ? ?/sec 1.00 75.0±1.93µs ? ?/sec
busy_systems/01x_entities_12_systems 1.03 100.6±6.68µs ? ?/sec 1.00 98.0±1.46µs ? ?/sec
busy_systems/01x_entities_15_systems 1.11 128.5±3.53µs ? ?/sec 1.00 115.5±1.02µs ? ?/sec
busy_systems/02x_entities_03_systems 1.16 50.4±2.56µs ? ?/sec 1.00 43.5±3.00µs ? ?/sec
busy_systems/02x_entities_06_systems 1.00 87.1±1.27µs ? ?/sec 1.05 91.5±7.15µs ? ?/sec
busy_systems/02x_entities_09_systems 1.04 139.9±6.37µs ? ?/sec 1.00 134.0±1.06µs ? ?/sec
busy_systems/02x_entities_12_systems 1.05 179.2±3.47µs ? ?/sec 1.00 170.1±3.17µs ? ?/sec
busy_systems/02x_entities_15_systems 1.01 219.6±3.75µs ? ?/sec 1.00 218.1±2.55µs ? ?/sec
busy_systems/03x_entities_03_systems 1.10 70.6±2.33µs ? ?/sec 1.00 64.3±0.69µs ? ?/sec
busy_systems/03x_entities_06_systems 1.02 130.2±3.11µs ? ?/sec 1.00 128.0±1.34µs ? ?/sec
busy_systems/03x_entities_09_systems 1.00 195.0±10.11µs ? ?/sec 1.00 194.8±1.41µs ? ?/sec
busy_systems/03x_entities_12_systems 1.01 261.7±4.05µs ? ?/sec 1.00 259.8±4.11µs ? ?/sec
busy_systems/03x_entities_15_systems 1.00 318.0±3.04µs ? ?/sec 1.06 338.3±20.25µs ? ?/sec
busy_systems/04x_entities_03_systems 1.00 82.9±0.63µs ? ?/sec 1.02 84.3±0.63µs ? ?/sec
busy_systems/04x_entities_06_systems 1.01 181.7±3.65µs ? ?/sec 1.00 179.8±1.76µs ? ?/sec
busy_systems/04x_entities_09_systems 1.04 265.0±4.68µs ? ?/sec 1.00 255.3±1.98µs ? ?/sec
busy_systems/04x_entities_12_systems 1.00 335.9±3.00µs ? ?/sec 1.05 352.6±15.84µs ? ?/sec
busy_systems/04x_entities_15_systems 1.00 418.6±10.26µs ? ?/sec 1.08 450.2±39.58µs ? ?/sec
busy_systems/05x_entities_03_systems 1.07 114.3±0.95µs ? ?/sec 1.00 106.9±1.52µs ? ?/sec
busy_systems/05x_entities_06_systems 1.08 229.8±2.90µs ? ?/sec 1.00 212.3±4.18µs ? ?/sec
busy_systems/05x_entities_09_systems 1.03 329.3±1.99µs ? ?/sec 1.00 319.2±2.43µs ? ?/sec
busy_systems/05x_entities_12_systems 1.06 454.7±6.77µs ? ?/sec 1.00 430.1±3.58µs ? ?/sec
busy_systems/05x_entities_15_systems 1.03 554.6±6.15µs ? ?/sec 1.00 538.4±23.87µs ? ?/sec
contrived/01x_entities_03_systems 1.00 14.0±0.15µs ? ?/sec 1.08 15.1±0.21µs ? ?/sec
contrived/01x_entities_06_systems 1.04 28.5±0.37µs ? ?/sec 1.00 27.4±0.44µs ? ?/sec
contrived/01x_entities_09_systems 1.00 41.5±4.38µs ? ?/sec 1.02 42.2±2.24µs ? ?/sec
contrived/01x_entities_12_systems 1.06 55.9±1.49µs ? ?/sec 1.00 52.6±1.36µs ? ?/sec
contrived/01x_entities_15_systems 1.02 68.0±2.00µs ? ?/sec 1.00 66.5±0.78µs ? ?/sec
contrived/02x_entities_03_systems 1.03 25.2±0.38µs ? ?/sec 1.00 24.6±0.52µs ? ?/sec
contrived/02x_entities_06_systems 1.00 46.3±0.49µs ? ?/sec 1.04 48.1±4.13µs ? ?/sec
contrived/02x_entities_09_systems 1.02 70.4±0.99µs ? ?/sec 1.00 68.8±1.04µs ? ?/sec
contrived/02x_entities_12_systems 1.06 96.8±1.49µs ? ?/sec 1.00 91.5±0.93µs ? ?/sec
contrived/02x_entities_15_systems 1.02 116.2±0.95µs ? ?/sec 1.00 114.2±1.42µs ? ?/sec
