Skip to main content

Crate openvm_cuda_backend

Crate openvm_cuda_backend 

Source

Re-exports§

pub use hash_scheme::*;

Modules§

base
data_transporter
hash_scheme
logup_zerocheck
Async frees and peak memory
merkle_tree
monomial
Monomial expansion for GPU zerocheck evaluation.
ntt
poly
prelude
sponge
GPU-accelerated duplex sponge with host/device state synchronization.
stacked_pcs
stacked_reduction
utils
whir

Structs§

AirDataGpu
ConstraintOnlyRules
Constraints only, no interactions
EvalRules
GenericGpuBackend
Generic GPU prover backend parameterised by a hash scheme HS.
GpuDevice
Stream-owning device handle. Wraps GpuDeviceConfig with a GpuDeviceCtx that carries the explicit CUDA stream.
GpuDeviceConfig
Pure device configuration — no stream, no CUDA runtime state. Renamed from the old GpuDevice.
GpuEngine
Generic GPU proving engine parameterised by a hash scheme.
GpuProverConfig
InteractionEvalRules
Used for GKR input evaluation and logup MLE sumcheck rounds.
InteractionMonomials
ZerocheckMonomials

Enums§

KernelError
Error type for functions that call CUDA kernels and involve some memcpy operations.
LogupZerocheckError
MerkleTreeError
ProverError
RsCodeMatrixError
StackTracesError
StackedReductionError
SumcheckError
WhirProverError

Type Aliases§

BabyBearPoseidon2GpuEngine
Concrete GPU engine using the default BabyBear-Poseidon2 hash scheme.
GpuBackend
Concrete GPU backend using the default BabyBear-Poseidon2 hash scheme.