All pull requests

Pull requests

• [BACKEND] Simplify ReduceOpHelper and reduction lowering

  #12013·Jokeren opened Sep 28, 2026Contributor·2/2·

  ·
  6

• [BACKEND] Fix reduction vectorization and layout reassociation

  #12011·Jokeren opened Sep 28, 2026Contributor·1/2·

  ·

• [NVIDIA] Reapply BF16 to FP32 conversion optimization

  #12010·lezcano opened Sep 28, 2026Contributor·

  ·
  5

• [Frontend] Narrow min/max reductions while preserving result types

  #12009·neildhar opened Sep 28, 2026Collaborator·

  ·

• [Gluon] Expose cumprod/cumsum

  #12007·peterbell10 opened Sep 28, 2026Contributor·

  ·
  1

• [CD] Raise minimum supported Python to 3.11. Max supported is 3.15

  #12006·atalman opened Sep 28, 2026Collaborator·

  ·
  1

• [Frontend] Support integers in tl.clamp

  #12005·peterbell10 opened Sep 28, 2026Contributor·

  ·
  1

• [RUNTIME] Store file cache group children relative to the cache dir

  #12001·iemAnshuman opened Sep 28, 2026·

  ·
  1

• [Gluon][NVIDIA] Support L2 cache policies on TMA transfers

  #12000·peterbell10 opened Sep 28, 2026Contributor·

  ·
  1

• [Frontend][Backend] Support uint8 inputs in tl.dot

  #11994·zihaomu opened Sep 28, 2026·

  ·
  1

• [AMD][KERNELS] Disable XCD swizzle on RDNA

  #11992·zihaomu opened Sep 28, 2026·

  ·
  2

• [NVIDIA] Fix TMA descriptor reduction regression

  #11991·chenmiaoming opened Sep 28, 2026·

  ·
  1

• [Hopper][WS] Fall back for unsupported atomic RMWs

  #11989·1sgtpepper opened Sep 28, 2026Contributor·

  ·
  1

• [Hopper][WS] Fall back when Gather intersects loop partitioning

  #11988·1sgtpepper opened Sep 28, 2026Contributor·

  ·
  1

• [BACKEND] Keep the dividend's sign on exact-zero float remainders

  #11987·he-yufeng opened Sep 27, 2026Contributor·

  ·
  1

• [AxisInfo] Do not keep contiguity across the wrap point of int casts

  #11982·dhairyashilRG opened Sep 27, 2026·

  ·
  3

• [BACKEND][NV] Fail warp-spec data partition gracefully on partitioned-dim reduce

  #11972·he-yufeng opened Sep 25, 2026Contributor·

  ·
  1

  4

• [TRITONGPU] Add verifier for NvidiaMmaEncodingAttr and handle unsupported versions in toLinearLayout

  #11969·agourakis82 opened Sep 25, 2026·

  ·
  1

  9

• [AMD] Add a pre-bound HIP kernel launcher

  #11968·adityakankariya opened Sep 25, 2026·

  ·

• [Layouts] Account for masks when choosing coalesced layouts

  #11967·peterbell10 opened Sep 25, 2026Contributor·

  ·

• [AMD][BACKEND] Fix TDM verifiers to use shaperPerCTA

  #11963·AlexAUT opened Sep 25, 2026Collaborator·

  ·

• [AMD][TEST] Update test_mxfp_gemm_tdm.py to align with Gluon test

  #11946·yiqian1 opened Sep 24, 2026Contributor·

  ·

• [AMD] Add GFX1250-strict Support

  #11945·amirumoAMD opened Sep 24, 2026·

  ·
  27

• [NVWS] Preserve accumulator lifetimes across conditional reads

  #11942·Jokeren opened Sep 24, 2026Contributor·

  ·
  1

• [FRONTEND][NV] Skip zero-result ops when picking the reduce combiner

  #11939·he-yufeng opened Sep 24, 2026Contributor·

  ·
  1

  11