contrived/03x_entities_03_systems 1.00 33.2±0.38µs ? ?/sec 1.01 33.6±0.45µs ? ?/sec
contrived/03x_entities_06_systems 1.00 62.4±0.73µs ? ?/sec 1.01 63.3±1.05µs ? ?/sec
contrived/03x_entities_09_systems 1.02 96.4±0.85µs ? ?/sec 1.00 94.8±3.02µs ? ?/sec
contrived/03x_entities_12_systems 1.01 126.3±4.67µs ? ?/sec 1.00 125.6±2.27µs ? ?/sec
contrived/03x_entities_15_systems 1.03 160.2±9.37µs ? ?/sec 1.00 156.0±1.53µs ? ?/sec
contrived/04x_entities_03_systems 1.02 41.4±3.39µs ? ?/sec 1.00 40.5±0.52µs ? ?/sec
contrived/04x_entities_06_systems 1.00 78.9±1.61µs ? ?/sec 1.02 80.3±1.06µs ? ?/sec
contrived/04x_entities_09_systems 1.02 121.8±3.97µs ? ?/sec 1.00 119.2±1.46µs ? ?/sec
contrived/04x_entities_12_systems 1.00 157.8±1.48µs ? ?/sec 1.01 160.1±1.72µs ? ?/sec
contrived/04x_entities_15_systems 1.00 197.9±1.47µs ? ?/sec 1.08 214.2±34.61µs ? ?/sec
contrived/05x_entities_03_systems 1.00 49.1±0.33µs ? ?/sec 1.01 49.7±0.75µs ? ?/sec
contrived/05x_entities_06_systems 1.00 95.0±0.93µs ? ?/sec 1.00 94.6±0.94µs ? ?/sec
contrived/05x_entities_09_systems 1.01 143.2±1.68µs ? ?/sec 1.00 142.2±2.00µs ? ?/sec
contrived/05x_entities_12_systems 1.00 191.8±2.03µs ? ?/sec 1.01 192.7±7.88µs ? ?/sec
contrived/05x_entities_15_systems 1.02 239.7±3.71µs ? ?/sec 1.00 235.8±4.11µs ? ?/sec
empty_systems/000_systems 1.01 47.8±0.67ns ? ?/sec 1.00 47.5±2.02ns ? ?/sec
empty_systems/001_systems 1.00 1743.2±126.14ns ? ?/sec 1.01 1761.1±70.10ns ? ?/sec
empty_systems/002_systems 1.01 2.2±0.04µs ? ?/sec 1.00 2.2±0.02µs ? ?/sec
empty_systems/003_systems 1.02 2.7±0.09µs ? ?/sec 1.00 2.7±0.16µs ? ?/sec
empty_systems/004_systems 1.00 3.1±0.11µs ? ?/sec 1.00 3.1±0.24µs ? ?/sec
empty_systems/005_systems 1.00 3.5±0.05µs ? ?/sec 1.11 3.9±0.70µs ? ?/sec
empty_systems/010_systems 1.00 5.5±0.12µs ? ?/sec 1.03 5.7±0.17µs ? ?/sec
empty_systems/015_systems 1.00 7.9±0.19µs ? ?/sec 1.06 8.4±0.16µs ? ?/sec
empty_systems/020_systems 1.00 10.4±1.25µs ? ?/sec 1.02 10.6±0.18µs ? ?/sec
empty_systems/025_systems 1.00 12.4±0.39µs ? ?/sec 1.14 14.1±1.07µs ? ?/sec
empty_systems/030_systems 1.00 15.1±0.39µs ? ?/sec 1.05 15.8±0.62µs ? ?/sec
empty_systems/035_systems 1.00 16.9±0.47µs ? ?/sec 1.07 18.0±0.37µs ? ?/sec
empty_systems/040_systems 1.00 19.3±0.41µs ? ?/sec 1.05 20.3±0.39µs ? ?/sec
empty_systems/045_systems 1.00 22.4±1.67µs ? ?/sec 1.02 22.9±0.51µs ? ?/sec
empty_systems/050_systems 1.00 24.4±1.67µs ? ?/sec 1.01 24.7±0.40µs ? ?/sec
empty_systems/055_systems 1.05 28.6±5.27µs ? ?/sec 1.00 27.2±0.70µs ? ?/sec
empty_systems/060_systems 1.02 29.9±1.64µs ? ?/sec 1.00 29.3±0.66µs ? ?/sec
empty_systems/065_systems 1.02 32.7±3.15µs ? ?/sec 1.00 32.1±0.98µs ? ?/sec
empty_systems/070_systems 1.00 33.0±1.42µs ? ?/sec 1.03 34.1±1.44µs ? ?/sec
empty_systems/075_systems 1.00 34.8±0.89µs ? ?/sec 1.04 36.2±0.70µs ? ?/sec
