All pull requests

Pull requests

• Warn when Hopper WGMMA layout constraints force fallback

  #3373·dajiaohuang opened Oct 2, 2026·

  ·
  1

  4

• [BugFix][Language] Accept integer per-axis shift flags

  #3372·Metis-dot opened Oct 2, 2026·

  ·
  2

• Report unsupported 16-bit integer atomic adds clearly

  #3370·dajiaohuang opened Oct 1, 2026·

  ·
  1

  2

• Reject unsupported atomic_addx4 destination dtypes

  #3369·dajiaohuang opened Oct 1, 2026·

  ·
  1

  2

• Validate mbarrier arrive-count upper bound

  #3368·dajiaohuang opened Oct 1, 2026·

  ·
  2

• Reject sparse MMA K tiles with unhandled tails

  #3367·dajiaohuang opened Oct 1, 2026·

  ·
  1

  2

• Preserve non-restrict buffers with handle suffixes

  #3366·dajiaohuang opened Oct 1, 2026·

  ·
  1

  2

• Fix constant row indices for TMA gather and scatter

  #3365·dajiaohuang opened Oct 1, 2026·

  ·
  1

  2

• fix: validate T.copy region bounds

  #3364·dajiaohuang opened Oct 1, 2026·

  ·
  2

• Fix warp specialization for while-wrapped pipelines

  #3363·dajiaohuang opened Oct 1, 2026·

  ·
  1

  5

• Document alignment requirements for shared PTX stores

  #3362·dajiaohuang opened Oct 1, 2026·

  ·
  1

  2

• Support bfloat16 scalar parameters

  #3361·dajiaohuang opened Oct 1, 2026·

  ·
  1

  5

• Report non-boolean if and assert conditions at the source

  #3360·dajiaohuang opened Oct 1, 2026·

  ·
  1

  4

• Honor current CUDA stream in NVRTC and CuTeDSL adapters

  #3359·dajiaohuang opened Oct 1, 2026·

  ·
  2

• [Carver] Remove unreachable tensor-core shared-scope branches

  #3358·dajiaohuang opened Oct 1, 2026·

  ·
  2

• [BugFix][Profiler] Use the current CUDA device when no device is specified

  #3357·Afloat16 opened Oct 1, 2026·

  ·
  3

• [BugFix][Transform] Unroll the epilogue of a pipeline with a runtime trip count

  #3356·FeathBow opened Sep 30, 2026·

  ·
  2

• [BugFix] Convert an over-range float to e5m2 as infinity

  #3355·173787247 opened Sep 30, 2026·

  ·
  6

• [BugFix] Run a T.ws(-1) region on every thread

  #3354·173787247 opened Sep 30, 2026·

  ·
  2

• [BugFix] Reject a constant rebound inside a nested region

  #3353·173787247 opened Sep 30, 2026·

  ·
  4

• [BugFix] Bound a sliced fill by the end of the region, not just its extent

  #3352·173787247 opened Sep 30, 2026·

  ·
  4

• [BugFix] Marshal narrow unsigned scalars on the cython backend

  #3351·173787247 opened Sep 30, 2026·

  ·
  2

• [BugFix] Reject round on a vectorized cast from a non-f32 source

  #3350·173787247 opened Sep 30, 2026·

  ·
  2

• [BugFix] Decode an unsigned sub-word without a zero-point bias

  #3348·173787247 opened Sep 30, 2026·

  ·
  2

• [BugFix] Reject an out-of-range constant negative index

  #3347·173787247 opened Sep 30, 2026·

  ·
  4