All pull requests

Pull requests

• Allow declared pre-atomic deterministic reads

  #2189·IAMSebyi opened Oct 31, 2026·5 tasks done·

  ·
  1

  12

• Fix tiled BVH query stack overflow (GH-2181)

  #2182·roto5296 opened Oct 29, 2026·5 tasks done·

  ·
  1

  4

• Fix unpacking typed tuple function arguments

  #2180·Fire162 opened Oct 28, 2026·

  ·
  1

  3

• Annotate the forward tile entry points CUDA_CALLABLE

  #2175·zhihuidu-amd opened Oct 28, 2026Contributor·5 tasks done·

  ·

  · 1.20.0
  7

• DO NOT REVIEW: Reduce stable-fluid pressure solve memory overhead

  #2173·vidurv-nvidia opened Oct 28, 2026·4 of 5 tasks·

  ·
  3

• DO NOT REVIEW: Cull redundant raymarch distance evaluations

  #2170·vidurv-nvidia opened Oct 28, 2026·5 tasks done·

  ·
  6

• Reduce SPH neighbor query stalls

  #2165·vidurv-nvidia opened Oct 27, 2026·5 tasks done·

  ·
  3

• Reject repeated axes in array.transpose()

  #2159·alphasafal opened Oct 25, 2026·5 tasks done·

  ·
  1

  4

• Preserve signed-zero order in CPU radix sort

  #2153·darkdi opened Oct 23, 2026·4 of 5 tasks·

  ·
  1

  4

• Allow array slices that start at the end of a dimension

  #2147·AkshayShah03 opened Oct 23, 2026·5 tasks done·

  ·
  1

  5

• Relocate fill_() handles in APIC captures (GH-2115)

  #2142·RyanL2 opened Oct 22, 2026·4 of 5 tasks·

  ·
  1

  10

• Key APIC kernels by overload symbol (GH-2140)

  #2141·RyanL2 opened Oct 22, 2026·5 tasks done·

  ·
  1

  3

• Fix CPU OpenGL buffer mappings (GH-2120)

  #2121·fusheng-ji opened Oct 20, 2026·5 tasks done·

  ·
  1

  5

• feat(example): add --resolution flag and CPU test to fft_poisson_navier_stokes_2d

  #2113·0x5t4l1n opened Oct 18, 2026·

  ·
  1

  4

• Fix flatten() silently accepting ragged input with leading scalars

  #2106·udsy19 opened Oct 18, 2026·

  ·
  3

• Fix arrays from lists of float16 and bfloat16 vectors

  #2105·usmanmateen opened Oct 17, 2026·4 of 5 tasks·

  ·
  1

  18

• Fix CPU erfinv and erfcinv precision

  #2103·usmanmateen opened Oct 16, 2026·5 tasks done·

  ·
  1

  6

• Bracket the tile device regions with an overridable marker

  #2087·zhihuidu-amd opened Oct 13, 2026Contributor·

  ·
  3

• Fix agent skills: normalize unofficial frontmatter fields per the Agent Skills spec

  #2074·janEbert opened Oct 10, 2026·

  ·
  2

• Add optional row-packed CUDA BSR products (GH-2069)

  #2073·fusheng-ji opened Oct 10, 2026·4 of 5 tasks·

  ·
  7

• Support mixed-precision BSR products (GH-2069)

  #2072·fusheng-ji opened Oct 10, 2026·4 of 5 tasks·

  ·
  11

• Fix zero-adjoint SVD gradients (GH-2054)

  #2064·fusheng-ji opened Oct 8, 2026·5 tasks done·

  ·
  1

  3

• Fix half-precision SVD3 NaNs (GH-2055)

  #2063·fusheng-ji opened Oct 8, 2026·4 of 5 tasks·

  ·
  1

  5

• Fix SGD for non-float32 parameter dtypes

  #2042·Zhuoxi2000 opened Oct 6, 2026·5 tasks done·

  ·
  1

  5

• [temp] Add forward-mode jets (wp.JetSpace / wp.JetSpace2) [GH-1813]

  #2037·alecjacobson opened Oct 5, 2026Contributor·3 tasks done·

  ·
  1

  12