empty_systems/080_systems 1.00 37.0±1.82µs ? ?/sec 1.05 38.7±1.37µs ? ?/sec
empty_systems/085_systems 1.00 38.7±0.76µs ? ?/sec 1.05 40.8±0.83µs ? ?/sec
empty_systems/090_systems 1.00 41.5±1.09µs ? ?/sec 1.04 43.2±0.82µs ? ?/sec
empty_systems/095_systems 1.00 43.6±1.10µs ? ?/sec 1.04 45.2±0.99µs ? ?/sec
empty_systems/100_systems 1.00 46.7±2.27µs ? ?/sec 1.03 48.1±1.25µs ? ?/sec
```
</details>
## Migration Guide
### App `runner` and SubApp `extract` functions are now required to be Send
This was changed to enable pipelined rendering. If this breaks your use case please report it as these new bounds might be able to be relaxed.
## ToDo
* [x] redo benchmarking
* [x] reinvestigate the perf of the try_tick -> run change for task pool scope
# Objective
- Safety comments for the `CommandQueue` type are quite sparse and very imprecise. Sometimes, they are right for the wrong reasons or use circular reasoning.
## Solution
- Document previously-implicit safety invariants.
- Rewrite safety comments to actually reflect the specific invariants of each operation.
- Use `OwningPtr` instead of raw pointers, to encode an invariant in the type system instead of via comments.
- Use typed pointer methods when possible to increase reliability.
---
## Changelog
+ Added the function `OwningPtr::read_unaligned`.
# Objective
Repeated calls to `init_non_send_resource` currently overwrite the old value because the wrong storage is being checked.
## Solution
Use the correct storage. Add some tests.
## Notes
Without the fix, the new test fails with
```
thread 'world::tests::init_non_send_resource_does_not_overwrite' panicked at 'assertion failed: `(left == right)`
left: `1`,
right: `0`', crates/bevy_ecs/src/world/mod.rs:2267:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test world::tests::init_non_send_resource_does_not_overwrite ... FAILED
```
This was introduced by #7174 and it seems like a fairly straightforward oopsie.
# Objective
I was reading through the bevy_ecs code, trying to understand how everything works.
I was getting a bit confused when reading the doc comment for the `new_archetype` function; it looks like it doesn't create a new archetype but instead updates some internal state in the SystemParam to facility QueryIteration.
(I still couldn't find where a new archetype was actually created)
## Solution
- Adding a doc comment with a more correct explanation.
If it's deemed correct, I can also update the doc-comment for the other `new_archetype` calls
# Objective
Speed up the render phase of rendering. An extension of #6885.
`SystemState::get` increments the `World`'s change tick atomically every time it's called. This is notably more expensive than a unsynchronized increment, even without contention. It also updates the archetypes, even when there has been nothing to update when it's called repeatedly.
