All pull requests

Pull requests

• [Loom] Correct gfx125x packed DOT admission

  #2106·benvanik opened Oct 20, 2026Collaborator·

  ·

• [Loom] Admit portable C++ kernel source forms

  #2105·benvanik opened Oct 20, 2026Collaborator·

  ·

• [Loom] Provision runtime services for selected executable roots

  #2103·benvanik opened Oct 20, 2026Collaborator·

  ·

• [Loom] Coordinate numerical work across model services

  #2083·benvanik opened Oct 20, 2026Collaborator·

  ·

• [HAL] Add libamdf-backed RDNA execution on Linux and Windows

  #2082·benvanik opened Oct 20, 2026Collaborator·

  ·

• [Loom] Multi-target composable FFT research prototype

  #1954·benvanik opened Oct 17, 2026Collaborator·

  ·

• [Loom] Validate native cooperative matrix types against target properties

  #1918·benvanik opened Oct 16, 2026Collaborator·

  ·

• [HAL] Diagnostic: capture AMDGPU completion carrier ownership

  #1762·benvanik opened Oct 15, 2026Collaborator·

  ·

• [Loom] Share the configured compiler through libloomc

  #1682·stellaraccident opened Oct 14, 2026Collaborator·

  ·

• [HAL/AMDGPU] Provision two queues for multi-queue tests

  #1134·Groverkss opened Sep 29, 2026·

  ·

• [HAL/AMDGPU] Map gfx1250 queue resources across XCDs

  #1133·Groverkss opened Sep 29, 2026·

  ·

• [AMDGPU] Support gfx1250-strict ISA across targets

  #1132·Groverkss opened Sep 29, 2026·

  ·

• [HIP] Fall back uncached virtual memory placement

  #1084·AWoloszyn opened Sep 28, 2026Collaborator·

  ·

• [HIP] Classify array storage overflow as exhaustion

  #1083·AWoloszyn opened Sep 28, 2026Collaborator·

  ·

• chore: add security scanning workflows

  #1081·haribabug opened Sep 28, 2026·

  ·
  2

• [HIP] Expand graph construction, execution, and stream capture

  #914·AWoloszyn opened Sep 23, 2026Collaborator·

  ·

• [Experiment] AMDGPU SDMA transfer/compute overlap

  #625·Groverkss opened Sep 17, 2026·

  ·

• [HIP] Implement IPC memory and event handles

  #624·zjgarvey opened Sep 17, 2026Collaborator·3/3·

  ·
  1

• [HRX] Align graph executable block storage

  #623·zjgarvey opened Sep 17, 2026Collaborator·2/3·

  ·

• [HAL/AMDGPU] Support both ROCr queue creation interfaces

  #617·zjgarvey opened Sep 17, 2026Collaborator·1/3·

  ·

• [HAL/AMDGPU] Add fixed masked execution queues

  #444·zjgarvey opened Aug 27, 2026Collaborator·

  ·
  9

• [HRX/CUDA] Fix inverted query results and build the binding under Bazel

  #383·zjgarvey opened Aug 17, 2026Collaborator·

  ·
  1

• [HAL/AMDGPU] Snapshot static indirect dispatch tables at queue issue

  #376·benvanik opened Aug 17, 2026Collaborator·

  ·

• [AMDGPU] Emplace implicit kernargs per field to support unaligned/partial hidden blocks

  #201·zjgarvey opened Jul 20, 2026Collaborator·

  ·

• [HIP] Null the module out-handle before any failed-load error path

  #200·zjgarvey opened Jul 20, 2026Collaborator·

  ·