All pull requests

Pull requests

• Keep PTXAS rejection policies in CUDA cache keys

  #34·LingYi-Liang opened Oct 22, 2026·

  ·

• Pass -std (JIT_CPP_STANDARD) to the Bisheng compiler options

  #33·dragonfive2002 opened Oct 21, 2026·

  ·

• Add and refine native Windows CUDA support

  #32·chinazhangchao opened Oct 21, 2026·

  ·

• docs: define CUDA runtime context affinity

  #31·lihaoran0314 opened Oct 21, 2026·

  ·

• Preserve zero Ascend launch timeouts

  #28·LingYi-Liang opened Oct 16, 2026·

  ·

• Prefer standalone pybind11 headers in stable ABI builds

  #23·cyanseek opened Oct 1, 2026·

  ·

• fix(cuda): preserve argv boundaries in compiler shell commands

  #22·Afloat16 opened Sep 30, 2026·

  ·
  1

• Allow the first CUDA device query during graph capture

  #17·LiRunGuo opened Sep 22, 2026·

  ·
  3

• Synchronize concurrent include hashing in Parser

  #16·CharlesXu-HQ opened Sep 18, 2026·

  ·

• Keep CUDA dynamic shared-memory limits monotonic across launches

  #15·CharlesXu-HQ opened Sep 18, 2026·

  ·

• Make LazyInit initialization single-flight

  #14·adenzhou1350 opened Sep 14, 2026·

  ·

• Add MIT license

  #13·adenzhou1350 opened Sep 14, 2026·

  ·
  1

• Make in-memory cache misses single-flight

  #12·adenzhou1350 opened Sep 14, 2026·

  ·

• ci: scan pull requests for credentials and injection with ThreatCrush

  #8·ralyodio opened Sep 11, 2026·

  ·
  1

• Add ROCm 10 backend for AMD GPUs

  #7·mehmetoguzderin opened Sep 11, 2026Contributor·

  ·
  6

• Make the libdw dependency of C++ backtraces optional

  #6·PerryLink opened Sep 10, 2026·

  ·

• Fix cache publication errors without accepting incomplete entries

  #5·morluto opened Sep 10, 2026·

  ·
  1

• Fix cache cleanup to unlink symlinks without traversing targets

  #4·morluto opened Sep 10, 2026·

  ·
  1