All pull requests

Pull requests

• fix(pto): accept tpop-derived views in set_validshape verifier

  #1725·jimmychou0 opened Nov 2, 2026Collaborator·

  ·
  1

• feat(pto): lower native A5 DMA L2 hint attributes

  #1722·Little-oil opened Oct 31, 2026·

  ·

• fix(vmi): support shared partial masks across mixed-width comparisons

  #1705·chenshengxin2026 opened Oct 28, 2026Collaborator·

  ·
  1

• fix(vecscope): preserve hazards with unknown memory footprints

  #1703·chenshengxin2026 opened Oct 28, 2026Collaborator·

  ·
  2

• fix(tilelib): use physical NZ pitches in A5 acc2gm templates (#1690)

  ci-slow
  #1698·jimmychou0 opened Oct 27, 2026Collaborator·

  ·
  1

  1

• fix(lowering): preserve repeated MX scale bindings

  #1692·chenshengxin2026 opened Oct 26, 2026Collaborator·

  ·
  1

• feat(a5): support RowPlusOne NZ for direct FP4-to-FP8 conversion

  #1685·parsifal-wang-1995 opened Oct 26, 2026·

  ·

• fix(pto): allow byte-preserving low-precision Mat views in pto.treshape

  #1681·jimmychou0 opened Oct 23, 2026Collaborator·

  ·
  1

• feat(costmodel): select preload before compiling native VF candidates

  #1676·TaoTao-real opened Oct 23, 2026Contributor·

  ·

• fix(insert-sync): prove GM column slices disjoint instead of barriering

  #1670·FangRui0 opened Oct 22, 2026Collaborator·

  ·
  1

• fix(emitc): preserve scalar i1 widening and float casts

  #1666·zhangstevenunity opened Oct 22, 2026Collaborator·

  ·
  1

• feat: add an option to emit device object file

  #1660·Melikano opened Oct 21, 2026Contributor·

  ·

• fix(emitc): preserve CFG parallel-copy semantics

  ci-slow
  #1642·zhangstevenunity opened Oct 20, 2026Collaborator·

  ·
  1

  2

• fix(vmi): lower small-shape f16 vexpdif through the f32 path (#1626)

  ci-slow
  #1641·jimmychou0 opened Oct 20, 2026Collaborator·

  ·
  1

  1

• feat(a5): support direct FP4 ND/NZ and safe MX pipeline synchronization

  #1637·parsifal-wang-1995 opened Oct 19, 2026·

  ·

• chore: drop accidentally committed workspace metadata

  #1576·mouliangyu opened Sep 23, 2026Collaborator·

  ·

• fix(vpto): optimize contiguous BF16/F32 casts

  #1574·Zhendong404 opened Sep 22, 2026Collaborator·

  ·
  1

• perf(sync): eliminate redundant rotating-slot dependencies

  #1572·zhangstevenunity opened Sep 21, 2026Collaborator·

  ·

• feat(costmodel): add PTOAS exchange and A5 mechanism acceptance

  ci-slow
  #1554·TaoTao-real opened Sep 18, 2026Contributor·

  ·
  1

• feat(ptodsl): expose missing A5 SIMD interfaces

  #1553·TaoTao-real opened Sep 18, 2026Contributor·

  ·
  1

• fix simt fp contraction semantics

  ci-slow
  #1550·ft1593000 opened Sep 17, 2026Collaborator·

  ·
  1

• vmi: layout decision layers + packed-store mask normalize pass

  #1539·mouliangyu opened Sep 16, 2026Collaborator·

  ·

• feat(mad): runtime unit_flag / acc-init / gemv / bias-init operands

  ci-slow
  #1525·jimmychou0 opened Sep 14, 2026Collaborator·

  ·
  1

• fix(sync): balance multi-tile prefetch event boundaries

  #1524·zhangstevenunity opened Sep 12, 2026Collaborator·

  ·
  1

  1

• feat(vpto): add pipe and cube bridge lowering

  #1505·kuri780 opened Sep 10, 2026Contributor·

  ·
  6