pulp::gpu_audio::GpuNeuralCpuKernel Struct Reference

One channel's CPU model step, supplied and owned by the caller. More...

#include <gpu_neural_model.hpp>

Public Attributes
void *	state = nullptr
bool(*	prepare )(void *state) noexcept = nullptr
	Host thread, before processing starts: clear the causal state.
void(*	process )(void *state, const float *input, float *output, std::uint32_t frames) noexcept = nullptr
	Runs on the audio thread (Inline shadow, CPU-only families) or on the shadow thread (Worker).
void(*	reset )(void *state) noexcept = nullptr
	Restarts the causal state; called off the audio thread.

Detailed Description

One channel's CPU model step, supplied and owned by the caller.

The node keeps one kernel per audio channel and never shares state between them. state must stay valid, and its storage allocated, until the model and every node that bound it are destroyed.

Member Data Documentation

prepare

bool(* pulp::gpu_audio::GpuNeuralCpuKernel::prepare) (void *state) noexcept = nullptr

noexcept

Host thread, before processing starts: clear the causal state.

Returning false refuses preparation.

process

void(* pulp::gpu_audio::GpuNeuralCpuKernel::process) (void *state, const float *input, float *output, std::uint32_t frames) noexcept = nullptr

noexcept

Runs on the audio thread (Inline shadow, CPU-only families) or on the shadow thread (Worker).

It must not allocate, lock or block. Input is always finite: the node replaces NaN and infinities with zero first.

reset

void(* pulp::gpu_audio::GpuNeuralCpuKernel::reset) (void *state) noexcept = nullptr

noexcept

Restarts the causal state; called off the audio thread.

state

void* pulp::gpu_audio::GpuNeuralCpuKernel::state = nullptr

---

The documentation for this struct was generated from the following file:

• /home/runner/work/pulp/pulp/core/gpu_audio/include/pulp/gpu_audio/gpu_neural_model.hpp