bevy_render/mesh/

allocator.rs

//! Manages mesh vertex and index buffers.

use alloc::vec::Vec;
use bevy_mesh::Indices;
use core::{
    fmt::{self, Display, Formatter},
    ops::Range,
};
use nonmax::NonMaxU32;

use bevy_app::{App, Plugin};
use bevy_asset::AssetId;
use bevy_derive::{Deref, DerefMut};
use bevy_ecs::{
    resource::Resource,
    schedule::IntoScheduleConfigs as _,
    system::{Res, ResMut},
    world::{FromWorld, World},
};
use bevy_platform::collections::{hash_map::Entry, HashMap, HashSet};
use bevy_utils::default;
use offset_allocator::{Allocation, Allocator};
use tracing::error;
use wgpu::{
    BufferDescriptor, BufferSize, BufferUsages, CommandEncoderDescriptor, DownlevelFlags,
    COPY_BUFFER_ALIGNMENT,
};

use crate::{
    mesh::{Mesh, MeshVertexBufferLayouts, RenderMesh},
    render_asset::{prepare_assets, ExtractedAssets},
    render_resource::Buffer,
    renderer::{RenderAdapter, RenderDevice, RenderQueue},
    Render, RenderApp, RenderSystems,
};

/// A plugin that manages GPU memory for mesh data.
pub struct MeshAllocatorPlugin;

/// Manages the assignment of mesh data to GPU buffers.
///
/// The Bevy renderer tries to pack vertex and index data for multiple meshes
/// together so that multiple meshes can be drawn back-to-back without any
/// rebinding. This resource manages these buffers.
///
/// Within each slab, or hardware buffer, the underlying allocation algorithm is
/// [`offset_allocator`], a Rust port of Sebastian Aaltonen's hard-real-time C++
/// `OffsetAllocator`. Slabs start small and then grow as their contents fill
/// up, up to a maximum size limit. To reduce fragmentation, vertex and index
/// buffers that are too large bypass this system and receive their own buffers.
///
/// The [`MeshAllocatorSettings`] allows you to tune the behavior of the
/// allocator for better performance with your application. Most applications
/// won't need to change the settings from their default values.
#[derive(Resource)]
pub struct MeshAllocator {
    /// Holds all buffers and allocators.
    slabs: HashMap<SlabId, Slab>,

    /// Maps a layout to the slabs that hold elements of that layout.
    ///
    /// This is used when allocating, so that we can find the appropriate slab
    /// to place an object in.
    slab_layouts: HashMap<ElementLayout, Vec<SlabId>>,

    /// Maps mesh asset IDs to the ID of the slabs that hold their vertex data.
    mesh_id_to_vertex_slab: HashMap<AssetId<Mesh>, SlabId>,

    /// Maps mesh asset IDs to the ID of the slabs that hold their index data.
    mesh_id_to_index_slab: HashMap<AssetId<Mesh>, SlabId>,

    /// The next slab ID to assign.
    next_slab_id: SlabId,

    /// Whether we can pack multiple vertex arrays into a single slab on this
    /// platform.
    ///
    /// This corresponds to [`DownlevelFlags::BASE_VERTEX`], which is unset on
    /// WebGL 2. On this platform, we must give each vertex array its own
    /// buffer, because we can't adjust the first vertex when we perform a draw.
    general_vertex_slabs_supported: bool,

    /// Additional buffer usages to add to any vertex or index buffers created.
    pub extra_buffer_usages: BufferUsages,
}

/// Tunable parameters that customize the behavior of the allocator.
///
/// Generally, these parameters adjust the tradeoff between memory fragmentation
/// and performance. You can adjust them as desired for your application. Most
/// applications can stick with the default values.
#[derive(Resource)]
pub struct MeshAllocatorSettings {
    /// The minimum size of a slab (hardware buffer), in bytes.
    ///
    /// The default value is 1 MiB.
    pub min_slab_size: u64,

    /// The maximum size of a slab (hardware buffer), in bytes.
    ///
    /// When a slab reaches this limit, a new slab is created.
    ///
    /// The default value is 512 MiB.
    pub max_slab_size: u64,

    /// The maximum size of vertex or index data that can be placed in a general
    /// slab, in bytes.
    ///
    /// If a mesh has vertex or index data that exceeds this size limit, that
    /// data is placed in its own slab. This reduces fragmentation, but incurs
    /// more CPU-side binding overhead when drawing the mesh.
    ///
    /// The default value is 256 MiB.
    pub large_threshold: u64,