## Solution
Piggyback off of #6885. Split `SystemState::validate_world_and_update_archetypes` into `SystemState::validate_world` and `SystemState::update_archetypes`, and make the later `pub`. Then create safe variants of `SystemState::get_unchecked_manual` that still validate the `World` but do not update archetypes and do not increment the change tick using `World::read_change_tick` and `World::change_tick`. Update `RenderCommandState` to call `SystemState::update_archetypes` in `Draw::prepare` and `SystemState::get_manual` in `Draw::draw`.
## Performance
There's a slight perf benefit (~2%) for `main_opaque_pass_3d` on `many_foxes` (340.39 us -> 333.32 us)
## Alternatives
We can change `SystemState::get` to not increment the `World`'s change tick. Though this would still put updating the archetypes and an atomic read on the hot-path.
---
## Changelog
Added: `SystemState::get_manual`
Added: `SystemState::get_manual_mut`
Added: `SystemState::update_archetypes`
# Objective
- Fixes#3158
## Solution
- clear columns
My implementation of `clear_resources` do not remove the components itself but it clears the columns that keeps the resource data. I'm not sure if the issue meant to clear all resources, even the components and component ids (which I'm not sure if it's possible)
Co-authored-by: 2ne1ugly <47616772+2ne1ugly@users.noreply.github.com>
# Objective
The trait `ReadOnlySystemParam` is not implemented for `Option<NonSend<>>`, even though it should be.
Follow-up to #7243. This fixes another mistake made in #6919.
## Solution
Add the missing impl.
# Objective
The trait `ReadOnlySystemParam` is implemented for `NonSendMut`, when it should not be. This mistake was made in #6919.
## Solution
Remove the incorrect impl.
# Objective
Complete the first part of the migration detailed in bevyengine/rfcs#45.
## Solution
Add all the new stuff.
### TODO
- [x] Impl tuple methods.
- [x] Impl chaining.
- [x] Port ambiguity detection.
- [x] Write docs.
- [x] ~~Write more tests.~~(will do later)
- [ ] Write changelog and examples here?
- [x] ~~Replace `petgraph`.~~ (will do later)
Co-authored-by: james7132 <contact@jamessliu.com>
Co-authored-by: Michael Hsu <mike.hsu@gmail.com>
Co-authored-by: Mike Hsu <mike.hsu@gmail.com>
# Objective
- We rely on the construction of `EntityRef` to be valid elsewhere in unsafe code. This construction is not checked (for performance reasons), and thus this private method must be unsafe.
- Fixes#7218.
## Solution
- Make the method unsafe.
- Add safety docs.
- Improve safety docs slightly for the sibling `EntityMut::new`.
- Add debug asserts to start to verify these assumptions in debug mode.
## Context for reviewers
I attempted to verify the `EntityLocation` more thoroughly, but this turned out to be more work than expected. I've spun that off into #7221 as a result.
# Objective
The usages of the unsafe function `byte_add` are not properly documented.
Follow-up to #7151.
## Solution
Add safety comments to each call-site.
# Objective
- The function `BlobVec::replace_unchecked` has informal use of safety comments.
- This function does strange things with `OwningPtr` in order to get around the borrow checker.
## Solution
- Put safety comments in front of each unsafe operation. Describe the specific invariants of each operation and how they apply here.
- Added a guard type `OnDrop`, which is used to simplify ownership transfer in case of a panic.
---
## Changelog
+ Added the guard type `bevy_utils::OnDrop`.
+ Added conversions from `Ptr`, `PtrMut`, and `OwningPtr` to `NonNull<u8>`.
# Objective
Fix#5248.
## Solution
Support `In<T>` parameters and allow returning arbitrary types in exclusive systems.
---
## Changelog
- Exclusive systems may now be used with system piping.
## Migration Guide
Exclusive systems (systems that access `&mut World`) now support system piping, so the `ExclusiveSystemParamFunction` trait now has generics for the `In`put and `Out`put types.
```rust
// Before
fn my_generic_system<T, Param>(system_function: T)
where T: ExclusiveSystemParamFunction<Param>
{ ... }
// After
fn my_generic_system<T, In, Out, Param>(system_function: T)
where T: ExclusiveSystemParamFunction<In, Out, Param>
{ ... }
```
# Objective
There are some utility functions for actually working with `Storages` inside `entity_ref.rs` that are used both for `EntityRef/EntityMut` and `World`, with a `// TODO: move to Storages`.
