Struct GpuProverConfig 

pub struct GpuProverConfig { /* private fields */ }

Low-level configuration for the GPU prover.

The default configuration is meant to be optimal in the vast majority of cases, so overriding it is only recommended if you either have specific constraints or you need to fine-tune the configuration for a specific and unusual workload. Each field left unset keeps the prover’s own default.

Implementations

impl GpuProverConfig

pub fn with_worker_threads(self, worker_threads: usize) -> Self

Number of worker threads for the prover’s thread pool and replay workers.

pub fn maybe_worker_threads(self, worker_threads: Option<usize>) -> Self

Like Self::with_worker_threads but a no-op when None.

pub fn with_max_device_memory_bytes(self, bytes: usize) -> Self

Caps the GPU device allocator at bytes. By default the allocator grabs all free VRAM, which leaves no room for a co-resident prover in the same process (e.g. the SNARK wrapper). Capping the FRI prover leaves headroom for it. A cap at or above the free amount is a no-op.

pub fn with_host_allocators_per_job(self, count: usize) -> Self

Pinned host transfer buffers pre-allocated per concurrent job (64 MiB each). Lowering this below the prover default reclaims committed RAM at the cost of less host<->device pipelining.

pub fn with_host_allocators_per_device(self, count: usize) -> Self

Pinned host transfer buffers pre-allocated per GPU device (64 MiB each). See Self::with_host_allocators_per_job.

Trait Implementations

impl Clone for GpuProverConfig

fn clone(&self) -> GpuProverConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GpuProverConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for GpuProverConfig

fn default() -> GpuProverConfig

Returns the “default value” for a type.

impl Copy for GpuProverConfig