All pull requests

Pull requests

• [cudax] Revisit binary and labeled partition const correctness

  #12613·davebayer opened Nov 12, 2026Contributor·

  ·
  1

• [libcu++] Make fmin/fmax work for all nvfp types

  #12612·davebayer opened Nov 12, 2026Contributor·

  ·
  1

• [cudax] Revisit mappings and synchronizers const correctness

  #12611·davebayer opened Nov 12, 2026Contributor·

  ·
  4

• [Backport branch/3.6.x] [libcu++] fpemu: drop the qualifier from the emulated int128 constructor

  #12610·github-actions[bot] opened Nov 12, 2026Bot·

  ·
  3

• # [CUB] Sort tiny segments during LRB partitioning and queue only the rest

  #12609·kaatish opened Nov 12, 2026Contributor·2 tasks·

  ·
  1

• [CI] Replace the pytest<9.2 cap with pytest-benchmark>=5.3.1

  #12607·pylaterreur opened Nov 11, 2026·2 tasks done·

  ·
  4

• [places] Do not query the stream's device during capture with a CTK 12 runtime

  #12604·caugonnet opened Nov 11, 2026Contributor·1 of 3 tasks·

  ·
  7

• [CUB] Take a block offset in LoadDirectStripedVectorized

  #12601·miscco opened Nov 11, 2026Contributor·1 of 2 tasks·

  ·
  6

• Use emoji status markers in benchmarks/scripts/compare.py and drop colorama

  #12600·bernhardmgruber opened Nov 11, 2026Contributor·

  ·
  4

• Ignore default subscription factor in DeviceReduce

  #12599·bernhardmgruber opened Nov 11, 2026Contributor·

  ·
  10

• [CUB] Skip inactive warp reduction steps for full warps

  #12598·miscco opened Nov 11, 2026Contributor·

  ·
  11

• Support non-default-constructible types in radix sort

  #12596·bernhardmgruber opened Nov 11, 2026Contributor·

  ·
  2

• Generalize __uninitialized_array into __uninitialized

  #12595·bernhardmgruber opened Nov 11, 2026Contributor·

  ·
  13

• docs: version each product's documentation independently

  #12592·sri-koundinyan opened Nov 11, 2026·2 tasks·

  ·
  21

• [libcu++] Experimental Multiply-With-Carry random engines

  #12590·fbusato opened Nov 11, 2026Contributor·

  ·
  1

  18

  

• Extend cuda::args::deferred to preserve compile-time information when run-time bounds are provided.

  #12589·srinivasyadav18 opened Nov 10, 2026Contributor·2 tasks·

  ·
  7

• [Backport branch/3.6.x] [libcu++] cuda::ptx::cp_async_bulk_tensor uses CTA smem from PTX ISA 86

  #12588·github-actions[bot] opened Nov 10, 2026Bot·

  ·
  3

• [cudax][cuco] Replace __index_type with size_t and drop __distance

  cucocudax
  #12584·PointKernel opened Nov 10, 2026Member·2 tasks done·

  ·
  1

  6

  

• Add CCCL agent skill with evals

  docs
  #12579·danielfrg opened Nov 10, 2026Contributor·2 tasks·

  ·
  11

  

• [cuda.compute] Add explicit device binding for cuda.compute algorithms

  #12575·oleksandr-pavlyk opened Nov 10, 2026Contributor·2 tasks done·

  ·
  1

  11

• Enable __float128 in cuda::std::lerp

  #12569·KassaSana opened Nov 10, 2026Contributor·

  ·
  1

  5

• [fpmp] Faster, platform-independent host seed for fp32mp2 cbrt and rcbrt

  #12568·rmlarsen opened Nov 10, 2026·2 tasks done·

  ·
  1

• Enable __float128 in cuda::std::fdim

  #12567·KassaSana opened Nov 10, 2026Contributor·

  ·
  1

  2

• Enable __float128 in cuda::std::ldexp

  #12566·KassaSana opened Nov 10, 2026Contributor·2 tasks done·

  ·
  1

  2

• Enable __float128 in cuda::std::modf

  #12565·KassaSana opened Nov 10, 2026Contributor·2 tasks done·

  ·
  1

  2