    /// The factor by which we scale a slab when growing it.
    ///
    /// This value must be greater than 1. Higher values result in more
    /// fragmentation but fewer expensive copy operations when growing the
    /// buffer.
    ///
    /// The default value is 1.5.
    pub growth_factor: f64,
}

impl Default for MeshAllocatorSettings {
    fn default() -> Self {
        Self {
            // 1 MiB
            min_slab_size: 1024 * 1024,
            // 512 MiB
            max_slab_size: 1024 * 1024 * 512,
            // 256 MiB
            large_threshold: 1024 * 1024 * 256,
            // 1.5× growth
            growth_factor: 1.5,
        }
    }
}

/// The hardware buffer that mesh data lives in, as well as the range within
/// that buffer.
pub struct MeshBufferSlice<'a> {
    /// The buffer that the mesh data resides in.
    pub buffer: &'a Buffer,

    /// The range of elements within this buffer that the mesh data resides in,
    /// measured in elements.
    ///
    /// This is not a byte range; it's an element range. For vertex data, this
    /// is measured in increments of a single vertex. (Thus, if a vertex is 32
    /// bytes long, then this range is in units of 32 bytes each.) For index
    /// data, this is measured in increments of a single index value (2 or 4
    /// bytes). Draw commands generally take their ranges in elements, not
    /// bytes, so this is the most convenient unit in this case.
    pub range: Range<u32>,
}

/// The index of a single slab.
#[derive(Clone, Copy, Default, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)]
#[repr(transparent)]
pub struct SlabId(pub NonMaxU32);

/// Data for a single slab.
#[expect(
    clippy::large_enum_variant,
    reason = "See https://github.com/bevyengine/bevy/issues/19220"
)]
enum Slab {
    /// A slab that can contain multiple objects.
    General(GeneralSlab),
    /// A slab that contains a single object.
    LargeObject(LargeObjectSlab),
}

/// A resizable slab that can contain multiple objects.
///
/// This is the normal type of slab used for objects that are below the
/// [`MeshAllocatorSettings::large_threshold`]. Slabs are divided into *slots*,
/// which are described in detail in the [`ElementLayout`] documentation.
struct GeneralSlab {
    /// The [`Allocator`] that manages the objects in this slab.
    allocator: Allocator,

    /// The GPU buffer that backs this slab.
    ///
    /// This may be `None` if the buffer hasn't been created yet. We delay
    /// creation of buffers until allocating all the meshes for a single frame,
    /// so that we don't needlessly create and resize buffers when many meshes
    /// load all at once.
    buffer: Option<Buffer>,

    /// Allocations that are on the GPU.
    ///
    /// The range is in slots.
    resident_allocations: HashMap<AssetId<Mesh>, SlabAllocation>,

    /// Allocations that are waiting to be uploaded to the GPU.
    ///
    /// The range is in slots.
    pending_allocations: HashMap<AssetId<Mesh>, SlabAllocation>,

    /// The layout of a single element (vertex or index).
    element_layout: ElementLayout,

    /// The size of this slab in slots.
    current_slot_capacity: u32,
}

/// A slab that contains a single object.
///
/// Typically, this is for objects that exceed the
/// [`MeshAllocatorSettings::large_threshold`]. This is also for objects that
/// would ordinarily receive their own slab but can't because of platform
/// limitations, most notably vertex arrays on WebGL 2.
struct LargeObjectSlab {
    /// The GPU buffer that backs this slab.
    ///
    /// This may be `None` if the buffer hasn't been created yet.
    buffer: Option<Buffer>,

    /// The layout of a single element (vertex or index).
    element_layout: ElementLayout,
}

/// The type of element that a slab can store.
#[derive(Clone, Copy, PartialEq, Eq, Hash)]
enum ElementClass {
    /// Data for a vertex.
    Vertex,
    /// A vertex index.
    Index,
}

/// The results of [`GeneralSlab::grow_if_necessary`].
enum SlabGrowthResult {
    /// The mesh data already fits in the slab; the slab doesn't need to grow.
    NoGrowthNeeded,
    /// The slab needed to grow.
    ///
    /// The [`SlabToReallocate`] contains the old capacity of the slab.
    NeededGrowth(SlabToReallocate),
    /// The slab wanted to grow but couldn't because it hit its maximum size.
    CantGrow,
}

