All pull requests

Pull requests

• Support Thrust and libcu++ benchmarks in PR benchmark CI

  #12066·bernhardmgruber opened Oct 18, 2026Contributor·

  ·
  1

  3

• [Thrust] Remove unused thrust/detail/event_error.h

  #12065·bernhardmgruber opened Oct 18, 2026Contributor·

  ·
  1

• [STF] Keep callback TLS key alive during static teardown

  #12064·fallintoplace opened Oct 18, 2026Contributor·

  ·
  4

• Reject unsupported tie_break requests in cub::DeviceTopK

  #12063·elstehle opened Oct 18, 2026Contributor·

  ·
  3

• [Docs] Fixes tied-element counts in the top-k requirements docs

  #12062·elstehle opened Oct 18, 2026Contributor·

  ·
  1

• [CUB][Test-only] Adds regression tests for narrow integral k truncating the segment size and writing too few items

  #12061·elstehle opened Oct 18, 2026Contributor·

  ·
  2

• Recoups performance for +/-0.0 float normalization in DeviceBatchedTopK

  #12060·elstehle opened Oct 18, 2026Contributor·

  ·
  1

  6

• Feature/device memcpy opts

  #12059·mfranzrebsal opened Oct 17, 2026·2 of 3 tasks·

  ·
  5

• c.parallel.v2: libnvccCompileProgramToHostIR — the host translation unit as LLVM IR

  #12058·caugonnet opened Oct 17, 2026Contributor·

  ·
  5

• Fix 'rapids_add_test could not infer a target to install' warnings

  #12056·Jacobfaib opened Oct 16, 2026Contributor·2 tasks·

  ·
  3

  

• Tune DeviceRadixSort pairs for sm100 with 64-bit keys, values and offsets

  #12053·gonidelis opened Oct 16, 2026Member·

  ·
  4

• Tune deterministic DeviceReduce for sm100 with 64-bit offsets

  #12052·gonidelis opened Oct 16, 2026Member·

  ·
  4

• [CUB] Allow vectorized loads of fancy iterators backed by contiguous storage

  #12049·miscco opened Oct 16, 2026Contributor·3/3·

  ·
  8

• Infer launch bounds for exact cuda.coop block specializations

  #12048·tpn opened Oct 16, 2026Contributor·5/11·

  ·
  1

• Add cuda.coop Block and Warp Scan with Numba-CUDA-MLIR

  #12047·tpn opened Oct 16, 2026Contributor·4/11·

  ·
  1

• Add cuda.coop group operations and Numba-CUDA-MLIR Reduce

  #12046·tpn opened Oct 16, 2026Contributor·3/11·

  ·
  1

• Collect third-party compile-time traces and single-dataset reports

  #12045·griwes opened Oct 16, 2026Contributor·

  ·
  30

• [CUB] Add machinery for striped vectorized loads

  #12043·miscco opened Oct 16, 2026Contributor·2/3·

  ·
  9

• [libcu++] Create machinery for vectorized loads

  #12042·miscco opened Oct 16, 2026Contributor·1/3·

  ·
  16

• c.parallel.v2: make libnvcc follow the build host (aarch64), fix PTX ISA for sm_103, add arm64/GB300 CI

  #12041·caugonnet opened Oct 16, 2026Contributor·2 tasks done·

  ·
  12

• Optimize isinf, isfinite, isnormal and fpclassify for double

  #12040·samridh10exe opened Oct 16, 2026·2 tasks done·

  ·
  18

• cuda.coop: shared-storage reservations [typed experiment]

  #12039·tpn opened Oct 16, 2026Contributor·

  ·
  1

• cuda.coop: scratch-storage walkthrough [typed experiment]

  #12038·tpn opened Oct 16, 2026Contributor·

  ·
  1

• cuda.coop: exact launch bounds [typed experiment]

  #12037·tpn opened Oct 16, 2026Contributor·

  ·
  1

• cuda.coop: Batched Warp Reduction [typed experiment]

  #12036·tpn opened Oct 16, 2026Contributor·

  ·
  1