All pull requests

Pull requests

• [Kernel][FA][AOT][Feature][Dialect][DSL][ROCDL][Perf][Doc][Test] Add …

  #1223·dimitri91209 opened Oct 1, 2026·

  ·

• [Build] Fall back to HIP runtime from rocm-sdk core wheels

  #1222·apinge opened Oct 1, 2026·

  ·
  1

• [dont merge] compare latest base main llvm

  #1221·jli-melchior opened Oct 1, 2026Collaborator·

  ·

• [Bugfix][Test] Fix atomic_fmin/fmax Out left at +inf on side-stream launch

  #1220·dimitri91209 opened Oct 1, 2026·

  ·

• feat(conv3d): align bf16 implicit GEMM with aiter #5370

  #1217·huizzhan opened Sep 30, 2026·

  ·

• [Kernel] Add dense Flash Attention for gfx1100

  #1216·vlluvia opened Sep 30, 2026Contributor·

  ·
  1

• [compiler][opt] add schedule_bank api for gfx1250

  #1215·jli-melchior opened Sep 30, 2026Collaborator·

  ·

• [Kernel][gfx1250] Add the GLM-5 MonoKernel

  #1214·jhinpan opened Sep 29, 2026Collaborator·

  ·

• [fix] Correct zipped_divide tiler hierarchy handling

  #1212·Peter9606 opened Sep 29, 2026Contributor·

  ·
  1

• [CI] Run full cooperative matrix before PyPI releases

  #1210·coderfeli opened Sep 28, 2026Collaborator·

  ·

• [JIT] Actionable diagnostic for dynamic layout field overflow

  #1194·mustafayildirim opened Sep 25, 2026·

  ·
  1

  8

• [LayoutLowering] Ignore size-one leaves in contiguous-segment detection

  #1193·mustafayildirim opened Sep 25, 2026·

  ·
  1

  2

• chore: add security scanning workflows

  #1192·haribabug opened Sep 24, 2026·

  ·

• [Fix][Expr] Honor allocate(..., alignment=) for static shared allocat…

  #1191·xudoyuan opened Sep 24, 2026Collaborator·

  ·

• [llvm] Extend LLVM for FlyDSL end-to-end performance work

  #1190·Phil-amd opened Sep 24, 2026Member·

  ·
  1

  

• [Compiler][Debug] Add a compile path for hand-edited assembly

  #1188·Phil-amd opened Sep 24, 2026Member·

  ·
  1

  

• fix(coop): preserve declared operator element types

  #1185·sjfeng1999 opened Sep 23, 2026Collaborator·

  ·
  3

• [fix] Support waves_per_eu min,max range for full VGPR utilization

  #1184·jli-melchior opened Sep 23, 2026Collaborator·

  ·
  1

• [Enh] Accept LLVM pointer types in ffi declarations

  #1179·LWenH opened Sep 22, 2026Contributor·

  ·

• [dont merge][llvm] cherry commits for unclausevmem patch

  #1166·jli-melchior opened Sep 18, 2026Collaborator·

  ·

• [DSL][Runtime][Kernel] Add in-kernel wave tracing (fx.ktrace) for gfx942/gfx950

  #1162·Phil-amd opened Sep 18, 2026Member·

  ·
  23

  

• feat(review): harden review engine for unattended runs

  #1157·jhinpan opened Sep 17, 2026Collaborator·

  ·

• gfx950 flex attention with user score/mask mods

  #1155·RichardChamberlain1 opened Sep 17, 2026Contributor·

  ·
  

• [LLVM] bump llvm which improve the sequence for initial unclaused vmem

  #1145·jli-melchior opened Sep 16, 2026Collaborator·

  ·

• [Kernel][Perf] Align gfx950 BF16 D128 attention with OPUS

  #1139·coderfeli opened Sep 15, 2026Collaborator·

  ·
  15