All pull requests

Pull requests

• fix: SwiGLU formal verifier counterexample

  #243·fiskrt opened Sep 22, 2026·

  ·

• Update tri_inv with varying doubling block_size

  #242·asobczyk opened Sep 16, 2026·

  ·
  1

• chore(kernels): use KERNEL_TASK_TYPE_DEFAULT

  #241·zouzias opened Sep 10, 2026·

  ·
  1

• examples: add mxfp4_matmul_a5

  #239·Mocchibird opened Sep 10, 2026·

  ·

• examples: Fast Hadamard + MXFP4 Quant fused kernel

  #237·Mocchibird opened Sep 3, 2026·

  ·

• examples/ptodsl: add fast_hadamard_vmi

  #230·MirkoDeVita98 opened Aug 18, 2026·

  ·

• feat: basic skill for a2a3-a5 port

  #222·fiskrt opened Aug 4, 2026·

  ·
  2

• fix(causal_conv1d): make K a dynamic parameter

  #212·zouzias opened Jul 3, 2026·

  ·

• Causal conv1d into wheel

  #208·raphael-s-steiner opened Jun 30, 2026·

  ·

• feature(causal_conv1d): import kernel from jit_cpp

  GDN
  #195·zouzias opened Jun 27, 2026·

  ·
  1

• Add paged attention highperf JIT example

  #193·MirkoDeVita98 opened Jun 26, 2026·

  ·
  1

• Update the skills.md for accuracy guidelines

  #175·asobczyk opened Jun 4, 2026·

  ·

• Idealized C-V data exchange kernels for A5

  A5
  #174·learning-chip opened Jun 1, 2026·

  ·
  

• Add Ascend950 pure-vector simulator examples for SiLU and SwiGLU.

  A5
  #172·learning-chip opened May 27, 2026·

  ·
  4

• Run abs kernel on A5 cannsim

  A5
  #167·zouzias opened May 15, 2026·

  ·
  1

  2

• (WIP) Add cpu execution mode for abs kernel (jit mode)

  #160·zouzias opened May 13, 2026·

  ·
  1

• Minimum demo to highlight cross-core sync API differences

  A5
  #158·learning-chip opened May 11, 2026·

  ·

• Add CPU simulation support

  #156·vloncar opened May 8, 2026·

  ·
  1

• WFPY optimized chunk_o

  #153·endrix opened May 5, 2026·

  ·

• [Feat] Implement doubly-stochastic Sinkhorn normalization kernel

  #134·Mocchibird opened Apr 21, 2026·

  ·
  6

• AIV implementation of histogram

  #89·vloncar opened Apr 7, 2026·

  ·
  1

• Chunkwise gated linear attention reaching 60~80 TFLOP/s, with step-by-step optimization records

  #88·learning-chip opened Apr 5, 2026·

  ·
  30

• compare host vs device-side chunk metadata computation

  #84·learning-chip opened Apr 1, 2026·

  ·
  1

• (cmake) use bisheng

  Infra
  #83·zouzias opened Apr 1, 2026·

  ·
  1

  5

• (pyproject) refactor extra-index-url and put it inside pyproject.toml

  #54·zouzias opened Mar 16, 2026·

  ·
  1