All pull requests

Pull requests

• Fix CuTeDSL object aliasing across dynamic loops

  #3746·knowrohit opened Oct 15, 2026·

  ·
  1

• [CuTeDSL] Lower device integer math.ipowi natively

  #3745·knowrohit opened Oct 15, 2026·

  ·
  1

• [Tests] Import running_on_ci from where pytest 9 defines it

  #3744·ConnorBaker opened Oct 14, 2026·

  ·

• [CuTeDSL] Don't add base_dsl to sys.path

  #3743·ConnorBaker opened Oct 14, 2026·

  ·

• [CuTeDSL] Add elementwise epilogues to the SM120 dense GEMM example

  #3742·CharlesXu-HQ opened Oct 14, 2026·

  ·

• [CuTeDSL] Add a host JIT tensor layout manipulation tutorial

  #3741·CharlesXu-HQ opened Oct 14, 2026·

  ·

• Bump pytest from 9.0.2 to 9.0.3 in /cutlass_cicd/cutedsl

  dependenciespython
  #3740·dependabot[bot] opened Oct 14, 2026Bot·

  ·

• [CuTeDSL] examples: add IKET profiling notebooks

  #3739·faradawn opened Oct 13, 2026·

  ·
  1

• [CuTe DSL] Expand integer ** in Python instead of emitting math.ipowi, issue #3735

  #3738·yunweili3 opened Oct 13, 2026·

  ·
  1

• Fix SM120 GEMM hang when K = 0

  #3737·vera-rubin opened Oct 13, 2026·

  ·
  1

• [CuTe DSL] Keep the width of Int64 tiler leaves in _pack_tile, issue #3734

  #3736·yunweili3 opened Oct 13, 2026·

  ·
  1

• feat(typing): add new Float8 data types for FNUZ and Float8E3M4

  #3731·Ammar-Alnagar opened Oct 11, 2026·

  ·
  1

  1

• Fix FP6 lowest()

  #3730·samridh10exe opened Oct 11, 2026·

  ·
  1

• [CuTe DSL] Reject cache_policy on invariant CopyG2ROp atoms instead of silently dropping it (#3726)

  #3729·knowrohit opened Oct 11, 2026·

  ·
  1

• [CuTeDSL] Apply preferred_smem_carveout in AOT loaders

  #3728·DQR-Happy opened Oct 10, 2026·

  ·
  1

• [CuTe DSL] Reject cache_policy on invariant CopyG2ROp Atoms instead of silently dropping it, issue #3726

  #3727·yunweili3 opened Oct 10, 2026·

  ·
  1

• [CUTLASS] SM90 cp.async + GMMA: add async-proxy fence before wgmma reads cp.async-filled smem

  #3724·yunweili3 opened Oct 8, 2026·

  ·
  1

• [CUTLASS] SM90 TMA GEMM: make the pipeline leader the TMA-issuing lane

  #3723·yunweili3 opened Oct 8, 2026·

  ·
  1

• Fix dangling iterator Params in grouped GEMM and grouped Rank2K kernels

  #3719·BwL1289 opened Oct 7, 2026Contributor·

  ·
  1

• [CuTe DSL] Support narrow float types (fp8/f16/bf16, f6/f4 vectors) in cute.arch.load / cute.arch.store

  #3714·yunweili3 opened Oct 5, 2026·

  ·
  1

  1

• [CuTe DSL] Reject memory attributes on CopyUniversalOp instead of silently dropping them

  #3713·yunweili3 opened Oct 5, 2026·

  ·
  1

  1

• Fix auxiliary and bias state updates in sparse GEMM

  #3710·atipre opened Oct 4, 2026·

  ·

• Fix GemmArray update grid shape arguments

  #3708·vampirechou opened Oct 4, 2026·

  ·

• Fix cross-rank offsets in specialized tile access iterators

  #3705·bishal0922 opened Oct 1, 2026·

  ·

• Fix cross-axis tile offsets in the pitch-linear access iterator

  #3704·md-info opened Oct 1, 2026·

  ·