All pull requests

Pull requests

• [XPU] Stage liblaunch.a that references try_launch_table

  mainxpu
  #1391·Cross2pro opened Oct 16, 2026Contributor·2 of 3 tasks·

  ·
  1

• [TLE][BACKEND] Add per-copy MUSA TME cache persistence

  COREDOCmainmthreadstle
  #1385·Kafka-Hatsune opened Oct 15, 2026Contributor·

  ·
  5

• [HCU] Support fp64 MMAC (v_mmac_16x16x4_f64) via LLVM intrinsic

  hcumain
  #1384·w1120029931-bit opened Oct 15, 2026·

  ·
  1

• [BUILD] packaging: depend on the Python development headers at runtime

  CI/CDmain
  #1382·shiptux opened Oct 15, 2026Contributor·

  ·

• [FlagMega] Emit the declared grid barrier before the qkv_rope_with_cache sequence advance

  COREfeature/flagmega
  #1374·zhangnju opened Oct 14, 2026Contributor·2 tasks·

  ·

• [FlagMega] Re-infer a slice dtype in the Qwen3 vLLM profile (fix prefill type mismatch)

  COREfeature/flagmega
  #1373·zhangnju opened Oct 14, 2026Contributor·2 tasks·

  ·

• [KMCompiler][TLERaw] tle_raw.call tensor fragment param demo

  COREDOCmaintle
  #1372·PixelLantern opened Oct 14, 2026·

  ·
  2

• ci: make the packaging workflows verify-only

  CI/CDmain
  #1370·shiptux opened Oct 14, 2026Contributor·

  ·

• [BUILD] packaging(deb): strip libtriton.so debug sections again

  CI/CDmain
  #1369·shiptux opened Oct 14, 2026Contributor·

  ·

• [BUILD] packaging: drop the bundled CUPTI static archives

  CI/CDmain
  #1368·shiptux opened Oct 14, 2026Contributor·

  ·

• [TLE][BACKEND] Add set_layout support for PPU

  COREmainpputle
  #1367·YOOkoishi opened Oct 14, 2026·

  ·
  1

• [FlagMega][AMD] Prefill (M>1) path: WMMA GEMM + split-K projections + last-token logits

  amdCOREDOCfeature/flagmegatle
  #1365·zhangnju opened Oct 13, 2026Contributor·

  ·
  2

• [FlagMega][AMD] Add amd-gfx1100 / amd-gfx1201 targets + cooperative megakernel (Qwen3-1.7B)

  amdCOREDOCfeature/flagmegatle
  #1364·zhangnju opened Oct 13, 2026Contributor·

  ·
  2

• [FlagMega][AMD] Lower TLE (pipe / warp_specialize / distributed_barrier) to RDNA

  amdCOREfeature/flagmegatle
  #1363·zhangnju opened Oct 13, 2026Contributor·

  ·
  2

• [FlagMega] Add nvidia-sm89 (Ada) target for the portable decode path

  COREfeature/flagmega
  #1362·zhangnju opened Oct 13, 2026Contributor·

  ·
  2

• [KMCompiler][Ascend] support custom types in kernel(kernel arguments, field access, etc)

  ascendCI/CDCOREDOCtriton_v3.5.x
  #1357·zhongsanming opened Oct 13, 2026Contributor·

  ·

• [TLE][CommonIR] Added CommonIR conversion integration on enflame/hcu/iluvatar/metax/mthreads

  amdCOREDOCenflamehcuiluvatarmainmetaxmthreadsnvidiatle
  #1353·henghengxiedaima opened Oct 12, 2026Contributor·5 tasks done·

  ·
  1

• [QC][TLE][PPU] Keep explicit swizzled TLE layouts in dot operand retagging and avoid replicated chained-dot warp layouts

  COREmainpputle
  #1348·monellz opened Oct 12, 2026Contributor·

  ·
  2

• [HCU][TLE] Add deferred tle.raw and DUSHMEM tutorials.

  COREhcumaintle
  #1345·i3wanna2 opened Oct 10, 2026Collaborator·

  ·

• [CI][Enflame] Enflame vLLM directory

  CI/CDenflamemain
  #1334·ShawnLiuxinghong opened Oct 9, 2026Collaborator·

  ·

• [BACKEND] Fix packed rotary output offsets

  cambriconCOREtriton_v3.2.x
  #1322·Cross2pro opened Sep 30, 2026Contributor·

  ·
  2

• [QC][TLE][MTHREADS] Skip the membar barrier between two async TME copies

  mainmthreadstle
  #1314·monellz opened Sep 29, 2026Contributor·

  ·

• [QC][TLE][MTHREADS] Lower explicit SQMMA to TCE commit/wait groups and allow wgmma_wait pendings

  COREmainmthreadstle
  #1313·monellz opened Sep 29, 2026Contributor·

  ·

• [QC][TLE][MTHREADS] Keep a zero accumulator init out of the use-C flag rewrite

  mainmthreadstle
  #1312·monellz opened Sep 29, 2026Contributor·

  ·

• [MThreads][TLE] Add persistent GEMM pipeline and SPMC synchronization support

  COREmainmthreadstle
  #1294·Zhang-kg opened Sep 27, 2026Contributor·

  ·
  12