Feat/local multi gpus#68

Open

blake-pro wants to merge 48 commits into

mainfrom

feat/local-multi-gpus

Commits on Dec 2, 2025

feat: given a compress proof, groth16 proof can be generated.
Blake
committed
write an empty public values
Blake
committed
refactor: replace unwrap with ? in shape_config::fix_shape
Blake
committed
feat: add from_input field to SnarkContext of v1
Blake
committed
update deps to new brand
Blake
committed
update commit id of Ziren
Blake
committed
feat: add elf_id in `GenerateProofRequest`
Blake
committed
feat: support compress in request
Blake
committed
feat: specify the maximum number of prover (#61)

VanhGer
authored and
Blake
committed
opt: max prover count
Blake
committed
Opt/db (#62)

blake-pro
authored and
Blake
committed
Refactor/mutex (#63)

blake-pro
authored and
Blake
committed
refactor: use OnceLock<Arc<_>> instead of Mutex for global prover; set DEFAULT_CACHE_SIZE = 5
Blake
committed
feat: add single node prover (#60)

VanhGer
authored and
Blake
committed
feat: add cycles for single-node tasks
Blake
committed
feat: add proving_time for each task
Blake
committed
dep: update ziren
Blake
committed
dep: update ziren
Blake
committed
add Cargo.lock
Blake
committed
feat: improve task dispatching logic in stage_worker
Blake
committed
refactor: simplify proof generation in single_node_prover
Blake
committed
refactor: streamline task handling in stage_worker
Blake
committed
refactor: reduce pre-allocation size of prove_tasks in stage.rs
Blake
committed
feat: add GPU support for prover components and update dependencies
Blake
committed
feat: enhance single node prover with multi-GPU support and input handling
Blake
committed
feat: implement GPU support for aggregation and root proving
Blake
committed
feat: add GPU job scheduling and dispatcher for aggregation and root proving
Blake
committed
feat: optimize GPU resource usage in executor and single node prover
Blake
committed
feat: enhance multi-GPU support with improved job handling and caching
Blake
committed
feat: improve program caching mechanism for enhanced performance
Blake
committed
feat: enhance key and program caching with entry initialization for improved performance
Blake
committed
feat: increase cache size for key and program to enhance multi-GPU performance, and add some logs
Blake
committed
feat: send raw ExecutionRecord
Blake
committed
update ziren-gpu dep
Blake
committed
add some logs
Blake
committed
feat: Update the scheduling method for aggregation.
Blake
committed
chrome: remove prove v1 (#66)

eigmax
authored and
Blake
committed
feat: refactor single node task to include proof and public values
Blake
committed
feat: update GPU prover to support local multi-GPU configuration
Blake
committed
feat: allow configurable GPU pool thread count in local provers
Blake
committed
feat: add vk to single node task and update related methods
Blake
committed
feat: enhance proof verification in aggregation with completion check
Blake
committed
update dep
Blake
committed
fix ci
Blake
committed

Commits on Dec 3, 2025