All pull requests

Pull requests

• docs(cutile): demonstrate recorded autotune dispatch

  #1740·shuqi2077 opened Nov 4, 2026·

  ·
  1

• feat(cuda-oxide): warn on rolled-loop local arrays

  #1737·praveenperera opened Nov 4, 2026Contributor·

  ·
  1

• fix(cuda-oxide): unroll libdevice loops like clang

  #1732·praveenperera opened Nov 4, 2026Contributor·

  ·
  1

  1

• feat(examples): capture SIMT kernel launches into a CudaGraph

  #1731·vyncint opened Nov 4, 2026Contributor·

  ·
  1

• test(cuda-core): gate stream_query on observed state instead of elapsed time

  #1729·vyncint opened Nov 4, 2026Contributor·

  ·
  1

• fix(mir-lower): name shared statics by identity

  #1728·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(mir-importer): emit promoted refs as globals

  #1726·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(intrinsics): admit scalar arithmetic on sm_75

  #1725·ssoj13 opened Nov 3, 2026·

  ·
  1

• feat(cuda-device): add BlockRows index space

  #1723·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• feat(cuda-device): add safe DisjointSlice::empty

  #1722·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(llvm-export): match call-site convergent to the intrinsic declaration

  #1721·praveenperera opened Nov 3, 2026Contributor·

  ·
  2

• fix(mir-importer): widen shared pointers stored through an address

  #1720·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• feat(unroll): add a per-loop max_ops budget

  #1719·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• feat(unroll): partially unroll runtime-step loops

  #1718·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• test(cuda-host,finalizer): replace timed contention and release with handshakes

  #1716·vyncint opened Nov 3, 2026Contributor·

  ·
  1

• docs(book): run cargo oxide doctor from the new project in hello-gpu

  #1715·vyncint opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(cute-layout): accept one-axis blocked and raked products

  #1713·0z5a opened Nov 3, 2026Contributor·3 of 6 tasks·

  ·
  1

• fix(cute-layout): reject compositions that leave a non-dividing mode

  #1712·0z5a opened Nov 3, 2026Contributor·3 of 6 tasks·

  ·
  1

• smoketest: skip optimizer-dependent code-shape gates when the environment forces full debug

  #1711·girithc opened Nov 3, 2026·

  ·
  1

• fix(mir-importer): undo an abandoned address walk

  #1710·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(mir-importer): pin borrowed pointer slots

  #1709·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(cargo-oxide): build standalone setup from deps

  #1708·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(unroll): fold branches to unreachable blocks

  #1707·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• docs(cuda-device): record 128-bit lane-read loads

  #1706·praveenperera opened Nov 3, 2026Contributor·

  ·
  1

  1

• fix(cuda-host): run sm_100f NVVM IR and LTOIR on sm_103 (B300/GB300)

  bugcuda-oxidehost-apis
  #1700·0z5a opened Nov 1, 2026Contributor·4 of 6 tasks·

  ·
  1