/// Information about the size of individual elements (vertices or indices)
/// within a slab.
///
/// Slab objects are allocated in units of *slots*. Usually, each element takes
/// up one slot, and so elements and slots are equivalent. Occasionally,
/// however, a slot may consist of 2 or even 4 elements. This occurs when the
/// size of an element isn't divisible by [`COPY_BUFFER_ALIGNMENT`]. When we
/// resize buffers, we perform GPU-to-GPU copies to shuffle the existing
/// elements into their new positions, and such copies must be on
/// [`COPY_BUFFER_ALIGNMENT`] boundaries. Slots solve this problem by
/// guaranteeing that the size of an allocation quantum is divisible by both the
/// size of an element and [`COPY_BUFFER_ALIGNMENT`], so we can relocate it
/// freely.
#[derive(Clone, Copy, PartialEq, Eq, Hash)]
struct ElementLayout {
    /// Either a vertex or an index.
    class: ElementClass,

    /// The size in bytes of a single element (vertex or index).
    size: u64,

    /// The number of elements that make up a single slot.
    ///
    /// Usually, this is 1, but it can be different if [`ElementLayout::size`]
    /// isn't divisible by 4. See the comment in [`ElementLayout`] for more
    /// details.
    elements_per_slot: u32,
}

/// The location of an allocation and the slab it's contained in.
struct MeshAllocation {
    /// The ID of the slab.
    slab_id: SlabId,
    /// Holds the actual allocation.
    slab_allocation: SlabAllocation,
}

/// An allocation within a slab.
#[derive(Clone)]
struct SlabAllocation {
    /// The actual [`Allocator`] handle, needed to free the allocation.
    allocation: Allocation,
    /// The number of slots that this allocation takes up.
    slot_count: u32,
}

/// Holds information about all slabs scheduled to be allocated or reallocated.
#[derive(Default, Deref, DerefMut)]
struct SlabsToReallocate(HashMap<SlabId, SlabToReallocate>);

/// Holds information about a slab that's scheduled to be allocated or
/// reallocated.
#[derive(Default)]
struct SlabToReallocate {
    /// The capacity of the slab before we decided to grow it.
    old_slot_capacity: u32,
}

impl Display for SlabId {
    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
        self.0.fmt(f)
    }
}

impl Plugin for MeshAllocatorPlugin {
    fn build(&self, app: &mut App) {
        let Some(render_app) = app.get_sub_app_mut(RenderApp) else {
            return;
        };

        render_app
            .init_resource::<MeshAllocatorSettings>()
            .add_systems(
                Render,
                allocate_and_free_meshes
                    .in_set(RenderSystems::PrepareAssets)
                    .before(prepare_assets::<RenderMesh>),
            );
    }

    fn finish(&self, app: &mut App) {
        let Some(render_app) = app.get_sub_app_mut(RenderApp) else {
            return;
        };

        // The `RenderAdapter` isn't available until now, so we can't do this in
        // [`Plugin::build`].
        render_app.init_resource::<MeshAllocator>();
    }
}

impl FromWorld for MeshAllocator {
    fn from_world(world: &mut World) -> Self {
        // Note whether we're on WebGL 2. In this case, we must give every
        // vertex array its own slab.
        let render_adapter = world.resource::<RenderAdapter>();
        let general_vertex_slabs_supported = render_adapter
            .get_downlevel_capabilities()
            .flags
            .contains(DownlevelFlags::BASE_VERTEX);

        Self {
            slabs: HashMap::default(),
            slab_layouts: HashMap::default(),
            mesh_id_to_vertex_slab: HashMap::default(),
            mesh_id_to_index_slab: HashMap::default(),
            next_slab_id: default(),
            general_vertex_slabs_supported,
            extra_buffer_usages: BufferUsages::empty(),
        }
    }
}

/// A system that processes newly-extracted or newly-removed meshes and writes
/// their data into buffers or frees their data as appropriate.
pub fn allocate_and_free_meshes(
    mut mesh_allocator: ResMut<MeshAllocator>,
    mesh_allocator_settings: Res<MeshAllocatorSettings>,
    extracted_meshes: Res<ExtractedAssets<RenderMesh>>,
    mut mesh_vertex_buffer_layouts: ResMut<MeshVertexBufferLayouts>,
    render_device: Res<RenderDevice>,
    render_queue: Res<RenderQueue>,
) {
    // Process removed or modified meshes.
    mesh_allocator.free_meshes(&extracted_meshes);

