Struct StreamBuffer 

pub struct StreamBuffer { /* private fields */ }

A growable, ring-reset GPU buffer for streaming vertex or index data across contiguous batches within a single frame.

Capacity growth requires &mut self; the per-batch write path only needs &self thanks to a Cell<u64> cursor, so it composes cleanly with methods that borrow other parts of the renderer immutably.

Implementations

impl StreamBuffer

pub fn new(usage: BufferUsages, label: &'static str) -> Self

pub fn reset(&self)

Reset the write cursor to the start of the buffer.

Call once at the top of render() — previous frame’s contents are orphaned and overwritten in place.

pub fn ensure_capacity(&mut self, device: &Device, required_bytes: u64)

Grow the underlying buffer if required_bytes exceeds the current capacity. Growth rounds up to the next power of two (minimum 1 KiB) to amortize reallocation. If the existing buffer already fits, this is a no-op.

pub fn write(&self, queue: &Queue, data: &[u8]) -> Option<(&Buffer, u64, u64)>

Upload data at the current write cursor and advance the cursor.

Returns Some((buffer, offset, len_bytes)) on success, or None if ensure_capacity was never called OR the write would overflow the buffer’s capacity. The caller should slice the returned buffer at offset..offset + len when binding.

An overflow means the frame-start sizing undercounted this pipeline’s quads (see stream_quad_counts in renderer.rs). Debug builds assert with the accounting details; release builds skip the batch — a dropped draw is recoverable, whereas forwarding an out-of-bounds write_buffer to wgpu is a validation error that kills the device.