All pull requests

Pull requests

• Fix cut mutating explicit bin edges

  Python
  #24679·fallintoplace opened Oct 18, 2026Contributor·

  ·
  1

  

• Fix cut(right=False) mutating CuPy inputs

  Python
  #24678·fallintoplace opened Oct 18, 2026Contributor·

  ·
  2

  

• ENH: Accept a host span of split points in contiguous_split

  libcudf
  #24677·Umang-projects opened Oct 18, 2026Contributor·2 of 3 tasks·

  ·
  1

  3

  

• Match PyArrow's row count for read_parquet with columns=[]

  bugnon-breakingpylibcudfPython
  #24663·bdice opened Oct 17, 2026Contributor·3 tasks done·

  ·
  1

  4

  

• Use cuda::std integral traits and concepts instead of std to avoid the __int128_t inconsistency

  improvementlibcudfnon-breaking
  #24662·vuule opened Oct 17, 2026Contributor·3 tasks done·

  ·
  2

  

• Use std::span and cudf::memory_resources in hybrid scan APIs

  JavalibcudfpylibcudfPython
  #24660·mhaseeb123 opened Oct 17, 2026Contributor·3 of 4 tasks·

  ·
  1

  

• Add sparse payload hybrid scan APIs

  2 - In Progressimprovementlibcudfnon-breakingpylibcudfPython
  #24659·mhaseeb123 opened Oct 16, 2026Contributor·3 of 4 tasks·

  ·
  1

  

• Split large JOIN and Parquet tests for finer-grained GPU scheduling

  CMakeimprovementlibcudfnon-breakingPerformancetests
  #24658·vyasr opened Oct 16, 2026Contributor·3 tasks done·

  ·
  3

  

• Remove unneeded reinterpret_cast<cudf::bitmask_type*>()

  improvementlibcudfnon-breaking
  #24656·KyleFromNVIDIA opened Oct 16, 2026Member·3 tasks·

  ·
  2

  

• Remove fatbin linker script from libcudf builds

  CMakefeature requestlibcudfnon-breaking
  #24654·robertmaynard opened Oct 16, 2026Contributor·3 tasks done·

  ·
  1

  

• Support streaming first and last groupby aggregations

  2 - In Progresscudf-polarsfeature requestnon-breakingPython
  #24653·rjzamora opened Oct 16, 2026Contributor·

  ·
  4

  

• Hotfix 26.10.1: Backport #24231, #24331, #24396, #24655 to reduce HashCSR hash join memory usage

  ! - Hotfixbuglibcudfnon-breakingPython
  #24652·bdice opened Oct 16, 2026Contributor·3 tasks done·

  ·
  6

  

• Convert nvtext scratch rmm::device_uvector<T> to cuda::device_buffer<T>

  3 - Ready for Reviewimprovementlibcudfnon-breaking
  #24651·davidwendt opened Oct 16, 2026Contributor·3 tasks done·

  ·
  1

  

• Allow for reseting the Quent context in StreamingEngine._reset

  cudf-polarsimprovementnon-breakingPython
  #24649·TomAugspurger opened Oct 16, 2026Contributor·3 tasks done·

  ·
  8

  

• wheels: use wildcards in auditwheel exclusions

  improvementnon-breaking
  #24648·jameslamb opened Oct 16, 2026Member·3 tasks done·

  ·
  2

  

• Update docs to match actual project requirements

  docnon-breaking
  #24647·robertmaynard opened Oct 16, 2026Contributor·2 tasks done·

  ·
  1

  

• Test cudf against cupy nightly in nightly test job

  improvementnon-breakingPython
  #24644·TomAugspurger opened Oct 16, 2026Contributor·3 tasks done·

  ·
  14

  

• Add and migrate to new hybrid scan APIs

  2 - In ProgressimprovementJavalibcudfnon-breaking
  #24641·mhaseeb123 opened Oct 16, 2026Contributor·3 of 4 tasks·

  ·
  1

  

• Add the DELTA_LENGTH_BYTE_ARRAY nz map consumer and a block-wide delta decode

  libcudf
  #24639·vyasr opened Oct 15, 2026Contributor·3 tasks done·

  ·
  1

  

• Convert quantiles' and reductions' scratch RMM vectors to CUDA buffers

  improvementlibcudfnon-breaking
  #24638·KyleFromNVIDIA opened Oct 15, 2026Member·3 tasks·

  ·
  2

  

• Improve matches reduce for hash joins

  3 - Ready for Reviewimprovementlibcudfnon-breakingPerformance
  #24637·scanhex12 opened Oct 15, 2026Contributor·3 tasks·

  ·
  2

  

• Make all scalar ctors explicit

  breakingbuglibcudf
  #24636·wence- opened Oct 15, 2026Contributor·3 tasks done·

  ·
  8

  

• Convert parquet's scratch rmm::device_uvector<T> to cuda::device_buffer<T>

  improvementlibcudfnon-breaking
  #24635·KyleFromNVIDIA opened Oct 15, 2026Member·3 tasks·

  ·
  9

  

• Convert json's scratch rmm::device_uvector<T> to cuda::device_buffer<T>

  improvementlibcudfnon-breaking
  #24627·KyleFromNVIDIA opened Oct 15, 2026Member·3 tasks·

  ·
  3

  

• Convert dictionary's scratch rmm::device_uvector<T> to cuda::device_buffer<T>

  improvementlibcudfnon-breaking
  #24626·KyleFromNVIDIA opened Oct 15, 2026Member·3 tasks·

  ·
  3