All pull requests

Pull requests

• Skip FP16 tensor core tuning on Mali GPUs

  #1268·jopdorp opened Oct 16, 2026·

  ·
  1

• OpenCL: give each NN server thread its own command queue

  #1267·Sagebati opened Oct 15, 2026·

  ·

• Pass the macOS deployment target to swiftc in Metal builds

  #1264·Slymaster opened Oct 13, 2026·

  ·

• Add kata-get-params and kata-set-params to known GTP commands

  #1263·Slymaster opened Oct 11, 2026·

  ·

• Allow negative humanSLChosenMovePiklLambda to prefer worse moves

  #1262·ChinChangYang opened Oct 10, 2026Contributor·

  ·

• Fix integer division and boundary errors in scored-SGF near-end detection

  #1259·22nsuk opened Oct 5, 2026·

  ·

• Fix visit-cap test invariants for terminal leaves

  #1258·22nsuk opened Oct 5, 2026·

  ·

• Add mirai to the GUI list

  #1257·zhaob1n opened Oct 4, 2026·

  ·

• Fix zobrist hash update for captured stones in python Board

  #1256·zliu1022 opened Sep 26, 2026·

  ·

• add support for loongarch64

  #1254·mogoweb opened Sep 14, 2026·

  ·

• [AMD NPU] Add a RyzenAI backend that runs the network on the AMD NPU

  #1240·Looong01 opened Aug 18, 2026Contributor·

  ·
  1

• Add MIGraphX backend for AMD GPUs (ROCm) — 2.94x over OpenCL on MI300X

  #1235·zhihuidu-amd opened Aug 12, 2026·

  ·
  49

• Serialize complete TensorRT handle initialization across GPU threads

  #1230·zsqdx opened Aug 6, 2026Contributor·

  ·

• Optimize TensorRT CUDA Graph/concurrency and add benchmark auto-tuning

  #1219·zsqdx opened Jul 28, 2026Contributor·

  ·

• Add Human-SL even-game ELO rank ladder configs (gtp_human<rank>.cfg, 8d→25k)

  #1209·ChinChangYang opened Jun 26, 2026Contributor·

  ·
  6

• TensorRT 11 support (Blackwell / strongly-typed networks), incl. FP16

  #1206·gweber opened Jun 6, 2026·

  ·

• Tune NN server threads and max batch size in genconfig

  #1201·zsqdx opened May 26, 2026Contributor·

  ·
  2

• Add MLX backend for Apple Silicon (GPU + ANE)

  #1199·ChinChangYang opened May 23, 2026Contributor·

  ·
  2

• Configure clang-tidy static analysis in CMake

  #1190·KvanTTT opened Apr 25, 2026Contributor·

  ·
  2

• [Intel NPU/GPU] Add Windows & Linux Intel NPU & GPU support

  #1171·Looong01 opened Mar 16, 2026Contributor·

  ·
  40

• Support ONNX models in TensorRT backend

  #1139·hzyhhzy opened Dec 31, 2025Contributor·

  ·

• Cuda graph implementation on trt backend

  #1071·zjuwyz opened Jun 6, 2025·

  ·
  7

• Add human SL value mixing

  #1004·lightvector opened Dec 3, 2024Owner·

  ·

• Eval cache

  #992·lightvector opened Oct 8, 2024Owner·

  ·
  3

• Android support

  #978·jopdorp opened Aug 23, 2024·

  ·
  1