Issues

• [Feature]: follow-up to PyTorch allocation provider to accept already rendezvoused tensors

  iris
  ROCm/iris#577 · Filter by author drprajap opened on Oct 8, 2026

• [Feature]: Add rooted collectives broadcast, scatter, gather, reduce or pull all-to-all

  benchmarksgluoniris
  ROCm/iris#561 · Filter by author drprajap opened on Sep 23, 2026

• [Bug]: 00_load and 02_all_load validate against arange but fill the source with ones/zeros — validation can never pass

  benchmarksbugexamplesiristest
  ROCm/iris#558 · Filter by author Sophia-Thickums opened on Sep 23, 2026

• [Bug]: 09_gemm_one_shot_all_reduce aborts on power-of-two CU counts (gemm_sms auto-detect equals total_sms)

  bugexamplesinvalidiris
  ROCm/iris#557 · Filter by author Sophia-Thickums opened on Sep 23, 2026

• [Bug]: SIGSEGV at multi-rank iris.iris() init on mixed-architecture AMD GPUs

  bugiristest
  ROCm/iris#556 · Filter by author Sophia-Thickums opened on Sep 22, 2026
  2

• [Feature]: Formalize allocator-agnostic symmetric tensor address translation

  iris
  ROCm/iris#546 · Filter by author drprajap opened on Aug 25, 2026
  2

  

• HSA fault on 8xMI300X all_gather at msg_per_rank >= 768 MiB (bf16, both persistent and partitioned variants)

  bugcoreiris
  ROCm/iris#532 · Filter by author ryanswann-amd opened on May 8, 2026

• Fix pre-existing kernel issues found during CCL split

  gluoniris
  ROCm/iris#524 · Filter by author mawad-amd opened on Apr 28, 2026

• test_amdsmi_query_helpers_do_not_manage_lifecycle fails in multi-rank CI

  bugiristest
  ROCm/iris#517 · Filter by author mawad-amd opened on Apr 24, 2026
  

• as_symmetric: deduplicate repeated imports of the same tensor

  iris
  ROCm/iris#499 · Filter by author mawad-amd opened on Apr 8, 2026

• get_device_context tensor goes stale if heap_bases change after init

  bugiris
  ROCm/iris#467 · Filter by author mawad-amd opened on Mar 19, 2026

• all_reduce_one_shot / all_reduce_two_shot use hardcoded lock value 1

  bugiris
  ROCm/iris#465 · Filter by author aamarnat opened on Mar 19, 2026

• matmul_all_reduce_preamble mixes collective allocation with per-call preparation

  bugiris
  ROCm/iris#464 · Filter by author aamarnat opened on Mar 19, 2026
  2

• Implement ring attention

  iris
  ROCm/iris#449 · Filter by author mawad-amd opened on Mar 12, 2026
  

• Modernize the flash decode example

  iris
  ROCm/iris#447 · Filter by author mawad-amd opened on Mar 12, 2026
  

• Implement power-of-two vmem allocator

  iris
  ROCm/iris#440 · Filter by author mawad-amd opened on Mar 9, 2026
  

• Introduce GEMM+AllScatter ops

  iris
  ROCm/iris#420 · Filter by author mawad-amd opened on Mar 4, 2026
  

• VMem allocator and atomics

  bugiris
  ROCm/iris#418 · Filter by author mawad-amd opened on Mar 4, 2026
  

• Gluon All2All

  gluoniris
  ROCm/iris#416 · Filter by author mawad-amd opened on Mar 4, 2026
  

• Test Suite Optimization Phase 1: Reduce Parametrization in Critical Path

  examplesiristest
  ROCm/iris#373 · Filter by author mawad-amd opened on Feb 15, 2026
  

• Unified benchmarking harness

  benchmarksexamplesiris
  ROCm/iris#367 · Filter by author mawad-amd opened on Feb 7, 2026
  

• Restructure the CI actions to take advantage of idle GPUs

  iristest
  ROCm/iris#363 · Filter by author mawad-amd opened on Feb 6, 2026
  

• Implement test suite optimization

  iristest
  ROCm/iris#355 · Filter by author mawad-amd opened on Feb 4, 2026
  

• Review all existing tests

  duplicateexamplesiristest
  ROCm/iris#353 · Filter by author mawad-amd opened on Feb 4, 2026
  

• Introduce other parameter

  iris
  Feature

  ROCm/iris#342 · Filter by author neoblizz opened on Feb 3, 2026