All pull requests

Pull requests

• Add Hexagon HMX pqs8-qc8w/qc4w GEMM and pqs8-qc8w IGEMM microkernels

  #12015·copybara-service[bot] opened Nov 13, 2026Bot·

  ·

• SSE, AVX, and F16C qd8-f32-qc2w and qd8-f16-qc2w GEMM use uint2

  #12014·copybara-service[bot] opened Nov 13, 2026Bot·

  ·

• Include num_k_dims, tile_k, tile_n, and max_block_n in

  #12012·copybara-service[bot] opened Nov 12, 2026Bot·

  ·

• Vectorize YNNPACK bf16 and fp8_e5m2 widening casts on x86 and HVX.

  #12011·copybara-service[bot] opened Nov 12, 2026Bot·

  ·

• Commute A @ B^T into (B @ A^T)^T during subgraph fusion when B is dynamic and M == 1 so B is streamed in-place as LHS without packing.

  #12008·copybara-service[bot] opened Nov 12, 2026Bot·

  ·

• Coalesce full tile_n blocks into a single transpose_fn call in packer::pack when packed panels are contiguous along each output row, and strip-mine n to 2 * N rows with L2 software prefetching in 16-row by 64-byte transpose tiles to avoid DTLB and L1 thrashing on large n.

  #12005·copybara-service[bot] opened Nov 11, 2026Bot·

  ·

• Skip dynamic pack_b in should_pack_b when tile_k == 1 and each microkernel panel spans whole cache lines, either unconditionally for single-panel tile_n-aligned B or up to 64 M-blocks when B fits within the per-core L2 cache budget. Keep m outermost in define_dot when A has an internal producer and unpacked B has none.

  #12004·copybara-service[bot] opened Nov 11, 2026Bot·

  ·

• Run INT4 FullyConnected with INT8 activations on ML Drift.

  #12000·copybara-service[bot] opened Nov 11, 2026Bot·

  ·

• Fixes failure handling when XNNPACK creates a batch matrix multiply operator

  #11995·Michael-McDonagh-ARM opened Nov 11, 2026·

  ·
  1

• Fix relative repository roots in compile command comparison

  #11992·NansKong opened Nov 11, 2026·

  ·

• Parallelize f32->qd8/qdu8 convert across a row when batch < threads

  #11990·gadgilrajeev opened Nov 11, 2026·

  ·

• Replace Python build identifier generator with POSIX shell sha256sum.

  #11987·copybara-service[bot] opened Nov 11, 2026Bot·

  ·

• Add YNNPACK runtime backend for text decoder models.

  #11986·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Add YNNPACK conv, pool, select, and transformer composite lowerings.

  #11985·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Support affine blockwise quantization and row-only gather on YNNPACK.

  #11983·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Optimize Qwen3.5-27B CPU memory usage, compile latency, and throughput.

  #11984·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Unpack GGUF sub-block, 5-bit, and IQ4 LUT weights losslessly.

  #11982·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Keep YNNPACK worker threads awake between short parallel loops.

  #11981·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Optimize Gemma 4 (E2B, E4B, 26B-A4B, 31B) prefill and decode performance across ML Drift GPU (Metal & OpenCL), XNNPACK CPU, YNNPACK CPU, and LiteRT-LM Tensor API executor:

  #11980·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Speed up YNNPACK quantized LLMs and make graph compilation stack-safe.

  #11979·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Bring up Qwen3.5-27B and dense Qwen3.5 family presets on Tensor API.

  #11978·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Lower GGUF Q8_0, Q4_1, and Q4_K weights natively on ML Drift GPUs.

  #11977·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Support Q5_K/Q6_K GGUF weights, lazy unpacking, and GPU weight cache.

  #11976·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Add 9p minmax s8 and u8 maxpool microkernels for AVX2 (u32), AVX-512

  #11975·copybara-service[bot] opened Nov 10, 2026Bot·

  ·

• Disable ARM SME ACLE microkernels by default behind build option.

  #11974·copybara-service[bot] opened Nov 10, 2026Bot·

  ·