All pull requests

Pull requests

• Derive page configs without a per-byte image

  #1006·jotabulacios opened Sep 23, 2026Collaborator·

  ·

• STARK recursion with BLAKE3 on the batched CPU pipeline: the August configuration, 104.2 min / 358.2 GiB

  #1005·MauroToscano opened Sep 22, 2026Contributor·

  ·

• WHIR recursion on GPU (RPX): full recursion of block 25368371 and three optimisation rounds, 128.3 s

  #1004·MauroToscano opened Sep 22, 2026Contributor·

  ·
  1

• spec: IS_B48

  spec
  #1003·erik-3milabs opened Sep 22, 2026Collaborator·

  ·
  

• fix(lfm,tests): the eleven pins the main-sync port left behind, each against its mover

  #1002·MauroToscano opened Sep 21, 2026Contributor·

  ·
  2

• perf: stream Keccak lookups across histogram workers

  #1001·diegokingston opened Sep 21, 2026Collaborator·

  ·
  2

• Write the keccak state in one literal

  #1000·jotabulacios opened Sep 21, 2026Collaborator·

  ·
  9

• perf(math-cuda): round 2 — evictable leaf-layer retention (−9.1s on the block wall)

  #999·MauroToscano opened Sep 20, 2026Contributor·

  ·

• feat(multilinear): the sumcheck, and the two arguments built on it

  #998·ColoCarletti opened Sep 18, 2026Collaborator·

  ·

• refactor(stark): share what the multilinear path needs

  #997·ColoCarletti opened Sep 18, 2026Collaborator·

  ·

• Prove-and-retire prover (Approach 1), per-table and batched

  #994·jotabulacios opened Sep 18, 2026Collaborator·

  ·
  1

• build(guest): remap the absolute paths out of the guest ELF, and pin its sha256

  #993·MauroToscano opened Sep 18, 2026Contributor·

  ·

• Make the WHIR path's hash a parameter, and add an RPX256 instantiation

  #992·MauroToscano opened Sep 16, 2026Contributor·

  ·

• feat(vm): add constrained SHA-256 compression precompile

  #990·diegokingston opened Sep 15, 2026Collaborator·

  ·
  5

• Multilinear whir

  #988·ColoCarletti opened Sep 14, 2026Collaborator·

  ·
  1

• Spec: unified MEMMOVE accelerator

  spec
  #986·nicole-graus opened Sep 11, 2026Collaborator·

  ·
  36

• Per-table GPU recursion, end to end: one root for block 25368371

  #985·MauroToscano opened Sep 11, 2026Contributor·

  ·

• Unify memcpy, memmove, memset and commit on one accelerator

  #984·nicole-graus opened Sep 10, 2026Collaborator·

  ·
  9

• Bump ethrex to 26.0.0 and rebuild the benchmark workload

  #983·jotabulacios opened Sep 10, 2026Collaborator·

  ·
  43

• docs(tables): correct the stale per-table census and name its derivation

  #982·MauroToscano opened Sep 10, 2026Contributor·

  ·

• Let the guest's memcpy use unaligned word loads

  #980·jotabulacios opened Sep 9, 2026Collaborator·

  ·
  19

• ci: bump the AI review workflows to actions v1.0.3

  #978·JuArce opened Sep 9, 2026Contributor·

  ·
  

• fix(stark): re-tune the Johnson gap, and stop calling a query budget "security"

  #976·MauroToscano opened Sep 9, 2026Contributor·

  ·

• spec: Field VM

  spec
  #971·RobinJadoul opened Sep 8, 2026Collaborator·2/2·

  ·
  85

  

• perf(math-cuda): one LDE buffer — transpose the fused commit's LDE in place

  #960·MauroToscano opened Sep 7, 2026Contributor·

  ·