    // Process newly-added or modified meshes.
    mesh_allocator.allocate_meshes(
        &mesh_allocator_settings,
        &extracted_meshes,
        &mut mesh_vertex_buffer_layouts,
        &render_device,
        &render_queue,
    );
}

impl MeshAllocator {
    /// Returns the buffer and range within that buffer of the vertex data for
    /// the mesh with the given ID.
    ///
    /// If the mesh wasn't allocated, returns None.
    pub fn mesh_vertex_slice(&self, mesh_id: &AssetId<Mesh>) -> Option<MeshBufferSlice<'_>> {
        self.mesh_slice_in_slab(mesh_id, *self.mesh_id_to_vertex_slab.get(mesh_id)?)
    }

    /// Returns the buffer and range within that buffer of the index data for
    /// the mesh with the given ID.
    ///
    /// If the mesh has no index data or wasn't allocated, returns None.
    pub fn mesh_index_slice(&self, mesh_id: &AssetId<Mesh>) -> Option<MeshBufferSlice<'_>> {
        self.mesh_slice_in_slab(mesh_id, *self.mesh_id_to_index_slab.get(mesh_id)?)
    }

    /// Returns the IDs of the vertex buffer and index buffer respectively for
    /// the mesh with the given ID.
    ///
    /// If the mesh wasn't allocated, or has no index data in the case of the
    /// index buffer, the corresponding element in the returned tuple will be
    /// None.
    pub fn mesh_slabs(&self, mesh_id: &AssetId<Mesh>) -> (Option<SlabId>, Option<SlabId>) {
        (
            self.mesh_id_to_vertex_slab.get(mesh_id).cloned(),
            self.mesh_id_to_index_slab.get(mesh_id).cloned(),
        )
    }

    /// Given a slab and a mesh with data located with it, returns the buffer
    /// and range of that mesh data within the slab.
    fn mesh_slice_in_slab(
        &self,
        mesh_id: &AssetId<Mesh>,
        slab_id: SlabId,
    ) -> Option<MeshBufferSlice<'_>> {
        match self.slabs.get(&slab_id)? {
            Slab::General(general_slab) => {
                let slab_allocation = general_slab.resident_allocations.get(mesh_id)?;
                Some(MeshBufferSlice {
                    buffer: general_slab.buffer.as_ref()?,
                    range: (slab_allocation.allocation.offset
                        * general_slab.element_layout.elements_per_slot)
                        ..((slab_allocation.allocation.offset + slab_allocation.slot_count)
                            * general_slab.element_layout.elements_per_slot),
                })
            }

            Slab::LargeObject(large_object_slab) => {
                let buffer = large_object_slab.buffer.as_ref()?;
                Some(MeshBufferSlice {
                    buffer,
                    range: 0..((buffer.size() / large_object_slab.element_layout.size) as u32),
                })
            }
        }
    }

    /// Processes newly-loaded meshes, allocating room in the slabs for their
    /// mesh data and performing upload operations as appropriate.
    fn allocate_meshes(
        &mut self,
        mesh_allocator_settings: &MeshAllocatorSettings,
        extracted_meshes: &ExtractedAssets<RenderMesh>,
        mesh_vertex_buffer_layouts: &mut MeshVertexBufferLayouts,
        render_device: &RenderDevice,
        render_queue: &RenderQueue,
    ) {
        let mut slabs_to_grow = SlabsToReallocate::default();

        // Allocate.
        for (mesh_id, mesh) in &extracted_meshes.extracted {
            let vertex_buffer_size = mesh.get_vertex_buffer_size() as u64;
            if vertex_buffer_size == 0 {
                continue;
            }
            // Allocate vertex data. Note that we can only pack mesh vertex data
            // together if the platform supports it.
            let vertex_element_layout = ElementLayout::vertex(mesh_vertex_buffer_layouts, mesh);
            if self.general_vertex_slabs_supported {
                self.allocate(
                    mesh_id,
                    vertex_buffer_size,
                    vertex_element_layout,
                    &mut slabs_to_grow,
                    mesh_allocator_settings,
                );
            } else {
                self.allocate_large(mesh_id, vertex_element_layout);
            }

            // Allocate index data.
            if let (Some(index_buffer_data), Some(index_element_layout)) =
                (mesh.get_index_buffer_bytes(), ElementLayout::index(mesh))
            {
                self.allocate(
                    mesh_id,
                    index_buffer_data.len() as u64,
                    index_element_layout,
                    &mut slabs_to_grow,
                    mesh_allocator_settings,
                );
            }
        }

        // Perform growth.
        for (slab_id, slab_to_grow) in slabs_to_grow.0 {
            self.reallocate_slab(render_device, render_queue, slab_id, slab_to_grow);
        }

