Struct SharedText 

pub struct SharedText(/* private fields */);

The device-scoped half of text rendering, shareable across Runners (issue #94): the font system + shaping cache, the CPU-side glyph and MSDF atlases, and their GPU page textures with bind groups. Everything here is independent of the swapchain format and MSAA sample count, so one SharedText per wgpu::Device can back every window — a multi-window host that passes the same handle to each Runner (Runner::with_shared_text) pays glyph rasterization, shaping, warm-up, and atlas VRAM once per device instead of once per window.

Cloning is cheap (an Arc); the inner state is mutex-guarded and locked per record/flush call, so windows can be prepared from one thread in any order. Each attached Runner widens the atlases’ LRU protection window (see MsdfAtlas::set_lru_protection_window) so a page referenced by one window’s in-flight frame can’t be recycled by another’s prepare.

The default Runner constructors create a private SharedText per runner — single-window behavior is unchanged.

Implementations

impl SharedText

pub fn new(device: &Device) -> Self

A fresh shared text pool for device. Pass the same handle to every Runner created on that device. Do not share one SharedText across devices — the page textures belong to the device that created them.

pub fn warm_default_glyphs(&self)

Pre-rasterize printable ASCII for the bundled default faces — see [TextPaint::warm_default_glyphs] for cost and rationale. On a shared pool this runs once per device: warm the pool before (or after) attaching runners, and every attached runner is warm. Runners attached to an already-warm pool skip the cost in their own warm_default_glyphs automatically (rasterized glyphs are cache hits).

Trait Implementations

impl Clone for SharedText

fn clone(&self) -> SharedText

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.