|
| void | clear_history () noexcept |
| | Render-owner reset of prepared audio/PDC storage without reallocating.
|
| | GraphRuntimeBufferPool ()=default |
| | GraphRuntimeBufferPool (const GraphRuntimeBufferPool &other) |
| GraphRuntimeBufferPool & | operator= (const GraphRuntimeBufferPool &other) |
| | GraphRuntimeBufferPool (GraphRuntimeBufferPool &&) noexcept=default |
| GraphRuntimeBufferPool & | operator= (GraphRuntimeBufferPool &&) noexcept=default |
| bool | reset (std::uint32_t slot_count, std::uint32_t max_frames) |
| bool | reset (std::uint32_t slot_count, std::uint32_t max_frames, std::span< const std::uint32_t > connection_delay_samples) |
| void | clear () noexcept |
| std::uint32_t | slot_count () const noexcept |
| std::uint32_t | max_frames () const noexcept |
| float * | slot_data (std::uint32_t index) noexcept |
| const float * | slot_data (std::uint32_t index) const noexcept |
| DelayRing | delay_ring (std::uint32_t index) noexcept |
| bool | can_adopt_delay_ring_state (std::uint32_t index, const GraphRuntimeBufferPool &source, std::uint32_t source_index) const noexcept |
| bool | adopt_delay_ring_state (std::uint32_t index, const GraphRuntimeBufferPool &source, std::uint32_t source_index) noexcept |
| bool | bind (const GraphRuntimeSnapshot &snapshot) noexcept |
| bool | bound_to (const GraphRuntimeSnapshot &snapshot) const noexcept |
| BoundNode & | bound_node (std::uint32_t index) noexcept |
| bool | alias_host_input (const BusBuffer *input, const BusBuffer *output, std::uint32_t frames) noexcept |
| void | release_host_input () noexcept |
| bool | fits (const GraphRuntimeSnapshot &snapshot, std::uint32_t frames) const noexcept |
Pre-allocated scratch buffers backing the routing process() path.
One slot is a mono buffer of max_frames floats — the same "slot" the snapshot's GraphRuntimeBufferAssignment indexes. The caller sizes the pool off-RT from a snapshot's buffer_slot_count() and the maximum block size, then reuses it across realtime blocks. The executor never resizes it.
Scope: this is the SHARED inter-node routing scratch for one serial walk. Per-worker scratch for a parallel executor is a separate concern and must not be overloaded onto this shared pool.