        // Copy new mesh data in.
        for (mesh_id, mesh) in &extracted_meshes.extracted {
            self.copy_mesh_vertex_data(mesh_id, mesh, render_device, render_queue);
            self.copy_mesh_index_data(mesh_id, mesh, render_device, render_queue);
        }
    }

    /// Copies vertex array data from a mesh into the appropriate spot in the
    /// slab.
    fn copy_mesh_vertex_data(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        mesh: &Mesh,
        render_device: &RenderDevice,
        render_queue: &RenderQueue,
    ) {
        let Some(&slab_id) = self.mesh_id_to_vertex_slab.get(mesh_id) else {
            return;
        };

        // Call the generic function.
        self.copy_element_data(
            mesh_id,
            mesh.get_vertex_buffer_size(),
            |slice| mesh.write_packed_vertex_buffer_data(slice),
            BufferUsages::VERTEX,
            slab_id,
            render_device,
            render_queue,
        );
    }

    /// Copies index array data from a mesh into the appropriate spot in the
    /// slab.
    fn copy_mesh_index_data(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        mesh: &Mesh,
        render_device: &RenderDevice,
        render_queue: &RenderQueue,
    ) {
        let Some(&slab_id) = self.mesh_id_to_index_slab.get(mesh_id) else {
            return;
        };
        let Some(index_data) = mesh.get_index_buffer_bytes() else {
            return;
        };

        // Call the generic function.
        self.copy_element_data(
            mesh_id,
            index_data.len(),
            |slice| slice.copy_from_slice(index_data),
            BufferUsages::INDEX,
            slab_id,
            render_device,
            render_queue,
        );
    }

    /// A generic function that copies either vertex or index data into a slab.
    fn copy_element_data(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        len: usize,
        fill_data: impl Fn(&mut [u8]),
        buffer_usages: BufferUsages,
        slab_id: SlabId,
        render_device: &RenderDevice,
        render_queue: &RenderQueue,
    ) {
        let Some(slab) = self.slabs.get_mut(&slab_id) else {
            return;
        };

        match *slab {
            Slab::General(ref mut general_slab) => {
                let (Some(buffer), Some(allocated_range)) = (
                    &general_slab.buffer,
                    general_slab.pending_allocations.remove(mesh_id),
                ) else {
                    return;
                };

                let slot_size = general_slab.element_layout.slot_size();

                // round up size to a multiple of the slot size to satisfy wgpu alignment requirements
                if let Some(size) = BufferSize::new((len as u64).next_multiple_of(slot_size)) {
                    // Write the data in.
                    if let Some(mut buffer) = render_queue.write_buffer_with(
                        buffer,
                        allocated_range.allocation.offset as u64 * slot_size,
                        size,
                    ) {
                        let slice = &mut buffer.as_mut()[..len];
                        fill_data(slice);
                    }
                }

                // Mark the allocation as resident.
                general_slab
                    .resident_allocations
                    .insert(*mesh_id, allocated_range);
            }

            Slab::LargeObject(ref mut large_object_slab) => {
                debug_assert!(large_object_slab.buffer.is_none());

                // Create the buffer and its data in one go.
                let buffer = render_device.create_buffer(&BufferDescriptor {
                    label: Some(&format!(
                        "large mesh slab {} ({}buffer)",
                        slab_id,
                        buffer_usages_to_str(buffer_usages)
                    )),
                    size: len as u64,
                    usage: buffer_usages | BufferUsages::COPY_DST | self.extra_buffer_usages,
                    mapped_at_creation: true,
                });
                {
                    let slice = &mut buffer.slice(..).get_mapped_range_mut()[..len];
                    fill_data(slice);
                }
                buffer.unmap();
                large_object_slab.buffer = Some(buffer);
            }
        }
    }

    /// Frees allocations for meshes that were removed or modified this frame.
    fn free_meshes(&mut self, extracted_meshes: &ExtractedAssets<RenderMesh>) {
        let mut empty_slabs = <HashSet<_>>::default();

        // TODO: Consider explicitly reusing allocations for changed meshes of the same size
        let meshes_to_free = extracted_meshes
            .removed
            .iter()
            .chain(extracted_meshes.modified.iter());

        for mesh_id in meshes_to_free {
            if let Some(slab_id) = self.mesh_id_to_vertex_slab.remove(mesh_id) {
                self.free_allocation_in_slab(mesh_id, slab_id, &mut empty_slabs);
            }
            if let Some(slab_id) = self.mesh_id_to_index_slab.remove(mesh_id) {
                self.free_allocation_in_slab(mesh_id, slab_id, &mut empty_slabs);
            }
        }

        for empty_slab in empty_slabs {
            self.slab_layouts.values_mut().for_each(|slab_ids| {
                let idx = slab_ids.iter().position(|&slab_id| slab_id == empty_slab);
                if let Some(idx) = idx {
                    slab_ids.remove(idx);
                }
            });
            self.slabs.remove(&empty_slab);
        }
    }

