All pull requests

Pull requests

• 优化GLM5.2关键算子的几处tile配置

  common-ir-feat-tensor-view
  #133·zhongsanming opened Oct 29, 2026Collaborator·

  ·

• [WarpIR] init warp IR infra

  commonir-triton36-main-syncnvidiatle
  #126·FelixCreator0 opened Oct 21, 2026·

  ·

• [TEST] TensorView success cases for six more Ascend operators

  common-ir-feat-tensor-view
  #125·Artemis6688 opened Oct 20, 2026·

  ·

• [BACKEND] Optimize dense PartitionView tile.load/tile.store

  ascendcommon-ir-feat-tensor-view
  #124·Artemis6688 opened Oct 20, 2026·

  ·

• [DOC][TLE] Document tle.scope in the CV inter-core pipeline doc

  common-ir-feat-tensor-viewDOC
  #122·zhongsanming opened Oct 15, 2026Collaborator·

  ·

• [BACKEND] Add Ascend O0 pipeline compile option

  ascendcommon-ir-feat-tensor-view
  #120·micropuma opened Oct 14, 2026·

  ·

• [TEST] TensorView success cases for ten Ascend operators

  common-ir-feat-tensor-view
  #119·Artemis6688 opened Oct 13, 2026·

  ·

• [LAYOUTS] Add DCU Gluon kernels with layout optimizations

  common-ir-feat-tensor-view
  #116·include-li opened Oct 9, 2026·

  ·

• [Ascend] feat: add plan memory strategy option

  ascendcommon-ir-feat-tensor-view
  #115·yuedenu opened Oct 9, 2026·

  ·

• test(ascend): add TensorView permute and topk cases

  common-ir-feat-tensor-view
  #113·oshy516 opened Oct 8, 2026·

  ·

• [TensorView] Add sparse attention example and padding/L1 tests

  ascendcommon-ir-feat-tensor-view
  #110·pengjh0111 opened Sep 29, 2026·

  ·
  1

• [Review Only] PartitionView: 迁移到 tile.load/tile.store，打通显式 dst_space/src_space

  ascendcommon-ir-feat-tensor-viewtle
  #105·Artemis6688 opened Sep 17, 2026·

  ·

• feat: Integrate PartitionView into CommonIR.

  ascendcommon-ir-feat-tensor-viewtle
  #100·togepro opened Sep 14, 2026Collaborator·

  ·

• add tle.dsa api

  common-ir-feat-tensor-viewDOC
  #99·Yangyang0906C opened Sep 14, 2026Collaborator·

  ·
  8

• benchFA.py

  common-ir-feat-tensor-view
  #97·jia-heng opened Sep 11, 2026Collaborator·

  ·
  14

• [BACKEND] Triton 3.6：重建 GPU CommonIR 并直接转换为 TTGIR

  nvidiatletriton_v3.6.x
  #92·macqueen09 opened Sep 8, 2026Collaborator·

  ·

• add fused_moe

  triton_v3.6.x
  #91·YY4994 opened Sep 3, 2026Collaborator·

  ·

• GPU CommonIR：适配冻结 FLIR main 并保持显式 TileIR 转换

  nvidiatletriton_v3.6.x
  #88·macqueen09 opened Sep 2, 2026Collaborator·

  ·
  2

• [TEST]: add triton-ascend fa perf test as refference

  common-ir-feat-tensor-view
  #81·103yiran opened Aug 25, 2026Collaborator·

  ·

• Route TLE GPU through TileIR conversion

  nvidiatletriton_v3.6.x
  #75·kateyijian opened Aug 18, 2026·

  ·
  9

• Support GPU local pointers and dataflow pipes in TileIR

  tletriton_v3.6.x
  #60·macqueen09 opened Aug 7, 2026Collaborator·

  ·
  11

• [TEST] add flash_attenton_forward and test

  tletriton_v3.6.x
  #56·103yiran opened Aug 3, 2026Collaborator·

  ·