Issues

• [BUG] sm120 GEMM hangs when K = 0 (grouped and non-grouped) after can_implement(), initialize() and run() return success

  NVIDIA/cutlass#3733 · Filter by author vera-rubin opened on Oct 11, 2026

• [QST] SM120 FP8 MMA with FP32 accumulate, plain form against block-scaled form with unit scales

  NVIDIA/cutlass#3732 · Filter by author emre570 opened on Oct 11, 2026

• [BUG] Cache policy from cute.copy is dropped when the copy atom is created with invariant=True

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3726 · Filter by author kainzhong opened on Oct 9, 2026
  1

• [BUG] SM90 TMA GEMM mainloops rely on elect_one_sync() electing lane 0 (TMA issuer vs. PipelineTmaAsync is_leader)

  CUTLASS C++
  NVIDIA/cutlass#3722 · Filter by author builtin-pb opened on Oct 8, 2026

• [BUG] SM90 cp.async + GMMA collectives: no fence.proxy.async between cp.async operand writes and wgmma reads

  CUTLASS C++
  NVIDIA/cutlass#3721 · Filter by author builtin-pb opened on Oct 8, 2026

• [BUG] Grouped GEMM and grouped Rank2K kernels build operand iterators from a temporary Params

  NVIDIA/cutlass#3718 · Filter by author BwL1289 opened on Oct 7, 2026

• [BUG] CuTe DSL from_dlpack rejects float8_e4m3fn tensors (dtype_map lists Float8E5M2 only)

  NVIDIA/cutlass#3716 · Filter by author Vinnie6167 opened on Oct 7, 2026
  1

• [BUG] CuTe DSL >=4.6.0: export_to_c does not apply explicit preferred_smem_carveout

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3715 · Filter by author naytiba opened on Oct 7, 2026
  1

• [BUG] [BUG] cute.arch.store doesn't work with fp8 vector

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3712 · Filter by author gau-nernst opened on Oct 5, 2026
  1

• [BUG] CuTe DSL CopyUniversalOp ignores l1c_evict_priority again (regression of #3199 since 4.6.0, loads and stores)

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3711 · Filter by author gau-nernst opened on Oct 5, 2026
  1

• [BUG] CuTe DSL >= 4.6: FlashAttention-4 SM100 backward with block-sparse masks spills registers, up to 1.8x slower than 4.5.3 (same kernel source)

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3707 · Filter by author janhuenermann opened on Oct 1, 2026
  5

• [BUG] CuTe DSL 4.7/4.8 regression: FlashAttention-4 SM100 forward kernel spills registers (0 -> 176 B) vs 4.5.1, ~3% slower on B300

  NVIDIA/cutlass#3701 · Filter by author zpcore opened on Sep 29, 2026
  1

• [BUG] CuTeDSL 4.8.0 nvpu-optimization fails without a diagnostic on a sum of 10 dynamic extents

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3699 · Filter by author allispaul opened on Sep 29, 2026
  1

• [BUG] mma_atom_call for mma.sync is eliminated if C is RZ

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3689 · Filter by author melonedo opened on Sep 27, 2026
  2

• [BUG] CuTe DSL: a % b gives different results for dynamic and constant operands when the signs differ

  bugCuTe DSL
  NVIDIA/cutlass#3687 · Filter by author Arthur031221 opened on Sep 26, 2026
  1

• [BUG] CuTe DSL: cute.compile stalls for minutes when no GPU is visible

  CuTe DSL
  NVIDIA/cutlass#3681 · Filter by author hwanseoc opened on Sep 26, 2026

• [BUG] CuTe DSL: export_to_c fails for a --gpu-arch other than the current GPU's

  bugCuTe DSL
  NVIDIA/cutlass#3680 · Filter by author hwanseoc opened on Sep 26, 2026
  1

• [BUG] Order-dependent wrong rmem values on Windows SM120 with CuTe DSL 4.8 CUDA12 stack; CUDA13 control passes

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3673 · Filter by author hisashi2026 opened on Sep 24, 2026
  1

• [FEA][CuTeDSL] Let make_tiled_tma_atom set the TMA L2 promotion

  CuTe DSL
  NVIDIA/cutlass#3672 · Filter by author zkyue opened on Sep 24, 2026
  1

• [BUG][CuTeDSL] cute.copy(..., tma_desc_ptr=p) with a gmem-space p passes the Python checks, then fails in libNVVM

  CuTe DSL
  NVIDIA/cutlass#3671 · Filter by author zkyue opened on Sep 24, 2026
  1

• [BUG][CuTeDSL] cute.copy(..., tma_desc_ptr=TensorMap.get_ptr()) is silently ignored on tiled TMA loads

  CuTe DSL
  NVIDIA/cutlass#3670 · Filter by author zkyue opened on Sep 24, 2026
  1

• [BUG][CuTe DSL] PipelineTmaUmma arrival masks include an out-of-cluster "v-peer" bit when the cluster's v-extent is 1, causing CGA out-of-range faults (Xid 13)

  NVIDIA/cutlass#3658 · Filter by author LST-Keen opened on Sep 22, 2026
  1

• [BUG] Instantiated tile sizes for sm120 fail to meet kernel's layout assertion during cutlass_profiler compile

  ? - Needs TriagebugCUTLASS C++
  NVIDIA/cutlass#3646 · Filter by author CYQFWang opened on Sep 20, 2026

• [BUG] Incorrect results in 2CTA MMA example (4.8.0dev0) due to barrier scope & missing SFB wait

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3643 · Filter by author type-shangshu opened on Sep 20, 2026
  1

• [BUG] Operand initialized with a constant doesn't work with inline_ptx as an operand

  ? - Needs TriagebugCuTe DSL
  NVIDIA/cutlass#3640 · Filter by author kainzhong opened on Sep 17, 2026
  2