This PR moves them to private methods on `World`, because that's the safest API boundary. On `Storages` you would need to ensure that you pass `Components` from the same world.
## Solution
- move get_component[_with_type], get_ticks[_with_type], get_component_and_ticks[_with_type] to `World` (still pub(crate))
- replace `pub use entity_ref::*;` with `pub use entity_ref::{EntityRef, EntityMut}` and qualified `entity_ref::get_mut[_by_id]` in `world.rs`
- add safety comments to a bunch of methods
# Objective
* `World::init_resource` and `World::get_resource_or_insert_with` are implemented naively, and as such they perform duplicate `TypeId -> ComponentId` lookups.
* `World::get_resource_or_insert_with` contains an additional duplicate `ComponentId -> ResourceData` lookup.
* This function also contains an unnecessary panic branch, which we rely on the optimizer to be able to remove.
## Solution
Implement the functions using engine-internal code, instead of combining high-level functions. This allows computed variables to persist across different branches, instead of being recomputed.
# Objective
Following #6681, both `TableRow` and `TableId` are now part of `EntityLocation`. However, the safety invariant on `EntityLocation` requires that all of the constituent fields are `repr(transprent)` or `repr(C)` and the bit pattern of all 1s must be valid. This is not true for `TableRow` and `TableId` currently.
## Solution
Mark `TableRow` and `TableId` to satisfy the safety requirement. Add safety comments on `ArchetypeId`, `ArchetypeRow`, `TableId` and `TableRow`.
# Objective
Improve safety testing when using `bevy_ptr` types. This is a follow-up to #7113.
## Solution
Add a debug-only assertion that pointers are aligned when casting to a concrete type. This should very quickly catch any unsoundness from unaligned pointers, even without miri. However, this can have a large negative perf impact on debug builds.
---
## Changelog
Added: `Ptr::deref` will now panic in debug builds if the pointer is not aligned.
Added: `PtrMut::deref_mut` will now panic in debug builds if the pointer is not aligned.
Added: `OwningPtr::read` will now panic in debug builds if the pointer is not aligned.
Added: `OwningPtr::drop_as` will now panic in debug builds if the pointer is not aligned.
# Objective
`MutUntyped` is a struct that stores a `PtrMut` alongside change tick metadata. Working with this type is cumbersome, and has few benefits over storing the pointer and change ticks separately.
Related: #6430 (title is out of date)
## Solution
Add a convenience method for transforming an untyped change detection pointer into its typed counterpart.
---
## Changelog
- Added the method `MutUntyped::with_type`.
# Objective
- Fixes#7066
## Solution
- Split the ChangeDetection trait into ChangeDetection and ChangeDetectionMut
- Added Ref as equivalent to &T with change detection
---
## Changelog
- Support for Ref which allow inspecting change detection flags in an immutable way
## Migration Guide
- While bevy prelude includes both ChangeDetection and ChangeDetectionMut any code explicitly referencing ChangeDetection might need to be updated to ChangeDetectionMut or both. Specifically any reading logic requires ChangeDetection while writes requires ChangeDetectionMut.
use bevy_ecs::change_detection::DetectChanges -> use bevy_ecs::change_detection::{DetectChanges, DetectChangesMut}
- Previously Res had methods to access change detection `is_changed` and `is_added` those methods have been moved to the `DetectChanges` trait. If you are including bevy prelude you will have access to these types otherwise you will need to `use bevy_ecs::change_detection::DetectChanges` to continue using them.
`Query`'s fields being `pub(crate)` means that the struct can be constructed via safe code from anywhere in `bevy_ecs` . This is Not Good since it is intended that all construction of this type goes through `Query::new` which is an `unsafe fn` letting various `Query` methods rely on those invariants holding even though they can be trivially bypassed.
This has no user facing impact
# Objective
- Fix#7103.
- The issue is caused because I forgot to add a where clause to a generated struct in #7056.
## Solution
- Add the where clause.
