icicle icon indicating copy to clipboard operation
icicle copied to clipboard

a GPU Library for Zero-Knowledge Acceleration

Results 54 icicle issues
Sort by recently updated
recently updated
newest added

The addition operation in projective coordinates implements operator overloading that computes a scalar multiplication using a naive double-and-add algorithm. Would be worth specializing the operation by switching it with Montgomery...

backlog

- [x] Remove gnark #130 - [ ] Remove arkworks

type:feature
backlog

Support 64 bit goldilocks ntt support according to [Ingonyama goldilocks](https://github.com/ingonyama-zk/papers/blob/main/goldilocks_ntt_trick.pdf) This issue is related to goldilocks implementation only.

type:feature
backlog

## Description Recommend Ingo to explore Metal acceleration as well. It seems to become something that requires a skillful player to work on it. ## Motivation Metal means the GPU...

backlog

- [ ] https://github.com/ingonyama-zk/icicle/blob/ef8beb8d0c7d481b8f3f3a47d48695889e4079b3/icicle/primitives/extension_field.cuh#L129 - [ ] https://github.com/ingonyama-zk/icicle/blob/ef8beb8d0c7d481b8f3f3a47d48695889e4079b3/icicle/primitives/extension_field.cuh#L135 - [ ] https://github.com/ingonyama-zk/icicle/blob/ef8beb8d0c7d481b8f3f3a47d48695889e4079b3/icicle/primitives/field.cuh#L859 - [ ] https://github.com/ingonyama-zk/icicle/blob/ef8beb8d0c7d481b8f3f3a47d48695889e4079b3/icicle/primitives/field.cuh#L865

lang:cuda/cpp
backlog

## Description Add support for Pasta curves (Pallas and Vesta) ## Motivation Pasta curves are great for recursive SNARKs and are actively used in modern proving systems (e.g. [Nova](https://github.com/microsoft/Nova))

type:feature
backlog

## Description There should be a flexible way to provide/override the compiler's chosen arch. Maybe the following would be best for that: ``` if(DEFINED CUDA_ARCH) // user defined arch takes...

type:feature
good first issue
lang:cuda/cpp
area:tests
backlog

## Describe the changes adds Blake2s cuda capability.

This PR adds the chunked msm feature to V3 cuda-backend-branch: chunked-msm