All pull requests

Pull requests

• Support disk-to-disk copies in buffer_copy

  improvementnon-breaking
  #1289·nirandaperera opened Oct 27, 2026Contributor·

  ·

• Collect metadata about memory reservations

  #1288·TomAugspurger opened Oct 27, 2026Contributor·

  ·
  1

• Make PausableThreadLoop's sleep interruptible

  improvementnon-breaking
  #1287·nirandaperera opened Oct 27, 2026Contributor·

  ·

• Allgather use exchange

  improvementnon-breaking
  #1286·madsbk opened Oct 27, 2026Member·

  ·
  1

  

• Sparse alltoall use exchange

  improvementnon-breaking
  #1285·madsbk opened Oct 27, 2026Member·

  ·
  2

  

• Extend MetadataPayloadExchange for use in the collectives

  improvementnon-breaking
  #1284·madsbk opened Oct 27, 2026Member·

  ·
  1

  

• [C++] Pass the source memory type to spill functions

  breakingimprovement
  #1283·nirandaperera opened Oct 26, 2026Contributor·

  ·
  1

• [C++] Enable spilling in SparseAlltoall

  improvementnon-breaking
  #1282·nirandaperera opened Oct 26, 2026Contributor·

  ·
  4

• [C++] Add per-memory-type spill targets and reservability to BufferResource

  breakingimprovement
  #1278·nirandaperera opened Oct 23, 2026Contributor·

  ·
  15

• [DNM]: Test CI

  #1272·TomAugspurger opened Oct 21, 2026Contributor·

  ·
  2

• Report progress function errors to an error callback

  breakingfeature request
  #1260·pentschev opened Oct 14, 2026Member·

  ·
  2

  

• Never let exceptions escape PausableThreadLoop

  bugnon-breaking
  #1257·pentschev opened Oct 13, 2026Member·

  ·
  16

  

• [C++] Fix get_current_numa_nodes() for non-MPOL_BIND memory policies

  bugnon-breaking
  #1256·nirandaperera opened Oct 9, 2026Contributor·

  ·
  1

  10

• [MINOR] Track per-memory-type data sizes in ReceivedChunks

  improvementnon-breaking
  #1252·nirandaperera opened Oct 7, 2026Contributor·

  ·
  5

  

• Evaluate per-rank/per-partition send ordering in the shuffler

  improvementnon-breaking
  #1247·nirandaperera opened Oct 4, 2026Contributor·

  ·
  1

• Make SpillableMessages spill destinations configurable

  improvementnon-breaking
  #1245·nirandaperera opened Oct 2, 2026Contributor·

  ·
  2

• Use CCCL memory resource aliases directly

  breakingimprovement
  #1238·bdice opened Sep 29, 2026Contributor·

  ·
  3

• Break UCXX communicator callback ownership cycles

  bugnon-breaking
  #1235·pentschev opened Sep 28, 2026Member·

  ·
  1

  2

  

• [DO NOT MERGE] Fix/python buffer resource spill dir

  #1233·quasiben opened Sep 26, 2026Member·

  ·
  1

• Add disk spilling to AllGather

  #1228·nirandaperera opened Sep 22, 2026Contributor·

  ·
  1

• Include rmm/cuda_stream_view.hpp in the CUDA stream tests.

  improvementnon-breaking
  #1218·miscco opened Sep 15, 2026·

  ·
  2

• Add a pre-partitioned shuffle benchmark

  improvementnon-breaking
  #1197·nirandaperera opened Sep 3, 2026Contributor·

  ·
  2

• Add disk spilling support to the shuffler

  improvementnon-breaking
  #1196·nirandaperera opened Sep 3, 2026Contributor·

  ·
  1

  42

  

• Add test for bind() and CUDA initialization.

  improvementnon-breaking
  #1147·TomAugspurger opened Jul 31, 2026Contributor·

  ·
  4

• Adding reservation aware resource adapter

  improvementnon-breaking
  #1142·nirandaperera opened Jul 25, 2026Contributor·

  ·
  10