All pull requests

Pull requests

• [CuTeDSL] Reject TMA global strides that are not 16B aligned (static layouts), #3866

  #3887·SycamoreQ opened Nov 12, 2026·

  ·

• docs: fix stride calculation variable in layout algebra

  #3886·D-D-H opened Nov 12, 2026·

  ·

• Reuse the division remainder in FastDivmodU64 setup

  #3885·ZenAlexa opened Nov 11, 2026·

  ·
  1

• [CuTeDSL] Reject IKET calls in host code

  #3884·0z5a opened Nov 10, 2026·

  ·
  1

  2

• [CuTeDSL] Accept boolean IKET payloads

  #3883·0z5a opened Nov 10, 2026·

  ·
  1

• fmha: gate the mbarrier wait to fp16 d128 non-causal and use fmax for the steady-state row max

  #3879·arvin-chou opened Nov 10, 2026Collaborator·

  ·
  1

• fix(cute): support building with -fno-exceptions

  #3877·riyaneel opened Nov 9, 2026·

  ·

• Fix scalar indexing of dynamic hierarchical layouts

  #3876·AckerMan5659 opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Convert while loop conditions to Boolean

  #3875·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

  1

• [CuTeDSL] Give integer TensorSSA true division a Float32 result

  #3872·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Fix TensorSSA slicing with an integer into a nested mode

  #3870·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Round BFloat16 scalar arguments to nearest even

  #3868·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Make red() default to relaxed ordering

  #3865·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Support MAX and MIN reduce on integer TensorSSA

  #3863·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Do not cache leading_dim on runtime tensors

  #3861·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Allow Int64 bounds in range with a runtime step

  #3859·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Make cute.math.abs of unsigned values the value itself

  #3857·ruiling-smartbear opened Nov 9, 2026·

  ·

• [examples/hopper] Fix the fp8 reference in dense_gemm

  #3856·ruiling-smartbear opened Nov 9, 2026·

  ·
  1

• [CuTeDSL] Keep unsigned element types on pointers

  #3853·ruiling-smartbear opened Nov 8, 2026·

  ·
  1

  1

• Fix integer-to-FP8 casts in CuTe DSL

  #3852·katepallewarprathmesh-sketch opened Nov 7, 2026·

  ·
  1

• [CuTeDSL] Wait for tcgen05.ld before TMEM handoff in the Blackwell examples

  #3851·0z5a opened Nov 7, 2026·

  ·

• [CuTeDSL] Keep -0.0 and 0.0 constants apart in the IR constant cache

  #3849·Zhonglin1231 opened Nov 6, 2026·

  ·
  1

• [CuTeDSL] Make abs() of an unsigned runtime value the value itself

  #3848·Zhonglin1231 opened Nov 6, 2026·

  ·
  1

• [CuTeDSL] Sign-extend signed to wider unsigned integer conversions

  #3847·Zhonglin1231 opened Nov 6, 2026·

  ·
  1

  1

• docs: fix moved repository paths referenced from the docs

  #3842·0z5a opened Nov 6, 2026·

  ·