    /// Given a slab and the ID of a mesh containing data in it, marks the
    /// allocation as free.
    ///
    /// If this results in the slab becoming empty, this function adds the slab
    /// to the `empty_slabs` set.
    fn free_allocation_in_slab(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        slab_id: SlabId,
        empty_slabs: &mut HashSet<SlabId>,
    ) {
        let Some(slab) = self.slabs.get_mut(&slab_id) else {
            return;
        };

        match *slab {
            Slab::General(ref mut general_slab) => {
                let Some(slab_allocation) = general_slab
                    .resident_allocations
                    .remove(mesh_id)
                    .or_else(|| general_slab.pending_allocations.remove(mesh_id))
                else {
                    return;
                };

                general_slab.allocator.free(slab_allocation.allocation);

                if general_slab.is_empty() {
                    empty_slabs.insert(slab_id);
                }
            }
            Slab::LargeObject(_) => {
                empty_slabs.insert(slab_id);
            }
        }
    }

    /// Allocates space for mesh data with the given byte size and layout in the
    /// appropriate slab, creating that slab if necessary.
    fn allocate(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        data_byte_len: u64,
        layout: ElementLayout,
        slabs_to_grow: &mut SlabsToReallocate,
        settings: &MeshAllocatorSettings,
    ) {
        let data_element_count = data_byte_len.div_ceil(layout.size) as u32;
        let data_slot_count = data_element_count.div_ceil(layout.elements_per_slot);

        // If the mesh data is too large for a slab, give it a slab of its own.
        if data_slot_count as u64 * layout.slot_size()
            >= settings.large_threshold.min(settings.max_slab_size)
        {
            self.allocate_large(mesh_id, layout);
        } else {
            self.allocate_general(mesh_id, data_slot_count, layout, slabs_to_grow, settings);
        }
    }

    /// Allocates space for mesh data with the given slot size and layout in the
    /// appropriate general slab.
    fn allocate_general(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        data_slot_count: u32,
        layout: ElementLayout,
        slabs_to_grow: &mut SlabsToReallocate,
        settings: &MeshAllocatorSettings,
    ) {
        let candidate_slabs = self.slab_layouts.entry(layout).or_default();

        // Loop through the slabs that accept elements of the appropriate type
        // and try to allocate the mesh inside them. We go with the first one
        // that succeeds.
        let mut mesh_allocation = None;
        for &slab_id in &*candidate_slabs {
            let Some(Slab::General(slab)) = self.slabs.get_mut(&slab_id) else {
                unreachable!("Slab not found")
            };

            let Some(allocation) = slab.allocator.allocate(data_slot_count) else {
                continue;
            };

            // Try to fit the object in the slab, growing if necessary.
            match slab.grow_if_necessary(allocation.offset + data_slot_count, settings) {
                SlabGrowthResult::NoGrowthNeeded => {}
                SlabGrowthResult::NeededGrowth(slab_to_reallocate) => {
                    // If we already grew the slab this frame, don't replace the
                    // `SlabToReallocate` entry. We want to keep the entry
                    // corresponding to the size that the slab had at the start
                    // of the frame, so that we can copy only the used portion
                    // of the initial buffer to the new one.
                    if let Entry::Vacant(vacant_entry) = slabs_to_grow.entry(slab_id) {
                        vacant_entry.insert(slab_to_reallocate);
                    }
                }
                SlabGrowthResult::CantGrow => continue,
            }

            mesh_allocation = Some(MeshAllocation {
                slab_id,
                slab_allocation: SlabAllocation {
                    allocation,
                    slot_count: data_slot_count,
                },
            });
            break;
        }

        // If we still have no allocation, make a new slab.
        if mesh_allocation.is_none() {
            let new_slab_id = self.next_slab_id;
            self.next_slab_id.0 = NonMaxU32::new(self.next_slab_id.0.get() + 1).unwrap_or_default();

            let new_slab = GeneralSlab::new(
                new_slab_id,
                &mut mesh_allocation,
                settings,
                layout,
                data_slot_count,
            );

            self.slabs.insert(new_slab_id, Slab::General(new_slab));
            candidate_slabs.push(new_slab_id);
            slabs_to_grow.insert(new_slab_id, SlabToReallocate::default());
        }

        let mesh_allocation = mesh_allocation.expect("Should have been able to allocate");

        // Mark the allocation as pending. Don't copy it in just yet; further
        // meshes loaded this frame may result in its final allocation location
        // changing.
        if let Some(Slab::General(general_slab)) = self.slabs.get_mut(&mesh_allocation.slab_id) {
            general_slab
                .pending_allocations
                .insert(*mesh_id, mesh_allocation.slab_allocation);
        };

        self.record_allocation(mesh_id, mesh_allocation.slab_id, layout.class);
    }

    /// Allocates an object into its own dedicated slab.
    fn allocate_large(&mut self, mesh_id: &AssetId<Mesh>, layout: ElementLayout) {
        let new_slab_id = self.next_slab_id;
        self.next_slab_id.0 = NonMaxU32::new(self.next_slab_id.0.get() + 1).unwrap_or_default();

        self.record_allocation(mesh_id, new_slab_id, layout.class);

        self.slabs.insert(
            new_slab_id,
            Slab::LargeObject(LargeObjectSlab {
                buffer: None,
                element_layout: layout,
            }),
        );
    }

    /// Reallocates a slab that needs to be resized, or allocates a new slab.
    ///
    /// This performs the actual growth operation that
    /// [`GeneralSlab::grow_if_necessary`] scheduled. We do the growth in two
    /// phases so that, if a slab grows multiple times in the same frame, only
    /// one new buffer is reallocated, rather than reallocating the buffer
    /// multiple times.
    fn reallocate_slab(
        &mut self,
        render_device: &RenderDevice,
        render_queue: &RenderQueue,
        slab_id: SlabId,
        slab_to_grow: SlabToReallocate,
    ) {
        let Some(Slab::General(slab)) = self.slabs.get_mut(&slab_id) else {
            error!("Couldn't find slab {} to grow", slab_id);
            return;
        };

        let old_buffer = slab.buffer.take();

        let mut buffer_usages = BufferUsages::COPY_SRC | BufferUsages::COPY_DST;
        match slab.element_layout.class {
            ElementClass::Vertex => buffer_usages |= BufferUsages::VERTEX,
            ElementClass::Index => buffer_usages |= BufferUsages::INDEX,
        };

        // Create the buffer.
        let new_buffer = render_device.create_buffer(&BufferDescriptor {
            label: Some(&format!(
                "general mesh slab {} ({}buffer)",
                slab_id,
                buffer_usages_to_str(buffer_usages)
            )),
            size: slab.current_slot_capacity as u64 * slab.element_layout.slot_size(),
            usage: buffer_usages | self.extra_buffer_usages,
            mapped_at_creation: false,
        });

        slab.buffer = Some(new_buffer.clone());

        let Some(old_buffer) = old_buffer else { return };

        // In order to do buffer copies, we need a command encoder.
        let mut encoder = render_device.create_command_encoder(&CommandEncoderDescriptor {
            label: Some("slab resize encoder"),
        });

        // Copy the data from the old buffer into the new one.
        encoder.copy_buffer_to_buffer(
            &old_buffer,
            0,
            &new_buffer,
            0,
            slab_to_grow.old_slot_capacity as u64 * slab.element_layout.slot_size(),
        );

        let command_buffer = encoder.finish();
        render_queue.submit([command_buffer]);
    }

    /// Records the location of the given newly-allocated mesh data in the
    /// [`Self::mesh_id_to_vertex_slab`] or [`Self::mesh_id_to_index_slab`]
    /// tables as appropriate.
    fn record_allocation(
        &mut self,
        mesh_id: &AssetId<Mesh>,
        slab_id: SlabId,
        element_class: ElementClass,
    ) {
        match element_class {
            ElementClass::Vertex => {
                self.mesh_id_to_vertex_slab.insert(*mesh_id, slab_id);
            }
            ElementClass::Index => {
                self.mesh_id_to_index_slab.insert(*mesh_id, slab_id);
            }
        }
    }
}

impl GeneralSlab {
    /// Creates a new growable slab big enough to hold a single element of
    /// `data_slot_count` size with the given `layout`.
    fn new(
        new_slab_id: SlabId,
        mesh_allocation: &mut Option<MeshAllocation>,
        settings: &MeshAllocatorSettings,
        layout: ElementLayout,
        data_slot_count: u32,
    ) -> GeneralSlab {
        let initial_slab_slot_capacity = (settings.min_slab_size.div_ceil(layout.slot_size())
            as u32)
            .max(offset_allocator::ext::min_allocator_size(data_slot_count));
        let max_slab_slot_capacity = (settings.max_slab_size.div_ceil(layout.slot_size()) as u32)
            .max(offset_allocator::ext::min_allocator_size(data_slot_count));

        let mut new_slab = GeneralSlab {
            allocator: Allocator::new(max_slab_slot_capacity),
            buffer: None,
            resident_allocations: HashMap::default(),
            pending_allocations: HashMap::default(),
            element_layout: layout,
            current_slot_capacity: initial_slab_slot_capacity,
        };

        // This should never fail.
        if let Some(allocation) = new_slab.allocator.allocate(data_slot_count) {
            *mesh_allocation = Some(MeshAllocation {
                slab_id: new_slab_id,
                slab_allocation: SlabAllocation {
                    slot_count: data_slot_count,
                    allocation,
                },
            });
        }

        new_slab
    }

    /// Checks to see if the size of this slab is at least `new_size_in_slots`
    /// and grows the slab if it isn't.
    ///
    /// The returned [`SlabGrowthResult`] describes whether the slab needed to
    /// grow and whether, if so, it was successful in doing so.
    fn grow_if_necessary(
        &mut self,
        new_size_in_slots: u32,
        settings: &MeshAllocatorSettings,
    ) -> SlabGrowthResult {
        // Is the slab big enough already?
        let initial_slot_capacity = self.current_slot_capacity;
        if self.current_slot_capacity >= new_size_in_slots {
            return SlabGrowthResult::NoGrowthNeeded;
        }

        // Try to grow in increments of `MeshAllocatorSettings::growth_factor`
        // until we're big enough.
        while self.current_slot_capacity < new_size_in_slots {
            let new_slab_slot_capacity =
                ((self.current_slot_capacity as f64 * settings.growth_factor).ceil() as u32)
                    .min((settings.max_slab_size / self.element_layout.slot_size()) as u32);
            if new_slab_slot_capacity == self.current_slot_capacity {
                // The slab is full.
                return SlabGrowthResult::CantGrow;
            }

            self.current_slot_capacity = new_slab_slot_capacity;
        }

        // Tell our caller what we did.
        SlabGrowthResult::NeededGrowth(SlabToReallocate {
            old_slot_capacity: initial_slot_capacity,
        })
    }
}

impl ElementLayout {
    /// Creates an [`ElementLayout`] for mesh data of the given class (vertex or
    /// index) with the given byte size.
    fn new(class: ElementClass, size: u64) -> ElementLayout {
        const {
            assert!(4 == COPY_BUFFER_ALIGNMENT);
        }
        // this is equivalent to `4 / gcd(4,size)` but lets us not implement gcd.
        // ping @atlv if above assert ever fails (likely never)
        let elements_per_slot = [1, 4, 2, 4][size as usize & 3];
        ElementLayout {
            class,
            size,
            // Make sure that slot boundaries begin and end on
            // `COPY_BUFFER_ALIGNMENT`-byte (4-byte) boundaries.
            elements_per_slot,
        }
    }

    fn slot_size(&self) -> u64 {
        self.size * self.elements_per_slot as u64
    }

    /// Creates the appropriate [`ElementLayout`] for the given mesh's vertex
    /// data.
    fn vertex(
        mesh_vertex_buffer_layouts: &mut MeshVertexBufferLayouts,
        mesh: &Mesh,
    ) -> ElementLayout {
        let mesh_vertex_buffer_layout =
            mesh.get_mesh_vertex_buffer_layout(mesh_vertex_buffer_layouts);
        ElementLayout::new(
            ElementClass::Vertex,
            mesh_vertex_buffer_layout.0.layout().array_stride,
        )
    }

    /// Creates the appropriate [`ElementLayout`] for the given mesh's index
    /// data.
    fn index(mesh: &Mesh) -> Option<ElementLayout> {
        let size = match mesh.indices()? {
            Indices::U16(_) => 2,
            Indices::U32(_) => 4,
        };
        Some(ElementLayout::new(ElementClass::Index, size))
    }
}

impl GeneralSlab {
    /// Returns true if this slab is empty.
    fn is_empty(&self) -> bool {
        self.resident_allocations.is_empty() && self.pending_allocations.is_empty()
    }
}

/// Returns a string describing the given buffer usages.
fn buffer_usages_to_str(buffer_usages: BufferUsages) -> &'static str {
    if buffer_usages.contains(BufferUsages::VERTEX) {
        "vertex "
    } else if buffer_usages.contains(BufferUsages::INDEX) {
        "index "
    } else {
        ""
    }
}